16th WCEAM Proceedings 3031254473, 9783031254475

Table of contents :
Foreword
Preface
Contents
Asset Management and Decision Support System
A Decision-Making Framework for Selecting an Optimum Package of Maintenance Improvement Projects in a Hospital
1 Introduction
2 Maintenance Multi-criteria Audit
3 Decision Framework for Selection of Optimum Package of Maintenance Improvement Projects
4 Selection of Projects for Maintenance Department
5 Conclusions
References
A Preliminary Implementation of Data-Driven TPM: A Real Case Study
1 Introduction and Background
2 Data-Driven TPM Approach
2.1 Data Collection and Pre-processing
2.2 Data Analytics
2.3 Decision Making
3 Data-Driven TPM Implementation
4 Conclusions
References
Assets’ Reliability Management Model for a Decision Making in Different Operational Contexts
1 Introduction
2 Methodology
3 Case Study
4 Conclusions
References
Realizing Sustainable Value from Engineering Innovation Ecosystems in EURope’s Outermost Regions
1 Introduction
2 Problem
3 Research Objective
4 Hypothesis and Thought Experiment
5 Research Method
6 Results
7 Discussion and Conclusions
References
RelOps – A Whole-of-Organisation Approach for Reliability Analytics
1 Introduction
2 Current Practice
3 Proposed Approach
4 Discussion
References
Methods for Comparing Asset Portfolio Reliability
1 Introduction
2 Background
2.1 Methods for Prioritizing Investment or Asset Management Activities
2.2 Criticality and Reliability Assessment Methods Comparison
3 Methodology
4 Results
5 Discussion
5.1 Contribution of Industry 4.0 and Artificial Intelligence
6 Conclusion
References
Industry 4.0 Tools and Its Impact in Asset Management
Digital Transformation in Maintenance
1 Introduction
2 Emerging Asset Management
3 The New System Architecture for Intelligent Asset Management
4 The Emerging Technologies
5 New DMM Framework
6 New DMM Framework
7 Conclusion
References
On the Definition of Requirements for a Digital Twin. A Case Study of Rolling Stock Assets
1 Introduction
2 The Digital Twin and Their Requirements
3 The Case Study: DT of Axle Bearings in Trains for Maintenance.
4 Conclusions
References
Review of Asset Digitalization Models in the Context of Intelligent Asset Management and Maintenance
1 Introduction
2 Review of References Related to Digital Asset Modelling
2.1 Building Information Model (BIM)
2.2 Asset Information Model (AIM)
2.3 Digital Twin (DT)
2.4 RAMI 4.0 and Asset Administration Shell (AAS)
2.5 Cognitive Digital Twin (CDT)
3 Comparison of References
4 Conclusion
References
An Immersive Virtual Reality Platform for Enablement and Assessment of Human-Robot Interactions for Intelligent Asset Management
1 Introduction
2 Related Research
3 Methodology
3.1 Development of HRI Scenarios
3.2 Development of the VR Platform
4 Results and Validation
4.1 Scenario 1: Manual Assembly Task
4.2 Scenario 2: Robot Assisted Assembly Task
4.3 Validation
4.4 Results
5 Discussion
6 Conclusion
References
Exploring Augmented Reality Applications to Support Maintenance Management in Hydroelectric Power Plants
1 Introduction
2 AR Technology within Maintenance Management
3 Methodology
4 Results and Discussion
5 Conclusions
References
Smart Water Dam Transformation in Industry 4.0
1 Introduction
2 Issues Faced by Dam Managers Who Rely on Traditional Catchment Management
2.1 Dynamic Asset Degradation
2.2 Variational Water Quality and Quantity
3 Transforming from Conventional Dam Management to an Industry 4.0-Driven Smart Catchment Environment
3.1 Ensuring the Dam Health and Sustenance
3.2 Ensure Water Quality and Quantity
4 Conservation of Natural Resources in Catchment Area
5 Conclusion
References
Can Industry 4.0 Keep Its Promises? A Literature-Based Comparison of Expectations and Experience
1 Introduction
2 Methods
3 What Topics Are Discussed in Industry 4.0
3.1 Industrial Revolution
3.2 Manufacture
3.3 Smart Manufacturing
3.4 Automation
3.5 Smart Factory
3.6 Cyber Physical Systems
3.7 Internet of Things
3.8 Embedded Systems
3.9 Big Data
3.10 Decision Making
3.11 Relationships Between the Top 10 Keywords
4 What is Expected of Industry 4.0
4.1 Domain Cluster
4.2 Execution Cluster
4.3 Technology Cluster
5 What is Experienced in Industry 4.0
6 Summary and Conclusion
References
Monitoring, Diagnostics and Prognostics for Smart Maintenance
Dynamic Maintenance Management Approach Based on Real Time Monitoring and Artificial Intelligence Using Digital Twins
1 Introduction
2 Smart Maintenance Based on RTM, RCM, CBM
3 Internet of Trains
3.1 Real-Time Monitoring System: TSMART
3.2 Machine Learning and Predictive Maintenance
4 Digital Twins Applied to Smart Maintenance
5 Dynamic Maintenance Scheduling and Management
6 Automatic Vehicle Inspection. TALVI
7 Conclusions
References
Heat Pumps Smart Asset Management Implementation Through Virtual Sensors
1 Introduction
2 Description of Facilities and Equipment
2.1 Equipment Used
3 Methodology
3.1 Basic Virtual Sensors
3.2 Derived Virtual Sensors
4 Summary and Conclusion
References
Driving Port Efficiency Through 5G-Enabled Condition Monitoring of Quay Cranes
1 Introduction
2 Driving Port Efficiency Through 5G-Enabled Condition Monitoring of Quay Cranes
2.1 Asset Identification and Hierarchisation in Systems
2.2 Failure Mode Analysis Associated with Asset Hierarchy and Physical Symptoms Associated
2.3 Define Monitoring Variables and Failure Mode State
2.4 Sensor Identification to Capture Physical Dimensions
2.5 Algorithm Development for Anomaly Detection and Failure Prediction
2.6 Data Pipeline for Algorithms and Applications Design
2.7 Sensor Installation on Physical Assets and 5G Testbed for Use Case Testing
2.8 Go On-Line
3 Concluding Discussion and Further Work
References
Remote Data Collection Motivational Drivers, Challenges, and Potential Solutions in Industrial SME Companies
1 Introduction
2 Methodology
3 Motivation for data collection
4 Selecting Data to Collect
5 Implementation Challenges and Potential Solutions
6 Conclusions
References
The Effect of Knowledge Based Feature Extraction on Failure Detection of Control Surface Failures of Fighter Aircraft
1 Introduction
2 Methods for Failure Detection
2.1 Knowledge Based Feature Extraction Process
2.2 Failure Detection Algorithms
3 Results of Failure Detection with Varying Feature Extraction
4 Conclusions
References
Advanced Maintenance of Distribution Assets Through the Application of Predictive Techniques Using GE'S APM System: Real Case in a Spanish DSO
1 Introduction
2 GE’s APM
3 APM Implementation Process and Results
4 Conclusions
References
Challenges on an Asset Health Index Calculation
1 Introduction
2 AHI Modelling Methodology
3 Case of Study
3.1 Application of the Methodology Proposed
3.2 Results
4 Conclusions
References
Asset Life Cycle Management
An Integrated Framework for Efficient Asset Life Cycle Costing in Case of Incomplete Historical Data
1 Introduction
2 Gas Turbine LCC Analysis Principle
2.1 The Acquisition Cost of Gas Turbine
2.2 The Ownership Cost of Gas Turbine
2.3 The Disposal Cost of Gas Turbine
3 Time Value of Costs
4 Combining Data Repository
5 Practical Implementation
6 Discussion
7 Sensitivity Analysis on Key Cost Driver
8 Conclusion
References
Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management
1 Introduction
2 Economic Health Index of the Asset
2.1 Calculation of the Economic Health Index (EHI)
3 Cost of Life Cycle Analysis Methodology (LCCA)
3.1 Definitions of the Alternatives to Be Analysed
3.2 Time Horizon of the Analysis
3.3 Selection of Cost Categories to Consider
3.4 Definition of the Cost Model
3.5 Data Collection for Analysis
3.6 Calculation of Life Cycle Costs
3.7 Risk Estimation of Each Alternative
3.8 Comparative Evaluation of Alternatives
4 Illustrative Case Study
4.1 Application of LCCA Methodology
5 Final Conclusions
References
Use Proposal of the Asset Health Index in the Public Health Sector. A Case Study in the Health Systems of the Republic of Costa Rica
1 Introduction
2 Scope
3 Application of the Asset Health Index (AHI) Calculation
3.1 Conclusions and Analysis of Results
References
Visual Quality Control via eXplainable AI and the Case of Human in the AI Loop
1 Introduction and Related Work
2 Methods
2.1 Industrial Case
2.2 Data Sets
2.3 Classification Models
2.4 Explainability and Interpretability Approaches
3 Results and Discussion
4 Conclusion
References
Start/stop Cost Evaluation of a Francis Turbine Runner Based on Reliability
1 Introduction
2 Start/Stop Cost Calculation: Non-reliability-Based Model
3 Start/Stop Cost Calculation of Francis Runner: Reliability-Based Model
4 Conclusion
References
Asset Management in the Industrial Sector
Machine Learning Supporting Maintenance Management: A Case Study in Scaffolding Industry's Servitization Process
1 Introduction
2 Motivation and Research Objectives
3 Industrial Use Case
3.1 Departing Situation and Proposed Approach
3.2 Data Description
3.3 Use Case Development
4 Conclusions
References
A Prescriptive Analysis Tool for Improving Manufacturing Processes
1 Introduction
2 Prescriptive Analysis Tool
3 Case Study
4 Conclusions
References
Managing Assets to Facilitate Circularity and Sustainability of Food Systems
1 Introduction
2 Food Systems Within Circular Economy and SDG 2
3 Managing assets within food systems
4 Concluding Remarks
References
A Reference Model for Engineering Asset Management Excellence
1 Introduction
2 Engineering Asset Management and Excellence
3 A Model for EAM Excellence
4 Case Study
5 Final Remarks
References
The Exploration of Digitalization and Digitalization Indicators Within the Scope of Asset Management
1 Introduction
2 Methodological Considerations
3 Theoretical Position
3.1 Engineering Asset Management
3.2 Digital Transformation in Manufacturing
3.3 Digitalization of EAM
4 Analysis
5 Conclusion
References
Audit Model for Asset Management, Maintenance and Reliability Processes: A Case Study Applied to Pulp Mill Sector
1 Introduction
2 General Background
2.1 Description of the Tool Used for the Audit
2.2 Background of the Audited Unit
3 General Results of the Audit Process
4 Specific Results
4.1 Analysis Behind the Results Achieved on Phase 5: Preventive Plan, Schedule and Resources Optimization
4.2 Analysis Behind the Results Achieved on Phase 6: Maintenance Execution Assessment and Control
4.3 Analysis Behind the Results Achieved on Phase 7: Asset Life Cycle Analysis and Replacement Optimization
4.4 Analysis Behind the Results Achieved on Phase 8: Continuous Improvement and New Tech
5 Conclusions and Future Work
References
Human Dimensions and Asset Management Performance
Using Wearable Sensors to Form a Relationship Between Driver Stress and Aggressive Driving Habits
1 Introduction
2 Study Method
3 Analysing EDA and G09 Data
4 Conclusion
References
Reshaping Industry Job Profiles to Better Meet Future Asset Management Needs
1 Introduction
2 Related Research
2.1 Occupations Databases
2.2 Projects
2.3 Publications
3 Profile Selection and Future Skill Definition Process
4 Definition of the Skills and Competences to be Developed by the Future Profiles
5 Sample Future Job Profile Related to Asset Management
6 Conclusions
References
Methods for the Criticality Assessment of Intangible Assets in a Knowledge Management Process
1 Introduction
2 Process Definition for the Management of Knowledge
2.1 Context and Interested Parties Analysis
2.2 Knowledge Areas Inventory
2.3 Knowledge Areas Assessment
2.4 Management Check-Point
2.5 Knowledge/Capabilities Management
2.6 Monitoring
3 Criticality Assessment Based on Risk Analysis
4 Criticality Assessment Based on AHP
4.1 AHP Brief Description
4.2 Case Study
5 Conclusions
References
What is Smart Maintenance in Manufacturing Industry?
1 Introduction
1.1 Smart Maintenance
1.2 Human Errors
2 State of Practice
2.1 Readiness in Manufacturing Industry
2.2 Considering the I-P Interval
2.3 Human Factors
3 Discussion
References
Audit Models for Asset Management, Maintenance and Reliability Processes: A Case Study Applied to the Desalination Plant
1 Introduction
2 General Background
2.1 Audit Models
2.2 Background of the Audited Unit
3 Audit Results
3.1 AMORMS Audit Result
3.2 AMS-ISO 55001 Audit Result
4 Conclusions and Future Work
References
Impact of Information Digitalization on Asset Availability - an Empirical Study
1 Introduction
2 Information Digitalization for Asset Availability
2.1 Availability
2.2 Digitalization within Asset and Maintenance Management
3 Research
3.1 Approach
3.2 Empirical Setting
3.3 Findings
4 Summary and Outlook
References
Infrastructure Asset Management
Linking Organisation Objectives with Asset Information Requirements for Highway Infrastructure Projects
1 Background
2 Methodology
3 Case Study
4 Conclusions
References
A Methodology for Ensuring Strategic Alignment of Railway Infrastructure Asset Management Processes
1 Introduction
2 Railway Infrastructure AM Current Frameworks and Developments
3 Proposed Methodology
4 Application Case
4.1 Company Overview and Propaedeutic Activities for Implementing the Methodology
4.2 Application and Results
5 Conclusions
References
Hierarchy Definition for Digital Assets. Railway Application
1 Introduction
2 Research Methodology.
3 Synthesis Review
3.1 Technical Scientific Literature
3.2 Research European Projects
3.3 Railway Administrators
4 Conceptual Scheme for Development a Digital Ar Chitecture: Railway Application
4.1 RDTA: Railway Digital Twin Architecture (Schematic)
4.2 Proof of Concept: Railway Application
5 Conclusions
References
Big Data Adoption in Strategic Decision-Making for Railway Infrastructure Asset Management
1 Introduction
2 Methodology and State of the Art
2.1 Research Methodology
2.2 State of the Art
3 Big Data and AM Decisions in Railways
3.1 Modelling of Domain-Related Concepts
3.2 Strategic AM Decisions for Railway IM Organizations
3.3 Expected Benefits from Big Data for Strategic Decisions
3.4 Data Types and Sources
4 Towards Big Data Driven AM Strategic Decisions
4.1 Opportunities and Challenges of the Adoption of Big Data
4.2 Recommendations and Roadmap
5 Conclusions
References
The Potential Value of Digital Twin in Rail and Road Infrastructure Asset Management
1 Introduction
2 Digital Twin
3 Methodology
4 Results and Discussion
4.1 Results
4.2 Discussion on the Rail and Road Context
5 Conclusions
References
Addressing Stakeholders Needs in Infrastructure Asset Management
1 Introduction
2 Background
2.1 Context
2.2 Theories on Decision Making
3 Stakeholder Analysis
4 From Stakeholders Interests to Indicators
5 Refining the Value Framework
6 Suggested Basic Set of Indicators
7 Practical Application
8 Conclusion
References
Asset Condition, Risk, Resilience, and Vulnerability Assessments
Assessment and Prioritization of Critical Assets for Updating Maintenance Plans in a Biomass Power Plant
1 Introduction
2 Criticality Analysis and Maintenance
3 The Biomass Power Plant
4 Assessment of Critical Assets
5 Conclusions
References
Application of Risk Management System for Intangible Assets in a Steel Company
1 Introduction
2 Risk Management System
2.1 Principles of Risk Management (According to ISO 31000)
2.2 Risk Management Steps
2.3 Maturity Model for the Risk Management Process in Standard Companies
2.4 Risk Treatment Options
3 Case Study
3.1 Risk Management for Intangible Assets
3.2 Risk Classification for Intangible Assets
3.3 Criticality Criteria
3.4 Criticality Matrix
3.5 Results
3.6 Priorization Criteria
4 Discussion and Conclusions
References
Case Studies on Condition Assessments of Infrastructure Assets
1 Introduction
2 Condition Assessment: Conventional Metrics
3 Case Studies
3.1 Case 1: Amusement Park Thrill Rides
3.2 Case 2: Waste Water Treatment Facility
3.3 Case 3: Primary and Secondary Schools Facilities
4 Discussion and Concluding Remarks
References
Identification of Emerging Safety and Security Risks in Drone Operations at Work Sites
1 Introduction
2 A Brief Overview to Current Status and Future Trends in Drone Applications
3 Legislation and Safety Requirements for Specific Category Drone Operations
4 Identification of Safety Risks According to the SORA Method
5 Safety and Security Challenges in Drone Operations at Work Sites
5.1 Drone Related Safety Risks at Construction Sites
5.2 Safety Challenges that are not Covered in SORA
5.3 Security Risks in Drone Operations
6 Conclusions and Future Work
References
Asset and Risk Management Approach in the Context of Complexity in Industry 4.0/5.0 Systems
1 Introduction
2 Literature Review
2.1 Asset Management Complexity and Uncertainty Associated with the Rising of Extreme, Rare and Disruptive Events
2.2 Industry 4.0/5.0 Challenges
2.3 Functional Resonance Analysis Method
2.4 System-Theoretic Accident Model and Processes
2.5 Risk-Informed Decision-Making
3 The Proposed Approach for Characterizing System Safety Risks in Asset Management
4 Future Case-Studies
5 Conclusion
References
Risk Assessment Using FMEA to Identify Potential Risks of Positive Displacement Pump Failure in Aluminum Industry: A Case Study
1 Introduction
2 Definitions and Methodology
3 Results
4 Conclusion
References
Asset Operations and Maintenance Strategies
Influence of the Income From the Use of an Asset on the Calculation of its Preventive Interval for a Planned Horizon. Use of Semi-Markov Processes and Degraded State
1 Introduction
2 Background
3 Material and Methods
3.1 Real Case. Returns and Weibull Distribution Data
3.2 Equations
4 Development of Mathematical Expressions
5 Analysis and Results
6 Conclusions
References
Factors Affecting the Quality of Network Services in Emerging Telecoms Operating Environment and Markets
1 Introduction
2 Literature Review
3 Research Method
4 Results
4.1 Summary of the Main Results
5 Discussion
6 Conclusions and Recommendations
Appendix A
Appendix B
Appendix C
References
Explaining Underlying Causes for the Degradation of Handover Information for Commercial Building Owners
1 Introduction
2 Literature Review
3 Methodology
3.1 Sampling
3.2 Data Collection
3.3 Data Analysis
4 Results
4.1 Lack of Data Governance
4.2 Consequences of Using Technological Solutions
4.3 Inadequate Management Support
5 Discussion
6 Conclusion
References
Perspectives on Smart Maintenance Technologies – A Case Study in Small and Medium-Sized Enterprises (SMEs) Within Manufacturing Industry
1 Introduction
2 Methodology
2.1 Data Collection
2.2 Data Analysis
3 Empirical Findings
4 Discussions and Conclusions
References
Improving Maintenance Data Quality: Application of Natural Language Processing to Asset Management
1 Introduction
2 Literature Review
2.1 Types of Maintenance
2.2 Machine Learning
2.3 Text Pre-processing
2.4 Word Representation
3 Methodology
3.1 Case Study and Problem Formalization
3.2 Data Acquisition and Preprocessing
3.3 Modelling, Validation and Testing
4 Results and Discussion
5 Conclusion
References
RQCM: Risk Qualitative Criticality Matrix. Case Study: Ophthalmic Lens Production Systems in Costa Rica
1 Introduction
2 Criticality Model RQCM: Risk Qualitative Criticality Matrix
3 Case Study: Application of the RQCM Model in Production Equipment of the PRATS Costa Rica Laboratory
3.1 Description of the Productive Process and Operational Context
3.2 Results and Analysis of the RQCM Application
4 Final Considerations
References
Economic and Environmental Indicators for Assessing Energy Efficiency Improvements in the Smart Manufacturing Processes
1 Introduction
2 Research Method
3 Proposed Indicators for Assessing Energy Efficiency Improvements
3.1 Economic Indicators
3.2 Environmental Indicators
4 Conclusions
References
Reliability and Resilience Engineering
Resilience Exposure Assessment Using Multi-layer Mapping of Portuguese 308 Cities and Communities
1 Introduction
2 Multi-level Mapping of Urban Resilience for the Portuguese Territory
3 Methodology
4 Result and Discussion
5 Conclusions
References
Use of Survival Analysis and Simulation to Improve Maintenance Planning of High Voltage Instrument Transformers in the Dutch Transmission System
1 Introduction
2 Data and Methodology
2.1 Description of Data
2.2 Survival Analysis and Failure Rate Modelling
3 Modelling in Cosmo Tech Asset and Simulations
3.1 Cosmo Tech Asset Platform
3.2 TenneT Asset Health Index
3.3 Simulation
4 Conclusion
References
Resilience Assessment of Public Treasury Elementary School Buildings in Lisbon Municipality
1 Introduction
2 Research and Method
3 Resilience Rating System
4 Case Study Implementation
5 Results Analysis
6 Conclusions
References
Disaster Risk Mitigation Through Capital Investment in Enhanced Building Resilience
1 Introduction
2 Conceptual Framework
3 Proposed Resilience Rating System
4 Case study implementation
5 Conclusions
References
Optimized Petri Net Model for Condition-Based Maintenance of a Turbine Blade
1 Introduction
2 Methodology
2.1 Monte Carlo Reinforcement Learning
2.2 Reinforcement Learning with Petri Net Model
3 Optimized Petri Net Model for the Blade of an Offshore Wind Turbine
4 Results and Conclusions
References
Multi-disciplinary and Dynamic Urban Resilience Assessment Through Stochastic Analysis of a Virtual City
1 Introduction
2 Methodology
3 Result and Discussion
4 Conclusion
References
Applications of International and Local Guidelines and Standards
Selecting KPIs in Asset Maintenance of Onshore Wind Farms Using Standard EN 15341:2019
1 Introduction
2 The Selection of KPIs in Asset Maintenance
2.1 Method for Classifying Indicators for Maintenance
2.2 The Selection of Indicators in Maintenance of Wind Farms
3 Case Study: Data Collection and Organization
4 Results and Conclusions
References
Mapping Maintenance Related Information Using the MIMOSA CRIS Standard: A Case Study Within Gravel Road Maintenance
1 Introduction
2 Applying MIMOSA for Gravel Road Maintenance
2.1 The MIMOSA CRIS Model and Selected Sub Models
2.2 Establishing the Context-Depended Information Model
2.3 Mapping of Gravel Road Information According to MIMOSA
2.4 Transferring the Conceptual Model into a Logical Model
3 Conclusions
References
Perceived Relevance of Asset Management Topics in Industry and Academia
1 Introduction
2 Surveys
3 Results and Discussion
3.1 Perceptions of the Portuguese Asset Management Community
3.2 Perceptions of the International Engineering Asset Management Community
4 Conclusions
References
The Concession Contract as an Instrument to Safeguard the Long-Term Condition of Logistics Infrastructure Assets
1 Introduction
2 Problem Statement
3 Methodology
4 Preliminary Results
4.1 Features of Concession Models and Analysis of International Best Practices
4.2 Analysis of the Chilean Context
5 Expected Results and Conclusion
References
Agile Methods in Industrial Maintenance
1 Introduction
2 Agile Methodology in Software Development
3 Agile Manufacturing
4 Industrial Maintenance
5 Discussion: To What Extent Do Agile Methods Fit Industrial Maintenance?
6 Conclusion
References
Standards-Based Interoperable Digital Twin in Industry 4.0 – A Pilot Demonstration
1 Introduction
2 Digital Twin for Asset Life Cycle – Pilot
3 Standards Used by the Pilot
4 Requirements for Digital Twins
5 Future Pilots
6 Conclusion
References
Author Index

Citation preview

Lecture Notes in Mechanical Engineering

Adolfo Crespo Márquez Juan Francisco Gómez Fernández Vicente González-Prida Díaz Joe Amadi-Echendu   Editors

16th WCEAM Proceedings

Lecture Notes in Mechanical Engineering Series Editors Fakher Chaari, National School of Engineers, University of Sfax, Sfax, Tunisia Francesco Gherardini , Dipartimento di Ingegneria “Enzo Ferrari”, Università di Modena e Reggio Emilia, Modena, Italy Vitalii Ivanov, Department of Manufacturing Engineering, Machines and Tools, Sumy State University, Sumy, Ukraine Editorial Board Members Francisco Cavas-Martínez , Departamento de Estructuras, Construcción y Expresión Gráfica Universidad Politécnica de Cartagena, Cartagena, Murcia, Spain Francesca di Mare, Institute of Energy Technology, Ruhr-Universität Bochum, Bochum, Nordrhein-Westfalen, Germany Mohamed Haddar, National School of Engineers of Sfax (ENIS), Sfax, Tunisia Young W. Kwon, Department of Manufacturing Engineering and Aerospace Engineering, Graduate School of Engineering and Applied Science, Monterey, CA, USA Justyna Trojanowska, Poznan University of Technology, Poznan, Poland Jinyang Xu, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China

Lecture Notes in Mechanical Engineering (LNME) publishes the latest developments in Mechanical Engineering—quickly, informally and with high quality. Original research reported in proceedings and post-proceedings represents the core of LNME. Volumes published in LNME embrace all aspects, subfields and new challenges of mechanical engineering. To submit a proposal or request further information, please contact the Springer Editor of your location: Europe, USA, Africa: Leontina Di Cecco at [email protected] China: Ella Zhang at [email protected] India: Priya Vyas at [email protected] Rest of Asia, Australia, New Zealand: Swati Meherishi at swati. [email protected] Topics in the series include: • • • • • • • • • • • • • • • • •

Engineering Design Machinery and Machine Elements Mechanical Structures and Stress Analysis Automotive Engineering Engine Technology Aerospace Technology and Astronautics Nanotechnology and Microengineering Control, Robotics, Mechatronics MEMS Theoretical and Applied Mechanics Dynamical Systems, Control Fluid Mechanics Engineering Thermodynamics, Heat and Mass Transfer Manufacturing Precision Engineering, Instrumentation, Measurement Materials Engineering Tribology and Surface Technology

Indexed by SCOPUS and EI Compendex. All books published in the series are submitted for consideration in Web of Science. To submit a proposal for a monograph, please check our Springer Tracts in Mechanical Engineering at https://link.springer.com/bookseries/11693

Adolfo Crespo Márquez Juan Francisco Gómez Fernández Vicente González-Prida Díaz Joe Amadi-Echendu •

•

•

Editors

16th WCEAM Proceedings

123

Editors Adolfo Crespo Márquez Department of Industrial Management, School of Engineering University of Seville Seville, Spain

Juan Francisco Gómez Fernández Department of Industrial Management, School of Engineering University of Seville Sevilla, Spain

Vicente González-Prida Díaz Department of Industrial Management, School of Engineering University of Seville Seville, Spain

Joe Amadi-Echendu Graduate School of Technology Management University of Pretoria Hatfield, Pretoria, South Africa

ISSN 2195-4356 ISSN 2195-4364 (electronic) Lecture Notes in Mechanical Engineering ISBN 978-3-031-25447-5 ISBN 978-3-031-25448-2 (eBook) https://doi.org/10.1007/978-3-031-25448-2 © ISEAM 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Foreword

The 16th World Congress on Engineering Asset Management (WCEAM) was held in Seville, Spain, from October 5 to 7, 2022. From all indications, attendees were delighted with the opportunity provided to meet in-person, especially to rekindle the warmth of face-to-face interactions and interpersonal relationships aftermath of the COVID-19 lockdowns and travel restrictions. The success of the events of 16th WCEAM provided proof of the remarkable collaboration between the International Society of Engineering Asset Management (ISEAM), the Spanish Association for the Development of Maintenance Engineering (INGEMAN), BCO Congresos Seville, and the University of Seville. The collaborating entities are grateful to the city of Seville and the hosting partners for the access granted to visit cultural and heritage sites, privileges granted to use the city’s facilities, and also for the general hospitality accorded to the 16th WCEAM delegates. Throughout the duration of the congress, the welcoming and safe environment of the city engendered enriching deliberations on ‘Value-Centered and Intelligent Asset Management in the 4th Industrial Revolution Era.’ The deliberations included business meetings, social engagements, and academic discourse between the participants. Much of the intellectual discussions, especially academic discourse, are summarily captured in the contents of this book of 16th WCEAM Proceedings. In this regard, the collaborating entities acknowledge the long-standing relationship between ISEAM and Springer with the primary objective of advancing education, research, training, and practice of the multidisciplinary body of knowledge in Engineering Asset Management. November 2022

Joe Amadi-Echendu Chair, ISEAM Board of Directors

v

Preface

The 16th World Congress on Engineering Asset Management (WCEAM www. wceam.com) was held in Seville, Spain, from October 5 to 7, 2022, under the auspices of the International Society of Engineering Asset Management (ISEAM www.iseam.org). The events were organized by ISEAM in collaboration with the Spanish Association for the Development of Maintenance Engineering (INGEMAN), coordinated by BCO Congresos Seville, and hosted by the University of Seville. The 16th instance of the WCEAM series took place as the global community increasingly relaxed lockdown and travel restrictions post the peak of the COVID-19 pandemic. The unprecedented and persistent effects of the pandemic, coupled with adverse influences of ongoing climate change events and geopolitical challenges, charge us to advance the principles, knowledge, and practice of the Engineering Asset Management body of knowledge within the context of 4IR technologies and the Society 5.0 ideal. 16th WCEAM 2022 provided opportunity for thought-leadership and research exchange in the multidisciplinary knowledge area of Engineering Asset Management. The Congress facilitated knowledge exchange between academics, researchers, industry practitioners, and policy makers in a friendly, multicultural, and transparent environment in Seville. The theme, ‘Value-Centered and Intelligent Asset Management in the 4th Industrial Revolution Era,’ is timely and relevant, since value-centered sustainable development is crucial for the future of human civilization. Interestingly, digitized and digitalized asset management provides new ways of looking at our world, encouraging us to utilize increasingly intelligent tools and methods to achieve more human-centric, resilient, and sustainable societies. Thus, this edition of 16th WCEAM Proceedings is a collection of high-impact discourses, serving as an important reference for Engineering Asset Management education, research, and practice. In addition to social activities, the events of 16th WCEAM included 7 plenary sessions, 16 academic sessions, 5 industry workshop, and business sessions. Although the activities and contributions provided a transversal view of the multidisciplinary body of knowledge of Engineering Asset Management, the contents

vii

viii

Preface

of this book are grouped into three thematic sections of ‘intelligence,’ ‘management,’ and ‘value’ as follows: • Section 1 on ‘intelligence’ contains three parts arranged as nineteen chapters that discuss matters regarding asset management and decision support systems based on the applications of 4IR technologies such as AR and VR, machine learning, digital twinning, etc for monitoring, diagnostics, prognostics. The section particularly includes methodologies and cases applied to different operational contexts. • Section 2 on ‘management’ contains four parts arranged as twenty-three chapters that discuss asset life-cycle management, especially, human dimensions on the management of infrastructure and industry-sector assets. • Section 3 on ‘value’ contains four parts structured into twenty-five chapters that deal with the applications of international standards, local regulations and industry guidelines to risk and resilience engineering, asset operations and maintenance, condition, risk, resilience and vulnerability assessments. The full editorial process for the WCEAM 2022 started in 2021 with the approval of the Proceedings book proposal by the publishing house Springer. Consequent upon the corresponding agreement between ISEAM and Springer, the qualification process involved two phases: Phase 1. Conference Review: • Initial call for papers in 2021, then a repeat call in 2022. • Abstracts submission and review using the BCO conference management system. The abstracts were reviewed to ensure consistency with the 16th WCEAM theme and topics. • Authors of accepted abstracts were invited to submit at least a 6-page version of their contributions formatted in accordance with the prescribed template. • The full-paper submissions were double-blind reviewed to ensure academic rigour. Phase 2. Editorial Review: • The authors of selected papers were notified to extend their manuscripts from 6 to 10/12 pages, formatted according to Springer template in order to qualify as book chapters. • The extended manuscripts were submitted and further reviewed using Springer’s Online Conference System (EquinOCS). • All reviews were conducted by the WCEAM technical review panel in accordance with ISEAM’s well-established guidelines. The panel of reviewers mostly comprised Members (M) and Fellows (F) of ISEAM (i.e., F/MISEAMs)

Preface

ix

The editors remain grateful to the global ISEAM community and INGEMAN associates for their participation during 16th WCEAM 2022. A special word of thank you to all the reviewers for ensuring the quality of the contributions published in this Proceedings. November 2022

Adolfo Crespo Márquez Juan F. Gómez Fernández Vicente González-Prida Díaz Joe Amadi-Echendu 16th WCEAM 2022 Editor

Contents

Asset Management and Decision Support System A Decision-Making Framework for Selecting an Optimum Package of Maintenance Improvement Projects in a Hospital . . . . . . . . . . . . . . . . . María Carmen Carnero, Aurora Martínez-Corral, and Javier Cárcel-Carrasco A Preliminary Implementation of Data-Driven TPM: A Real Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sara Antomarioni, Laura Lucantoni, Filippo Emanuele Ciarapica, and Maurizio Bevilacqua Assets’ Reliability Management Model for a Decision Making in Different Operational Contexts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jone Uribetxebarria, Ainhoa Zubizarreta, Ángel Rodríguez, and Urko Leturiondo

3

14

23

Realizing Sustainable Value from Engineering Innovation Ecosystems in EURope’s Outermost Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oliver Schwabe and Nuno Almeida

34

RelOps – A Whole-of-Organisation Approach for Reliability Analytics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Melinda Hodkiewicz, Tyler Bikaun, and Michael Stewart

45

Methods for Comparing Asset Portfolio Reliability . . . . . . . . . . . . . . . . Gabrielle Biard and Georges Abdul Nour

56

Industry 4.0 Tools and Its Impact in Asset Management Digital Transformation in Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . Adolfo Crespo Márquez

67

xi

xii

Contents

On the Definition of Requirements for a Digital Twin. A Case Study of Rolling Stock Assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adolfo Crespo Márquez, Urko Leturiondo, José A. Marcos, Antonio J. Guillén, and Eduardo Candón Review of Asset Digitalization Models in the Context of Intelligent Asset Management and Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . Pilar Jiménez Alonso, Antonio J. Guillén, Juan Fco. Gómez, and Eduardo Candón An Immersive Virtual Reality Platform for Enablement and Assessment of Human-Robot Interactions for Intelligent Asset Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sören Dominik Sonntag, Windo Hutabarat, Vinayak Prabhu, John Oyekan, Ashutosh Tiwari, and Chris Turner

76

87

98

Exploring Augmented Reality Applications to Support Maintenance Management in Hydroelectric Power Plants . . . . . . . . . . . . . . . . . . . . . . 108 Renan Favarão da Silva and Gilberto Francisco Martha de Souza Smart Water Dam Transformation in Industry 4.0 . . . . . . . . . . . . . . . . 118 Gowrishankar Sabapathipillai, Srijeyanthan Kuganesan, and Thanansan Kuganesan Can Industry 4.0 Keep Its Promises? A Literature-Based Comparison of Expectations and Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 Lasse Metso and Nils E. Thenent Monitoring, Diagnostics and Prognostics for Smart Maintenance Dynamic Maintenance Management Approach Based on Real Time Monitoring and Artificial Intelligence Using Digital Twins . . . . . . . . . . . 145 José Antonio Marcos-Alberca Carriazo and Adolfo Crespo Márquez Heat Pumps Smart Asset Management Implementation Through Virtual Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 Pedro Barandier, Alexandre Miranda, and Antonio João Marques Cardoso Driving Port Efficiency Through 5G-Enabled Condition Monitoring of Quay Cranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 Adolfo Crespo del Castillo, Manuel Herrera, Manu Sasidharan, Jorge Merino, Ajith Kumar Parlikad, Loretta Liu, Richard Brooks, and Karen Poulter Remote Data Collection Motivational Drivers, Challenges, and Potential Solutions in Industrial SME Companies . . . . . . . . . . . . . . . . . 172 Teemu Mäkiaho, Topias Kallio, Henri Vainio, Jouko Laitinen, and Kari Koskinen

Contents

xiii

The Effect of Knowledge Based Feature Extraction on Failure Detection of Control Surface Failures of Fighter Aircraft . . . . . . . . . . . 182 Tauno Toikka, Jouko Laitinen, and Kari T. Koskinen Advanced Maintenance of Distribution Assets Through the Application of Predictive Techniques Using GE’S APM System: Real Case in a Spanish DSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 Sergio Bustamante, Mario Manana, Alberto Arroyo, Antonio González, and Richard Maurice Challenges on an Asset Health Index Calculation . . . . . . . . . . . . . . . . . 205 E. Candón, Adolfo Crespo Márquez, A. Guillén, and U. Leturiondo Asset Life Cycle Management An Integrated Framework for Efficient Asset Life Cycle Costing in Case of Incomplete Historical Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Mohammad Baharshahi, Mostafa Yousofi Tezerjan, and Saeed Ramezani Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 Virginia Montiel Use Proposal of the Asset Health Index in the Public Health Sector. A Case Study in the Health Systems of the Republic of Costa Rica . . . . 240 B. Picado Arguello Visual Quality Control via eXplainable AI and the Case of Human in the AI Loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252 Christos Emmanouilidis and Elena Rica Start/stop Cost Evaluation of a Francis Turbine Runner Based on Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261 Abdossaber Peiravi, Denis Thibault, Michel Blain, Mustapha Nourelfath, and Masoumeh Kazemi Zanjani Asset Management in the Industrial Sector Machine Learning Supporting Maintenance Management: A Case Study in Scaffolding Industry’s Servitization Process . . . . . . . . . . . . . . . 273 Juan Izquierdo, Urko Lopez, and Eduardo Castellano A Prescriptive Analysis Tool for Improving Manufacturing Processes . . . Ana Gómez González, Estela Nieto, and Urko Leturiondo

283

Managing Assets to Facilitate Circularity and Sustainability of Food Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292 Anthea P Amadi-Echendu, Nonceba Ntoyanto-Tyatyantsi, and Joe Amadi-Echendu

xiv

Contents

A Reference Model for Engineering Asset Management Excellence . . . . 301 Oliver Schmiedbauer and Hubert Biedermann The Exploration of Digitalization and Digitalization Indicators Within the Scope of Asset Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312 Lucas Peter Høj Brasen and Torben Tambo Audit Model for Asset Management, Maintenance and Reliability Processes: A Case Study Applied to Pulp Mill Sector . . . . . . . . . . . . . . 323 Andrés Aránguiz, Félix Pizarro, Carlos Parra, Pablo Duque, and Emanuel Vega Human Dimensions and Asset Management Performance Using Wearable Sensors to Form a Relationship Between Driver Stress and Aggressive Driving Habits . . . . . . . . . . . . . . . . . . . . . . . . . . . 333 Tahrim Zaman Tila and Turuna S. Seecharan Reshaping Industry Job Profiles to Better Meet Future Asset Management Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 Aitor Goti, Tugce Akyazi, Aitor Oyarbide, and Elisabete Alberdi Methods for the Criticality Assessment of Intangible Assets in a Knowledge Management Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354 Vicente González-Prida, Matías Velásquez, Antonio Guillén, Carlos Parra, Pablo Viveros, and Fredy Kristjanpoller What is Smart Maintenance in Manufacturing Industry? . . . . . . . . . . . 366 Antti Salonen Audit Models for Asset Management, Maintenance and Reliability Processes: A Case Study Applied to the Desalination Plant . . . . . . . . . . 375 Pablo Duque, Carlos Parra, Felix Pizarro, Andrés Aránguiz, and Emanuel Vega Impact of Information Digitalization on Asset Availability - an Empirical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385 Katja Gutsche and Santina Schlögel Infrastructure Asset Management Linking Organisation Objectives with Asset Information Requirements for Highway Infrastructure Projects . . . . . . . . . . . . . . . . 397 Georgios Hadjidemetriou, Nicola Moretti, James Heaton, Manu Sasidharan, Ajith Parlikad, and Jennifer Schooling A Methodology for Ensuring Strategic Alignment of Railway Infrastructure Asset Management Processes . . . . . . . . . . . . . . . . . . . . . 405 Irene Roda, Donatella Fochesato, Adalberto Polenghi, Margherita Luciano, Isabella Tordi, Lorenzo Di Pasquale, and Ivan Cavaiuolo

Contents

xv

Hierarchy Definition for Digital Assets. Railway Application . . . . . . . . . 416 Mauricio Rodríguez Hernández, Adolfo Crespo Márquez, Antonio Guillen López, and Eduardo Candon Fernandez Big Data Adoption in Strategic Decision-Making for Railway Infrastructure Asset Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428 Irene Roda, Adalberto Polenghi, and Vesa Männistö The Potential Value of Digital Twin in Rail and Road Infrastructure Asset Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439 João Vieira, Hugo Patrício, João Poças Martins, João Gomes Morgado, and Nuno Almeida Addressing Stakeholders Needs in Infrastructure Asset Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448 Ype Wijnia, John de Croon, and Rhys Davies Asset Condition, Risk, Resilience, and Vulnerability Assessments Assessment and Prioritization of Critical Assets for Updating Maintenance Plans in a Biomass Power Plant . . . . . . . . . . . . . . . . . . . . 463 Daniel Gaspar, Odete Lopes, João Costa, and Elson Grilo Application of Risk Management System for Intangible Assets in a Steel Company . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474 Manuel González Case Studies on Condition Assessments of Infrastructure Assets . . . . . . 487 Joe E. Amadi-Echendu, Jedial O. Mvele, and Refiloe R. Lapshe Identification of Emerging Safety and Security Risks in Drone Operations at Work Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499 Risto Tiusanen, Eetu Heikkilä, Tero Välisalo, and Emrehan Öz Asset and Risk Management Approach in the Context of Complexity in Industry 4.0/5.0 Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508 Issa Diop, Georges Abdul-Nour, and Dragan Komljenovic Risk Assessment Using FMEA to Identify Potential Risks of Positive Displacement Pump Failure in Aluminum Industry: A Case Study . . . . 521 Hamid Ahmadi, Meysam Esmaeilzadeh Mofrad, and Abolfazl Sedghi Asset Operations and Maintenance Strategies Influence of the Income From the Use of an Asset on the Calculation of its Preventive Interval for a Planned Horizon. Use of Semi-Markov Processes and Degraded State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533 Antonio Sánchez-Herguedas, Adolfo Crespo Márquez, and Francisco Rodrigo-Muñoz

xvi

Contents

Factors Affecting the Quality of Network Services in Emerging Telecoms Operating Environment and Markets . . . . . . . . . . . . . . . . . . . 544 Charles Okeyia and Nuno Marques de Almeida Explaining Underlying Causes for the Degradation of Handover Information for Commercial Building Owners . . . . . . . . . . . . . . . . . . . . 561 Janet Chang, Jorge Merino Garcia, Xiang Xie, Nicola Moretti, and Ajith Parlikad Perspectives on Smart Maintenance Technologies – A Case Study in Small and Medium-Sized Enterprises (SMEs) Within Manufacturing Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571 San Giliyana, Marcus Bengtsson, and Antti Salonen Improving Maintenance Data Quality: Application of Natural Language Processing to Asset Management . . . . . . . . . . . . . . . . . . . . . . 582 Mathieu Payette, Georges Abdul-Nour, Toualith Jean-Marc Meango, and Alain Côté RQCM: Risk Qualitative Criticality Matrix. Case Study: Ophthalmic Lens Production Systems in Costa Rica . . . . . . . . . . . . . . . . . . . . . . . . . 590 Carlos Parra, Juan Rodríguez, Adolfo Crespo Márquez, Vicente González-Prida, Pablo Viveros, Fredy Kristjanpoller, and Jorge Parra Economic and Environmental Indicators for Assessing Energy Efficiency Improvements in the Smart Manufacturing Processes . . . . . . 602 Minna Räikkönen, Teuvo Uusitalo, Saara Hänninen, Andrea Barni, Claudio Capuzzimati, Alessandro Fontana, and Marco Pirotta Reliability and Resilience Engineering Resilience Exposure Assessment Using Multi-layer Mapping of Portuguese 308 Cities and Communities . . . . . . . . . . . . . . . . . . . . . . . . . 615 Seyed M. H. S. Rezvani, Nuno Almeida, Maria João Falcão Silva, and Damjan Maletič Use of Survival Analysis and Simulation to Improve Maintenance Planning of High Voltage Instrument Transformers in the Dutch Transmission System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624 Swasti R. Khuntia, Fatma Zghal, Ranjan Bhuyan, Erik Schenkel, Paul Duvivier, Olivier Blancke, and Witold Krasny Resilience Assessment of Public Treasury Elementary School Buildings in Lisbon Municipality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636 João Garcia, Seyedi Rezvani, Maria João Falcão Silva, Nuno Almeida, Cláudia Pinto, Rui Gomes, Mónica Amaral Ferreira, Filipe Ribeiro, Filipa Salvado, and Carlos Sousa Oliveira

Contents

xvii

Disaster Risk Mitigation Through Capital Investment in Enhanced Building Resilience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645 Maria João Falcão Silva, Filipa Salvado, and Nuno Almeida Optimized Petri Net Model for Condition-Based Maintenance of a Turbine Blade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657 Ali Saleh, Manuel Chiachio, and Juan Chiachio Multi-disciplinary and Dynamic Urban Resilience Assessment Through Stochastic Analysis of a Virtual City . . . . . . . . . . . . . . . . . . . . 665 Seyed M. H. S. Rezvani, Nuno Almeida, and Maria João Falcão Silva Applications of International and Local Guidelines and Standards Selecting KPIs in Asset Maintenance of Onshore Wind Farms Using Standard EN 15341:2019 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 677 Daniel Gaspar, Odete Lopes, Carlos Rodrigues, and Serafim Oliveira Mapping Maintenance Related Information Using the MIMOSA CRIS Standard: A Case Study Within Gravel Road Maintenance . . . . . 688 Mirka Kans and Jaime Campos Perceived Relevance of Asset Management Topics in Industry and Academia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 697 Nuno Almeida, Joe Amadi-Echendu, Daniel Gaspar, Edmea Adell, Joana Torcato, João Vieira, and Eduardo Leite The Concession Contract as an Instrument to Safeguard the LongTerm Condition of Logistics Infrastructure Assets . . . . . . . . . . . . . . . . . 708 Monica Lopez-Campos, Raúl Stegmaier, and Eduardo Candón Agile Methods in Industrial Maintenance . . . . . . . . . . . . . . . . . . . . . . . . 716 Lasse Metso and Nils E. Thenent Standards-Based Interoperable Digital Twin in Industry 4.0 – A Pilot Demonstration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726 Karamjit Kaur, Matt Selway, Markus Stumptner, Alan Johnston, and Joseph Mathew Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737

Asset Management and Decision Support System

A Decision-Making Framework for Selecting an Optimum Package of Maintenance Improvement Projects in a Hospital María Carmen Carnero(B) , Aurora Martínez-Corral, and Javier Cárcel-Carrasco University of Castilla-La Mancha, Escuela Técnica Superior de Ingeniería Industrial de Ciudad, Real. Avda. Camilo José Cela s/n, 13071 Ciudad Real, Spain [email protected]

Abstract. This Chapter presents a decision framework for the selection of the most satisfactory combination of projects in the maintenance department of a healthcare organisation, using a multi-criteria audit. A number of projects were designed with the aim of improving the current state of the maintenance department of a Spanish hospital. To select the optimum package of projects, a multi-criteria additive model was constructed by means of MACBETH, using two types of benefit (internal and external) and the cost of implementing each project. From the cost/benefit ratio, the resulting efficient frontier with all the possible combinations of projects to be implemented, and the particular conditions in the maintenance department under study, the first package of projects was selected.

1 Introduction Maintenance is particularly important in healthcare organisations, since there are facilities and devices that must operate with 100% availability, safety and quality, or serious consequences will ensue for the quality of the service (patient waiting time, reprogramming of appointments, incorrect diagnoses, etc.), putting patients’ safety, and even lives, at risk. For this reason, maintenance managers need to monitor the efficiency of their departments to prevent deficiencies, and then to introduce improvement projects to tackle any deficiencies found. Ideally, they should do this through a process of continuous improvement. Project selection is a strategic decision (Liesio et al. 2007). Thus, the decisionmaking process is complicated, with a large number of stages, decision-making groups, and conflicting objectives, and a high risk and uncertainty (Ghasemzadeh and Archer 2000). There are many methodologies in the literature for selecting the optimal project portfolio that best aligns with the strategic priorities of the organization. For example, Bai et al. (2021) proposed a model based on the past record of the projects. Dou et al. (2019) gave two methodologies, one based on a single objective, and another on multiple objectives, to choose the set that maximises the values of these objectives. Kornfeld and Kara (2013) used lean and six sigma for their project ranking methodology. However, these approaches are not always suitable, as pointed out by Pérez et al. (2018): a) The strategy should regulate the criteria for selecting the project porfolio, but a defective © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 3–13, 2023. https://doi.org/10.1007/978-3-031-25448-2_1

4

M. C. Carnero et al.

choice of criteria may make it hard to introduce the portfolio in such a way as to make the strategy work; b) there may be complex synergies between the candidate projects that might not be accounted for. The best project to select individually may not form part of the best set of projects when a decision is taken with respect to a group of projects. Thus, Zhang and An (2016) show that, if a project has the highest degree of synergy, it should be preferred to the others; to calculate the degree of synergy, they applied a scoring method based on expert opinions. Optimal project selection depends on many factors, and therefore the multi-criteria technique could be very useful in the evaluation of different criteria simultaneously. Although project selection by means of multi-criteria techniques has been analysed in the literature (Ghasemzadeh and Archer 2000; Salo et al. 2004; Bana e Costa et al. 2005; Mohanty et al. 2005; Phillips and Bana e Costa 2007; Singh et al. 2021), in the area of maintenance, the use of analytical techniques has historically been rare, and is even rarer in hospitals, despite the important potential repercussions for patient care. Consequently, this research combines different techniques, including maintenance audits, Measuring Attractiveness by a Categorical Based Evaluation Technique (MACBETH), and the Taguchi loss function, with the aim of continuous improvement in a hospital maintenance department. In order to apply it, it is thus necessary first to carry out a multi-criteria audit, based on the judgements of experts, to weight each subject and fundamental point of view to be considered. On the other hand, the evaluation of each project is objective, providing a methodology that guarantees the objectivity of the solutions produced. The structure of this Chapter is as follows. Section 2 examines the characteristics of the internal audit carried out. Section 3 discusses the criteria used to select the most suitable packages of maintenance projects. Section 4 describes the procedure for building the efficient frontier for the selection of the first project package. The Chapter ends with the reference section.

2 Maintenance Multi-criteria Audit The maintenance audit is structured into the following subjects: maintenance strategy, attitude of maintenance staff and other hospital personnel, resources and installations, human resources, records, planning, scheduling, work orders, purchases, warehouse stock, maintenance documentation, calibration, technical aspects, effectiveness and control. Each subject comprises a number of Fundamental Points of View (FPV). A FPV is an aspect that is relevant to the evaluation of the alternatives. The FPVs group into a hierarchical structure, and each is associated with a descriptor. The descriptor is a set of impact levels that can measure, either quantitatively or qualitatively, the level of compliance of an FPV (Bana e Costa and Carvalho 2002). The audit is based on Bana e Costa et al. (2012). Table 1 shows, as an example, the FPVs “Quality control of spare parts, tools and machinery that enter a warehouse, measures to prevent stock breakage and obsolescence check” in the subject Warehouse in the audit carried out. The descriptors in this audit are constructed and generally qualitative, although some are quantitative. Descriptors usually have five scale levels (L1 –L5 ), among which the reference level is known as “Neutral” and the best level is known as “Good”.

A Decision-Making Framework for Selecting an Optimum Package

5

Table 1. FPV Quality control of spare parts, tools and machinery that enter a warehouse, measures to prevent stock breakage and obsolescence check. Scale level

Description

L1 (Good)

Quality control is carried out on each spare part, tool or machine that enters the warehouse. Action protocols are in place in the event of stock breakage to guarantee optimum availability of hospital facilities as a function of their criticity. Obsolescence checks are carried out periodically

L2

Quality control is carried out on each spare part, tool or machine that enters warehouse. Action protocols are in place in the event of stock breakage. Obsolescence checks are carried out periodically

L3 (Neutral) Quality control is carried out on each spare part, tool or machine that enters the warehouse. Action protocols are in place in the event of stock breakage. Obsolescence checks are not carried out periodically L4

Quality control is carried out on machines that enter the warehouse. No action protocols are in place in the event of stock breakage. Obsolescence checks are not carried out

L5

No quality control is carried out on machines that enter the warehouse. No action protocols are in place in the event of stock breakage. Obsolescence checks are not carried out

For each FPV a matrix of MACBETH judgements must be built. The matrices built are all consistent. Then a numerical scale is constructed based on the qualitative judgements with value scores of 100 and 0 arbitrarily assigned to the Neutral and Good reference levels, respectively (Bana e Costa and Chagas 2004). Figure 1 shows the numerical scale and the value function corresponding to the FPV “Classification of spare parts according to criticity”. To obtain the weights for the FPVs in each subject, the following procedure is adopted. First, the possibility of an alternative with all criteria or FPVs at a Neutral level is considered. The system calculates how much a swing from Neutral to Good in all FPVs would improve the preference of this alternative, using the semantic categories of MACBETH. This ranks the FPVs in the matrix of judgements. Then, the system compares how much more preferable would a swing from Neutral to Good be in the first FPV in comparison to in the second FPV. The comparison repeats for the first FPV and the third FPV, and so on. This process continues row-by-row until the matrix of judgements is complete. Each subject can be in one of the following states: excellent, satisfactory, acceptable, poor and very poor. In addition, the states “All excellent” and “All very poor” define the best and worst possible states of each subject. The limits between each pair of states in a subject are defined as: Excellent/Satisfactory limit; Satisfactory/Acceptable limit; Acceptable/Poor limit; and Poor/Very poor limit. The current state of the hospital maintenance department is labelled “current state”. The methodology for establishing the limits between states by subject is a variation on the bottom-up and top-down procedures described in Bana e Costa and Carvalho (2002). The limits between states obtained using these procedures may differ, in which case the limits are reanalysed. Table 2 shows the

6

M. C. Carnero et al.

final limits obtained by the subject, and the current state of the maintenance department assessed (in a quantitative and qualitative manner).

Fig. 1. Judgement matrix, scale and value function for FPV “Classification of spare parts according to criticity”.

3 Decision Framework for Selection of Optimum Package of Maintenance Improvement Projects The objective is to calculate an efficient frontier with packages of maintenance projects offering the optimum cost-benefit relation. In order to identify the most efficient investment in the whole portfolio, all projects must be considered simultaneously. This produces an envelope graph that shows every possible combination of investments. The upper line in the graph is known as the efficient frontier. This line represents the investment portfolios that generate the most benefit at a particular cost. An area is built by grouping the investment opportunities into projects. The areas defined are: undergraduate projects, outsourcing, and internal development. Only one project can come from each area to form any one package due to the time constraints of the head of the hospital’s maintenance department, who has to manage or control each project adopted in the maintenance department. Table 2. Limits between states by subject and current state of Maintenance Department. Subject

Excellent/satisfactory Satisfactory/acceptable Acceptable Poor/very Current state limit limit poor limit poor limit

Strategy

50.00

19.23

−11.54

−50.00

38.46 (Satisfactory)

Attitude

50.00

25.00

−12.50

−28.13

12.50 (Acceptable)

(continued)

A Decision-Making Framework for Selecting an Optimum Package

7

Table 2. (continued) Subject

Excellent/satisfactory Satisfactory/acceptable Acceptable Poor/very Current state limit limit poor limit poor limit

Resources

61.11

19.45

−2.78

−50.00

−38.89 (Poor)

Human Resources

53.66

31.70

−31.71

−53.66

−36.59 (Poor)

Records

71.79

28.20

−53.85

−97.44

−10.26 (Acceptable)

Planning

63.64

33.33

−6.06

−39.40

18.18 (Acceptable)

Scheduling

38.46

15.38

−34.62

−61.54

−23.08 (Acceptable)

Work orders

65.38

36.54

−3.85

−46.16

71.43 (Excellent)

Purchases

50.00

22.23

−11.11

−50.00

61.11 (Excellent)

Warehouse

65.38

36.54

−3.85

−46.16

−30.77 (Poor)

Maintenance 54.16 documentation

18.75

−20.84

−39.59

−33.33 (Poor)

Calibration

58.14

25.58

−32.56

−62.79

−18.60 (Acceptable)

Technical aspects

64.28

35.71

−28.57

−75.00

−21.43 (Acceptable)

Effectiveness

51.91

23.02

−20.24

−40.40

9.52 (Acceptable)

Control

50.00

23.96

−16.67

−43.75

10.42 (Acceptable)

The criteria are the elements for evaluating each project: cost and benefit. The benefit criterion has the sub-criteria (for a detailed explanation see Phillips (2004)): • Internal benefit. This benefit is related with the different FPVs analysed in the multicriteria audit. In each maintenance project, the potential benefit is estimated by considering the change in the scale level of one or various FPVs in one subject of the audit. Figure 2 shows the internal benefits of the projects in each area, which are calculated from the estimated value of a subject after implementing the project less the current value of the subject. • External benefit. This benefit is related to the quality of the service provided, in other words, how a project can improve the satisfaction of the hospital’s patients and healthcare staff. This increase in satisfaction is due to improvements attributable to the project in maintenance aspects that have a direct or indirect effect on quality of healthcare (e.g., decrease in breakdowns of a medical device, with a consequent

8

M. C. Carnero et al.

decrease in customer waiting time for a diagnostic test). The external quality relates to modifications in the opportunity costs, defined as the losses as a result of the failure of machinery and medical devices to carry out their functions properly. The Taguchi loss function (see Eq. 1) is used to evaluate the external quality of the maintenance department. L(y) is the benefit for the quality of healthcare expected from the implementation of a project, k the effect on the quality of healthcare of the improvements in the maintenance department, y the current value obtained from the maintenance audit in the subject, and τ the target value in the subject desired for a project. L(y) = k(y − τ )2

(1)

This study uses an additive model, following Eq. (2), where vij is the value associated with alternative i in criterion j, and wj is the weight of criterion j. Vi =

 j

wj vij

(2)

As a result of the double weighting, the additive model takes on the form of Eq. 3, where wjk is the within-criterion weight, vijk the value of alternative i in criterion j in area k, and wj is the weight of criterion j. The multiplication by 10 is so that the weighted values obtained are above 1000, rather than 100 as in the starting data. The final value Vik of a project is the sum of the scores of alternative i in all criteria (Phillips 2004).    Vik = 10(wj wjk vijk )/( wj wjk ) (3) j

j

k

The cost-benefit ratio rik is then calculated by dividing the difference of values between one level and the next by the difference in costs, as shown in Eq. 4. rik = (Vik − V(i−1)k )/(Cik − C(i−1)k )

(4)

The doubly-weighted mean of the two benefit scales means that each project is characterised by two numbers: the cost, and one single benefit. This enables the calculation of the cost and total benefit of all the combinations or packages of projects. The efficient frontier is the curve of the best set of investments or most beneficial package of projects for each level of total cost. The shaded area in the graph represents all the possible combinations of packages (each one consisting of three projects).

4 Selection of Projects for Maintenance Department Looking at the results of the multi-criteria audit carried out reveals which subjects or PFVs are in a worse situation. Fourteen projects were proposed to improve the subjects of the internal audit carried out. The projects fall into the following areas: Undergraduate projects, Outsourcing, and Internal development, as shown in Fig. 2.

A Decision-Making Framework for Selecting an Optimum Package

9

Total Productive Maintenance (TPM) Cost: 60,000€ Internal benefit: 19.52 External benefit:19051.52 Alert

Acceptable

Benefit control Cost: 20,000 € Internal benefit: 28.12 External benefit: 15814.69 Alert

Acceptable

Stock control

Technical Training

Document management system

Cost: €7,200 Internal benefit: 46.14 External benefit:191600.96

Cost: €60,000 Internal benefit: 9.73 External benefit:5715.46

Cost: €20,000 Internal benefit: 16.65 External benefit:27722.25

Alert

Acceptable

Alert

Acceptable

Acceptable

Satisfactory

Technical procedures

Vibration analysis

Warehouse review

Cost: €7,200 Internal benefit: 20.85 External benefit:39125.03

Cost: €60,000 Internal benefit: 33.33 External benefit:111088.89

Cost: €12,000 Internal benefit: 15.38 External benefit:4730.89

Alert

Acceptable

Alert

Alert

Alert

Alert

Maintenance manual

Thermography

Data analysis

Cost: €7,200 Internal benefit: 16.65 External benefit:8316.68

Cost: €50,000 Internal benefit: 33.33 External benefit:111088.89

Cost: €12,000 Internal benefit: 10.26 External benefit:4210.70

Alert

Acceptable

Alert

Alert

Acceptable

Acceptable

CMMS control

Standardisation of maintenance activities

Calibration

Cost: €2,700 Internal benefit: 12.5 External benefit:3906.25

Cost: €12,000 Internal benefit: 20.52 External benefit:21053.52

Cost: €12,000 Internal benefit: 37.2 External benefit: 3459.60

Acceptable

Acceptable

Acceptable

Acceptable

Acceptable

Acceptable

Fig. 2. Maintenance improvement projects by areas (adapted from Carnero 2015).

Projects offering the most benefits generally also have the highest costs. But because of the particular characteristics of the maintenance area, in this study a number of the projects proposed here are exceptions to this rule, having very low costs despite potentially offering considerable benefits. Each combination of projects aiming to improve maintenance is called a package and consists of three projects, one from each area. Each project aims to improve only one single subject, and in some cases only a single FPV in that subject. The number of potentially buildable packages is 4 × 4 × 6 = 96. Figure 2 shows the projects in each area, the costs and internal and external benefits of each project, and the qualitative current state of the subject in which the project would be implemented before (left cell) and after (right cell) its implementation. To obtain the total benefits of each project a multi-criteria additive model considering internal and external benefits has been produced. Relative scales are used ranging from 0 to 100; thus, for internal benefits, the undergraduate projects area becomes 100, 70.15 for outsourcing and 53.09 for internal development, and for external benefits 100, 55.91 and 12.87 respectively. The internal benefits obtained with each project are estimated in the audit. For example, implementing the project Stock control leads to an improvement in the descriptor

10

M. C. Carnero et al.

for the FPV “Inventory policy for each spare part and periodic review” from Neutral to Good, and in the descriptor for the FPV “Classification of spare parts according to criticity” from L5 to Good. With this, Warehouse would improve from the current state of -30.77 (Poor) to 15.37 (Acceptable) (see Table 2). Thus the potential internal benefit of this project is 46.14. To obtain the external benefits, the current state of the subject is obtained in the audit. A target value for each project is determined. The value of the loss coefficient is estimated using the Smart method. Using this method, the influence of the projects on healthcare is obtained via swing weights. A swing weight of 100 is assigned to the project that is most important to swing from its least preferred level to its most preferred level with respect to healthcare quality. The levels of influence of each project on healthcare quality are: none, low, medium and high. Lower weights are assigned to the projects on the basis of the relative importance of swinging them in comparison to the most important project. Figure 2 shows the costs and internal and external benefits attributable to each project. The figure also shows the current state of the subject in which the project would be implemented (left cell) and the estimated final state after implementing the project (right cell). To obtain the total benefits of each project, a multi-criteria additive model that considers the internal and external benefits has been produced. Relative scales are used, so the values of the projects in each area and in each type of benefit are converted into scales ranging from 0 to 100. That is, the value 100 is assigned to the project with the most internal (or external) benefit in each area, and 0 to the project with the least benefit. The rest of the projects in each area are assigned values corresponding to a linear conversion between 0 and 100. To calculate the within-criterion weights, the difference between the projects with the best and worst internal benefit in each area is evaluated. For the Undergraduate projects area, this calculation is 46.14 − 12.50 = 33.64; for the Outsourcing area, 33.33 − 9.73 = 22.54; and for the Internal development area, 28.12 − 10.26 = 17.86. Subsequently, the area with the biggest difference between projects is assigned 100, with the other two areas being assigned values in proportion. Thus for the internal benefits the value 33.64 becomes 100, while the value becomes 70.15 for Outsourcing and 53.09 for Internal development. For external benefits a similar procedure results in 100 for the Undergraduate projects area, 55.91 for Outsourcing and 12.87 for Internal development. The cross-criteria weights demonstrate the relative importance of one criterion with respect to another. By means of cross-criteria weights it is possible to obtain the equivalence of a scale in one criterion in comparison to another scale belonging to a different criterion. When assessing the cross-criteria weights the swing in preference from 0 to 100 is considered (this swing is considering the difference in value from the least to the most preferred level when the criteria are compared). In this study, the external benefit is assigned double the swing from 0 to 100 than the internal benefit because external benefits are closer to the ultimate aim of achieving improvement that patients and healthcare staff can appreciate. The external or internal benefit of each project resulting from the additive model is then calculated. The project Technical procedures returns an external benefit of 39125.03 and an internal benefit of 28.85. After applying the concept of relative scales these values become 19.10 and 48.60 for external and internal benefit, respectively. The weighted external benefit is: 10 ∗ 19.10 ∗ 100 ∗ 100/(100 ∗ 100 + 100 ∗ 50 + 55.91 ∗ 100 + 70.15 ∗ 50 + 12.87 ∗ 100 + 53.09 ∗ 50) = 68.12

A Decision-Making Framework for Selecting an Optimum Package

11

and the weighted internal benefit is: 10 ∗ 48.60 ∗ 100 ∗ 50/(100 ∗ 100 + 100 ∗ 50 + 55.91 ∗ 100 + 70.15 ∗ 50 + 12.87 ∗ 100 + 53.09 ∗ 50) = 86.66.

The total benefit for this project is therefore: 68.12 + 86.66 = 154.78 ≈ 155 Figure 3 shows the efficient frontier that links the packages with the greatest benefits for a given cost. These packages define the efficient frontier and will always be on the upper surface of the envelope. The shaded area of the graph represents the position of all the possible packages of projects. Equity software is used to calculate the efficient frontier. The head of the maintenance department proposed the package called P (see Fig. 3), comprising the projects Stock control, Technical training, and Document management system. These projects would improve the subjects in the Poor state: Warehouse, Human resources and Maintenance documentation, respectively, and would move them all into the Acceptable state. The cost of this package is e87,200, and the benefit 614.86. Nevertheless, a better package than the one proposed is B, which consists of the projects Stock control, Thermography and Benefit control. These projects would improve the subjects Warehouse, Resources and Control. The first two subjects are in a Poor state and would become Acceptable and Poor, respectively, while Control would improve to Satisfactory. The cost of this package is e77,200 and the benefit 977.43. Package F consists of the projects Stock control, Standardisation of maintenance activities and Benefit control. The cost is e39,200 and the benefit is 739. A cheaper package than the one proposed (C) consists of the projects Stock control, Standardisation of maintenance activities and Calibration. The cost would be e31,200 and the benefit 665.30. Both F and C would improve one subject from Poor to Acceptable. The projects with the highest cost-benefit ratio make up package B. Finally, package B is selected for implementation, since its benefits are 58.97% higher than those of the proposed project, while its cost is 11.47% less. The head of the maintenance department considered the increase in benefit resulting from this project more important than the potential cost reductions attributable to packages F and C. Benefits

Costs

Fig. 3. Efficient frontier with packages P, B, C and F (adapted from Carnero 2015).

12

M. C. Carnero et al.

5 Conclusions This Chapter describes a methodology for project portfolio selection in a hospital maintenance department. This area of application is historically lacking in contributions, and so it could be useful in service companies, and in particular, to healthcare organisations, although it could also be applied in manufacturing companies. The use of a maintenance audit via MACBETH allows objective evaluation of all the benefits potentially to be obtained from each proposed improvement project. The limitations of the methodology set out here relate to the need to carry out a prior audit of the state (qualitative and quantitative) of maintenance in different subjects, and the need to relate each improvement project to improvements in a single subject, and in some cases only a single FPV in that subject. Furthermore, it must be borne in mind that only one project can be selected in each area, in order to make up a package. This is, however, due to constraints specific to the case study described here. Future work could consider evolving criteria, and how one project might improve a number of FPVs from different subjects in the maintenance audit.

References Bai, L., Han, X., Wang, H., Zhang, K., Sun, Y.: A method of network robustness under strategic goals for project portfolio selection. Comput. Industr. Eng. 161, 107658 (2021). https://doi.org/ 10.1016/j.cie.2021.107658 Bana e Costa C.A., Chagas, M.P.: A career choice problem: an example of how to use MACBETH to build a quantitative value model based on qualitative value judgments. Eur. J. Oper. Res. 153, 323–331 (2004) Bana e Costa, C.A., Carnero, M.C., Oliveira, M.D.: A multi-criteria model for auditing a predictive maintenance programme. Eur. J. Oper. Res. 217(2), 381–393 (2012) Bana e Costa, C.A., Carvalho, R.: Assigning priorities for maintenance, repair and refurbishment in managing a municipal housing stock. Eur. J. Oper. Res. 138, 380–391 (2002) Bana e Costa, C.A., Fernandez, T.G., Correia, P.V.D.: Prioritisation of public investments in social infra-structures using multicriteria value analysis and decision conferencing: a case-study. In: Operational Research working papers, LSEOR 05.78. Operational Research Group, Department of Management, London School of Economics and Political Science, London, UK (2005) Carnero, M.C.: Methodology for selection of optimal portfolio in maintenance departments. Int. J. Industr. Eng. 22(5), 549–574 (2015) Dou, Y., Zhao, D., Xia, B., Zhang, X., Yang, K.: System portfolio selection for large-scale complex systems construction. IEEE Syst. J. 13(4), 3627–3638 (2019) Ghasemzadeh, F., Archer, N.P.: Project portfolio selection through decision support. Decis. Support Syst. 29, 73–88 (2000) Kornfeld, B., Kara, S.: Selection of lean and six sigma projects in industry. Int. J. Lean Six Sigma 4(1), 4–16 (2013) Liesio, J., Mild, P., Salo, A.: Preference programming for robust portfolio modeling and project selection. Eur. J. Oper. Res. 181, 1488–1505 (2007) Mohanty, R.P., Agarwal, R., Choudhury, A.K., Tiwari, M.K.: A fuzzy ANP-based approach to R&D project selection: a case study. Int. J. Prod. Res. 43(24), 5199–5216 (2005) Pérez, F., Gómez, T., Caballero, R., Liern, V.: Project portfolio selection and planning with fuzzy constraints. Technol. Forecast. Soc. Chang. 131, 117–129 (2018)

A Decision-Making Framework for Selecting an Optimum Package

13

Phillips, L.: The Mathematics of Hiview and Equity (2004). http://www.catalyze.co.uk/resources/ docs/pdf/Catalyze_White_Paper_Mathematics_of_Hiview_and_Equity.pdf. Accessed 28 Sept 2022 Phillips, L.D., Bana e Costa, C.A.: Transparent prioritisation, budgeting and resource allocation with multi-criteria decision analysis and decision conferencing. Ann. Oper. Res. 154(1), 51–68 (2007) Salo, A., Gustafsson, T., Mild, P.: Prospective evaluation of a cluster program for Finnish forestry and forest industries. Int. Trans. Oper. Res. 11, 139–154 (2004) Singh, K., Swarnakar, V., Singh, A.R.: Lean Six Sigma project selection using best worst method. Mater. Today Proc. 47(17), 5766–5770 (2021) Zhang, Z., An, J.: Comparative study on the order degree of organization structure synergy in coal mine under-well project. J. Mines Metals Fuels 209, 803–818 (2016)

A Preliminary Implementation of Data-Driven TPM: A Real Case Study Sara Antomarioni, Laura Lucantoni, Filippo Emanuele Ciarapica, and Maurizio Bevilacqua(B) Università Politecnica delle Marche, Ancona, Italy {s.antomarioni,l.lucantoni,f.ciarapica,m.bevilacqua}@univpm.it

Abstract. Total Productive Maintenance (TPM) is one of the methodologies widely used to increase the availability of existing facilities reducing downtimes, stops, and defects, by improving manufacturing methods, usage, and maintenance equipment. Considering the large amount of data currently available thanks to the Industry 4.0 (I4.0) digitization processes, extending the analysis performed in TPM with the support of new techniques and tools is interesting. More in detail, in this work, Association Rule Mining (ARM) is used to identify the hidden relationships between different failure events, allowing their monitoring and prediction and improving the continuity and resilience of the production flow. The final aim pursued by the proposed approach is the development of a maintenance strategy in order to improve the Overall Equipment Effectiveness (OEE) of the selected process. An example case based on real data from an automotive company is used to present the approach and extract proper knowledge from its implementation.

1 Introduction and Background Maintenance is nowadays considered one of the main strategic business activities for company performance improvement within Lean Production (LP) (Lucantoni et al. 2019): it is particularly useful to ensure the continuity of the production flow from a resilient perspective. Among LP practices, TPM is one of the most widely applied methodologies to increase the availability of existing facilities. Specifically, TPM has a relevant role in reducing stoppages, wastes, and defects and promoting employee participation in operation and maintenance (Au-Yong et al. 2022). TPM is usually combined with OEE assessment to find the cause of low values and provide suggestions for improvements (Sukma et al. 2022), paving the way towards perfect production. The current maintenance management systems, however, need a certain degree of personalization since their main features do not meet the requirements of each company when dealing with a wide amount of data (Lopes et al. 2016). Within TPM, Planned Maintenance is widely regarded in the literature as the main pillar (Morales Méndez and Rodriguez 2017): its main weakness can be recognized in the fact that it relies on the historical failure rate of the equipment but does not include any probability measure (Adesta et al. 2018). However, Predictive Maintenance is nowadays extensively used for failure prediction, equipment cost reduction, and performance improvement (Sahal et al. 2020). Novel data-driven techniques are required due to the large amount of data © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 14–22, 2023. https://doi.org/10.1007/978-3-031-25448-2_2

A Preliminary Implementation of Data-Driven TPM

15

available for knowledge extraction (Antomarioni et al. 2021). In parallel, some authors highlighted how Lean Automation can be applied for the concomitant implementation of I4.0 technologies into LP practices, even though the complexity of IT infrastructure necessary to fully integrate I4.0 into TPM could make such adoption less desirable (Tortorella et al. 2021). In line with this perception, despite I4.0 being one of the primary paradigms of the current industrial context (Marcucci et al. 2021), the current literature appears poor on how I4.0 techniques can really support LP principles and practices (Ferreira et al. 2022) showing that more research is needed in this area. One of the few examples in the existing literature presents that data mining techniques, such as ARM, can be integrated with traditional Pareto Chart and Ishikawa diagrams or network analysis in order to assess the magnitude of the production losses and identify the related causes within TPM (Djatna and Alitu 2015; Antomarioni et al. 2022). Considering the existing research gap and the opportunities related to the importance of this research field, the focus of the proposed application is based on relating a metric derived from the well-known Failure Modes Effects Analysis – namely, the Risk Priority Number (RPN) - and ARM: from a practical point of view, they will be used to prioritize failure events; from a theoretical point of view, the aim of the proposed research approach is bridging the existing lack of research in this area through a novel data-driven approach. More in detail, RPN is used to identify the risk associated with each failure mode, considering the current best practices implemented in the company object of the study. Through ARM, instead, the hidden relationship existing between the occurrence of different failure events will be investigated. The last goal is to propose improvement actions that benefit the TPM strategy, improving the OEE and the continuous process flow. A case study from the automotive industry has been used as a pilot project to explain the proposed research approach. In the rest of the paper, a general explanation of the proposed approach is provided in Sect. 2, while Sect. 3 contains its application to the case study. Conclusions and future research directions are drawn in Sect. 4.

2 Data-Driven TPM Approach In order to introduce an effective data-driven TPM strategy in manufacturing, the proposed methodology can be summarized as in Fig. 1. Three main steps can be identified in carrying out such an application, as explained in the following sub-sections. 2.1 Data Collection and Pre-processing Data collection and pre-processing: data represent the basis for an effective maintenance strategy; thus, this module is the foundation of the developed approach. It is fundamental to be able to access data from different sources and integrate them into a unique and reliable dataset. Indeed, the quality of the whole process relies on the quality of data, and the correctness of the decision that will be made is strictly related to them.

16

S. Antomarioni et al.

2.2 Data Analytics Data analytics: The analytics phase is carried out on the integrated dataset produced in the previous step. It mainly consists of two further sub-steps: RPN calculation and Association Rule Mining. Firstly, indeed, Failure Mode and Effect Analysis (FMEA) is carried out to identify any possible failure modes in the production processes and the related RPN values. At this point, ARM is implemented to identify the failure events often occurring concurrently. The analysis can be limited to those failure modes having a high value of RPN, i.e., the ones that are considered more critical by the company or could be extended to the whole set of the identified failure modes. The objective will be, at this point, determining which are the failure modes frequently occurring concurrently: indeed, ARM aims to identify the relations among attributes and values stored in large datasets that frequently co-occur (Buddhakulsomsiri et al. 2006). A SSOCIATION RULE M INING In the following, a formal definition of the ARM process is provided: given a set of items (i.e., Boolean data) I = {ι1 , ι2 ,…, ιn } and given a set of transactions T = {τ 1 , τ 2 ,· · · , τ m } each of whom is composed by an itemset included in I. An Association Rule (AR) α → β can be defined as an implication between itemsets - α and β - belonging to I (α, β ⫅ I) and having no elements in common (α ∩ β = ∅). ARs’ quality is determined through the calculation of different metrics. The Support (Supp) (1) and the Confidence (Conf) (2). Basing on them, ARM aims to identify relationships between failure events and select the ones requiring urgent and essential interventions. The association rules reported in Table 1 have been extracted through the ARM application. The co-occurrence of the failure events is obtained and, through them, decisions benefiting the continuous flow of the production can be made, prioritizing the rules having the highest Supp and Conf. #(α, β) (1) Supp(α → β) = #T Supp(α → β) Conf (α → β) = (2) Supp(α) 2.3 Decision Making As a third step of the proposed approach, the main criticalities of the production process can be identified through the information provided by the ARM. Thus, an Eisenhower matrix is filled to classify the main criticalities and, most importantly, prioritize them. Such a matrix is built considering the RPN of the failure modes and the relationships identified through the ARM; its aim is to allow to classify the critical failure modes and prioritize components for maintenance intervention, defining appropriate preventive strategies. In addition, when a failure event occurs, the occurrence of the related failure modes should be inspected as a priority, in order to be able to intervene promptly.

3 Data-Driven TPM Implementation The production system is an assembly line of an automotive company, composed of twelve fully automated stations. In standard conditions, the line operates during the whole

A Preliminary Implementation of Data-Driven TPM

17

Fig. 1. Research approach

day, considering three shifts of eight hours each. It daily produces about 3,400 pieces (140 units per hour). The preventive maintenance system currently in place has optical sensors controlling parts position, manual operator checks, and planned maintenance for early equipment replacement every 1,000 parts produced. However, this strategy is currently not effective since unwanted failure events and stoppages of the production flow often occur, requiring immediate corrective interventions. In the proposed case study, two data frames regarding daily production data and failure events are used to build the dataset. In all, it contains 1,122 integrated records referring to a time interval of six months. An extraction of the data frames with only the main columns is reported below.

Fig. 2. Dataset for the analytics process

In order to build a reliable dataset for analysis, cleaning and standardization processes are carried out, removing all the inconsistencies and missing values. Specifically, empty columns (e.g., lack of operators, blackout, strike, etc.) were directly removed, while attributes with empty or negative values were well-analyzed and corrected through

18

S. Antomarioni et al.

brainstorming with line experts. In addition, the difference between the Current availability and the Operating time values of DataFrame1 was compared with the sum of all values in the Min/pcs column of DataFrame2 having the same merging key, also considering any further intermediate values in the first dataframe. Finally, both the Failure/downtime and Description columns of Dataframe2 were analyzed in order to ensure consistent nomenclature. Lastly, 919 failure or downtime events are recorded during the monitored time interval. The resulting time dedicated to maintenance interventions corresponds to 28,417 min. In order to verify whether the implementation of the approach can be considered successful, the Overall Equipment Effectiveness (OEE) will be monitored. The as-is OEE is calculated daily with an average of 67%. The analytics phase is carried out on the integrated dataset (see Fig. 2 for an excerpt). It mainly consists of two further steps: RPN calculation and Association Rule Mining. Firstly, Failure Mode and Effect Analysis (FMEA) is carried out to identify any possible failure modes in the production processes and RPN values associated with them. Taking into consideration the production process and the company’s best practices, numerical ranges have been defined to classify the criticality of the RPN: excellent-good from 1 to 10, good-sufficient from 10 to 100, and sufficient-poor from 100 to 1000. According to this classification, 53% of the fault events showed an RPN higher than 10, which is the threshold value and required further investigation. At this point, ARM is implemented to identify the failure events often occurring concurrently when the RPN is higher than the identified threshold. It should be noticed that the selected threshold resulted in being valid for the proposed application, while other case studies or different processes could increase it, decrease it or extend the analysis to the whole dataset. The ARs are then mined using the integrated dataset as a starting point, but, as mentioned before, excluding the events concerning failure modes having an RPN under the threshold. Table 1 shows an excerpt of the results obtained: it can, for example, be noticed that failures 43 and 49 occur together in 3% of cases since the support of both the rules 43 → 49 and 49 → 43 is 3%. Conversely, when failure 43 occurs, 49 verifies in 24% of cases (Conf(43 → 49) = 24%)); the opposite rule, instead, indicated that, when failure 49 occurs, the probability of occurrence of failure 43 is 29%. Once the ARs have been mined, the Eisenhower matrix can be filled: Table 2 displays how the association rules are represented in it, so that prioritization can be made. First of all, the failure modes whose RPN is under the defined threshold are inserted into the non-important and non-urgent quadrant, since they are not object of the current study: forty failure modes are then excluded from the rest of the analysis. For the classification of the remaining ones, the ARs are used to fill the matrix: if a failure mode has an RPN above the defined threshold and it appears in the ARs mined both on the left- and right-hand side, then it is classified as urgent and important; if it appears in the left- or right-hand side of the ARs, then it is considered important but not urgent; if it does not appear in any rule, then it is not considered important, even though it can be urgent. The non-importance of the failure modes is related to the fact that they do not appear to be triggered by the occurrence of other failure events and they do not trigger others

A Preliminary Implementation of Data-Driven TPM

19

Table 1. Some of the association rules among failure events with RPN above the threshold value X

Y

Supp

Conf

26

→

46

5%

32%

43

→

47

4%

35%

43

→

49

3%

24%

46

→

26

5%

29%

46

→

47

4%

25%

49

→

43

3%

29%

49

→

47

4%

36%

59

→

26

5%

24%

59

→

46

4%

21%

59

→

47

9%

41%

91

→

46

5%

41%

91

→

47

4%

35%

either. The most critical failure modes in the upper red area are identified, namely, those requiring urgent action. Then, the extraction of the graphical results of the methodology is shown in Fig. 3. Table 2. Criteria for the failure modes prioritization RPN

ARs

Eisenhower matrix

Above threshold

“X” AND “Y”

Urgent and Important

Above threshold

“X” OR “Y”

Important and Non-Urgent

Above threshold

No

Urgent and Non-Important

Under threshold

Not analyzed in this work

Non-Urgent and Non-Important

In the upper right area, urgent and important failure modes are prioritized, namely, those with an RPN index above the threshold revealed by the FMEA analysis and closely related to each other as revealed by the ARs. In this way, a data-driven strategy is defined to select which improvement strategies should be prioritized. TPM pillars are first applied to the failure modes considered urgent and important in order to mitigate their occurrence and anticipate their causes. Table 3 shows the improvement actions taken. From a continuous improvement perspective, the research approach is expected to be iteratively implemented, in order to gradually improve the quality of the overall process. After the implementation of such measures and an observation period of two months, an average improvement of the OEE by 2% has been achieved and a significant reduction of the failure events (about 70%). They can both be related to the actions taken and also to the monitoring of the process through

20

S. Antomarioni et al.

Fig. 3. Eisenhower matrix: failure modes classification and prioritization

the ARs results: indeed, when a failure event happens, the occurrence of the related failure modes should be inspected as a priority, in order to be able to intervene promptly. For instance, in the event of the occurrence of failure mode 59, from Table 1 we can see that three further failure modes could happen, i.e., 26, 46, and 47. Considering the confidence values of the association rules, it is evident that failure mode 47 is the most likely one (Conf(59 → 47) = 41%), followed by number 26 and 46 (Conf(59 → 26) = 24%; Conf(59 → 46) = 21%). In this way, preventive replacement of components can be performed when the probability is high (e.g., in case of number 47), while an inspection could be enough for the remaining ones. Going into more detail of the results obtained, the 70% reduction in failure events is due to addressing those in the urgent and important quadrant (28–43–46–47–49–59–91). Our interest was to reduce the failure events of these modes being classified as the most urgent so 70% can be considered as satisfactory. OEE is, however, also affected by all other events, as well as other parameters such as, the item “cycle time different from standard” which is part of Performance losses. Longer monitoring would have allowed a broader view of the results obtained. In general, based on these considerations, while the reduction of mode-specific failure events is visible right away, the improvement of OEE definitely requires longer time and one with the continuous implementation of the methodology, as well as for continuous learning by operators. That said, the implementation of the proposed methodology was also matched by a change of suppliers related to the current difficulties in the supply environment, which caused a deterioration in the quality of delivered materials and, consequently, a further

A Preliminary Implementation of Data-Driven TPM

21

OEE penalty. Excluding this additional OEE penalizing factor, the overall improvement achieved as a result of the implemented approach was 10%. Table 3. Improvement actions to mitigate the Important & Urgent failure modes Failure mode code

Solution

FM 28

Production line refilling improvement

FM 43

Substitution of robot gripper sensors

FM 46

Production line refilling improvement

FM 47

Production line refilling improvement

FM 49

Programmed maintenance interval reduction

FM 59

Programmed maintenance interval reduction

FM 91

Training for workplace organization

4 Conclusions The final goal of the proposed methodology is the identification of improvement actions for failure events prioritization in the field of lean automation. The added value deal with the application of a new data-driven approach as the I4.0 technique in a real case study supporting the TPM implementation and OEE improvement allowing a continuous production flow. After implementing the proposed data-driven TPM methodology, the results have been monitored for two months and four of the eight TPM pillars have been achieved in the short term: Planned Maintenance, avoiding recurrent failure events of FM49 and FM 59 keeping the equipment more operational, Quality Management reducing defects caused by FM 43 keeping the system more performant, Education and Training empowering maintenance operators about the new data management system and the workplace organization to avoid FM91, Autonomous Maintenance actively involving operators in minor maintenance tasks for the regular management of equipment as to cope with the FM 28, 46 and 47. In addition, it should be emphasized that Continuous Improvement may also be achieved through the periodic application of the proposed methodology. In conclusion, an average OEE improvement and reduction in the occurrence of failure modes important and urgent have been obtained due to the actions taken. When considering the improvement actions, the focus should surely be on the technical arrangements of the production systems, as well as on improving the policies currently in place. However, the training of the maintainers is fundamental in carrying out this approach since it ensures an improvement in operations quality. Future research direction will focus on iteratively extending the proposed case study, also focusing on the remaining TPM pillars and on the failure modes not addressed in this study. With a view to exploiting association rules for monitoring the production system, a time dimension will be added in future developments to provide a more precise indication of the time at which it is necessary to be ready to perform preventive replacements.

22

S. Antomarioni et al.

Acknowledgment. This work was supported in part by the European Commission under the Project Reference No. 2021-1-IT01-KA220-VET-000033287. The authors would like to acknowledge the support of the researchers participating in the project “Business Continuity Managers Training Platform” (CONTINUITY) (business-continuity-project.eu).

References Adesta, E.Y.T., Prabowo, H.A., Agusman, D.: Evaluating 8 pillars of Total Productive Maintenance (TPM) implementation and their contribution to manufacturing performance. In: IOP Conference on Series: Materials Science and Engineering (2018) Antomarioni, S., et al.: Data-driven decision support system for managing item allocation in an ASRS: a framework development and a case study. Expert Syst. Appl. Pergamon 185, 115622 (2021). https://doi.org/10.1016/J.ESWA.2021.115622 Antomarioni, S., Ciarapica, F.E., Bevilacqua, M.: Association rules and social network analysis for supporting failure mode effects and criticality analysis: framework development and insights from an onshore platform. Saf. Sci. (2022) Au-Yong, C.P., Azmi, N.F., Myeda, N.E.: Promoting employee participation in operation and maintenance of green office building by adopting the total productive maintenance (TPM) concept. J. Clean. Prod. (2022) Buddhakulsomsiri, J., Siradeghyan, Y., Zakarian, A., Li, X.: Association rule-generation algorithm for mining automotive warranty data. Int. J. Prod. Res. 44(14), 2749–2770 (2006) Djatna, T., Alitu, I.M.: An application of association rule mining in total productive maintenance strategy: an analysis and modelling in wooden door manufacturing industry. Procedia Manuf. (2015). https://doi.org/10.1016/j.promfg.2015.11.049 Ferreira, W.D.P., et al.: Extending the lean value stream mapping to the context of Industry 4.0: an agent-based technology approach. J. Manuf. Syst. (2022) Lopes, I., et al.: Requirements specification of a computerized maintenance management system - a case study. In: Procedia CIRP. Elsevier B.V., vol. 52, pp. 268–273 (2016). https://doi.org/ 10.1016/J.PROCIR.2016.07.047 Lucantoni, L., Bevilacqua, M., Ciarapica, F.E.: An enhanced approach for implementing riskbased maintenance in a total productive maintenance perspective. In: XXIV Summer School “Francesco Turco” – Industrial Systems Engineering (2019) Marcucci, G., et al.: The impact of operations and IT-related Industry 4.0 key technologies on organizational resilience. Prod. Plann. Control (2021). https://doi.org/10.1080/09537287.2021. 1874702 Morales Méndez, J.D., Rodriguez, R.S.: Total productive maintenance (TPM) as a tool for improving productivity: a case study of application in the bottleneck of an auto-parts machining line. Int. J. Adv. Manuf. Technol. 92(1–4), 1013–1026 (2017). https://doi.org/10.1007/s00170-0170052-4 Sahal, R., Breslin, J.G., Ali, M.I.: Big data and stream processing platforms for Industry 4.0 requirements mapping for a predictive maintenance use case. J. Manuf. Syst. Elsevier 54, 138–151 (2020). https://doi.org/10.1016/J.JMSY.2019.11.004 Sukma, D.I., Prabowo, H.A., Setiawan, I., Kurnia, H., Fahturizal, I.M.: Implementation of total productive maintenance to improve overall equipment effectiveness of linear accelerator synergy platform cancer therapy. Int. J. Eng. Trans. A: Basics (2022) Tortorella, G., et al.: Towards the proposition of a lean automation framework: integrating industry 4.0 into lean production. J. Manuf. Technol. Manag. (2021)

Assets’ Reliability Management Model for a Decision Making in Different Operational Contexts Jone Uribetxebarria, Ainhoa Zubizarreta, Ángel Rodríguez, and Urko Leturiondo(B) Ikerlan Technology Research Centre, Basque Research and Technology Alliance (BRTA), 20500 Arrasate-Mondragon, Spain {juribetxebarria,azubizarreta,arodriguez,uleturiondo}@ikerlan.es

Abstract. This article describes a general framework that supports a holistic reliability management model to ensure optimal performance of critical assets over their lifetime in different operational contexts. This allows to achieve the expected levels of availability and consequently to optimize the total cost of ownership. The proposed approach focuses on studying the behaviour of systems/components in different operational contexts and identifying the critical parameters of the context that most impact on the expected reliability. To support the decision-making process an agent-based simulation model has been developed, which enables to build a virtual model and assess each maintenance strategy’s impact on availability and lifecycle costs. It will allow to make a more accurate prediction of the asset’s performance, strengthen systems in more aggressive operational contexts and optimising maintenance strategies to real needs, avoiding oversizing or redundancy units.

1 Introduction Industry 4.0 has brough innovative sensing and detection technologies to equipment and systems but improving Overall Equipment Efficiency (OEE) remains a must in industrial plants or services such as transport. Assets are designed to operate at optimal levels of reliability and performance but are often unable to achieve them because they operate in demanding operational contexts with high levels of stress not considered at the design stage. This can lead to significant cost overruns in production processes and services. According to Rocquigny et al., (2008) important factors such as real usage or the operational context in which an asset operates (temperature, humidity, etc.) affect its reliability, and not considering them can lead to wrong decisions on asset maintenance or replacement lead times. In this context, the asset management framework (Crespo et al. 2009) enables improving availability levels by considering different technical solutions and operational contexts (Izquierdo et al. 2019). In addition, virtual models created from simulation techniques and analytical models and optimization algorithms help the manufacturer to analyse in detail the risks and opportunities of a service offering (Erguido et al. 2018). This article presents an integrated reliability management framework where the actors of the mobility value chain collaborate in a cyclic process of learning and managing reliability. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 23–33, 2023. https://doi.org/10.1007/978-3-031-25448-2_3

24

J. Uribetxebarria et al.

2 Methodology The reliability of an asset is the consequence of the reliability of systems and components of which it is composed, so its management requires the collaboration of all actors in a value chain, from service operators and manufacturers to component suppliers. The Reliability Management Framework introduced in this article helps value chain actors to study systems/components behaviour in real operating context through the Reliability Learning & Assessment phase. From the acquired knowledge, the Reliability Management phase enables them to improve decision for optimizing the performance of an equipment /asset along its lifetime.

Fig. 1. Reliability along the value chain.

Figure 1 presents both phases as part of a cyclic process, in which the output of each phase acts as input of the other continuously. Value chain actors play a key role in this framework, as they must be closely involved in both phases. The actors showed in the figure belong to a tram service value chain, which will serve as the use case present in a later section of the article. The actors of the value chain collaborate in gathering the knowledge, and they apply this acquired knowledge in decision making during the Reliability Management phase. They may improve the reliability of systems/components in each operational context, acting in different stages of the life cycle: from the early design stage (where redesigns can be released), through the use stage (defining the optimal maintenance strategies), and onward. The proposed methodology studies and models the behaviour of the system/component in a real context and generates virtual models that evaluate the impact of different strategies that could be implemented. The Reliability Management Framework presented in Fig. 2 shows the stages covered by each phase. Phase 1 is divided in 4 sequential stages. The first one (1.1) involves the acquisition of operational data, specific sensor data and historical failure data which will be used later to fit the reliability models. The integrity and quality of the data collected at this stage will have a direct bearing on the results provided by the final models developed. Data gathered from several systems/components with the same design and technical characteristics often show highly variable behaviours when they operate in different

Assets’ Reliability Management Model for a Decision Making

25

Fig. 2. Reliability management framework.

context. That is why it is essential to identify the context parameters that most impact on the intrinsic basic reliability of the design, which is performed in the second stage (1.2) by applying machine learning algorithms (regression, decision trees, etc.). In the third stage (1.3), the knowledge acquired on the behaviour of systems/components allows the development of accurate reliability models customized to each operational context. They can be used to continuously assess the risk of failure and, at any onset of degradation/deterioration, they can be used to calculate the Remaining Useful Life (RUL) of the component. Two types of models are employed. Regression survival models are used to estimate the probability of failure in each context and predictive models (based on physics, data or a mixture of them) improve the diagnostic and prognostic processes and assess the RUL. The last stage (1.4) deals with the strategic viewpoint. Operations and Maintenance (O&M) strategies that could be employed under both normal and unforeseen conditions are identified, considering resources and other constraints. Reliability models together with O&M strategies allow the creation of a virtual model of the whole system in phase 2. It mirrors the industrial process under study and enables learning from observed behaviour in historical and/or simulated scenarios. Optimization techniques and sensitive analysis are applied to ensure optimal balance between availability and Total Cost of Ownership (TCO). Behavioural patterns analysis and O&M strategies impact assessment are critical in this phase.

3 Case Study The methodology presented in the previous section has been used to improve the performance of a transport service. In phase 1, reliability models have been generated based on field data gathered from the monitoring of several geographically distributed tram services. The data consists of sets of failure events associated to the components of different trams. For each event, the information from the operating context is recorded as well. Two types of reliability models have been developed to assess systems/components

26

J. Uribetxebarria et al.

health condition: a predictive model, used for critical failures that can cause serious consequences, and a more generalist risk model, based on statistical techniques and regression methods. The predictive model assesses damage based on physical degradation and estimates the RUL of a system/component. Predictive models, based on physical degradation, data driven or a combination of both, allow to continuously assess the health of the components of a tram and as a result to estimate their Remaining Useful Life (RUL). There is a wide range or predictive models’ development techniques that are beyond the scope of this article. In this work, a simple exponential degradation model shown in Fig. 3 has been employed. It has been assumed that damage becomes measurable after a period of time since the last replacement (excluding initial or random failures) and that the error in the prediction of the RUL is normally distributed. One focus of the use case deal with the importance of prediction error in making replacement decisions.

Fig. 3. Predictive model

On the other hand, statistical risk models provide less accurate failure risk information, but they only need historical field data gathered in different operational contexts as input. Based on a preliminary analysis, the variability in the failure behaviour of tram services that share the same design suggests that the operational context has a significant impact in reliability. Therefore, the analysis of the relationship between the variables of the operational context and the failure of components requires the use of survival regression models. The Cox’s hazard model (Fig. 4) is a survival regression model that considers that the log-hazard of an individual is a linear function of some covariates and a population-level baseline hazard that changes over time. The model is semiparametric, so it can be fit without assuming or knowing a particular distribution. This model allows to estimate the risk of failure of components considering its initial design reliability and the impact of the parameters of the operational context. Additionally, the most important parameters can be obtained by fitting univariate models to the experimental dataset.

Assets’ Reliability Management Model for a Decision Making

27

Fig. 4. Cox’s hazard reliability model

The operational context characterisation has been performed considering several external and internal environmental parameters to the critical components. In addition, several alternative O&M strategies have been envisaged for the tram service. Based on failure risk and RUL information and other tram service factors, decision on replacement action can be taken. In phase 2, Reliability Management, a cost analysis is performed. Only Lifecycle Costs (LCC) directly impacted by reliability that are critical for TCO are considered: costs related to preventive and corrective maintenance actions, and costs concerning penalties (delays, low comfort and stoppage on service). An agent-based simulation model has been developed to study the impact that the reliability models have on tram services under different operating contexts and so to optimize the balance between availability and LCC over a 4-year period. Figure 5 shows the developed virtual environment that emulates the behaviour of a city tram service. A set of trams fulfil the services mobility requirements, transporting people between stops along a specific route according to a timetable. The service performance is calculated from availability (extra tram utilisation rate), service delays, comfort level and cost deviations. Each tram has the capability to calculate the RUL of its critical components based on degradation models and to estimate the failure risk based on the statistical model. Based on the RUL and risk calculations, the simulation model estimates potential service impacts in different scenarios, improving decision making on O&M strategies.

28

J. Uribetxebarria et al.

Fig. 5. Virtual simulation model.

As shown in Fig. 6, the virtual simulation model architecture considers different types of agent populations that replicate the specific behaviour of the tram service elements and can interact between them.

Fig. 6. Simulation model architecture

Tram agents and stations agents interact through a rail network. Station agents have a passive role, setting requirements and needs and controlling tram arrivals and departures. Tram agents can assess their status and risk of failure in a comprehensive manner based on predictive models and risk models (which emulate the reliability observed in the data gathered from the tram services). That information allows that the tram agents implement

Assets’ Reliability Management Model for a Decision Making

29

local strategies to ensure passenger safety: delay tram departures to avoid collisions, or slowdown in case of minor failures or stops in case of major failures. The coordinator agent can decide when a tram agent needs maintenance during or after daily service. In case of major failure, the damaged tram must be replaced with an extra tram. Each of these actions impacts on the global cost and the performance metrics of the service. The application of Monte Carlo experiments & Sensitivity Analysis on the virtual model allows estimating a range of output key performance indicators (service performance and LCC), mainly due to uncertainty arising from reliability models. As shown in Fig. 7, multiple challenges can be addressed by considering and simulating different scenarios for each of them.

Fig. 7. Study of alternative scenarios

Estimating the impact of different O&M strategies in alternative scenarios improves decision making by allowing the adoption of strategies that optimise the balance between service performance and LCC at any given stage. In the use case, three challenges have been addressed: Challenge 1: Estimate the variability of LCC in different operational contexts. An operational context change may lead to significant changes in the behaviour of a component. Decision making on component’s replacement based on generic reliability model’s risk (not customized to operational context) may lead to cost overruns due to well in advance in component replacements (risk estimated higher than the actual) or non-expected major correctives (risk estimation lower than the actual). Based on a Monte Carlo experiment, a set of simulations have been launched tuning the reliability models to different operating contexts.

30

J. Uribetxebarria et al.

Fig. 8. LCC variability analysis.

Figure 8 shows that costs fluctuate significantly under different operation contexts. Therefore, for optimal comprehensive risk assessment it is critical to make decisions on customized reliability models that take into account the impact of the most important parameters of the operational context. Challenge 2: Optimize maintenance strategies linked to predictive models and statistical risk models. In this challenge, predictive maintenance strategies have been compared with maintenance strategies based on statistical risk models. The overall goal is to strike a balance between cost overruns or savings due to replacements made too early and consequences caused by major correctives.

Fig. 9. Replacement strategy impact (Predictive model).

Figure 9 shows the variability in LCC produced by changing the component with different lead time windows from an RUL estimation by the predictive model. According to the results, a very conservative maintenance strategy should be avoided, scheduling

Assets’ Reliability Management Model for a Decision Making

31

component replacement near the end of its useful life, without causing service disruptions (maintenance action taking place only at the end of the daily tram service). The optimal value for replacement is at 80% of its expected lifetime from time of detection (20% of RUL), since exceeding this threshold can cause unexpected cost overruns.

Fig. 10. Replacement strategy impact (Failure risk model).

Figure 10 shows the results obtained by the strategy based on the failure risk model. The optimal value for replacement is when a preventive replacement is triggered before exceeding 40–50% probability of failure (based on the survival curves modelled from the park data). As the statistical model describes the survival of a set of individuals, the physical damage of each individual asset is not known, so decision making must be more conservative than in the previous case (predictive models). Challenge 3: Evaluate and select the O&M strategies that best balance functionality and LCC, considering operational needs. This challenge is aimed at optimising O&M strategies at tram service level, to ensure its optimal performance level beyond local optimisations focused on each failure type. Any decision on a replacement or maintenance affects the whole service, so it may be necessary to adopt smarter strategies to ensure optimal availability and LCC level of the service. Service O&M strategies must avoid major failure events and optimize the replacement decision, considering the individual failure risk, the possibilities of synchronisation with other maintenance actions and other operational needs of the service. In this last challenge, different strategies are implemented and compared based on the costs obtained: • Non-optimized replacement strategies for the predictive and risk models. • Optimized replacement strategies for the predictive and risk models. • Flexible strategy that synchronizes predictive and risk models to achieve LCC reduction.

32

J. Uribetxebarria et al.

Fig. 11. LCC improvement based on O&M strategies.

Figure 11 shows that the LCC linked to tram service can be significantly reduced by improving replacement decision making based on failure risk and RUL information, using flexible strategies to leverage the use of critical resources or other service-related factors.

4 Conclusions New detection and monitoring capabilities and maintenance field data analysis enable the development of more accurate reliability models to describe systems and components behaviour in variable operating contexts. Nowadays, due to asset complexity and resource constraints, different reliability modelling techniques can be applied and combined: from advanced and accurate predictive models based on physics laws or specific field knowledge to more generic statistical models. This article presents a methodology to manage assets in a comprehensive way, considering all the available information about the behaviour of the systems under study. Once the reliability behaviour of the system is analysed and a comprehensive health assessment is performed, optimal O&M strategies can be implemented during its lifetime to improve the LCC of the service. In this sense, optimisation through virtual models, using simulation and optimisation techniques, is becoming an increasingly important tool in the LCC calculation and critical resource sizing of a service or for the improvement of its performance. Funding. This topic has been addressed under Horizon 2020 research program in iREL (Intelligent Reliability 4.0), a European co-funded innovation project that has been granted by the ECSEL Joint Undertaking (JU) under grant agreement No 876659.

Assets’ Reliability Management Model for a Decision Making

33

References Crespo, A., Moreu, P., Gómez, J.F., Parra, C., López, M.: The maintenance management framework: a practical view to maintenance management. J. Qual. Maintenace Eng. 15(2), 167–168 (2009) Erguido, A., Crespo, A., Castellano, E., Flores, J.L.: After-sales services optimisation through dynamic opportunistic maintenance: a wind energy case study. Proc. Inst. Mech. Eng. Part O: J. Risk Reliab. 232(4), 352–367 (2018) Izquierdo, J., Crespo, A., Uribetxebarria, J.: Dynamic artificial neural network-based reliability considering operational context of assets. Reliab. Eng. Syst. Saf. 188, 483–493 (2019) Rocquigny, E., Devictor, N., Tarantola, S.: Uncertainty in Industrial Practice: A Guide to Quantitative Uncertainty Management. Wiley, Hoboken (2008)

Realizing Sustainable Value from Engineering Innovation Ecosystems in EURope’s Outermost Regions Oliver Schwabe1(B) and Nuno Almeida2 1 Instituto Superior Téchnico (IST), Lisbon, Portugal

[email protected] 2 CERIS, IST, Lisbon, Portugal [email protected]

Abstract. The Outermost Regions are remote territories of EU Member States and emergent innovators seeking to derive sustainable value from the 4th industrial revolution through their Engineering Innovation Ecosystems (EIEs). They face unique challenges for creating sustainable value, which is realized when late adopters within such EIEs are reached, and their intent aligns with Citizen Good Health and Well-Being. This paper presents the initial results of experimental inductive case-study research examining the conditions under which sustainable value can be achieved in the EIEs of the Azores and Madeira (Portugal), the Canary Islands (Spain) and La Réunion (France). The research applies Qualitative Comparative Analysis with Ecosystem, Innovation Diffusion, Sustainable Development, and Intellectual Capital assessment methods.

1 Introduction Specific EU (sub-) territories that struggle to significantly to improve their economic capability due to their geographic isolation from continental Europe have a special status as “Outermost Region” (OR)1 . The ORs are parts of three EU Member States with territories that are geographically very dispersed and isolated from continental Europe. There are nine of them: French Guiana, Guadeloupe, Réunion, Mayotte, Martinique, and Saint-Martin (France), the Azores and Madeira (Portugal) and the Canary Islands (Spain), spread across two oceans: the Atlantic and the Indian. They are an integral part of the EU and the European Strategy for the ORs encourages these regions to make use of their assets through research and development in growth-enhancing areas (such 1 Of similar character are EU Overseas Countries and Territories (OCTs); Aruba (NL), Bonaire

(NL), Curação (NL), French Polynesia (FR), French Southern and Antarctic Territories (FR), Greenland (DK), New Caledonia (FR), Saba (NL), Saint Barthélemy (FR), Sint Eustatius (NL), Sint Maarten (NL), St. Pierre and Miquelon (FR), Wallis and Futuna Islands (FR). The (scattered) island context is also shared by Cyprus (CY), Malta (MT), and Iceland (IS), the islands of the EU B7 Baltic Islands Network (Åland (FI), Bornholm (DK), Gotland (SE), Hiiumaa (EE), Öland (SE), Rügen (DE), and Saaremaa (EE), as well as the Mediterranean regions Baleares (ES), Corsica (FR), Crete (Greece), Notio Aigaio (GR), and Sardinia (IT). © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 34–44, 2023. https://doi.org/10.1007/978-3-031-25448-2_4

Realizing Sustainable Value from Engineering Innovation Ecosystems

35

as the blue economy) and the areas defined in their Smart Specialisation Strategies. The ORs give Europe a geostrategic, economic, and cultural position alongside several continents and provide it with unique benefits. These benefits have intrinsic added value and underpin EU cooperation with third countries, even outside their geographical areas. They contribute to Europe’s international influence and offer unique potential for implementing solutions to the challenges facing the EU (Destom 2021). Despite having a population of only almost five million inhabitants, the ORs add significantly to the maritime dimension of the European Union, making it the world’s largest maritime area with over 25 million km2 of Exclusive Economic Zone. 80% of Europe’s biodiversity is in the ORs and the European Overseas Countries and Territories. The ORs have been identified as emerging innovative regions, a classification given to regions whose performance in terms of innovation is below 50% of the European Union average. Because of their insularity, they also suffer more from marine pollution than any other mainland region of Europe and this affects their environmental and economic development. The school drop-out rate among young people in the ORs, regrettably, is significantly higher than the EU average. There are limited training and education opportunities, a large share of young people that are not graduated and unemployment rates. ORs thus present major economic opportunities and have an extremely rich cultural heritage that helps to extend Europe’s influence in the world but also major challenges that need to be addressed. The complex and reactive business environments of these regions call for tailored approaches to the local innovation ecosystems, which can capitalise on these opportunities and overcome these challenges and help strengthen partnerships with the EU outermost regions”. Outermost Regions (ORs) are autopoietic place-based nested Innovation Ecosystems (IE) (Allee and Schwabe 2015; Amidon et al. 2005; Mercier-Laurent 2015; Varela et al. 1974). The research explores the mechanism of value realization for the Citizens in these regions. Value is understood as the (tacit) intent to improve the Good Health and WellBeing of Citizens as measured in terms of all Sustainable Development Goals (SDGs) achievement. Figure 1 shows the progress of Azores (1), Canaries (2), La Réunion (3) and Madeira (4) against the exemplary SDG 3 “Good Health and Well-Being” in comparison to all other EU regions. The end value is the aspired position in 2030.

Fig. 1. Progress towards SDG 3 (OECD Webtool)

The IEs of this research are seen as “Transformative Enterprises” (TEs). TEs are the most competitive, sustainable, resilient, and successful type of IE with intensely collaborative, caring, bottom-up and highly moral way of working (Formica 2022). TEs

36

O. Schwabe and N. Almeida

are often invisible to performance measures and thus escape management attention. They are usually aligned around individuals and highly successful in the often hierarchical and conservative OR context. The more individuals in a TE exchange knowledge, the greater its sustainability and positive impact on Good Health and Well-Being of citizens (Allee and Schwabe 2015; Amidon et al. 2005). An exemplary view of participants and their connectivity in a TE is shown in Figs. 2a and 2b. Each node represents an individual and each connector a knowledge exchange (Homans 1958). The “nexus” of collaborative linkages provides the “unfair advantage” of rapid diffusion of innovations needed for high performance (Correia et al. 2021; Rogers 2003; Schwabe et al. 2020a, 2020b).

Fig. 2. a: Un-sustainable TE, b: Sustainable TE

2 Problem Progress of ORs towards SDG goals is significantly below EU average mainly due to their geographical context and invisibility of their TEs. This paper presents the initial results of ongoing experimental inductive case-study based research effort investigating the TEs in four ORs to determine under what conditions the nexus creates the most sustainable value. Case studies were identified by co-authors and are seen as acceptably representative since TEs are of fractal nature. The paper specifically focuses on Engineering Transformative Enterprises (ETEs). Figure 3 illustrates the nested hierarchy of ETEs as a subset of IEs in the ORs within the EU as part of global society. Important to note is that “at the heart” of these social systems there is always a number of individuals interacting in archetypal patterns. Furthermore, important to remember is that value is created for the social systems regardless of the legal entities that individuals might be associated with (Almeida et al. 2022), which by default will point to tensions between the social need to create Good Health and Well-Being of Citizens participating at all levels versus the pressure to realize value within legally bounded commercial entities.

3 Research Objective The objective of the research is to use experimental inductive case-study research to examine the conditions under which sustainable value can be achieved in the EIEs of the Azores and Madeira (Portugal), the Canary Islands (Spain) and La Réunion (France). The research hypothesis will be tested by a thought-experiment based on a research method applied to multiple case studies investigated through primary research activities. The experimental nature of the approach is chosen to help ensure all possible outcomes of

Realizing Sustainable Value from Engineering Innovation Ecosystems

37

Global EU OR IE ETE Indi vidu als Fig. 3. Nested IE hierarchy

the experiment are considered as valid so that a “normalization” process is not (unconsciously) applied. The researchers consider that the unique context of the investigation, as well as its small/scarce data conditions must be assumed to preclude the uncritical extension of previous research.

4 Hypothesis and Thought Experiment The hypothesis investigated is that the denser the exchange of knowledge between individuals in an ETE, the greater the sustainability and positive impact on Good Health and Well-Being of citizens related to that ETE will be. To test the hypothesis a thought-experiment is used as shown in Fig. 4.

Fig. 4. Thought-experiment

38

O. Schwabe and N. Almeida

The thought-experiment considers the diffusion of innovation path from ideation to late market adoption from an initial point in time (Time = 1) to a future point in time (Time = n). An intelligent agent is assumed to open and close a “door” between the time periods to pass changes in density of knowledge exchange between the participants of the ETE between these different points in time, whereby this is a bi-directional phenomenon. The agent nominally automates the activity of the orchestrator of an ETE, whereby this orchestrator also actively facilitates the changes in density of knowledge exchanges during the time periods themselves. The assessment of potential sustainability is then performed through the research method as described in the following section of the paper. Important to note is that the thought-experiment consciously leaves the question open whether the diffusion of knowledge/innovation is a linear process moving only forward in time. The reason for this is that while significant efforts are made to assess value creation in quantitative forms, the robustness of such evaluations depends significantly on the “perceived” (and therefore “qualitative”) perception of change by individual members of the ETE and additionally their collective perception of such. This perception is subject to change over time, so that the assessment at Time = 1, when revisited at a later point in time, may indeed be “perceived” to be very different.

5 Research Method The research method is “Qualitative Comparative Analysis” (QCA) (Mello 2022) which analyses cases in (dynamic) complex (adaptive) situations to help explain why change occur. Its application of set theory respects small data limitations. The change theory is: “IF an ETE has a unique causal condition, THEN it will achieve sustainable valorisation”, therefore continuously improving the Good Health and Well-Being of Citizens over a 25-year strategic time horizon (Sustainability Potential (SP)). Cases of interest are ETE of organizations in an OR. The change factors Solution Maturity (S), Role Maturity (R), Intent (I), and Performance (P) are assessed on a %-scale (Schwabe 2022) and for assessor confidence of the answer accuracy. The combination of the crisp scores gives Causal Codes (CC). SP is determined judgementally and relatively through interviews with experienced investors (3 = High, 2 = Medium, and 1 = Low). Table 1 shows the generic scenarios (s) for all potential CC and SP for the four Change Factors with Fig. 5 showing their relative location in the corresponding set diagram.

Realizing Sustainable Value from Engineering Innovation Ecosystems

39

Table 1. General logic table s

Truth value

CC

SP

S

R

I

P

1

0

0

0

0

0000

1

2

1

0

0

0

1000

2

3

0

1

0

0

0100

2

4

1

1

0

0

1100

1

5

0

0

0

1

0001

2

6

1

0

0

1

1001

1

7

0

1

0

1

0101

1

8

1

1

0

1

1101

3

9

0

0

1

0

0010

1

10

1

0

1

0

1010

3

11

0

1

1

0

0110

2

12

1

1

1

0

1110

2

13

0

0

1

1

0011

1

14

1

0

1

1

1011

3

15

0

1

1

1

0111

3

16

1

1

1

1

1111

3

Fig. 5. QCA framework Venn

6 Results Results derive from assessing the S, R, I & P of 16 cases of interest (12 were ETE) through their organisational websites and selected semi-structured interviews on a (fuzzy) %scale. S, R, I & P scores were then turned into crisp values (default threshold of 1 being

40

O. Schwabe and N. Almeida

> = 0.5). Crisp values were turned into a Causal Code (CC). Case numbers (#) marked with a “x” are SP = High, those marked with “†” are engineering focused and those marked with “*” were assessed based on their websites and interviewees familiar with them only. Case 16 is a comparative study from Germany. Table 2 shows the analysis results. Table 2. Case study analysis resultsa Case#

Fuzzy

Crisp

CC

s

S

R

I

P

S

R

I

P

1*

0.6

0.8

0.2

0.2

1

1

0

0

1001

4

2 x*†

0.6

0.8

0.5

0.4

1

1

1

0

0101

12

3†

0.2

0.2

0.2

0.2

0

0

0

0

1101

1

4 x†

0.8

0.8

0.2

0.7

1

1

0

1

1001

8

5*†

0.7

0.4

0.5

0.5

1

0

1

1

1010

14

6*†

0.7

0.4

0.5

0.5

1

0

1

1

1111

14

7*†

0.7

0.4

0.5

0.5

1

0

1

1

1110

14

8*†

0.7

0.4

0.6

0.5

1

0

1

1

0010

14

9 x†

0.8

0.6

0.5

0.7

1

1

1

1

1001

16

10

0.8

0.2

0.4

0.4

1

0

0

0

1000

2

11 x†

0.8

0.8

0.2

0.7

1

1

0

1

1110

3

12†

0.8

0.6

0.2

0.6

1

1

0

1

1111

3

13 x

0.8

0.6

0.2

0.4

1

1

0

0

1111

4

14

0.6

0.2

0.4

0.2

1

0

0

0

0011

2

15*†

0.8

0.4

0.5

0.4

1

0

1

0

0111

10

16†

0.2

0.5

0.7

0.2

0

1

1

0

1111

11

a Case Studies: (1) Startup Madeira – Societal Actor (Madeira, PT), (2) Oceanográfica - SME

(Canaries, ES)*, (3) Porvalor - SME (Azores, PT), (4) La Palma Research Centre - SME (Canaries, ES), (5) FETCH – SME (Reunion, FR)*, (6) GIP Centre Sécurité Requin – Societal Actor (Reunion, FR)*, (7) IDocean – SME (Reunion, FR), (8) Efuzif – SME (Reunion, FR), (9) Scubanana – SME (Canaries, ES), (10) Pepe Jose – SME (Reunion, FR), (11) TriSolaris - SME (Azores, PT), (12) GeoPlan – SME (Mainland PT), (13) Ponteditora - SME (Madeira, PT), (14) Local Food Culture - SME (Azores, PT), (15) Elitoral – SME (Canaries, ES), (16) SMARTUP – SME (DE)

Five cases (2, 4, 9, 11, and 13) corresponded to SP = High, representing five different scenarios (s) namely 3, 4, 8, 12 and 16. Four scenarios were related to ETEs. The equality of s to the number of cases examined emerged through the research effort and has no methodological reason; indeed, as with any research effort, the more cases of interest that are examined, the more robust any results may be.

Realizing Sustainable Value from Engineering Innovation Ecosystems

41

7 Discussion and Conclusions The change theory put forward was that (E)TE meeting a unique CC with the highest SP (3) will contribute most significantly to the OR IEs accelerating the progress towards aspired end value. The research results however show that each identified case with SP = High corresponded to a different CC, whereby the researcher effort suggests that such may change over time as well. The important insight appears to be that an ETE may be sustainably successful under different CC over time and that this then begs for variable interventions to accelerate the appropriate diffusion to late adopters and sustainable valorisation. The identification of multiple relevant CC also suggests that any data analysis should assume the presence of multiple data centres which precludes the use of general statistical approaches which usually rely on a single data centre. In this respect IE and TE analysis, evaluation and assessment demonstrate that in such emergent environments the foundations of OR innovation development policies and procedures need to be revisited. (E)TEs contribute significantly to SDG achievement. By strengthening the overall scoring for S, R, I & P towards achieving the causal conditions deemed as having the highest SP (and ensuring alignment to and visibility of) the relevant unique indicators, such movement should be sustainably enabled. Improvement of scoring occurs both on the level of individual assessment question scores (i.e., regarding the novelty of a product/service) and regarding the subjective confidence of the assessor in determining this score. It is especially this subjective scoring confidence that is relatively straightforward to improve and usually based on more intensive collaboration with (E)TE stakeholders. Comparative analysis of case study results also suggests that it is less the absolute than the relative scoring that leads to the more robust identification of pragmatic beneficial actionable interventions. While QCA is considered as eminently suitable for the existing small data context, the increase of the number of case studies may provide the opportunity to develop a parametric dependency model of assessment factors for further investigation and simulation. Specific further research is ongoing into applying mathematical decryption approaches to the causal code approach, while monitoring the evolution of the (E)TEs on a semi-annual basis. Key Terms and Definitions See Table 3.

42

O. Schwabe and N. Almeida Table 3. Key terms and definitions

Key Term

Definition

Acceleration

Increase of the speed with which an idea evolves from ideation to sustainable market adoption. Measured against either the specific planned business model and/or the relative progress to other ideas (Schwabe et al. 2020a)

Innovation

The rise of an idea, its research and early application in various forms, the socialization of the idea to a wider community which leads to a market validation phase followed by a commercialization phase which ends in sustainable market adoption (Rogers 2003)

Innovation Ecosystems

Any group of individuals aiming to diffuse ideas from ideation to late adopters and assuming the roles of various actors negotiating/performing the sustainable exchange of (in-) tangible assets Mercier-Laurent 2015)

Resilience

The adaptive capacity of an organization in a complex and changing environment (ISO Standard 73:2009)

Transfor-mative Enterprises (TEs)

(Non-) Formal organizations which are (a) holistic (using experimentation, TEs are concerned with the social and economic community in which they live), (b) heuristic (TEs prefer bottom-up learning), and (c) moral (not laws, but high standards of fairness and honesty shape the behaviour of members). TEs are living (eco-) systems (Formica 2022)

Transfor-mation

Capacity-building over time from within the TE. It goes beyond performing tasks to changing mindsets and attitudes. Major examples are related to economic globalization and use of new technologies (SDG 2021)

Value

Improvements in indicators of the European Innovation Scorecard (EU Innovation Scorecard 2021) and the European Union Eurostat Regional Performance Measures (Eurostat 2021)

Well-being

Goal of the EU (Article 3 Lisbon Treaty) stating that “The Union’s aim is to promote peace, its values and the well-being of its peoples”. Well-Being exists within a subjective and an objective dimension. This includes the life experience of an individual, and the comparison of life circumstances with social norms and values (European Union, 2021)

Acknowledgements. The authors gratefully acknowledge the support of the EIT Cross-KIC HEI Initiative Project INCORE (P.6076: Innovation Capacity Building for Higher Education in Europe’s Outermost Regions) and the Foundation for Science and Technology’s support through funding UIDB/04625/2020 from the research unit CERIS. The authors also express their deepest

Realizing Sustainable Value from Engineering Innovation Ecosystems

43

gratitude to the support received from M. Amaral (Instituto Superior Téchnico (IST, PT), E. Leite and E. Fernandes (Universidade da Madeira, PT), J. Arquillo and C. Oliveira (Universidad Europea de Canarias SL, ES), K. Addi (Université de La Réunion, FR), V. Correia (Trisolaris Lda., PT), B. Bodo and L. Lopes with M. Pinto (La Palma Research Centre SL, ES), L. Schneider (Entovation International, USA), as well as Y. Yehorova and P. Formica (Innovation Value Institute, Maynooth University, IE).

References Allee, V., Schwabe, O.: Value Networks and the True Nature of Collaboration. Meghan-Kiffer Press, Tampa (2015) Almeida, N., Trindade, M., Komljenovic, D., Finger, M.: A conceptual construct on value for infrastructure asset management. Util. Policy 75, 101354 (2022). https://doi.org/10.1016/J. JUP.2022.101354 Amidon, D., Formica, P., Mercier-Laurent, E. (eds.): Knowledge Economics: Emerging Principles, Practices and Policies. Tartu University Press, Estonia (2005) Correia, L., Schwabe, O., Almeida, N.: Speed of Innovation Diffusion in Green Hydrogen Technologies, World Congress on Engineering Asset Management (WCEAM). 5–7 October 2022. Sevilla Proceeding, Accepted for Publication) Destom, J.X.: The benefit of the outermost regions for the EU. European Union, ECO/567 – EESC2021-05077-00-00-APA-TRA (FR) 7 (2021). See https://www.eesc.europa.eu/en/agenda/ourevents/events/hearing-benefits-outermost-regions-european-union/opinions. Accessed 27 Aug 2022 European Union, Lisbon Treaty: European Union Lisbon Treaty. Official Journey of the European Union (2012). https://eur-lex.europa.eu/resource.html?uri=cellar:2bf140bf-a3f8-4ab2b506-fd71826e6da6.0023.02/DOC_1&format=PDF. Accessed 27 Aug 2022 European Union, Regional Policy: EU Regional Policy (2021). http://ec.europa.eu/regional_pol icy/en/policy/themes/outermost-regions/. Accessed 27 Aug 2022 Eurostat: Eurostat Regional Performance Measures (2021). https://ec.europa.eu/eurostat/statisticsexplained/index.php?title=Main_Page. Accessed 27 Aug 2022 Formica, P.: IDEATORS. Their Words and Voices. Emerald Group Publishing, Bingley (2022) Homans, G.: Social behavior as exchange (T. U. Press, Ed.) Am. J. Sociol. 63(6), 597–606. https:// www.jstor.org/stable/2772990. Accessed 27 Aug 2022 Mello, P.A.: Qualitative comparative analysis: an introduction to research design and application. Georgetown University Press. https://www.barnesandnoble.com/w/qualitative-compar ative-analysis-patrick-a-mello/1139218712. Accessed 27 Aug 2022 Mercier-Laurent, E.: The Innovation Biosphere - Planet and Brains in Digital Era. Wiley, San Francisco (2015) Rogers, E.: Diffusion of Innovations, 5th edn. The Free Press A Division of Macmillan Publishing Co., Inc., New York (2003) Schwabe, O., Schneider, L., de Almeida, N.M., Salvado, A.F.: A framework for accelerating innovation through innovation webs. In: Rodrigues, H., Gaspar, F., Fernandes, P., Mateus, A. (eds.) Sustainability and Automation in Smart Constructions. ASTI, pp. 205–210. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-35533-3_23 Schwabe, O., et al.: A Maturity Model for Rapid Diffusion of Innovation in High Value Manufacturing. CIRPe 2020b – 8th CIRP Global Web Conference – Flexible Mass Customization. https://doi.org/10.1016/j.procir.2021.01.074. Accessed 27 Aug 2022

44

O. Schwabe and N. Almeida

Schwabe, O., Almeida, N.: Healthcare innovation ecosystems as transformative enterprises improving the good health and well-being of citizens in Europe’s Outermost Regions. Chapter in Formica, P. (ed.) (2022) One Health. Emerald Publishing (Forthcoming) Standards, ISO: ISO Guide 73:2009, Risk Management, Section 3.8.1.7. See https://www.iso.org/ standard/44651.html. Accessed 27 Aug 2022 United Nations Sustainable Development Goals (SDGs): United Nations Capacity Building for SDG 17 (2021). https://www.un.org/en/academic-impact/capacity-building. Accessed 27 Aug 2022 Varela, F.G., Maturana, H.R., Uribe, R.: Autopoiesis: the organization of living systems, its characterization and a model. Biosystems 5(4), 187–196 (1974). https://doi.org/10.1016/0303-264 7(74)90031-8. Accessed 27 Aug 2022

RelOps – A Whole-of-Organisation Approach for Reliability Analytics Melinda Hodkiewicz1(B) , Tyler Bikaun2 , and Michael Stewart2 1 School of Engineering, University of Western Australia, Perth, WA, Australia

[email protected]

2 Department of Computer Science, University of Western, Perth, WA, Australia

{tyler.bikaun,michael.stewart}@uwa.edu.au

Abstract. Reliability analysis on in-service assets uses well-established methods to, for example, determine mean-time-between-failure (MTBF) estimates or identify failure modes. However, the data inputs to these calculations depend on how the raw data from maintenance repair records have been processed. Furthermore, processes to extract and clean raw maintenance data are often ad hoc and performed differently by each engineer. As a result, calculations for asset reliability measures and identification of historical events and failure modes are difficult to replicate. Currently, the process is manual, time-consuming and not scalable. As a solution we present RelOps, a process to achieve standardised, scalable, and efficient end-to-end data handling and processing for organisation-wide reliability analysis. The process is illustrated with a case study showing current practice in MTBF estimation and the opportunities for technical language processing (TLP) to infer MTBF from maintenance work orders raised against a slurry pump.RelOps draws on DevOps and MLOps practices widely used in the software engineering and machine learning communities. The aim of RelOps is to shorten the reliability analysis development lifecycle and provide continuous delivery of quality outputs using a standardised and repeatable process.

1 Introduction Asset management has become an important capability for board members and executives of asset-intensive industries as the costs and risks of poor capital allocation, and day-to-day asset decision-making rise and the competitive advantages of being ‘good’ at asset management are appreciated. Metrics for assessing asset management performance are now standard in executive reporting. Metrics include historical values for asset/plant availability, throughput and costs, that cascade from organisational level down to asset level. Alongside this growth in executive interest in asset management has been a rapid transformation in information management in infrastructure, resources, and process plant industries. The widespread adoption of standards like the ANSI/ISA-95 EnterpriseControl System Integration set of models and terminology (ANSI/ISA-95.00.01:2010) has enabled organisation-wide systems integration. This has vastly increased the data available as input to analysts and decision-makers. As access to information held in manufacturing execution systems (MES) and enterprise resource planning (ERP) systems © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 45–55, 2023. https://doi.org/10.1007/978-3-031-25448-2_5

46

M. Hodkiewicz et al.

has increased, so have expectations about the timeliness of reporting. There has also been a move away from individually configured software on personal computers and locally stored data files to cloud-based, corporate-managed systems and software as a service (SaaS). We are now seeing a proliferation of corporate dashboards, developed in business intelligence software such as Microsoft Power BI, Tableau and Spotfire, which produce real-time or near real-time metrics and graphs from data pulled directly from MES and CMMS systems. It is no longer necessary to wait for the end of the month for engineers and analysts to produce reports, as much information on operational performance is now readily available on these corporate dashboards. However, what is missing are timely statistical measures for reliability performance. The development of measures for asset reliability described in reliability textbooks and online teaching materials is described as the straightforward application of statistical models to data sets (Meeker et al. 1998). Very little attention is given to how difficult it is to curate meaningful data sets, especially for in-service assets. Most illustrations come from accelerated life tests or other well-resourced reliability programs on critical assets. However, for most reliability engineers working on in-service assets in industry today, building, maintaining and updating data sets of failures and suspension data for each asset is unmanageable. In reality, engineers perform reliability analysis on a reactive basis, asset by asset, as failures or critical decisions arise. With the exception of simple point estimates (based on an arithmetic mean) of mean life, we have seen few examples where engineers have priors for current reliability measures for their assets and can update them as new information arise, and can do this at scale. Producing these measures is still a time-consuming, manual task for a trained reliability engineer. In the following sections, we describe why this situation is how it is and what can be done to make reliability measures as readily available on dashboards as other operational measures on our assets.

2 Current Practice The challenges engineers face in acquiring data to support reliability analysis are illustrated in Fig. 1, a swimlane diagram developed from the authors, and others in industry, experience of dealing with industry data. This figure identifies common steps performed each time an engineer wants to calculate a reliability measure (for example mean-timebetween-failure) from semi-structured data held in an organisation’s computerised maintenance management system (CMMS). We display this process as six stages: i) extract, ii) filter (structured), iii) filter (unstructured), iv) transform, v) compute and vi) verify. Only the last two stages, compute and verify, are commonly discussed in textbooks (Meeker et al., 1998). The same criticism is levelled at academic papers where the many decisions made in extract, filter (structured), filter (unstructured), and transform stages of data processing are seldom adequately described (Astfalck et al. 2016). The swimlane diagram (Fig. 1) shows several interconnected workflows. The first column shows the engineer’s required tasks (Engineer Task Steps) that must be repeated for every data set. The second column identifies the domain knowledge required to do the step, and the third captures data processing decisions/operations. The fourth column identifies the process to perform the work identified prior while the fifth and

RelOps – A Whole-of-Organisation Approach

47

Fig. 1. Steps taken, data and metadata generated in MTBF estimation from MWO.

sixth describe the data set and metadata produced in the preceding steps. The end-to-end process produces six data sets d1 , d2 , d3 , d4 , d5 , and d6 . Each data set is associated with implicit metadata capturing decisions made and operations performed; we number these as m1 , m2 , m3 , m4 , m5 , and m6 . In practice, Excel spreadsheets are used for much of this manipulation and, without version control, is impossible to replicate.

48

M. Hodkiewicz et al.

Extract: To start the process, the engineer must have selected the metric of interest; in this paper, we use the mean time between event as an example. First, they must identify which reports in the CMMS or other databases contain data fields of interest. For example, many of the partners we work with use the SAP Enterprise Resource Planning systems, which means downloading a specific report (such as an IW39). Other information may also be required, such as operating times if the asset is in a primary/standby configuration or part of a fleet and for this access to OSI-PI or similar operational records or fleet management system data is required. Alignment must then be made on functional location and date/time records between the two systems. Filter-Structured: In the CMMS, the information for each maintainable item of interest must be recovered from individual work order records. Required information includes the event date the maintainable item was installed, the event date it was removed from service, whether it was removed pre-emptively (indicative of a suspension) or, as a result of failure and, if available, why it was removed (its failure mode). None of this information is readily available in machine extractable fields. Instead, each MWO must be read to see if the unstructured text field contains clues to an installation, removal, repair, inspection or another event (Hodkiewicz and Ho 2016). Drawing on the data captured in SAP, one immediate decision is what date and cost fields to include. Because of the needs of maintenance planning, there are several fields to choose from. For the end of life, do we choose the date the maintenance notification was raised,or the date the work order was created or actioned? For costs, we have fields for planned costs, estimated costs and actual costs. Each time an engineer does this, he/she makes their own selections. Different selections lead to different values being used in subsequent calculations. We have never seen a corporate business process or guidance document setting out a standardised process for how to do this. Table 1. Examples of fields in maintenance work orders No

Actual start date

Description

MWO type

Actual labour costs ($)

Actual material costs ($)

Actual total costs($)

1

2012-10-18

Remove spools and unbog pump

A3

227

0

113

2

2013-01-05

26W A2 Overhaul wet end

728

5891

0

3

2013-01-06

Not pumping to capacity

A3

1682

159

1842

4

2013-01-08

Replace discharge valve

A1

354

323

636

(continued)

RelOps – A Whole-of-Organisation Approach

49

Table 1. (continued) No

Actual start date

Description

MWO type

Actual labour costs ($)

Actual material costs ($)

Actual total costs($)

5

2013-02-12

Investigate pumps keeps tripping

A1

409

0

409

6

2013-05-13

Not pumping to capacity

A1

394

10914

11734

7

2013-08-08

Replace packing

A1

491

123

574

8

2013-10-21

Replace discharge spool

A1

354

3464

3818

9

2015-07-28

Wet end needs inspecting

A1

1129‘

2239

5434

10

2016-09-23

Wet end rebuild

A1

2199

819

2818

11

2017-09-29

Not pumping

A3

3741

4065

6334

12

2017-11-06

Pumps U/S liner holed

A3

1836

13198

14249

13

2017-11-21

Pump bogged A3

2892

11688

16111

14

2018-06-19

Pump A3 jammed/ bogged. Trips

1116

60

1258

15

2018-08-14

Gland seal A3 leaking slurry

4514

14480

16633

16

2018-12-03

Wet end overhaul

A3

2819

4548

5395

17

2019-02-06

Gland repack and wet end checks

A3

1500

1689

2529

18

2019-04-23

Not achieving A3 required flow

2250

1559

2993

19

2019-06-03

Not pumping to capacity

A3

1512

0

1332

20

2020-06-30

Pump housing needs repair

A3

3064

8223

12300

50

M. Hodkiewicz et al.

An example of 20 records of data extracted for analysis for a single slurry pump for the period 2012–2020 is shown in Table 1. We show fields commonly of interest, namely the date the work was actioned on the unit (Actual Start Date), an unstructured short text description of desired work (Description), the order type (A1 indicates corrective work sent to planning, A2 work based on maintenance strategy, and A3 corrective work requiring immediate attention), the actual labour costs, actual material costs and total actual costs. These are only 6 of the 30 fields available in the SAP report downloaded for this example. These 20 records are only a fragment of the 136 work orders on this pump which has total historical maintenance costs over $250,000. This is not unusual in chemical/process plants with pumps moving material that is corrosive, abrasive, and often at elevated temperatures and pressures. Annual costs of maintaining the pumps assets alone in this environment at a moderate-sized operation can exceed $4M per year. Obviously, improved reliability is a key goal. Filter Unstructured: A key question is to decide which of the 136 MWO records available for this asset contain information relevant to the measure we want to calculate. We have chosen these 20. What guided our decision, and where is this documented? Why did we ignore 116 others? Every reliability engineer will do this in their own way. Our thinking, in this case, is that we are concerned with the time-between-significant event on this slurry pump system. The pump system is a critical production item, and while it operates in an operating/standby configuration it must be available at all times. Any failure of a major component (motor, coupling, pump housing, pump wet end liners) is a failure of the system. It is unclear whether the gland seal packing should be considered a failure or not since this has a lower expected service life than the other elements. Once again, another decision the engineer needs to consider, but is seldom captured. The 20 records were selected for one of the following reasons: a) a verb indicating a replace/remove activity in the description, and b) the materials cost indicates a part with a significant cost has been replaced. There are two ways of doing this work. The first and by far the most common is with keyword filters. The engineer looks at a spreadsheet file extracted from SAP and manually filters records using keyword/string-matching for words such as replace or remove. This approach has the potential to both identify records that are not relevant and miss records that are. Specifically, a keyword filter would include records #1 and #8 for the spool, and #4 for the valve, even though both the spool and valve are not within our designated system boundary for analysis. More importantly, the costs associated with MWOs #6, #12, #13 and #14 indicate significant parts were replaced, but there are no replace synonyms in their descriptions. The alternate approach is to use the tools of natural language processing (NLP), an emerging area of interest for the reliability community.Several groups are now working on its application to MWO’s, an area coined Technical Language Processing (TLP) (Dima et al. 2021). The field is evolving rapidly from early work demonstrating the issues of applying outof-the-box NLP packages on maintenance work order texts (Brundage et al. 2021), to the development of annotation tools to enable the building of fit-for-purpose annotated training and test sets for machine learning (Stewart et al. 2019; Sexton et al. 2019), bespoke tools for pre-processing technical texts (Bikaun et al. 2021; Sexton et al. 2018;

RelOps – A Whole-of-Organisation Approach

51

Stewart et al. 2018), development of gazetteers (conceptual dictionaries mapping words to entities) (Bikaun et al. 2021), and pipelines for end-to-end processing (Gao et al. 2020). Transform: Once we have a list of candidate MWOs that may describe the event of interest, we need to consider if the unstructured text in the MWO description field is consistent with the text in the structured fields (costs and WO type) and then tag it as an event for future analysis. This requires engineering expertise. For example, a MWO might contain, “Not pumping to capacity”. This could be for operational issues; for example, a valve in the system is partially closed, the strainer is plugged, or something is stuck in the inlet. Alternatively, it could be an issue with the pump, such as excessive suction recirculation due to worn internal seals or wear on the impeller. If the latter proved to be the case, the impeller would need to be changed out, incurring a parts cost, and the event would be noted as an end-of-life event for the impeller. If the former, there would be no parts cost charged to the work order. To make this determination, the numbers in the labour and materials cost fields need to be considered, and engineering knowledge of pumps and their failure modes is required. In our experience, there is limited documentation of business logic to guide engineers in this task. In 2021, the application of a TLP pipeline to 14,508 and 89,259 MWOs produced MTBF estimates for 93 and 669 pumps required only 10 min end-to-end (Bikaun et al. 2021). Equivalently, this would have taken a reliability engineer days or weeks to complete. A key focus of this work was to illustrate how different technical decisions made by a reliability engineer can impact the MTBF estimates produced (Bikaun et al. 2021). Particularly whether the engineer uses data from structured fields such as cost (or not) in their decision making. These decisions were shown to produce variations in MTBF estimates of more than 100%. Compute and Verify: The engineer has now produced a data set, usually an ordered list of time values (for example, the time between the start and end of the event of interest, e.g. the life of the asset) and a matching categorical variable indicating if the event is a failure or a suspension. This data is processed using standard statistical techniques. The resulting statistical distributions provide insight into failure behaviour (wear in, wear out, ‘random’) and mean life estimates (Meeker et al. 1998). In industry these calculations are often aided by use of specialist software commercial packages such as Isograph (Availability Workbench) and Reliasoft (Weibull++). The last decade has seen an increased acceptance of open-source packages in programming languages such as R and Python in the workplace. Many graduating engineers are now proficient using these packages, and the statistical analysis, previously done by generations of reliability engineers in stand-alone commercial packages (with expensive annual licences) can now be done entirely in open-source alternatives. R and Python packages for Weibull analysis are free accessible from several popular GitHub sites. An additional advantage of using code is that this analysis can be integrated into the same codebaseas centralised data repositories, limiting the need to move between data systems. The final stage of the process is to check that the statistical distribution on which the parameter estimates are based, these parameter estimates are used in subsequent decisions such as inputs for determining the timing for fixed interval replacements, are sensible. This involves knowledge by the engineer of concepts such as IID (independent and identically distributed

52

M. Hodkiewicz et al.

random variable), trend tests, goodness of fit tests, appropriate practice for dealing with outliers, etc. Problems with the Current Approach: Figure 1 shows that the end-to-end ExtractFilter-Transform-Compute-Verify process produces six data sets d1 , d2 , d3 , d4 , d5 , d6 and five sets of metadata m1 , m2 , m3 , m4 , m5 , m6 . To replicate this process, the original data set d1 and code to generate each of the following versions (or their diffs) need to be captured and stored.This is not done routinely in organisations. However, suppose TLP processing is done in one end-to-end code pipeline using best practice software development principles, such asversion control, the engineer could automatically produce the metadata necessary to recreate any pipeline stage. Without version control systems, it is not possible to replicate the process (Bryan 2018). Compounding this is the use of third-party software that require manual handling of data between systems. Manipulation of data while in different systems, such as the removal of outliers, is also not usually a documented process. Even if individual engineers are using version control and can replicate the pipelines they develop, there are still issues at an organisational level with different engineers making different decisions unless they are using the same code for the Filtering stages and the same business rules in the Transform stages. To enable this, the TLP pipelines need to be moved off individual computers and into an enterprise layer, with the process managed by a reliability business process group in the organisation. This is described in the following section.

3 Proposed Approach Continuous software engineering practices, such as DevOps, are well established and used by software development teams in organisations. DevOps practices advocate for automation and monitoring of all stages of development and deployment, defined as “a process whose aim is to shorten the system development life cycle and provide continuous delivery with high software quality” (Bass et al. 2015). More recently, there has been increasing interest in the rapid, but controlled, deployment of machine learning algorithms in support of operational improvements called MLOps (Sculley et al. 2015). MLOps builds on DevOps practices and there are several formalisations (Makinen et al. 2021). Data pre-processing and statistical analysis for reliability analysis in an organisational setting is no different from the application of machine learning models, and both need to be actioned in a structured and controlled environment. The infrastructure and processes to do this exist in the software engineering and data science teams, already embedded in asset-owning organisations, and similar processes and practices need to be adopted by reliability engineers. This is now possible as the end-to-end pipeline can be achieved using open-source packages in R and Python without the need to export data to third party packages (such as Weibull++, Availability Workbench and Asset Performance Management). We suggest a definition of RelOps as follows ‘a process whose aim is to shorten the reliability analysis development lifecycle and provide continuous delivery of quality outputs using a standardised and repeatable process’.

RelOps – A Whole-of-Organisation Approach

53

Fig. 2. The RelOps process.

RelOps is envisaged as a set of interlocking, continuously improving, steps that are conceptually similar to both DevOps and MLOps but with a ‘REL’ loop as shown in Fig. 2. These steps are managed at an organisational level, with software hosted and managed centrally, nowadays, this usually means in a cloud environment. RelOps follows the principles of the Cross Industry Standard Process for Data Mining (CRISP-DM) process model (Chapman et al. 2000) and the emerging CRISP-ML (Studer et al. 2021) model by incorporating steps for business and data understanding, data preparation, modelling, evaluation, and deployment. However, we have incorporated adaptations to deal with factors specific to the reliability context. Each stage is described briefly below. Data Pipelines (DATA): RelOps organisations transition from manual, ad hoc, collect methods to automatic processes leveraging data pipelines for data management. Data for reliability analysis is made available in a standardised format after integrating and transformating data sources such as CMMS and downtime accounting reports. Data pipelines are executed on a routine basis to maintain relevance and freshness. Pipelines are pre-configured with data cleaning processes that are tested to ensure a high standard of data quality tailored to the desired output. Essential elements of data cleaning include filtering and text normalisation. A data engineer manages this stage. Reliability Models (REL): This stage supports the building, storage and evaluation of models allowing for experimentation and evaluation, and is supported by a code repository and version control, metadata storage and a model repository (John et al. 2021). When the desired output is a reliability measure, as described in Fig. 1, this stage involves the Filter-Transform-Compute-Verify steps. Of particular note is the reliance in this stage on machine learning using language models. Given the input to this stage include both structured fields and unstructured text documents, the use of natural language processing will always be necessary; hence, traditional MLOps approaches should be used. Reliance on processes using technically sophisticated ML models such as transformers and off-the-shelf embedding models means metadata management needs close attention (John et al. 2021). This stage requires close collaboration between reliability engineers and data scientists. Development (DEV): Once an appropriate pipeline has been tested and validated, it needs to be instantiated in a development environment by software engineers where it can be socialised with stakeholders, links to data sources and other software assessed, triggers for model updates identified, and performance measures developed (John et al. 2021). The software engineering team does not have deep expertise in data processing

54

M. Hodkiewicz et al.

and reliability analysis, and data scientists do not want to manage the infrastructure supporting the analysis, so it is essential to have bridges between the groups at this stage. Deployment, Monitoring and Control (OPS): The final stage is moving the pipeline to the production environment and will usually be managed by the software engineering team. An important part of this is establishing performance monitoring, triggers, and feedback processes to identify if the pipeline’s performance is deteriorating or if certain tasks are not producing ‘good’ results. Performance monitoring for the pipeline needs to be sensitive to changes such as the emergence of previously unidentified failure modes, changing business rules, and the availability of new data, with any triggering a need to revisit either the Data Ingestion or Reliability stages or both.

4 Discussion The proposed RelOps process aims to support end-to-end reliability analysis at an organisational level by implementing a version-controlled set of decisions for pre-processing structured and unstructured data necessary for the analysis. RelOps implements source and processing control to maintain line-of-sight to decisions made in processing, validation, deployment strategies, and monitoring protocols. It follows the principles of the Cross Industry Standard Process for Data Mining (CRISP-DM) process model and the emerging CRISP-ML (Studer et al. 2021) model by incorporating steps for business and data understanding, data preparation, modelling, evaluation, and deployment. RelOps replaces the current practice of ad-hoc pre-processing decisions, and manual data handling for reliability analysis-based decisions.

References ISA: ANSI/ISA-95.00.01-2010: Enterprise-control system integration - part 1: models and terminology. Technical report, ISA-95.00.01, International Society of Automation (2010) Meeker, W.Q., Escobar, L.A., Pascual, F.G.: Statistical Methods for Reliability Data. Wiley, Hoboken (1998) Astfalck, L., Hodkiewicz, M., Keating, A., Cripps, E., Pecht, M.: A modelling ecosystem for prognostics. In: Annual Conference of the PHM Society (2016). http://www.papers.phmsoc iety.org/index.php/phmconf/article/view/2568 Hodkiewicz, M., Ho, M.T.W.: Cleaning historical maintenance work order data for reliability analysis. J. Qual. Maint. Eng. 22, 146–463 (2016) Dima, A., Lukens, S., Hodkiewicz, M., Sexton, T., Brundage, M.P.: Adapting natural language processing for technical text. Appl. AI Lett. 2(3) (2021). https://doi.org/10.1002/ail2.33 Brundage, M.P., Sexton, T., Hodkiewicz, M., Dima, A., Lukens, S.: Technical language processing: unlocking maintenance knowledge. Manuf. Lett. 27, 42–46 (2021) Stewart, M., Liu, W., Cardell-Oliver, R.: REDCOAT: a collaborative annotation tool for hierarchical entity typing. In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP): System Demonstrations, pp. 193–198 (2019)

RelOps – A Whole-of-Organisation Approach

55

Sexton, T.B., Brundage, M.P., et al.: NESTOR: a tool for natural language annotation of short texts. J. Res. Nat. Inst. Stand. Technol. 124, 1–5 (2019) Bikaun, T., French, T., Hodkiewicz, M., Stewart, M., Liu, W.: LEXICLEAN: an annotation tool for rapid multi-task lexical normalisation. In: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pp. 212–219 (2021) Sexton, T., Hodkiewicz, M., Brundage, M.P., Smoker, T.: Benchmarking for keyword extraction methodologies in maintenance work orders. In: Annual Conference of the Prognostics and Health Management Society, pp.1–10 (2018) Stewart, M., Liu, W., Cardell-Oliver, R., Wang, R.: Short-text lexical normalisation on industrial log data. In: 2018 IEEE International Conference on Big Knowledge (ICBK), pp. 113–122 (2018) Bikaun, T., Hodkiewicz, M.: Semi-automated estimation of reliability measures from maintenance work order records. In: PHM Society European Conference (2021). https://doi.org/10.36001/ phme.2021.v6i1.2950 Gao, Y., Woods, C., Liu, W., French, T., Hodkiewicz, M.: Pipeline for machine reading of unstructured maintenance work order records. In: Proceedings of the 30th European Safety and Reliability Conference and 15th Probabilistic Safety Assessment and Management Conference (2020) Bryan, J.: Excuse me, do you have a moment to talk about version control? Am. Stat. 72(1), 20–27 (2018) Bass, L., Weber, I., Zhu, L.: DevOps: A Software Architect’s Perspective. Addison-Wesley Professional, Boston (2015) Sculley, D., et al.: Hidden technical debt in machine learning systems. In: Cortes, C., Lawrence, N., Lee, D., Sugiyama, M., Garnett, R. (eds.) Advances in Neural Information Processing Systems, vol. 28 (2015) Mäkinen, S., Skogström, H., Laaksonen, E., Mikkonen, T.: Who needs MLOps: what data scientists seek to accomplish and how can MLOps help? In: 2021 IEEE/ACM 1st Workshop on AI Engineering-Software Engineering for AI (WAIN), pp. 109–112 (2021) Chapman, P., et al.: CRISP-DM 1.0: step-by-step data mining guide. SPSS inc 9, 13 (2000) Studer, S., et al.: Towards CRISP-ML (Q): a machine learning process model with quality assurance methodology. Mach. Learn. Knowl. Extract. 3(2), 392–413 (2021) John, M.M., Olsson, H.H., Bosch, J.: Towards MLOPS: a framework and maturity model. In: 2021 47th Euromicro Conference on Software Engineering and Advanced Applications (SEAA), pp. 1–8. IEEE (2021)

Methods for Comparing Asset Portfolio Reliability Gabrielle Biard(B) and Georges Abdul Nour Université du Québec à Trois-Rivières, Trois-Rivières, QC, Canada {gabrielle.biard,georges.abdulnour}@uqtr.ca

Abstract. Investment optimization is a challenge for companies that hold asset portfolios of multiple categories. Companies in the power industry fall into this category of organizations. Therefore, prioritizing AM investments in order to maintain or improve the network reliability level is a challenge. Thus, the objective of this article is to help investment decision-making in the field of asset management. To do so, the methods used to assess asset reliability are analyzed in order to develop a method for ranking assets based on their criticality. The benefits of using Industry 4.0 tools, such as artificial intelligence in the reliability assessment process are also highlighted. The results are based on a literature review compiling the analysis of 85 publications.

1 Introduction Market globalization has forced organizations to constantly improve their performance to ensure that they remain competitive in a continuously evolving market. This phenomenon has an impact on productivity, efficiency, and quality of products or services in all industry sectors. There are also acknowledged impacts on monopolistic organizations, such as those found in the electrical industry. Regulatory agencies constrain expenditures and raise revenue expectations. However, power system assets are aging and require additional maintenance and replacement. At the same time, industrial and residential electricity consumers are integrating more technological tools that require increased continuity of service. Generation, transmission, and distribution networks therefore require investment optimization to maintain or improve the reliability level of these respective asset portfolios. Investment optimization is a challenge for companies that hold asset portfolios of multiple categories, such as in the power industry. In this type of organization, there are several heterogeneous asset classes, with various components and management strategies. There is therefore a dependency issue in terms of performance and resource allocation. The impact of investments on performance is also uneven depending on the assets involved. Therefore, as traditional performance measures are not compatible with all assets, advanced analytical methods must be used for decision-making and investment prioritization. (Petchrompo and Parlikad 2019). Power systems are also large complex systems given the set of interrelationships between each of their component. (Mahmood et al. 2019; Xu et al. 2010). The reliability © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 56–64, 2023. https://doi.org/10.1007/978-3-031-25448-2_6

Methods for Comparing Asset Portfolio Reliability

57

of a complex system results from the interactions between the elements that make up the system, rather than from the sum of each individual results. The assessment of the probability of failure is then difficult to predict and the modelling of reliability is complex. Therefore, prioritizing AM investments in the power industry is a challenge. In this respect, Industry 4.0 tools offer significant advantages (Biard and Abdul-Nour 2021). This revolution encourages, among other things, the use of Big Data and Artificial intelligence (AI) in decision-making. The integration of innovative technologies is an essential element to ensure the sustainability of organizations. For the electrical industry, these technologies provide significant benefits (Biard and Abdul-Nour 2021). Thus, the objective of this paper is to assist investment decisions in power system asset management (AM). To do this, assets must be prioritized. This hierarchy must identify most critical assets for achieving the organization’s goal, which is to provide electricity to customers. Therefore, the methods used to assess asset criticality and reliability are analyzed and a method for ranking assets applicable to a portfolio of assets of multiple categories is developed. Benefits of using AI are also highlighted. The results are based on a literature review compiling 85 publications. The next section presents the background. Section 3 details the methodology, while Sect. 4 presents the results. Finally, there is a discussion in Sect. 5 and a conclusion at Sect. 6.

2 Background Investment decisions in power system asset management, from a asset portfolio perspective is a recent topic in the literature. Thus, there are few publications on this topic. First, Alvarez et al. (2021) proposes a detailed methodology for optimal decision-making for electrical systems based on Monte Carlo simulation and reliability distributions. However, their methodology is applied to only one asset category, i.e., power transformers. On the other hand, an asset management decision support tool, based on simulation and asset reliability data, has been developed by Hydro-Quebec (Côté et al. 2019; Gaha et al. 2021). Those publication presents a comprehensive methodology or tool to support decision-making, but do not provide a comparison of methods used to assess asset reliability or criticality, nor do they propose a method to rank assets based on their criticality. Then, publications in the literature that have a similar objective to the one in this article can be divided into two categories. These categories are distinguished by the objective of the article, i.e.: 1. To Propose a Methodology for Prioritizing Investment or Asset Management Activities 2. To compare methods for assessing criticality and/or reliability for the purpose of prioritizing asset management activities. 2.1 Methods for Prioritizing Investment or Asset Management Activities In terms of prioritizing asset management activities, there are several publications. First, particularly for investment decisions related to the “maintenance” phase of the asset

58

G. Biard and G. A. Nour

life cycle, Fleckenstein and Balzer (2014) demonstrated that a risk-based maintenance model can reduce the overall transmission system risk by 8%, with an equivalent level of investment. Also, Clements and Mancarella (2018) provide a simulation model for prioritizing maintenance activities that is applicable in a resource-constrained environment. This prioritization is based on a simulation including the probability of equipment failure and the repair time. The output of this simulation is a priority indicator for maintenance activities. However, both of these publications are specific to maintenance activities rather than the entire asset life cycle. Similarly, Johnson, Strachan, and Ault (2012) developed a prioritization model based on a health index (HI) that integrates the assessment of asset degradation and risk impacts. However, employing the HI in these contexts does not address issues related to the interrelationships of assets in complex systems. In this regard, Gómez et al. (2019) studied multi-criteria decision-making in AM for complex systems in public utilities. Their methodology is based on the ISO 55000 standard and establishes a criticality level using the Analytic Hierarchy Process (AHP). The results show a 0.73% network availability improvement and a 10% decrease in maintenance costs. Butans and Orlovs (2016) also propose an optimization tool based on modelling and simulation, but their model focuses on the optimal configuration based on future demand. Lastly, it can be observed that those publications offer a specific efficient tool but does not provide a global understanding of potential methodologies. Also, the contribution of AI in decision-making for AM investments is not widely discussed. 2.2 Criticality and Reliability Assessment Methods Comparison The latter proposed prioritization methods require inputs related to the criticality and reliability of assets. A few authors have provided comparative analyses of asset criticality and/or reliability assessment methods. First, Bharadwaj and Polyviou (2013) analyze the risk-based asset management methods used from a survey completed by 169 assets managers. Their results demonstrate the rate of use of these methods and identify that RBI/RBM, HAZOP, and FMECA are the most widely used methods. However, their research focuses on the oil and gas sector. Then, in a more specific context, to establish the HI, some authors propose a comparison of methods applicable to power systemspecific equipment. This is the case, among others, of Azmi et al. (2017). This author analyzes the mathematical or expert judgment-based methods that are used to establish the HI for power transformers specifically. However, it is observed that, despite similarities with other research, there is a lack of analysis of the methods used to prioritize assets in the entire power system asset portfolio in the literature. Studies often focus on the impact of one method on investment prioritization in a very specific context. In addition, the contribution of artificial intelligence to decision-making for asset management investments is still a topic that is not widely discussed.

Methods for Comparing Asset Portfolio Reliability

59

3 Methodology The Scopus and IEEE Explore databases were selected for their relevance to the field of study, i.e., the electrical industry. Then, the combination of the following keywords was used to target relevant publications: “Asset” AND (“reliability” OR “health” OR “probability of failure”) AND (“evaluation” OR “calculation” OR “assessment”). Then specific keywords related to electrical industries were used to narrow the search for the area of application. From this first extraction, 226 results were retrieved from Scopus and 132 from IEEE Explore. Duplicate publications or papers concerning the same study and written by the same authors were withdrawn. This resulted in 259 unique papers. The relevance of these results was evaluated by reading the abstracts. The publications were analyzed to assess if the application area covered physical asset management in the electrical industry. The paper also had to specify a method for assessing asset reliability or criticality. Ultimately, 85 publications met these requirements. Of these 85 publications: • • • • •

40% were specific to distribution networks, 37% to transmission networks, 6% to generation assets. 8% combined two or three of these asset classes 9% did not specify any asset type.

4 Results The methods used to assess asset reliability or criticality were grouped by the scope of the study. Thus, among the selected publications, there are distinctions in the scope of the analysis, which are: • Entire network (11% of publications), • Multiple asset categories (26% of publications), • Single asset category (63% of publications). Then, Fig. 1 shows the combined analysis of the number of publications according to their scope and the type of network. In this graph, publications that combine two or three of these asset classes (e.g., distribution and transmission network) are represented individually in each category. Next, the literature review made it possible to group publications by the following objectives: • Assess the probability of failure (PoF) • Evaluate the HI • Evaluate the criticality Based on the objectives of the analysis, the specific methods used were categorized. Figure 2 provides the results of that observation. It can be observed that the most appropriate method for assessing asset reliability or criticality for prioritizing investments in an asset portfolio depends on the scope of the

60

G. Biard and G. A. Nour

100%

1

2

6

0 2

24

18

2

8

3

80% 60% 40%

10

6

20% 0% Generation network

8 1 Transmission network

4

5

Distribution Combination of Not specified network 2 or 3 network type

Entire network

Multiple asset categories

Single asset category

Not specified

Fig. 1. Number of publications by scope and type of network

Fig. 2. Methods used based on identified objectives

analysis and the available data. These two elements make it possible to identify the ideal and achievable objective, in the context of prioritization of investments. Table 1 presents the required or optional data by methods and the related objective. Boxes marked with an R* indicate that only one of these data is sufficient. This table makes it possible to identify the most appropriate method according to the available data and objective. The monitoring method is not identified because the data required varies depending on the application context. Also, in an iterative way, the results of one analysis can be used as input to others, which serve a distinct purpose. For example, the results of the HI may be used as input to

Methods for Comparing Asset Portfolio Reliability

61

the risk matrix in order to compare asset criticality. Thus, it can be determined that assets can be ranked according to their PoF, their HI and their criticality evaluation, depending on the scope and available data. Table 1. Data required based on the method selected Method

Required data

Objective

Historical/external Failure Reliability Expert Failure Criticality HI reliability data rate distribution judgment/Inspection causes data identification FMECA

R*

R*

R*

Classification R*

R*

R*

x

Criticality index

R*

R*

x

R*

R

x

Risk matrix

R*

R*

R*

Multicriteria analysis

R*

R*

R*

Simulation

R*

R*

R*

Algorithm

R*

R*

R*

R*

x

Clustering model

R*

R*

R*

R*

x

Regression model

R*

R*

R*

Weighted sum

R*

R*

R*

Success diagram

R*

R*

R*

Reliability distribution

R

Petri network

R

Baysian network

R

R*

PoF

x x Op

x

x x

x

x R

Op

x x

R R

x x

R

x

R = Required, R* = One of them required, Op = Optional.

5 Discussion The process of identifying the most appropriate ranking method is described in Fig. 3. The process to identify the most appropriate method includes 3 intermediate steps, which are: • Evaluate the HI by system/equipment category • PoF assessment for each equipment/system category • Assess criticality Each of these steps includes the process of identifying the most appropriate method based on specific required data. For this, reference should be made to Table 1, which

62

G. Biard and G. A. Nour

Sufficient data? Single category/equipment Asset prioritization

No

Yes

Evaluate the HI by system/equipment category

PoF assessment for each equipment/ system category

Level

Evaluate the impact of failure

Asset hierarchy

Yes

Portfolio

PoF available for all categories?

No

Assess criticality

Fig. 3. Asset hierarchy global development steps

suggests multiple methods depending on the objective and specific available data. Also, in this proposal, considering the methods included in calculating asset criticality, we consider that the impact of failure is considered. The PoF assessment and the HI calculation, however, must be combined with the failure impact assessment because decision-making in AM must include a risk analysis. Similarly, human reliability assessment must be considered in the criticality, HI and PoF analysis for highly critical and high-risk systems (Rozuhan et al. 2020). 5.1 Contribution of Industry 4.0 and Artificial Intelligence In terms of the contribution of Industry 4.0 and AI to asset prioritization for prioritizing investments, there are few use cases (12 publications). The most common application (5 out of 12 cases) is to use the Internet of Things to perform real-time asset monitoring to establish the PoF or HI. The HI can also be evaluated using the k-means algorithm or neural networks (Islam et al. 2017). However, these strategies are applied to individual equipment families, rather than a portfolio of assets from multiple categories. Machine learning algorithms can also improve the quality of the data needed for HI evaluation methods. (Manninen et al. 2022) When attempting to establish an asset hierarchy for a portfolio of multiple categories, modelling, and simulation methods applicable to complex systems must be employed. To this end, AI and Digital Twins offer considerable impact for data processing and analysis. The challenge of this approach is to adequately represent all of the interrelationships between assets and to determine a common performance indicator for all assets, allowing them to be compared with a uniform measure.

6 Conclusion In conclusion, prioritizing investments is a challenge for power system assets. Thus, this article aims to inform decision-making. A process for identifying the most appropriate asset prioritization method, based on the scope and available data, is proposed. This contribution is distinguished by identifying the contribution of AI and Industry 4.0

Methods for Comparing Asset Portfolio Reliability

63

tools in asset prioritization. The next step is to develop a case study to prioritize asset investments in all phases of the assets’ life cycle for a generation, transmission, and distribution network. Complex system modelling and simulation methods then apply to rank assets based on their criticality.

References Côté, A., et al.: Élaboration d’un systeme d’aide à la décision pour la gestion des actifs à TransÉnergie. Paper presented at the Congrès 2019 CIGRÉ Canada, Montreal (2019) Alvarez, D.L., et al.: Optimal decision making in electrical systems using an asset risk management framework. Energies 14(16), 4987 (2021). https://www.mdpi.com/1996-1073/14/16/4987 Azmi, A., Jasni, J., Azis, N., Kadir, M.Z.A.A.: Evolution of transformer health index in the form of mathematical equation. Renew. Sustain. Energy Rev. 76, 687–700 (2017) Bharadwaj, U.R., Polyviou, P.: Assessing industry trends in risk-based asset management practices. Paper presented at the Proceedings of the International Offshore and Polar Engineering Conference (2013) Biard, G., Abdul-Nour, G.: Industry 4.0 contribution to asset management in the electrical industry. Sustainability 13(18), 10369 (2021). https://www.mdpi.com/2071-1050/13/18/10369 Butans, J., Orlovs, I.: Diagnostics and long term prognostics for investment decision support in smart grids. IFAC-PapersOnLine 49(28), 13–18 (2016). https://doi.org/10.1016/j.ifacol.2016. 11.003 Clements, D., Mancarella, P.: Systemic modelling and integrated assessment of asset management strategies and staff constraints on distribution network reliability. Electr. Power Syst. Res. 155, 164–171 (2018). https://doi.org/10.1016/j.epsr.2017.09.029 Fleckenstein, M., Balzer, G.: Outage cost oriented maintenance strategies of outgoing feeders in transmission systems. Paper presented at the 2014 International Conference on Probabilistic Methods Applied to Power Systems (PMAPS), 7–10 July 2014 Gaha, M., et al.: Global methodology for electrical utilities maintenance assessment based on risk-informed decision-making. Sustainability (Switzerland) 13(16) (2021). https://doi.org/10. 3390/su13169091 Gómez, J.F., Fernández, P.M.G., Guillén, A.J., Márquez, A.C.: Risk-based criticality for network utilities asset management. IEEE Trans. Netw. Serv. Manag. 16(2), 755–768 (2019). https:// doi.org/10.1109/TNSM.2019.2903985 Islam, M.M., Lee, G., Hettiwatte, S.N.: Application of a general regression neural network for health index calculation of power transformers. Int. J. Electr. Power Energy Syst. 93, 308–315 (2017). https://doi.org/10.1016/j.ijepes.2017.06.008 Johnson, A., Strachan, S., Ault, G.: A framework for asset replacement and investment planning in power distribution networks. Paper presented at the IET & IAM Asset Management Conference 2012, 27–28 November 2012 Mahmood, I., Kausar, T., Sarjoughian, H.S., Malik, A.W., Riaz, N.: An integrated modeling, simulation and analysis framework for engineering complex systems. IEEE Access 7, 67497– 67514 (2019). https://doi.org/10.1109/ACCESS.2019.2917652 Manninen, H., Kilter, J., Landsberg, M.: A holistic risk-based maintenance methodology for transmission overhead lines using tower specific health indices and value of loss load. Int. J. Electr. Power Energy Syst. 137, 107767 (2022). https://doi.org/10.1016/j.ijepes.2021.107767 Petchrompo, S., Parlikad, A.K.: A review of asset management literature on multi-asset systems. Reliab. Eng. Syst. Saf. 181, 181–201 (2019). https://doi.org/10.1016/j.ress.2018.09.009

64

G. Biard and G. A. Nour

Rozuhan, H., Muhammad, M., Niazi, U.M.: Probabilistic risk assessment of offshore installation hydrocarbon releases leading to fire and explosion, incorporating system and human reliability analysis. Appl. Ocean Res. 101, 102282 (2020). https://doi.org/10.1016/j.apor.2020.102282 Xu, Q., Jia, X., He, L.: The control of distributed generation system using multi-agent system. Paper presented at the 2010 International Conference on Electronics and Information Engineering, 1–3 August 2010

Industry 4.0 Tools and Its Impact in Asset Management

Digital Transformation in Maintenance Adolfo Crespo Márquez(B) Department of Industrial Management, School of Engineering, Universidad de Sevilla, Camino de los Descubrimientos, 41092 Seville, Spain [email protected]

Abstract. The study of the digital transformation of maintenance management in the context of industry and infrastructure is highly topical and interesting. In fact, maintenance is expected to be one of the business areas where this transformation is expected to be most significant. It is important to analyze why, and how, maintenance can benefit from this transformation: What are the new technologies and tools with the greatest potential impact on maintenance? How can this transformation process be accomplished? What is the impact of emerging asset management platforms and new intelligent maintenance Apps? Etc. Clearly, digital transformation is both an organizational challenge and a major technical challenge, needing a strategic planning process to guide it. To increase asset performance using 4.0 technologies, it is also necessary to face new technical problems and challenges: the non-ergodicity of data processes in many assets, the selection of the dimension of the number of data needed to explain their performance, the way to consider and interpret risks, the way to use such risk assessment for dynamic maintenance scheduling, etc. This document addresses each of these topics, providing the reader with keys to further explore each of them.

1 Introduction The use of digital data processing is revolutionizing industrial and infrastructure maintenance in many ways: • From a strategic and tactical standpoint, digital asset data can be used to quantify the overall health of complex assets by identifying their end of life. • From a more operational point of view, sensors on critical asset components can provide huge amounts of data points for analysis to detect anomalies, diagnose component failure modes and accurately estimate their remaining useful life. An increasing number of systems will be interconnected within industrial plants and infrastructures, and this will lead to sophisticated scenarios in which the principle of collaborative learning can help to gain intelligence in various areas of maintenance management, both strategic and operational. The expected high volume of maintenance data will require new business processes and ICT systems. Computing needs will now be distributed, close to the assets on the edge, or on the cloud, always with the aim of increasing the efficiency of advanced apps. The servitization of traditional manufacturers’ products (OEMs) and the development of new digital services, such as Predictive © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 67–75, 2023. https://doi.org/10.1007/978-3-031-25448-2_7

68

A. Crespo Márquez

Maintenance as a Service (PMaaS) or Simulation as a Service (SaaS) will also be here to stay.

2 Emerging Asset Management Nowadays companies are complementing the original Enterprise Asset Management Systems (EAM) or Computer Aided Maintenance Management System (CMMS), with different solutions to gain more capabilities in different management areas. Gartner (Gartner 2019) distinguishes between three different types of asset management systems on the market: EAM, APM and AIP solutions (Fig. 1).

Fig. 1. Data flow in advanced asset management systems.

APM and AIP benefit from connectivity and digital technologies, often sharing the same data and using similar predictive analytics, but with a different objective, as follows (See Fig. 2): • EAM investments are made to manage asset inventory, configuration, and maintenance execution. • Organizations are willing to invest in APM tools and technologies to reduce corrective maintenance, increase availability and reduce the risk of failure (especially when assets are critical). In addition, APM tools can increase the organization’s ability to comply with regulations related to asset inspections and maintenance. Data capture, analysis, and visualization enable these tools to improve the synchronization of operations and maintenance and identify which maintenance and inspection activities should be performed on industrial assets (Gartner 2019). • Investments in AIP software are made to improve long-term complex strategic and tactical decisions related to CAPEX/OPEX budget allocations and overall asset management planning (Voulgaropoulos 2021). AIP tools predict current and future asset performance and link that expected performance to different investment options over a

Digital Transformation in Maintenance

69

predefined medium- or long-term horizon. Investment planning for asset- and processintensive industries is now more relevant than ever. The economic downturn has exacerbated this situation and executives are constantly relying on AIP software to improve these decision-making processes.

Fig. 2. Asset management decision areas and the EAM, APM & AIP Systems (Crespo 2022).

3 The New System Architecture for Intelligent Asset Management With respect to the implementation of APM and AIP systems, companies currently find themselves in one of the following three basic scenarios: • Companies that take on solutions developed by asset providers (Crespo et al. 2018), • Other companies are developing customized solutions in-house (Crespo et al. 2020), and • Others adopt developments from other companies, supported by intelligent big data solutions configured on top of cloud ICT solutions (Kolevski and Gusev 2010).

70

A. Crespo Márquez

Based on the definition of Cloud Computing (Hamdaqa and Tahvildari 2012), the reader can expect these systems to be implemented using cloud-based PaaS (Platform as a Service: a class of cloud computing services that provides a platform that supports the development, execution, and management of applications for asset management) that provides a standard way to connect sensing elements, instruments and other devices together with industrial applications and computer people. One of the main advantages of a PaaS environment is that software developers do not have to worry about some of the lower-level details of the environment (Metcalfe and Winter 2018; Abualkibash et al. 2012). The applications provided by different industrial and ICT vendors to end users at the SaaS level of the pyramid are structured according to the principles of software hierarchy design (Maciaszek et al. 2014). Typically, today, an advanced asset management system. or platform, encompasses “everything needed” for “digital asset management”. Examples of such names are GE’s Predix, SAP’s IAM or SIEMENS’ MindSphere.

4 The Emerging Technologies A brief list of technologies to be considered for significant maintenance improvement is as follows: • Internet of Things (IoT) Technologies. New IoT platforms introduce a multitude of exciting possibilities. They provide components for data capture (front-end) and analysis, on-device data processing and cloud-based implementation. The features of a platform to consider for your choice are the possibilities it offers for (Mineraud et al. 2016; Borgia 2014; Perry 2016): Device management, Data collection and storage, Virtual machine creation, Remote device configuration and control, Data reliability and security, among others. • Big Data. In modern IoT platforms, health monitoring and predictive analytics data are stored and processed in a different way because (ISO 2014): they are too large in volume, with very high velocity, high variability, low veracity (high level of noise) and are data from very diverse as well. Big Data technologies provide analytical functions that can facilitate data collection and processing. • Predictive Analytics. Depending on the desired results and the available input data, different analyses can be performed: The basic analysis is to detect anomalies but can be extended to failure diagnostic & prognostics. • Digital twins (Digital Twin - DT). They are virtual replicas of physical devices or processes built to run simulations of real entities that can help to understand, predict, design, etc. This improved decision making throughout the lifecycle of an asset leads to the possibility of creating applications based on DTs of physical assets [18,19] (Boschert et al. 2016; Roda and Macchi 2018). • Augmented reality (AR). It is a form of human-machine interaction that superimposes computer-generated information (virtual data) on the real environment (real objects) (Zhu et al. 2013). AR prototypes have shown that this technology can improve the implementation of maintenance activities (Martinez et al. 2013), especially when users require additional information for the development of their tasks (Webel et al. 2013).

Digital Transformation in Maintenance

71

• Business Intelligence and Data Visualization (BI & DV) tools. BI tools are a set of technologies and processes that provide business metrics for their users (See for instance clicdata.com 2022). They are often needed because many transactional systems do not have adequate reporting built in, so producing a particular report required by the business is difficult. These BI processes require the creation of a data warehouse that in turn necessitates the creation of interfaces from source data systems or transactional systems. In addition to BI processes, data visualization (DV) processes allow information to be transmitted quickly, efficiently, and intuitively. • Other complementary technologies. For example: mobility systems and GIS systems for better scheduling, planning, and routing, etc.

5 New DMM Framework The new Digital Maintenance Management (DMM) framework (Crespo 2022) is represented using an IPO (input-process-output) diagram in Fig. 3. The process is divided into systems: First, the data is going to be extracted, transformed, and converted into different databases using ETL systems, there are therefore then database systems, intelligent asset management (IAM) systems, artificial intelligence (AI) systems and business intelligence (BI) systems. Each of these systems will have a certain function (or group of functions) to ensure effective and efficient digital maintenance management. Finally, the different systems generate an output for different purposes: identify the risk in the assets, assess that risk, mitigate the risk, and so on. Whatever is necessary to control business risk in the normal operations of the assets. Each of these outputs is related and assigned to a different IAM application. ERPIntegrated CMMS GIS Dispatching system

Simulaon & Machine Learning Apps (Digital Twin Apps) Data Extracon Apps

Current “data in streaming” data base Apps

Plant DCS Plant Informaon system

Data Transformaon Apps

Data Base Apps

Equipment embedded monitoring Historical data in other formats Legacy Systems Other

MM Strategy (BSC)

Cricality Analysis

Improve MM (e-M)

IAMS Apps: Data Models Business Rules Funconality

Ass. Inv. Control

Data Loading Apps

“Data in Storage” data base Apps

(LCC-AHI)

(CA)

Maint. Exec. Control (RAMS)

Week Points Analysis

External and Internal context analysis for risk idenficaon (Cricality model) Cricality assessment for risk assessment (Cricality Matrix)

(RCA) PM Plan Design (RCM) PM

Defined PM Policy for risk migaon (PM Plans Master) Defined CM strategy for risk migaon (CM resources allocaon)

Scheduling

(CRBA)

Business Intelligence Apps (BI Apps)

Maintenance control for risk control (Maintenance KPIs & Funconal Ind.) Asset cost, health and performance control for risk control (LCC & Health Monitoring)

Fig. 3. IPO diagram of the new digital asset management and maintenance framework.

In this new framework, the data model becomes critical for each of the processes that support the various Apps. An appropriate asset data model for a business requires

72

A. Crespo Márquez

that all applications needed to support the asset management processes in that business are identified, and their data requirements fulfilled.

6 New DMM Framework The use of data to improve maintenance management will be possible if we are able to conveniently address several challenges, among which the following are highlighted here: • Non-ergodicity of assets. Different assets, by definition, have non-ergodicity characteristics (Crespo 2022; Pathria and Beale 2011) (see Fig. 4). If a machine component fails several times, the mean properties of these failures do not necessarily converge to the mean properties of all failures of that component in the fleet. This means that asset fleets are often not ergodic (Palau 2020). When considering the possibility of using artificial intelligence tools for asset maintenance, this can seriously affect the ability of assets to use information from other assets to update their own fault detection and predictions.

Fig. 4. Non-ergodicity of assets and collaborative learning

• The curse of dimensionality. This name is given to a phenomenon that appears in Machine Learning models when algorithms must learn from a large volume of features, with abundant values within each one (Koutroumbas and Theodoridis 2008) (see Fig. 5). Reaching samples with every combination of values when training the models

Digital Transformation in Maintenance

73

would be very complicated. Therefore, it may happen that the accuracy of the classifier or regressor first improves by including more dimensions, but then may even decrease. This is a common problem when selecting, for instance, the right subset of features for the design of a CBM strategy.

Fig. 5. A view of the dimensionality curse problem (Magdy 2022).

• Dynamic risk assessment. In highly digitized operational contexts, a large amount of information related to the operation and condition of assets emerges. Linking this information to a reasonable measure of risk of failure becomes fundamental to the maintenance decision-making process. To take advantage of digitization requires a more sophisticated way of scheduling and performing maintenance activities, relying on a highly dynamic assessment of failure risk (Zio 2018). Risk management is explicitly introduced in the asset management principles (Sola and Crespo 2016) that present maintenance engineering as a main tool for management. • Dynamic scheduling of maintenance activities. The massive use of CBM generates dynamic and highly complex decision making. The management of all these complex scenarios is a barrier to the practical implementation of CBM solutions within preventive maintenance plans. How to facilitate the understanding of the information provided by these solutions and the simple interconnection with maintenance action scheduling processes, which include human participation, becomes a scientifictechnical problem to be solved of great impact in this new industrial environment (Chemweno et al. 2017).

74

A. Crespo Márquez

7 Conclusion Digital transformation is changing the way maintenance is managed. This article attempts to answer some fundamental questions such as: What is the process to follow and what are the tools to use to achieve excellence in management? How can we adapt our current maintenance management system to an intelligent asset management platform? What difficulties will we encounter? It has been described how this process is not trivial, the sequence of implementation of the different tools and technologies requires analysis. Many manufacturers and asset users and managers must prepare for different possible scenarios according to the asset’s digital configuration, leading to different maintenance plans and possible management needs. One of the biggest challenges will be providing people with the right support and training in change management. The best algorithms for detection, diagnosis and prognosis will be in vain if maintenance personnel do not change the way they work. At the same time, and to achieve impact, the entire maintenance value chain needs to be addressed; monitoring and predictive analytics represent only the first step. In the new maintenance ecosystem, critical attention must be paid to the ownership of data masters and the possibilities these masters offer for management. Developing analytical capabilities that enable a successful operating model may be a matter of negotiating access to data. In short, digitalization will change the way we conceive, develop, and industrialize maintenance. This must be understood as a long, complex, and formal process. A process that must not be implemented all at once, but must be consolidated in a continuous, orderly, and secure manner. Please refer to the book “Digital Maintenance Management” (Crespo 2022) to deepen each of the aspects discussed in this paper. Acknowledgements. This paper has been written within the framework of the projects INMA “Asset Digitalization for INtelligent MAintenace” (Grant PY20 RE014 AICIA, founded by Junta de Andalucía PAIDI 2020, Andalucía FEDER 2014–2020) and Geminhi (Digital model for Intelligent Maintenance based on Hybrid prognostics models), (Grant US-1381456, founded by Junta de Andalucía, Andalucía FEDER 2014–2020).

References Abualkibash, M., Elleithy, K. Cloud Computing. The Future of IT industry. Int.J.Distrib. Parallel Syst. 2012. Volume 3, No. 4, 1–12 Borgia, E.: The Internet of Things vision: Key features, applications and open issues. Comput. Commun. 54, 1–31 (2014) Boschert, S., & Rosen, R. Digital twin-the simulation aspect. In Mechatronic Futures: Challenges and Solutions for Mechatronic Systems and Their Designers. 2016. https://doi.org/10.1007/ 978-3-319-32156-1_6 Chemweno, P., Pintelon, L., De Meyer, A.M., Muchiri, P.N., Van Horenbeek, A., Wakiru, J.: A Dynamic Risk Assessment Methodology for Maintenance Decision Support. Qual Reliab Eng Int 33, 551–564 (2017). https://doi.org/10.1002/qre.2040 https://www.clicdata.com/blog/what-are-bi-data-visualization-data-analytics/

Digital Transformation in Maintenance

75

Crespo, M.A.: Digital Maintenance Management. Guiding Digital Transformation in Maintenance. Springer, Cham (2022). ISBN 978-3-030-97659-0. https://doi.org/10.1007/978-3-03097660-6 Crespo, M.A., Rosique, A.S., Moreu de León, P., Gómez Fernández, J.F., Diego, A.G., Fernández, E.C.: Exploiting EAMS, GIS and dispatching systems data for criticality analysis. In: Value Based and Intelligent Asset Management (2020). https://doi.org/10.1007/978-3-030-20704-5_7 Crespo, A., et al.: Criticality analysis for improving maintenance, felling and pruning cycles in power lines. IFAC-PapersOnLine (2018). https://doi.org/10.1016/j.ifacol.2018.08.262 Crespo, M.A., de la Fuente Carmona, A., Marcos, J.A., Navarro, J.: Designing CBM plans, based on predictive analytics and big data tools, for train wheel bearings. Comput. Ind. 122, 103292 (2020). ISSN 0166-3615, https://doi.org/10.1016/j.compind.2020.103292 Gartner’s Market guide for Asset Performance Management Software, dated 26th June 2019 (ID G00388410) Hamdaqa, M., Tahvildari, L.: Cloud computing uncovered: a research landscape. Adv. Comput. (2012). https://doi.org/10.1016/B978-0-12-396535-6.00002-8 ISO/IEC JTC 1 Information technology. Big data Preliminary Report 2014. ISO/IEC (2015) Kolevski, G., Gusev, M.: Analysis of cloud solutions for asset management. In: Gusev, M. (ed.) Conference: ICT Innovations 2010, Web Proceedings (2010). ISSN 1857-7288 Koutroumbas, K., Theodoridis, S.: Pattern Recognition, 4th edn. Burlington (2008). ISBN 978-159749-272-0. Retrieved 8 Jan 2018 Maciaszek, L.A., Skalniak, T., Biziel, G.: Architectural principles for service cloud applications. In: Shishkov, B. (ed.) BMSD 2014. LNIP, vol. 220, pp. 1–21. Springer, Cham (2020). https:// doi.org/10.1007/978-3-319-20052-1_1 Martinez, H., Laukkanen, S., Mattila, J.: A new hybrid approach for augmented reality maintenance in scientific facilities. Int. J. Adv. Robot. Syst. 10, 1–10 (2013) Metcalfe, D., Winter, S.: Operational Risk Management Software Market Size And Forecast 2018–2038. Verdentix (2018) Mineraud, J., Mazhelis, O., Su, X., Tarkoma, S.: A gap analysis of internet-of-things platforms. Comput. Commun. 89–90, 5–16 (2016) Pathria, R.K., Beale, P.D.: Statistical mechanics - computer simulations. In: Elsevier (ed.) Statistical Mechanics, 3rd edn. pp. 637–652. Academic Press, Cambridge (2011). https://doi.org/10. 1016/B978-0-12-382188-1.00016-5 Perry, M.: Evaluating and Choosing an IoT Platform. O’Reilly Media, Inc., Sebastopol (2016) Roda, I., Macchi, M.: A framework to embed asset management in production companies. Proc. Inst. Mech. Eng. Part O: J. Risk Reliab. (2018). https://doi.org/10.1177/1748006X17753501 Magdy, S.: The Curse of Dimensionality. IME Company. http://www.infme.com/curse-of-dimens ionality-ml-big-data-ml-optimization-pca/ Palau, S.: Distributed collaborative prognostics. Ph.D. dissertation, University of Cambridge (2020) Sola, R.A., Crespo, M.A.: Principles and Frameworks of Asset Management (2016) Voulgaropoulos (2021). https://www.verdantix.com/blog/asset-investment-planning-what-is-itand-how-is-it-useful) Webel, S., Bockholt, U., Engelke, T., Gavish, N., Olbrich, M., Preusche, C.: An augmented reality training platform for assembly and maintenance skills. Robot. Auton. Syst. 61, 398–403 (2013) Zhu, J., Ong, S.K., Nee, A.Y.C.: An authorable context-aware augmented reality system to assist the maintenance technicians. Int. J. Adv. Manuf. Technol. 66, 1699–1714 (2013) Zio, E.: The future of risk assessment. Reliab. Eng. Syst. Saf. 177, 176–190 (2018). https://doi. org/10.1016/j.ress.2018.04.020

On the Definition of Requirements for a Digital Twin. A Case Study of Rolling Stock Assets Adolfo Crespo Márquez1(B) , Urko Leturiondo2 , José A. Marcos3 , Antonio J. Guillén1 , and Eduardo Candón1 1 Department of Industrial Management, University of Seville, Seville, Spain

{adolfo,ajguillen,ecandon}@us.es

2 Control and Monitoring Area, Ikerlan Technology Research Centre, Basque Research and

Technology Alliance (BRTA), Arrasate-Mondragón, Spain [email protected] 3 Smart Maintenance Department, Patentes Talgo, Madrid, Spain [email protected]

Abstract. The definition of a Digital Twin (DT) and its requirements are not yet fully established, and many researchers consider that there is still a clear lack of conceptual basis. Some authors sustain that a DT must have some specific characteristics, and they aim to narrow this research gap by proposing an initial synthesis of DT requirements for connecting the asset models and the real conditions of the asset, others describe DT as a Virtual replica of a system to simulate its real behavior. DT characteristics have been described, such as: Real-time data (ability to deal with on-line data for the fast optimization of products and production processes); scalability (ability to analyze different scales of information); Integration (ability to integrate different models); interoperability (ability to convert, match and establish equivalence between representation models); expansibility (ability to integrate models); fidelity (ability to conform to the physical model); and/or interaction (possibility to show its own behavior in interaction with the environment in the real world). However, the requirements a DT model should possess to be widely used in industry remains an open question in the literature. In this paper we first do a literature review concerning DT features and requirements and then we justify and test the set of selected requirements in a case studie regarding high-speed train digitalization.

1 Introduction We are experiencing rapid advances in digital technologies, data analytics and artificial intelligence applied to maintenance. These approaches have the potential to transform the way maintenance is managed. The Fourth Industrial Revolution have equipped industry with tools that help generate a deeper understanding of how complex industrial systems behave and perform, thus enabling us to manage them better. In this context, data plays a pivotal role to enhance maintenance management processes, but data need integration models as DTs to convert isolated data or data silos into valuable information to support maintenance decision-making process, using DTs to generate synthetic variables to enhance and improve machine learning models. DT models involve a complete design © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 76–86, 2023. https://doi.org/10.1007/978-3-031-25448-2_8

On the Definition of Requirements for a Digital Twin

77

architecture and data model that generates a virtual entity of connectivity, processing, and digital functionalities, which can model and process data and information, calculate new data using data analysis and simulation tools, optimise equipment performance and make informed decisions. This improved decision-making throughout the lifecycle of an asset leads to the possibility of creating new high-value applications for the enterprise based on DT. The postulates of AM are truly ambitious holistically addressing management. Regarding digitization, it can be stated that it is the technologies and capabilities it brings (DT in this case) that will allow asset management to be fully developed. Conversely, it can be understood that asset management provides a complete and structured vision of the organization (value, assets, processes, information, etc.) that can be used precisely as a model for digitization. However, what is important is to understand the complementarity between digitization and asset management. The idea of an intelligent asset management system (IAMS) tries to combine both approaches. In this way, IAMS would be a digitized system that envisages the progressive and systematic integration of intelligence and knowledge management through the processes defined for optimized asset management and maintenance.

Fig. 1. A generic vision of the basic IAMP functional elements (Marquez et al. 2020).

Data can now be extracted, prepared, and recorded, for specific decision-making maintenance processes, automatically (this is named ETL extraction-transformationand-load of data). Then, the functions of intelligent assets management systems (IAMS) support the different decision-making processes organizing the collection and the analysis of data. In a more general approach, any advanced digitalization solution deal with the functional problem and the process/architecture problem. IAMS Apps work together with ETL Apps to generate ad hoc data repositories with a certain data structure model (See Fig. 1). Apps will ensure that data extracted from selected data sources, are transformed according to information requirements, formats, etc. These data repositories are expected to be placed in databases that are available in the cloud, using IaaS and PaaS Tools. Then, it is just a question of intelligence, i.e., selecting the specific data for each occasion and purpose, and the proper business rules for each specific decision-making. IAMS Apps may also interact with additional tools such as simulation tools, providing extra analytical services, and they may add complementary data to the database

78

A. Crespo Márquez et al.

records with results provided by these software elements. In addition to this asset knowledge discovering, creation, and storing (Marquez et al. 2020), these IAMS Apps are provided by vendors together with business intelligence features or Apps (BI Apps). A BI App is designed the interaction with the end user and extract database records to present the information according to the reporting needs and end-user requirements, on-demand or at the time needed by the business. A simple basic block architecture and data flow of the process are presented in Fig. 1 (adapted from (Marquez et al. 2020).

Fig. 2. An input—process—output diagram of the DMM framework (Crespo 2022).

This complex scenario has different visions and demands different forms of representation. The most IT vision (or one connected with the management of concepts such as CPS, Ciber Physical System) leads us to represent the architecture (Fig. 1). However, a vision is needed that in a complementary way outlines the functionalities in terms of maintenance improvement and, beyond that, its connection with the global asset management and maintenance model (Fig. 2). A graphical tool that eases the visualization of the overall digital maintenance management framework is a data Input—Process— Output diagram, which can be synthesized and represented as in Fig. 2. In Fig. 2, input is raw data from different business systems to be transformed or converted. Systems can be, for instance, ERP systems, dispatching systems, GIS systems, DCS, etc. Wherever relevant asset information is stored in the business (Crespo et al. 2018, 2020a, 2020a). The process is then divided into different building blocks (similar to what is done in the standard ISO 14224:2016), each of the blocks representing a system: ETL systems, Database Systems, IAM Systems, AI systems, and BI systems. Each one of these systems will have a certain function (or group of functions) to ensure effective and efficient digital maintenance management. The different systems will handle a very precise and predetermined data model to generate an output. These outputs can be for different purposes: to identify risk in assets, to assess that risk, to mitigate the risk, etc. Whatever is needed to control risk for the business in assets normal operations. Each one of these outputs is related and assigned to a different IAM App. DT Apps and BI Apps are very important supporting Apps within this framework. They may interact with the IAM Apps to allow the introduction of powerful data analytics

On the Definition of Requirements for a Digital Twin

79

and visualization tools. Figure 2 is a simple schematic of a complex digital maintenance management system.

2 The Digital Twin and Their Requirements In this context of the fourth industrial revolution, we have also seen the emerging popularity of the concept of DTs, which aim to replicate physical equipment and systems in the digital world through effective integration of data, models, and decision-support systems, promising a step change in productivity and sustainable performance. Despite encouraging developments in digital solutions, several challenges remain to be addressed before the potential opportunities they present can be realized effectively for maintenance. The challenges include a general lack of awareness of which techniques and technologies are suitable to tackle specific maintenance management problems. In the case of the DT, this is also generated by the fact that the definition of a DT and its requirements are not yet fully established and there is a lack of conceptual basis. Thus, many companies and organizations have defined the concept of DT in different ways: • “Virtual representation of real-world entities and processes, synchronized at a specified frequency and fidelity” (Digital Twin Consortium). • “Virtual representation of an object or system that spans its lifecycle, is updated from real-time data and uses simulation, machine learning, and reasoning to help decision-making” (IBM). • “Digital representation of a real-world entity or system” (Gartner). • “Virtual representation of a physical product or process used to understand and predict the physical counterpart’s performance characteristics” (Siemens). • A digital twin is a formal digital representation of some asset, process, or system that captures attributes and behaviors of that entity suitable for communication, storage, interpretation or processing within a certain context. (Industrial Internet Consortium. IIC). Some authors sustain that in order to be considered a DT, a model must have some specific characteristics (Schleich et al. 2017) such as: Scalability (ability to analyze different scales of information); Interoperability (ability to convert, match and establish equivalence between representation models); Expansibility (ability to integrate models); and Fidelity (ability to conform to the physical model). However, the requirements a DT model should possess to be widely used in industry remains an open question in the literature. Some contributions aims to narrow this research gap by proposing an initial synthesis of DT requirements (Durão et al. 2018). In Durão et al. (2018) research, they found that the most frequent requirements of DTs in the context of Industry 4.0 are real-time data, integration, and fidelity. These are crucial requirements for connecting the Product Model and the real conditions of a product. Real-time data is used for the optimization of products and production processes (Zhang et al. 2017) and it is important for knowing the status of the product and to focus on the management and optimization of processes through monitoring and data analytics (Konstantinov et al. 2017).

80

A. Crespo Márquez et al.

Integration is the most important value creation in the DT world (Uhlemann et al. 2017). A real- world object is represented by different models. The integration of the different models is essential for creating valuable data (Canedo 2016). Regarding fidelity, the DT allows the description of different operations in the real world. It is the fidelity of the model that provides the closeness to the physical product (Schleich et al. 2017). But besides the definition of the DT as a very realistic model of the current state of the process, this model must have the possibility to show its own behavior in interaction with the environment in the real world. Therefore, in the framework where DTs are considered, their interaction with the real world is another important requirement for our digital and intelligent maintenance management (Rosen et al. 2015). Having a look at industry, in a survey concerning the difficulties to implement DTs in industry (Durão et al. 2018) the main obstacles found are: robust integration of data and real-time control of the assets. The fact is, therefore, that most of the companies use the DT model as a conventional simulation model. IIC (IIC 2020) proposes a complementary approach. It lists the following characteristics or technical aspect that converge on the DT implementation or are provide by DT modelling: Information modelling as an fundamental key of DT, Information population to automatically add information of different sources trough or supported by DT modelling, Information synchronization include means, mechanisms and processes for the connection DT-information sources and DT-DT; APIs, Digital twins interact with other components; Deployment regarding how application requirements determine the edge-cloud system support design (requirement include: latency, interoperability and integration, control, complexity and power); Security as fundamental aspect to allow the information access and interchange between DTs (role control, secure interaction between digital entities, authenticity, secure deployment and version management, methods for disputes resolution); and Interoperability (syntax of information, semantics of information, expected behavior and information exchange policies). This is a paper to study, in a practical manner, how to cover the existing gap in many digital configurations for maintenance management, designed to benefit of DTs. For this aim, it is proposed to reduce to six only requirements the analysis of DT solution. These have been included in Fig. 3. In the following case study, we try to ensure that the DT model is built fulfilling the requirements and how interaction with the real maintenance world is reached through the proposed framework.

3 The Case Study: DT of Axle Bearings in Trains for Maintenance. This case study covers a DT developed for the maintenance of axle bearings in trains. This analysis focuses on the axle bearings of a fleet of high-speed trains. Each axle of the train to be modelled is equipped with four axle bearings, two inner and two with four axle bearings, two inner and two outer. The temperature. The temperature of each bearing is controlled by a temperature probe that allows continuous monitoring of the temperature of the bearings by which is integrated as an input signal to the TCMS (Train Control Monitoring System). For the development of the case, we have access to a fleet of 16 trains which, although they correspond to the same car model, can operate

pa ns ib ilit y gr at io n In te

ct io n

Ex

In te ro er a In t

Fi de lity

y ilit ab l a Sc

pe ra bi lity

On the Definition of Requirements for a Digital Twin

81

Applicable to more equipment or failure Combinable with different models and data Extendable with new models development Precise tracking of system behavior and status Decision making and human interaction General process and system architecture integration

Fig. 3. Six DT requirements

with different configurations. The TCMS system that extracts the signals is integrated as part of an edge/cloud system that the company deploys as a general service to all its train operations and maintenance worldwide. This system allows the extraction and processing of information, as well as the representation of this information in different interfaces that allow the representation of the state of the assets and the monitoring of the fundamental parameters that the company uses in making maintenance decisions. This edge/cloud infrastructure supports the construction of digital twins. In this case, the aim is to obtain an digital twin of bearings, supported by predictive models based on the capture of temperature signals is proposed, the final objective of which is to facilitate predictive maintenance decision-making.

Fig. 4. Factors (physical model inputs) conditioning train axle bearing temperature.

A set of models were developed to understand and mimic the bearing temperatures when the trains and bearings were running under certain conditions, in specific travels where the railway and infrastructure changed as well as ambient temperature. The idea was to use this information to guide bearings maintenance, once it could be possible to detect bearing anomalies, to classify them and to offer an estimation of the remaining useful life of the bearing. A train axle bearing temperature depends on a set of factors, the operational regime, the type and dimensions of bearings, the antifrictional and hydrodynamic properties of the lubricant, the spaces between the bearing rollers and rings,

82

A. Crespo Márquez et al.

the static and dynamical loads of the bearing, the train running speed, the duration of travel without stops, the ambient air temperature, and the road curves (Lunys et al. 2015; Mironov 2008). (See Fig. 4). In a recently published article, Crespo et al. (2020a, 2020b) considered these temperatures to build the required predictive analytics for each bearing temperature. This is presented in that paper as an innovative approach (See Fig. 5). The DT presented in this case incorporates a machine learning model. Bearing temperature modelling is possible using a wide range of techniques. In this case, an ANN model was chosen. Applying the available data to other modelling options, namely: generalised linear models (GLM), decision trees (DT), random forest (RF), gradient boosted trees (GBT) and support vector machines (SVM) (GBT) and support vector machines (SVM), the company decided to use ANN because it gave the best correlation coefficient ratio vs. total time, with 0.9600 being the minimum correlation allowed.

Fig. 5. Crespo et al., approach to predict axle bearing temperatures.

The initially proposed predictive model is the basis of the digital twin in terms of fidelity in tracking the behavior of the bearings it represents. But in terms of decisionmaking, two additional lines of modelling are included. On the one hand, the definition of a system data model that allows the representation of state and degradation information at the failure mode level. The failure mode is the key element in maintenance management and the information generated by the basic predictive model is connected to the dynamic risk representation. This is developed in detail in the reference Martínez-Galan et al. (2022). Despite all of the above, it is essential to highlight the usefulness of the twin. What it is for. And not only to declare it but to model it. In this sense, the description of the services or, more precisely, micro-services, to which the WP gives rise, appears. These micro-services are a fundamental part of the DT, as are the data, the ML model and the architecture/processes that allow the DT to function. The IIC (IIC 2020) distinguishes or proposes 6 micro-services or service groups: descriptive/detection, diagnostics, predictive, visualization, AR (augmented reality). Using a similar approach Guillen et al. (2016) describe how to use detection, diagnostics and prediction as principal element of CBM (Condition Based Maintenance) Design and the interpretation of such elements within the six data-processing levels proposed by ISO 13374. Following a similar philosophy but adapted to more practical from the point of view of the most immediate needs of the company, only 3 are detailed in this use case. Results in detection, diagnosis and predictive (RUL) obtained with these models are presented now:

On the Definition of Requirements for a Digital Twin

83

Detection. One of the most relevant conclusions of the referred paper is the possibility to predict abnormal bearing temperature behavior with maximum precision and specificity, when selecting the Absolute Error descriptor with a threshold of 10 ºC in the rule, and regardless the speed of the train. Diagnosis. A sorting algorithm was selected to attempt to separate bearings with internal deterioration from those with overtemperature caused by external causes, mainly train axle guidance problems. To that end, it was necessary: 1) to generate a specific ETL process to reduce a variable stress spectrum into a simpler, equivalent set of stresses; 2) to know the final diagnosis of all the bearings observed to have suffered overtemperature cycles. It is essential to have data on whether the bearing was replaced or not, and if it was replaced, whether the analysis performed by the quality department found it with internal deterioration or not. Bearings in the train that were not replaced, but which had overtemperature cycles recorded, were obviously classified as “non-deteriorating” bearings. Basically, most of these bearings went back to normal temperature conditions when the guidance problem was solved. All these records helped to better train the classification algorithm.

Fig. 6. A comprehensive description of the axle bearings maintenance DT.

Prognosis/RUL. In this paper a statistical approach is followed to estimate the RUL (of any bearing of a train), once a positive (or anomaly detected for a failure mode) appears in a train axle bearing. A positive (according to the company existing Procedure for the Design and Implementation of CBM Plans) is defined as the occurrence of an absolute error (AE) of prediction greater than 10ºC between the actual bearing temperature and that predicted by the ANN designed for detection, when the train is running at more than 90 km/h (i.e., AE ≥ 10 ºC, V ≥ 90 km/h) and for more than one minute. RUL is defined as a random variable that, estimated from the appearance of the first positive, offers a good prediction of the life of the element until its replacement due to over temperature or noise. This replacement is nowadays performed after the activation of the safety alert in the train monitoring and control system (TCMS) and/or because of a certain inspection (probably during a weekly train inspection in the workshop). It

84

A. Crespo Márquez et al.

should be noted that this solution does not interfere with safety control system of the train which always acts as a safeguard and independently of the systems for predictive maintenance. Company’s objective through the analysis included in this part of the paper is to foresee the recommended time of bearing replacement, after its first positive, even without prior inspection, according to statistical estimates. After reviewing the comprehensive description of the DT in Fig. 6, let’s discuss about the DT requirements fulfillment. Table 1 summarize the main aspect of the study of each requirement regarding the case study. Table 1. Requirements analysis for the case study. Requirement

Case study description

Scalability

The DT model has been scalable to all train bearings requiring only the development of models per axle bearing position, regardless the axle in the train nor the train in the fleet

Interoperability Data used to train the three different types of models came from the same source and there is a procedure explaining how original data is converted and matches the different predictive analytics data models. Real time data is now used to generate an on-line output; Expansibility

There is a clear possibility to integrate new models. For instance, RUL models based on machine learning models have been introduced to replace statistical models in some applications with more consistent data

Fidelity

The ML models for anomalies detection replace in this case, with high tested precision, the very complex physical models related to the calculation of the dynamic behavior of loads in the train per axle bearings in each railway point at a certain speed

Interaction

This part has been found a very interesting requirement to fulfill. When modeling a given failure mode (FM) different risk levels or states are proposed: low, medium, high and fault. At the same time two different types of events may show up: monitoring and preventive maintenance events. It is considered that both monitoring events and PM events (with human intervention) may lead to a change in the risk level of one or more failure modes of the asset. This is because these events trigger a new risk assessment of the affected FMs. A given event may affect different failure modes and in different ways. It is also assumed that reaching a new failure mode state triggers a maintenance action (the release of an algorithm for detection or prediction, an inspection, a replacement, etc.). This human supervision of the model’s performance and interaction with the DT resulted to be critical for the DT success (continued)

On the Definition of Requirements for a Digital Twin

85

Table 1. (continued) Requirement

Case study description

Integration

The DT is to be integrated in the App in place, to control the trains fleet dynamic maintenance. Axle bearing DT must be incorporated into the comprehensive train CBM App. In this App, a total of 10 train critical systems are monitored to generate an on-line train risk assessment and to suggest an immediate action. Understanding the implications of each system risk, according to each system criticality, is critical to establish an effective dynamic maintenance strategy. In this case this DT has been integrated within Google cloud infrastructure/services

4 Conclusions Starting the path of DTs is a strategical decision for organizations. The real necessity of maintenance departments, working with assets that currently are in the middle of its life cycle with very different digitation levels, is not only to obtain a single DT model of a specific equipment, but mainly to learn modelling and understanding DTs. In this context it is critical to justify the digitalization effort that must be harmonized with the real transformation rate of traditional maintenance structures. The introduction of the six requirements of reference not only promote a suitable and accurate level of detail for description of DT solutions but also connect what could be initially appreciated as an isolate solution with a more general view, where the organization finally is aware of the real potential of the DT modeling. Results of a real use case have been presented. In this case it has been an important tool to establish a current DT development program as one work line of intelligent maintenance department. This analysis approach has been combined with other initiatives as a CBM design and implementation process that include the use of DTs and analysis methodology for new CBM project elections. This paper tries to contribute to the current efforts of the community to establish references and standards for the analysis of advanced digitisation solutions such as the digital twin. The proposed general requirements analysis is a simplified view that nevertheless highlights the key aspects to be considered and balanced in digital twin technology applications. It can be applied both in the design phase of new solutions and in the analysis phase to establish aspects such as the effectiveness, efficiency, or maturity of solutions and their scalability for larger applications. In this sense, it should also be noted that the simple model proposed can be applied to any sector or industrial application and could therefore serve as a reference base for cross-sectoral comparative analyses of the implementation of digital twin technology and its impact.

References Canedo, A.: Industrial IoT lifecycle via digital twins. In: 2016 International Conference on Hardware/Software Codesign and System Synthesis, CODES+ISSS 2016 (2016). https://doi.org/ 10.1145/2968456.2974007

86

A. Crespo Márquez et al.

Crespo, A., et al.: Criticality Analysis for improving maintenance, felling and pruning cycles in power lines. IFAC-PapersOnLine (2018). https://doi.org/10.1016/j.ifacol.2018.08.262 Crespo Márquez, A., de la Fuente Carmona, A., Marcos, J.A., Navarro, J.: Designing CBM plans, based on predictive analytics and big data tools, for train wheel bearings. Comput. Ind. 122, 103292 (2020). https://doi.org/10.1016/j.compind.2020.103292 Crespo, M.A., Rosique, A.S., Moreu de León, P., Gómez Fernández, J.F., Diego, A.G., Fernández, E.C.: Exploiting EAMS, GIS and dispatching systems data for criticality analysis. In: Value Based and Intelligent Asset Management (2020b). https://doi.org/10.1007/978-3-03020704-5_7 IIC: Industrial Internet Consortium, 2020. Digital Twins for Industrial Applications Definition, Business Values, Design Aspects, Standards and Use Cases (2020) Guillén, A.J., Crespo, A., Gómez, J.F., Sanz, M.D.: A framework for effective management of condition based maintenance programs in the context of industrial development of E-maintenance strategies. Comput. Ind. 82, 170–185 (2016) Durão, L.F.C.S., Haag, S., Anderl, R., Schützer, K., Zancul, E.: Digital twin requirements in the context of industry 4.0. In: Chiabert, P., Bouras, A., Noël, F., Ríos, J. (eds.) PLM 2018. IAICT, vol. 540, pp. 204–214. Springer, Cham (2018). https://doi.org/10.1007/978-3-030-01614-2_19 Konstantinov, S., Ahmad, M., Ananthanarayan, K., Harrison, R.: The cyber-physical e-machine manufacturing system: virtual engineering for complete lifecycle support. Procedia CIRP (2017). https://doi.org/10.1016/j.procir.2017.02.035 Lunys, O., Dailydka, S., Bureika, G.: Investigation on features and tendencies of axle-box heating. Transp. Probl. 10(1) (2015). https://doi.org/10.21307/tp-2015-011 Marquez, A.C., Fernandez, J.F.G., Fernández, P.M.G., Lopez, A.G.: Maintenance management through intelligent asset management platforms (IAMP). Emerging factors, key impact areas and data models. Energies 13(15) (2020). https://doi.org/10.3390/en13153762 Martínez-Galán, F.P., Guillén, L.A.J., Márquez, A.C., Gomez, F.J.F., Marcos, J.A.: Dynamic risk assessment for CBM-based adaptation of maintenance planning. Reliab. Eng. Syst. Saf. 223(C) (2022) Mironov, A.A.: Virtual model of the contactless thermal control axle-boxes units of a rolling stock. Transp. Urals 3(18), 59–65 (2008) Rosen, R., Von Wichert, G., Lo, G., Bettenhausen, K.D.: About the importance of autonomy and digital twins for the future of manufacturing. IFAC-PapersOnLine (2015). https://doi.org/10. 1016/j.ifacol.2015.06.141 Schleich, B., Anwer, N., Mathieu, L., Wartzack, S.: Shaping the digital twin for design and production engineering. CIRP Ann. Manuf. Technol. (2017). https://doi.org/10.1016/j.cirp.2017. 04.040 Uhlemann, T.H.J., Schock, C., Lehmann, C., Freiberger, S., Steinhilper, R.: The Digital twin: demonstrating the potential of real time data acquisition in production systems. Procedia Manuf. (2017). https://doi.org/10.1016/j.promfg.2017.04.043 Zhang, H., Liu, Q., Chen, X., Zhang, D., Leng, J.: A digital twin-based approach for designing and multi-objective optimization of hollow glass production line. IEEE Access (2017). https:// doi.org/10.1109/ACCESS.2017.2766453

Review of Asset Digitalization Models in the Context of Intelligent Asset Management and Maintenance Pilar Jiménez Alonso, Antonio J. Guillén(B) , Juan Fco. Gómez, and Eduardo Candón Department of Industrial Management, Universidad de Sevilla, 41092 Seville, Spain {ajguillen,jfgomez,ecandon}@us.es

Abstract. Assets digitalization is currently one of industrial companies’ main challenges in finally getting access to Industry 4.0 promises. Over the years industry and academia have been involved on the development of numerous references such as BIM, AIM, AAS or a great variety of DT approaches among others, to manage data and give support to organize the use of tools allowing to improve the intelligence of the asset management and maintenance. The study of these models, the fundamental elements they handle, and the application framework that each of them proposes, gives a clear idea about the complexity of the common problem that, from different approaches, these models address. At the same time, there is a clear need today to standardise the way assets are defined and managed in the digital world. Not just from the point of view of an isolated asset, but by laying the foundations for the interaction of different digitised assets in complex systemic environments and broader management levels. The fact is that digitisation is changing the nature of assets themselves, and the development of methods and strategies for asset modelling, including consideration of the asset data model and architecture model, is a necessity for developing and exploiting more complex digitisation solutions. This paper aims to review the state of the art of digital asset modelling approaches, analysing the description and the fundamental elements considered by each of them, as well as their differences and similarities.

1 Introduction Digital transformation has a particular impact on asset maintenance and management, due to the increased availability of data/information/knowledge about asset condition and performance and the increased possibilities for management and risk control, in real time and throughout the lifetime of the assets. However, the reality today is that asset management and intelligent maintenance are often reduced, to very particular and short-range applications. Modern asset management and maintenance must necessarily be data-driven and digitized. This will require an appropriate digital asset model, which structures and organizes all available data and knowledge of the asset(s). The fundamental idea of this approach is that in a context of high level of digitization the information model of an asset must be part of the asset itself. Today, there is no single vision of the digital asset but applications or partial models that try to take advantage of the available © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 87–97, 2023. https://doi.org/10.1007/978-3-031-25448-2_9

88

P. J. Alonso et al.

data, so that isolated and partial technological solutions with a low level of interaction are superimposed. The integration of different systems or technologies ends up being very complex and involves high costs and high consumption of hours of qualified personnel. This asset digitization model involves a certain complexity in its conception and management. Giving it an entity involves simultaneously managing aspects ranging from the configuration of IoT/Cloud platforms, development of analytical and simulation models, database models and anthologies, etc., to complex functional analysis and maintenance techniques. Defining the digital asset means defining its data model but also the software/hardware architecture and the processes that run through this architecture to feed and bring to life this data model. In this sense, a great research effort is being made to generate more integrated solutions. However, it is still a broad concept and therefore lacks concreteness, which is used in the most diverse ways depending on the approach or the specific industrial problem to which it is applied. There is now a consensus for generalization of models and methods. There are several international institutional initiatives such as ISO 23247 [28] or RAMI 4.0 [11] that seek to standardise this issue to facilitate the level of interaction and interconnection of complex systems. The IMF initiative [14], set out a philosophy focused on the creation of models that allow a complete integration of data and information, which supports decision-making with regard to the development, operation, maintenance and use of infrastructures and services. In any case, the reality is that there are many initiatives that can be used as a basis in the digitization of assets. This paper aims to review the main ones, comparing them and establishing basic conclusions for new research initiatives in maintenance and asset management.

2 Review of References Related to Digital Asset Modelling 2.1 Building Information Model (BIM) The Building Information Model, known by its abbreviation as BIM model, is one of the most widely used models in the industry, especially in the construction sector, as it is especially focused on the design and construction phases of assets, in this case buildings. It is also one of the most standardised models in the industry. The standard ISO 19650-1 [8] define BIM model as: “use of a shared digital representation of a built asset to facilitate the design, construction and operation processes, and provide a reliable basis for decision making”. For this type of model [31] proposes a structuring through the generation of sub-models following two different approaches, as shown in Fig. 1. On the one hand, it proposes the need to create different models according to the specific phase of the life cycle in which the asset is located. These models, which [31] defines as a three-dimensional digital representation of a constructive intention, could be called life cycle evolution models, among which we could find existing infrastructure models, construction project model, construction tender model or maintenance model; all of them should be classified into the three main stages of the life cycle. On the other hand, [31] and [13] coincide in the proposal to create IFC language models to carry out a structuring of the asset into elements according to the parent-child relationships between them, which will be associated with a particular class for better

Review of Asset Digitalization Models

89

Fig. 1. Model management in BIM technology: model hierarchy (from model of IFC basic classes to asset BIM model) & evolution of model with time

management, facilitating the exchange of information. Finally, there are many references [13, 24, 31] that propose the development within BIM technology of a data extraction platform, from which to carry out the development of diagrams, for the creation of a database that can store the information exported. The exported databases are used for their incorporation in management systems (e.g., CMMS); however, there are many occasions in which there are limitations in the integration of the information with these asset management systems. 2.2 Asset Information Model (AIM) The asset information model is defined among other references by [24] as “Data model that contains all digital data required to operate an asset or portfolio of assets”. This model focuses on the management of quality information, and the available access and visualisation of such information. The literature [13, 24] proposes its decomposition mainly into two components. Firstly, it presents the electronic document management system (EDMS) to store models and documents (graphical and non-graphical), secondly, it entails a data and information store of the asset itself (sensors data, cost, etc.); both parts are encompassed in a common data environment (CDE). This environment sometimes leads to problems of information integration, due to the lack of integration of the systems, whose data are not always compatible in type, format or quantity with the rest of the systems or with AIM itself. Asset Information Model uses a technology based on BIM and, like this model, proposes a data export system through a database, this is why both models are usually combined [24, 26]. This combination does not always optimizes the process, as there are often interoperability limitations within the CDE itself. To this end [12] the concept of federated model appears, understood as the composition

90

P. J. Alonso et al.

or relationship of different models, such as IFC models or databases, which allow the integration of the asset’s information and its complete visualization. 2.3 Digital Twin (DT) This model is defined by [4] as “A digital model of a given physical element or process with data connections that allow convergence between physical and virtual states at an appropriate rate of synchronisation”. It involves a digital and dynamic representation of an asset, process, or system, mimicking its actual behaviour. It is understood as a model based on the advanced use of data, as set out by [18]. A digital twin will collect data centrally for each entity, and then make that information available through integration interfaces. This model [18] proposes within the scientific literature a hierarchy of integration of discrete digital twins, i.e. single entity models, understood as base elements, into composite digital twins, as a combination of discrete digital twins. These in turn will be integrated as a set into a composite digital twin system. Establishing relationships and connectivity between the digital twins of various entities in an automatic way is one of the main challenges facing the design of digital twins. This reference also presents a division of the DT into three main components [18]: • Data: information necessary to represent and understand the states and behaviours of the twin in the real world. • Models: necessary to allow the system to describe, understand, and predict states and behaviours of the twin. • Services: which can be understood as the functions that a digital twin has with respect to the asset. However, despite being one of the most entrenched, this is only one of the many visions that can be found, as the Digital Twin is one of the models with the most different visions in the literature today as shown by E. Negri [21]. 2.4 RAMI 4.0 and Asset Administration Shell (AAS) The Asset Administration Shell is known as the interoperability basis of RAMI 4.0 and arises from the need to transform information into other formats due to the lack of interoperability of DT. RAMI 4.0 is a [11] service-oriented reference architecture, driven by the Platform Industrie 4.0, consisting of a three-dimensional map that encompasses what are the fundamental concepts of Industry 4.0 (Fig. 2). The AAS is defined as “Digital representation of a relevant asset, providing an interface to an Industry 4.0 network that enables connection to the physical element, and communication with other assets, as well as information exchange” [6]. Its main objective is to provide information about an asset throughout its life cycle. Each asset must have its own AAS associated. This is why they have two identifiers, one for the specific asset, meaning for the physical location of the model; and a second identifier for the Asset Administration Shell, as the digital constituent of the model. The literature [6, 25, 32] proposes the AAS model as a metamodel based on a class diagram, with the aim of structuring and simplifying the representation of asset

Review of Asset Digitalization Models

91

Fig. 2. RAMI 4.0 dimensions taken from Schweichhart, 2016 [27].

Fig. 3. Example of AAS logic interpretation taken from Wagner et al. 2017 [32].

information. This metamodel sets out a hierarchical structure so that an AAS is composed of submodels, and each of them in turn by submodel elements, such as those data, properties or functions understood as entities. Figure 3 shows an example of the AAS of a particular asset [32]. This model include information to identify not only the physical asset itself but the identification of the AAS that represent or can be related to the physical asset. Models that allow to. 2.5 Cognitive Digital Twin (CDT) The concept of Cognitive Digital Twin (CDT) arises as an evolution of what we know as a traditional Digital Twin, towards a more intelligent, integral and full life cycle representation of complex systems [2, 34]. This evolution is carried out through the development

92

P. J. Alonso et al.

of certain capabilities such as the ability to evolve the digital twin throughout the life cycle of the asset, in addition to a greater capacity for cognition and autonomy with respect to the traditional DT. Kiritsis [34], proposes a structure composed mainly of three dimensions, following a similar structure to the RAMI 4.0, so that it covers the main aspects of Industry 4.0 (Fig. 4).

Fig. 4. CDT dimensions proposed by Zheng et al. 2021 [34].

• Full lifecycle phases: dimension focuses on full lifecycle management. Throughout the lifecycle of a system, numerous digital models will be created to support the different lifecycle phases. • System hierarchy levels aims to specify the structure and boundaries of a CDT. For this model, a hierarchy is proposed that, starting with a system of systems as the highest level of the hierarchy, is decomposed into systems, subsystems, and components down parts as the lowest level. • Functional layers: specifies the different functions that a CDT can provide. Finally, it should be noted that the main challenges to be faced are related to the complexity of the system itself. That includes the complexity of data representation, and the complexity involved in the process of updating the data throughout the entire life cycle of the asset; as well as problems of information interoperability.

3 Comparison of References Once the different references have been analysed, a comparison is made by means of a summary table in which some of their main aspects are compared (Table 1).

Review of Asset Digitalization Models

93

Table 1. Comparison of references on digital asset modelling. Reference

Review

Lifecycle treatment

Asset definition

IMF elements

BIM

[5, 8, 13, 30, 31]

Technical-graphical description of physical assets from classes (IFC) integrated into a complete design

Different Classification of Data library model for parent-child data model each relationships architecture phase

AIM

[9, 13, 19, 24]

Asset quality information repository based on simplified BIM & Different model for each phase

Different Classification of Data model model for parent-child each relationships phase

DT

[18, 23, 34]

Digital representation of physical assets & Unique model generally for a single phase

Unique model generally for a single phase

ISO 23247

[1, 28]

Framework that provides a generic guideline, reference architecture, methods and approaches for the application of DT

RAMI

[7, 11, 20, 27]

Service-oriented, three-dimensional system (life cycle, hierarchy and functional layers)

Unique and evolving model

Smart product - Architecture smart factory connected world

AAS

[6, 25, 32–34]

Information record that digitally represents the asset and evolves over time. It is the basis for RAMI interoperability

Unique and evolving model

Sheet for single active

Generally single Architecture asset

Observable manufacturing element

Data model architecture

Data model

(continued)

94

P. J. Alonso et al. Table 1. (continued)

Reference

Review

Lifecycle treatment

Asset definition

IMF elements

CDT

Evolution of traditional DT, forming a set of multi-asset twins interacting with each other with cognitive capabilities

Unique and evolving model

SoS - sys. subsys. - comp. - part

Data library data model architecture

[2, 27, 34]

Starting from the philosophy of the IMF, which stems from the proposal to create a digital twin infrastructure with the objective of providing quality information to support decision-making, this reference raises the need to standardise a complete system that supports the secure exchange of information, interoperability, and the integration of data and models in any environment. To this end, it presents three components as the main components at the core, for which the need for standardisation arises: data model, reference data library and integration architecture. The last column of the table refers to the mention made by each of the references analysed to these concepts, with the aim of showing the purpose of each reference to satisfy the need for a system that allows complete integration of data and information. Another point that is important to review is the use that each of them makes of the functions and services of the model itself. While some of the references do not even mention these concepts, such as AAS, AIM or BIM; others, such as DT or CDT, do mention their existence and even [34] state the need for a location in its proposed architecture for the model’s services. However, none of them considers differentiating the two concepts. Consequently, this paper proposes to follow a similar line to two of the models studied, the DT and the CDT, due to the philosophy of the use of these models, and therefore to the importance they give to the fact of carrying out integration of information and data, beyond the simple exchange of data, reaffirming what was stated by the IMF.

4 Conclusion Digital asset modelling has become a major challenge in the development of Industry 4.0 or, if preferred, 5.0. Addressing the design of the digital asset or digital model of the asset or digital part of the asset is a need that appears explicitly in any digitisation development, but especially in the digitisation of maintenance. A general question emerges about what an asset is after digital transformation, and what should be the content of the digital part of the asset and how to design it. This is even more critical talking about the digitalization of legacy assets. According to the literature review, there is a variety of previous references, frameworks, or models that users have when it comes to managing the digitalization of their company’s assets. These should be known and can be used by companies as

Review of Asset Digitalization Models

95

digital asset modelling starting points. But it is evident the lack of general agreement of approaches, basic concepts, and proposed methodologies to be applied between them. However, what can be established as a general conclusion is the idea that the design of the digital asset cannot be approached as an isolated technical solution, but that a holistic and multidisciplinary design approach is necessary. It is necessary to consider as part of the design: aspects of systems architecture, aspects of internal functionalities linked to the internal functioning of this architecture, aspects related to the definition of the data model adapted to the asset management needs, aspects related to the definition and management of models (analytical, representation, simulation, etc.) that enrich the information and knowledge of the asset, aspects of microservices that make it possible to describe the real usefulness of these digital models for the company. It is also key that the digital model evolves with the evolution of the physical asset itself. Throughout its lifecycle the asset changes, and the digital model must change too, recording and tracking these changes to always be a faithful representation of the original physical asset. This evolution of the model throughout the lifecycle must also include the fact that the management needs and the different stages are also different and must necessarily condition the digital asset. Regarding maintenance, there is a huge gap between the practical application of technologies like digital twins and maintainers’ day-to-day, who take complicated decisions (maintenance planning and execution) every day using traditional methods and with low trust in technologies results. Even very mature technologies such as CMMS (Computer Maintenance Management System) or EAM (Enterprise Asset Management) are not fully exploited. Maintenance vision must condition digital asset modelling. On the other hand, maintenance will be one of the most impacted areas by digital asset modelling development. Therefore, it is worth highlighting the need to deep into the implementation of any of these models during the maintenance phase of the asset. Even though most of the references analysed in this work mention it -as they are models that manage the asset throughout its life entire cycle and they include the operation and maintenance phase - and there the analysis the maintenance information requirement I well-known topic - especially from the software or app design - there is a lack of studies that shows relevant results about advantages and problems of developing and implementing digital asset models for maintenance. Acknowledgements. This paper has been written within the framework of the projects INMA “Asset Digitalization for INtelligent MAintenace” (Grant PY20 RE014 AICIA, founded by Junta de Andalucía PAIDI 2020, Andalucía FEDER 2014–2020) and Geminhi (Digital model for Intelligent Maintenance based on Hybrid prognostics models), (Grant US-1381456, founded by Junta de Andalucía, Andalucía FEDER 2014–2020).

References 1. ISO/DIS 23247-2: Automation systems and integration—Digital Twin framework for manufacturing—Part 2: Reference architecture (2021) 2. Abburu, S., Berre, A.J., Jacoby, M., Roman, D., Stojanovic, L., Stojanovic, N.: Cognitive digital twins for the process industry. In: Proceedings of the Twelfth International Conference on Advanced Cognitive Technologies and Applications, Nice, France, pp. 25–29 (2020)

96

P. J. Alonso et al.

3. Alwan, Z., Gledson, B.J.: Towards green building performance evaluation using asset information modelling. In: Built Environment Project and Asset Management (2015) 4. Bevilacqua, M., et al.: Digital twin reference model development to prevent operators’ risk in process plants. Sustainability (Switzerland) 12(3), 1–17 (2020) 5. BSI: PAS 1192-3:2014 Specification for information management for the operational phase of assets using building information modelling. BSI Standards (2014) 6. Cavalieri, S., Salafia, M.G.: Asset administration shell for PLC representation based on IEC 61131–3. IEEE Access 8, 142606–142621 (2020) 7. International Electrotechnical Commission et al.: IEC PAS 63088: 2017. Smart Manufacturing–Reference Architecture Model Industry 4.0 (RAMI4. 0). IEC, Genf, pp. 1–35 (2017) 8. CTN 41/SC 13: Organización y digitalización de la información en obras de edificación e ingeniería civil que utilizan BIM. Gestión de la información al utilizar BIM. Parte 1: Conceptos y principios. (UNE-EN ISO 19650-1:2018) (2018) 9. Erguido, A., Crespo, A., Castellano, E., Kumar, A., Izquierdo, J.: Asset management framework and tools for facing challenges in the adoption of product-service systems. IEEE Trans. Eng. Manag. (2019) 10. Gavrikova, E., Volkova, I., Burda, Y.: Implementing asset data management in power companies. Int. J. Qual. Reliab. Manag. (2021) 11. Hankel, M., Rexroth, B.: The reference architectural model industrie 4.0 (rami 4.0). ZVEI, 2(2), 4–9 (2015) 12. Heaton, J., Parlikad, A.K.: Asset information model to support the adoption of a digital twin: west Cambridge case study. IFACPapersOnLine 53(3), 366–371 (2020) 13. Heaton, J., Parlikad, A.K., Schooling, J.: Design and development of BIM models to support operations and maintenance. Comput. Ind. 111, 172–186 (2019) 14. Hetherington, J., West, M.: The pathway towards an Information Management Framework (2020). https://www.repository.cam.ac.uk/handle/1810/305579 15. Une-En Iso: Norma Española Manual de entrega de la información Parte 1: Metodología y formato (2018) 16. Konstantinov, S., Assad, F., Azam, W., Vera, D., Ahmad, B., Harrison, R.: Developing webbased digital twin of assembly lines for industrial cyber-physical systems. In: 2021 4th IEEE International Conference on Industrial Cyber-Physical Systems (ICPS), pp. 219–224. IEEE, May 2021 17. Lu, Q., Xie, X., Heaton, J., Parlikad, A.K., Schooling, J.: From BIM towards digital twin: strategy and future development for smart asset management. In: Borangiu, T., Trentesaux, D., Leitão, P., GiretBoggino, A., Botti, V. (eds.) SOHOMA 2019. SCI, vol. 853, pp. 392–404. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-27477-1_30 18. Malakuti, S., et al.: Digital twins for industrial applications. IIC J. Innov. 1–19 (2020) 19. Metso, L., Kans, M.: An ecosystem perspective on asset management information. Manag. Syst. Prod. Eng. 25(3), 150–157 (2017) 20. Modularidad en la industria de procesos con I4.0 (Parte 1) (2017). https://www.infoplc.net/ plus-plus/tecnologia/item/104784-modularidad-en-la-industria-de-procesos 21. Negri, E., Fumagalli, L., Macchi, M.: A review of the roles of digital twin in CPS-based production systems. Procedia Manuf. 11, 939–948 (2017) 22. Ouertani, M.Z., Parlikad, A.K., McFarlane, D.: Asset information management: research challenges. In: Second International Conference on Research Challenges in Information Science, pp. 361–370. IEEE (2008) 23. Contributing Partners, Lead Partner, and Quality Controllers: First report on standards relevant for digital twins (2021)

Review of Asset Digitalization Models

97

24. Patacas, J., et al.: Supporting building owners and facility managers in the validation and visualisation of asset information models (AIM) through open standards and open technologies. J. Inf. Technol. Constr. 21, 434–455 (2016) 25. PlateniusMohr, M., et al.: File-and API-based interoperability of digital twins by model transformation: an IIoT case study using asset administration shell. Future Gener. Comput. Syst. 113, 94–105 (2020) 26. Raslan, A., et al.: A framework for assembling asset information models (AIMs) through permissioned blockchain. In: 2020 IEEE 44th Annual Computers, Software, and Applications Conference (COMPSAC). IEEE (2020) 27. Schweichhart, K.: Reference architectural model industrie 4.0 (rami 4.0). An Introduction (2016). https://www.plattform-i40 de I 40 28. Shao, G., et al.: Use Case Scenarios for Digital Twin Implementation Based on 11 ISO 23247. National Institute of Standards, Gaithersburg (2021) 29. TC 184/SC 4: Automation systems and integration. Digital twin framework for manufacturing. Part 1: Overview and general principles (UNE-EN ISO 23247-1:2021) (2021) 30. BSI: Specification for information management for the capital/delivery phase of construction projects using building information modelling: PAS 1192-2:2013. BSI Standards Publication 1, pp. 1–68 (2013). http://www.carillionplc.com/media/105185/building_information_mod elling.pdf 31. Ferrocarrils de la Generalitat Valenciana: Manual BIM Metodología Ferrocarrils de la Generalitat Valenciana 32. Wagner, C., et al.: The role of the Industry 4.0 asset administration shell and the digital twin during the life cycle of a plant. In: 2017 22nd IEEE International Conference on Emerging Technologies and Factory Automation (ETFA), pp. 1–8 (2017) 33. Ye, X., Seung, H.H.: Toward industry 4.0 components: insights into and implementation of asset administration shells. IEEE Industr. Electron. Mag. 13(1), 13–25 (2019) 34. Zheng, X., Lu, J., Kiritsis, D.: The emergence of cognitive digital twin: vision, challenges and opportunities. Int. J. Prod. Res. 1–23 (2021)

An Immersive Virtual Reality Platform for Enablement and Assessment of Human-Robot Interactions for Intelligent Asset Management Sören Dominik Sonntag1 , Windo Hutabarat2 , Vinayak Prabhu3 , John Oyekan2 , Ashutosh Tiwari2 , and Chris Turner4(B) 1 Manufacturing Department, Cranfield University, Bedford MK43 0AL, UK 2 Faculty of Engineering (ACSE), University of Sheffield, Sheffield S1 3JD, UK 3 Nanyang Polytechnic, 180 Ang Mo Kio Ave 8, Singapore 569830, Singapore 4 Surrey Business School, University of Surrey, Guildford GU2 7XH, UK

[email protected]

Abstract. Human-robot collaboration (HRC) is at the heart of intelligent automation and is essential for increasing efficiencies in industrial systems by combining the strength, dexterity and precision of robots with the intelligence and adaptation skills of humans. HRC is enabled by Human-Robot Interaction (HRI) in which humans and robots co-work on a task communicating and physically interacting with each other. However, the principles of HRI are neither well understood nor well established and detailed investigation is difficult because real-life experimentation involving human and industrial robot interactions carry the risk of detrimental encounters. This paper presents an opportunity to investigate and assess human-robot interactions for collaboration efficacy and safety within a realistic Virtual Reality (VR) environment A simple maintenance activity involving the assembly of two blocks was designed with two scenarios: manual assembly; robot assisted assembly.

1 Introduction The collaboration between humans and robots can enhance the efficiency of partly automated maintenance activities by combining the performance and precision of robots fulfilling repetitive tasks with the human skills of dexterity and creative problem solving. Therefore, Human-Robot Interaction (HRI) will be essential for increasing the efficiency of asset lifecycle and management activities (Rückert et al. 2020). The field of HRI focusses on understanding robotic behaviours and systems that are related to interaction with humans through communication. However, the principles of HRI are at present poorly understood. For instance, robots are commonly placed in cages in order to reduce the risk of harmful human-robot encounters. The ISO 15066 standard provides safety guidance for collaborative robot application but still refers to a previous standard (ISO 10218-2) for verification and validation. There is an opportunity to use Virtual Reality (VR) scenarios to investigate interaction strategies with virtual industrial robots in a safe © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 98–107, 2023. https://doi.org/10.1007/978-3-031-25448-2_10

An Immersive Virtual Reality Platform for HRI

99

environment. Within the virtual reality environment, the user is able to perform a set of actions, which can involve manipulation of the environment and the creation of new objects. Processes for the maintenance of assets, while being a target for automation are still likely to require human input in the near future; the proposed platform will aid the further understanding of how humans and robots can safely cooperate in the completion of asset maintenance tasks.

2 Related Research Industrial robots are part of many modern manufacturing systems. HRI becomes an important issue when humans and robots share the same workspace. Especially in assembly line production systems, robots are utilised to support human workers executing tasks requiring high precision or the handling of heavy loads (Bicchi, et al. 2008). The latter is a common application in automotive assembly lines, where industrial robots carry heavy and unwieldy loads such as car seats for the human operator, who can move the seats almost without restrictions (Krüger, et al. 2009). Literature presents two different approaches for safety in HRI. In the first approach, collisions are considered avoidable and the movement is tracked by cameras. The second approach investigates the situation in which a collision is considered unavoidable. Using the first approach, Kuli´c and Croft (2006) introduced a method to ensure real-time safety in HRI by continuously calculating a level of danger index. This index feeds into the robot motion algorithm, which tries to minimise the danger while determining the robot’s movement. This is augmented by incorporating a face tracking mechanism that identifies the direction in which the user is facing and therefore knows if the robot is addressed and if the operator is paying attention to what is happening. The approach is based on system logic in which the robot and its human operator never touch as so to prevent collisions (Kuli´c and Croft 2007). (Haddadin et al. 2008). Haddadin et al. (2008) investigate the scenario in which a collision is unavoidable or has already happened. Haddadin et al. (2008) focus on the detection of collisions and have developed strategies to prevent serious injuries, such as switching the robot to a “zero gravity” torque control mode to reduce impact forces. With regards to human-robot interaction, Richer and Drury (2006) developed a framework in order to characterise HRI platforms based on video games. The framework was used to evaluate and compare HRI systems, with multiple input and output devices and limited immersiveness. The examples described in Richer and Drury (2006) are in the field of unmanned aerial vehicles, comparing a “Virtual Cockpit Screen”, which only displays data and a very simple graphic of the UAVs position compared to the horizon, with an “Augmented Virtuality Interface”, providing a live video stream and an animated environment as output. “Furhat”, is an interactive robot developed by Skantze et al. (2015) to enable a card game that involves collaboration between two humans and a humanoid robot, investigating the extent to which the humans use the robot’s AI-based model in their decision making. However, it does not include any specific or certain knowledge of the robot, which would be a more likely scenario for industrial robots in a manufacturing shop floor environment. The work of Sosa et al. (2015) demonstrates an approach with a robot utilising a decision making process for navigating around obstacles. In this research, the robot does not incorporate

100

S. D. Sonntag et al.

any autonomous features and is fully guided by the humans during the collaborative task. Additionally Guerra-Zubiaga et al. (2021) provide a narrative on the capture of tacit knowledge to aid the further development of platforms that support human-robot interaction. The use of mixed reality in HRI is a growing field of research. Sandygulova et al. (2012) introduced a non-traditional interface for “Immersive Human-Robot Interaction” and developed a prototype where a robot and a human user both operate in a mixed reality environment. It builds on three systems from the literature incorporating the “LAIR: Lightweight Affordable Immersion Room” (Denby et al. 2009), the “Virtual Robotic Workbench” and the “NeXuS Mixed Reality Framework. The system tracks the user’s movement and actions through infrared sensors and a track pad within LAIR. While the physical robot operates in its real environment, the user has access to a 3D virtual replica monitoring the robot’s actions simultaneously. Jarvis et al. (2015) utilise the Oculus Rift DK2 in order to create a platform for training and safety assessment of employees in an industrial environment in which the user can for example face a gas leakage in a power plant.

3 Methodology An outline schematic of the proposed VR platform was developed and is shown in Fig. 1. The two actors in the schematic are the human, who will interact with the virtual robot while performing a manual task and the computer, which will provide the immersive virtual environment to enable and assess HRI.

Fig. 1. Conceptual diagram of the immersive VR platform for HRI

Fig. 2. Manual assembly task in Virtual Reality

The two scenarios are designed and delivered via the immersive environment and compared for the assessment and validation of the VR platform for HRC. The Two scenarios are as follows: 1. Manual assembly task: This scenario requires the operator to assemble two blocks manually in the virtual environment. The idea is to grip each block using the handheld VR controllers that symbolise the operator’s hands in the virtual environment, positioning and orientating the blocks for assembly using the visual markers on the blocks and mating the two blocks together to complete the assembly (Fig. 2).

An Immersive Virtual Reality Platform for HRI

101

2. Robot assisted assembly task: This scenario is similar to the first one with the difference being that there is an industrial robot arm placed in the centre of the work table to assist in the assembly task. In order to complete the task, the human operator must pick up any one of the two blocks and wait for the robot to react. The robot arm will pick up the other part and orient it in the correct position for the human to clearly see the visual markers. The human is always the first mover while the robot is always a follower reacting to the human’s actions. An industrial robotic arm with six degrees of freedom was chosen as the collaborative robot in the platform. The 3D model for the robot was developed using AutoCAD 3DS Max. 3DS Max was chosen because the CAD and animation files created and exported by this software can be easily imported in to Unity 3D, the game development engine that was used to develop the virtual environment and the interaction scenarios. A prefabricated CAD model of a robotic arm, which was obtained from a CAD model database, was imported into 3DS Max. The model required simplifications such as combining nonessential moving parts and decreasing the number of vertices of the parts’ surfaces in order to reduce the processing time for the platform later on in the development and demonstration stages. Furthermore, the robot end effector was disregarded in this work due to the complexity involved in its modelling and animation for the grip and release functions, which could be subsequently developed as part of the extended sophistication of the platform. 3.1 Development of HRI Scenarios The primary component of the proposed immersive VR platform for HRI is the set of interaction scenarios that will assist in enabling and assessing HRC. A range of possible interaction scenarios relevant to manual tasks that would benefit from HRC such as welding, surgery, assemblies and composite ply lay-ups were developed as part of this research. The two interaction scenarios were then further described and rendered using a state machine created with the PlayMaker plugin, within Unity 3D. An additional add-on enabled the use of HTC Vive with the PlayMaker plugin, getting quick access to the controller actions. The use of a state machine improved the ease of organising the scenarios for the VR platform, triggering certain events (e.g. change of state) on controller actions performed by the user (human). It is a graphical means of coding and therefore is simpler for less experienced programmers to manage events using a state machine rather than coding the scenarios into scripts (Fig. 3). 3.2 Development of the VR Platform Agile methodology for software development, including the four stages PLAN, BUILD, LAUNCH (used as ‘TEST’ for this work) and FEEDBACK, was adopted in this work in order to develop the platform with short test and refinement cycles to accommodate changing technical requirements resulting from the constantly evolving scenarios during their journey from concept to demonstration. Both internal and external participants were used in the TEST and FEEDBACK stages of platform development.

102

S. D. Sonntag et al.

Fig. 3. State machine for manual assembly scenario (scenario 1)

Fig. 4. HTC Vive development kit and the handheld controller configuration

The current second wave of virtual reality has significantly cut down prices of VR devices such as HMDs and the commercialisation has also made VR devices affordable

An Immersive Virtual Reality Platform for HRI

103

for academic research purposes (Anthes et al. 2016). Two immersive virtual reality development kits were evaluated in this work, namely, Oculus Rift and HTC Vive (the Vive is shown in Fig. 4). Both use head-mounted displays for immersive visualisation and are relatively affordable compared to industrial solutions such as VR caves. The Oculus can be used with the Leap Motion controller, which allows the users to utilise their hands to manipulate objects in the virtual environment (Webel et al. 2013). However, the Oculus restricts users to remain seated or standing in one location, which is not preferred. The HTC Vive allows the users to move around in a 5 m × 5 m square (maximum) and provides wireless handheld controllers with multiple options for interaction and navigation. After evaluating the two devices, the HTC Vive was chosen because it provides a higher level of immersion, supports the opportunity for room-scale interactions without the risk of the user developing motion sickness, and is provided with two wireless handheld controllers for user interactions with haptic feedback for the user (Fig. 4). This VR platform was developed using Unity3D, a freely available game development engine. It was selected over industrial tools for robot simulation such as CM Labs’ Vortex, as it is compatible with the HTC Vive. Unity 3D links game objects such as CAD and animation files of the robotic arm and other physical components of the task scenarios with the HRI scripts, which are coded in C#. The scripts contain code that drives the robot’s actions and reactions based on the operator’s behaviour within the task scenario. Industrial robot simulation software tends to be limited to traditional industry use cases with basic visualisation capabilities and do not support human-robot interaction scenarios.

4 Results and Validation The two scenarios were tested with users (humans) holding the two controllers, one in each hand, and wearing the headset to immerse themselves in the virtual assembly environment. Within the virtual environment, they see the assembly work table in front of them with the two blocks placed on the table ready for assembly. In the robot assisted assembly scenario, the users also see the industrial robot arm placed in the centre of the table in between the two blocks. The results are in the form of expert feedback received in running the two scenarios. 4.1 Scenario 1: Manual Assembly Task In order to complete the manual assembly task in this scenario, the operator reaches out for the first block, presses the controller’s grip button (Fig. 2), which is the same as grabbing the part in VR, and positions it within the assembly zone on the table in front of him. They then grab the second block by following the same method and correctly positions it with respect to the first part using visual markers. As soon as the two parts make contact in the correct orientation, the assembly snaps into position to complete the task.

104

S. D. Sonntag et al.

4.2 Scenario 2: Robot Assisted Assembly Task In order to complete the assembly task, the human positions their hand via the controller to grab and pick up any one of the two blocks on the table. The robot arm recognises the block picked up by the human and picks the other one up orienting it to a position where the human can clearly see the visual markers before positioning it for assembly. The human then orients their own block to match the one held by the robot before joining the blocks for assembly (Fig. 5). 4.3 Validation Validation was undertaken to assess the usefulness of the developed platform for enabling HRC. The chosen validation methodology is illustrated in Fig. 6. The methodology consisted of the four steps: Introduction; Testing; Questionnaire; Evaluation - undertaken by subjects with an academic background in manufacturing and basic knowledge of virtual reality. In the introduction phase, the subjects were introduced to the platform hardware and the associated controls. A VR environment was created so subjects could familiarise themselves with the immersive virtual setting. In the testing phase, the subjects completed the assembly task in scenario 1. They performed the task multiple times, ascertaining the quickest way to assemble the blocks correctly. Then the subjects performed the same assembly task in scenario 2 assisted by the virtual robot arm (Fig. 5). In the questionnaire phase, the subjects were asked to provide feedback on the entire VR platform experience. The questionnaire was aimed towards obtaining feedback concerning ergonomics, intuitiveness, possible enhancements and the usefulness of the VR platform for enabling HRC. In the evaluation phase, the feedback from the subjects was analysed to assess the platform’s ability to enable HRC and to obtain areas of improvement for the platform.

Fig. 5. Robot assisted assembly

Fig. 6. Validation methodology

4.4 Results The platform evaluation questionnaire was undertaken by fifteen subjects drawn from manufacturing industry and academia. Over 80% of the subjects rated the VR platform as

An Immersive Virtual Reality Platform for HRI

105

good, ‘very intuitive’, ‘quite close to reality’ and rated the ergonomics between average and good. All subjects encountered no problems during the robot assisted assembly with most stating that they felt safe and that the robot was interacting with them as expected. Their awareness for the robot’s movement increased with multiple repetitions of the scenario. These results indicate that the developed platform could work as a tool for training an operator on a previously unknown task, raising their awareness and teaching them how to fulfil the task in an environment where the operator would feel safe with the robot. A few subjects reported problems with the physics engine (e.g. assembly became stuck in the table, inaccurate collisions) and the object grabbing accuracy and identified a need for the further development of the assembly environment towards a real manufacturing setting with the use of richer graphics.

5 Discussion The main objective of this research was to develop a VR platform for HRC to provide a safe environment for investigating human-robot interactions and to give an indication of human perception should co-working between humans and robots become a possibility in real world manufacturing. This objective was achieved by building the platform and embedding it with assembly scenarios that facilitated a comparison between a completely manual task and a robot assisted task. The platform was tested by fifteen subjects and their feedback was obtained to assess the feasibility of the platform to enable HRC in the future. The results obtained were primarily positive indicating that the platform, once fully developed to the standards expected by industry, would be a useful tool for the manufacturing industry to acclimatise its workforce to HRC scenarios. In the VR platform, industrial robots can be selected from a wide variety of 3D models, task parameters can be changed and workers can run through the various task scenarios. The effects of these changes can be visualised within a few minutes in VR while realising such amendments in the real-world would be impractical with the time, effort and costs involved. The overall VR experience was perceived by most as quite close to reality, more than 50% of the subjects considered the platform as a toy rather than an industrial tool. This could be due to the close association of virtual reality to games and the perception of games as a medium of entertainment rather than a medium of training. Inclusion of more real-life scenarios, with higher fidelity graphics, may help in achieving the appropriate user perceptions. It is also the case that the current script-based implementation of inverse kinematics and the dimensions of the robot arm within it are specific to the model of the industrial robot arm used in this research. This is due to the position that at the time of writing Unity 3D has not yet released an inverse kinematics solver for generic non-human avatars, which would enable the embedding of inverse kinematics to any robot arm model that is imported into the software. It has also been found that while handheld controllers helped in impersonating human hands in the VR environment, they do not provide a natural experience of grabbing and manoeuvring objects virtually. In addition interaction opportunities with the handheld controllers are limited and the full potential is therefore not exploited in this work. Combining natural gesture recognition using

106

S. D. Sonntag et al.

depth imaging or wireless inertial sensors such as Perception Neurons would eliminate the need to use handheld controllers and make the experience more natural for users, making it more intuitive for higher complexity tasks. Finally, in the validation phase, the subject could compare the completion of a manual assembly task virtually with the performance of the same task with virtual robotic assistance. Comparison with the actual manual assembly task in a real manufacturing setting was not completed which could have provided an additional insight into how workers perceived HRC as a medium to enhance their productivity and to further gauge whether the proposed VR platform is effective in acclimatising them to real-world HRC.

6 Conclusion This paper presents an immersive virtual reality platform to enable and assess the potential of human-robot collaboration in manufacturing. The platform was developed in Unity 3D and was delivered using the HTC Vive headset as the immersive visualisation and the interaction medium. The platform included a manual assembly task with two task scenarios which enabled the users to compare the completion of a manual assembly task with and without robotic assistance in the virtual environment. In the scenario with robot assistance, the 3D model of the industrial robot arm interacted with the human by following human actions with robot actions that assisted the human with the assembly. The tasks were designed to create perceptions of human-robot collaboration effectiveness and safety into the minds of the users who performed the tasks. These perceptions were later analysed using a feedback exercise during the validation stage. The platform was evaluated as useful for the manufacturing industry in providing a safe environment for simple HRC tasks. Even with simple and unsophisticated scenarios, the differences between manual and robot-assisted tasks were perceivable, which meant that this platform could be used for acclimatisation of workers in future manufacturing scenarios, where co-working of humans and robots is anticipated to be the norm. The main contributions to knowledge as a result of this work are: (1) development of an immersive VR platform to enable and assess the potential of human-robot collaboration for the manufacturing industry, (2) development of interaction methodologies between humans and virtual robots in a virtual setting that eliminates the risk of experimenting with people and real robots, (3) creating an integrated development environment within Unity3D that combines game development with human action recognition enabling first-person interactivity between the human and game elements within a manufacturing context using affordable technology from the world of gaming, The positive feedback received in the validation of this approach indicates that the proposed platform could be a useful contribution to the manufacturing industry considering the increased demand for co-working solutions between humans and robots and the need to enhance manufacturing flexibility and productivity. Moreover, the platform has succeeded in creating a sense of safety and security for workers interacting with industrial robots, indicating that HRC could receive a positive reception when adopted by manufacturing industry.

An Immersive Virtual Reality Platform for HRI

107

References Anthes, C., García-Hernández, R.J., Wiedemann, M., Kranzlmüller, D.: State of the art of virtual reality technology. In: IEEE Aerospace Conference 2016, pp. 1–19 (2016) Bicchi, A., Peshkin, M.A., Colgate, J.E.: Safety for physical human–robot interaction. In: Siciliano, B., Khatib, O. (eds.) Springer Handbook of Robotics. pp. 1335-1348. Springer Berlin (2008). https://doi.org/10.1007/978-3-540-30301-5_58 Denby, B., Campbell, A.G., Carr, H., O’Hare, G.P.: The lair: lightweight affordable immersion room. Teleoper. Virtual Environ. 8(5), 409–11 (2009) Guerra-Zubiaga, D.A., Nasajpour-Esfahani, N., Phan, N.Q., Gupta, S., Block, L.: Tacit knowledge capture using digital tools in a human-robot interaction: a case study. In: ASME, vol. 85567, pp. V02BT02A001. American Society of Mechanical Engineers, November 2021 Haddadin, S., Albu-Schaffer, A., De Luca, A., Hirzinger, G.: Collision detection and reaction: a contribution to safe physical human-robot interaction. In: IEEE/RSJ International Conference on Intelligent Robots and Systems. IROS 2008, pp. 3356–3363, IEEE (2008) Jarvis, C., Løvset, T., Patel, D.: Revisiting virtual reality training using modern head mounted display and game engines. In: Proceedings of the 8th International Conference on Simulation Tools and Techniques, ICST (Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering), pp. 315–318 (2015) Krüger, J., Lien, T.K., Verl, A.: Cooperation of human and machines in assembly lines. CIRP Ann. Manuf. Technol. 58(2), 628–646 (2009) Kuli´c, D., Croft, E.: Pre-collision safety strategies for human-robot interaction. Auton. Robot. 22(2), 149–164 (2007) Kuli´c, D., Croft, E.A.: Real-time safety for human–robot interaction. Robot. Auton. Syst. 54(1), 1–2 (2006) Richer, J., Drury, J.L.: A video game-based framework for analyzing human-robot interaction: characterizing interface design in real-time interactive multimedia applications. In: Proceedings of the 1st ACM SIGCHI/SIGART Conference on Human-Robot Interaction. pp. 266–273. ACM (2006) Rückert, P., Tracht, K., Herfs, W., Roggendorf, S., Schubert, V., Schneider, M.: Consolidation of product lifecycle information within human-robot collaboration for assembly of multi-variant products. Procedia Manuf. 49, 217–221 (2020) Sandygulova, A., Campbell, A.G., Dragone, M., O’Hare, G.M.: Immersive human-robot interaction. In: Proceedings of the Seventh Annual ACM/IEEE International Conference on Human-Robot Interaction, pp. 227–228. ACM (2012) Skantze, G., Johansson, M., Beskow, J.: A collaborative human-robot game as a test-bed for modelling multi-party, situated interaction. In: Brinkman, W.-P., Broekens, J., Heylen, D. (eds.) IVA 2015. LNCS (LNAI), vol. 9238, pp. 348–351. Springer, Cham (2015). https://doi.org/10. 1007/978-3-319-21996-7_37 Sosa, A., Stanton, R., Perez, S., Keyes-, C., Gonzalez, S., Toups, Z.O.: Imperfect robot control in a mixed reality game to teach hybrid human-robot team coordination. In: Proceedings of the 2015 Annual Symposium on Computer-Human Interaction in Play, pp. 697–702. ACM (2015) Webel, S., Olbrich, M., Franke, T., Keil, J.: Immersive experience of current and ancient reconstructed cultural attractions. In: Digital Heritage International Congress (DigitalHeritage), vol. 1, pp. 395–398. IEEE (2013)

Exploring Augmented Reality Applications to Support Maintenance Management in Hydroelectric Power Plants Renan Favarão da Silva(B) and Gilberto Francisco Martha de Souza Department of Mechatronics and Mechanic Systems Engineering, University of São Paulo, Av. Professor, Mello de Moraes, 2231, São Paulo, SP 05508-900, Brazil [email protected]

Abstract. Several enabling technologies were adopted to transform industrial sectors and deliver benefits in Industry 4.0. Within an advanced Human-Machine Interface, Augmented Reality (AR) is a technology that can be applied to this purpose in physical asset management. Nevertheless, AR applications for maintenance are still emerging and not widely used, especially in the power generation industry. Hence, this paper proposes AR applications to support maintenance management in hydroelectric power plants (HPPs). To this end, the methodology comprised five steps: a review of AR technology, identification of needs and opportunities for maintenance management, propositions of AR applications, demonstration through AR prototyping, and discussion of the applications. As the main result, a set of proposals for AR applications were presented and discussed to support maintenance management in HPPs. The proposed AR applications mainly addressed activities of the maintenance planning and execution processes and showed the potential to contribute to human reliability aspects in the plant and boost digital transformation within maintenance.

1 Introduction The increasing pursuit of productivity and competitiveness has contributed to the recognition of the importance of maintenance management in organizations. Maintenance has a key role in ensuring the success of a manufacturing company due to its impact on both productivity and quality (Silvestri et al. 2020). In asset-intensive industries such as hydroelectric power plants, the performance of maintenance to realize value from physical assets is even more critical. Within the development of the Fourth Industrial Revolution, several enabling technologies were adopted to transform industrial sectors and deliver benefits, also known as the nine technological pillars (Bona et al. 2021). Industry 4.0 enables intelligent and flexible manufacturing control using IT-based intercommunicating and interacting machines, products, services, equipment, and tools (Wang 2016). Accordingly, there is a growing interest in investigating how these technologies can improve maintenance management. Augmented Reality (AR) is one of the Industry 4.0 technologies that can be integrated with maintenance for digital transformation enabling advanced Human-Machine © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 108–117, 2023. https://doi.org/10.1007/978-3-031-25448-2_11

Exploring Augmented Reality Applications

109

Interface applications. AR is a way to augment the real environment with virtual objects (Milgram and Kishino 1994). In other words, it is a real-time interaction that superimposes digital content on the reality that can be accessed by using devices that support AR applications. Although AR technology has been around for over 50 years, there are still limited examples of its concrete application industry (Palmarini et al. 2018), especially in the power generation industry. In this context, this paper proposes AR applications to support maintenance management in hydroelectric power plants. It intends to explore the use of augmented reality to maintenance from the needs and opportunities of this industry sector. Hence, AR applications for maintenance are proposed and discussed considering the operational context of a case study hydroelectric power plant (HPP). The remainder of this paper is organized as follows: Sect. 2 presents a brief description of the AR technology and its potential for maintenance management. Section 3 describes the methodology for achieving the objective of this research. Then, Sect. 4 presents and discusses the proposed AR applications and demonstrates some of them through AR prototyping. Finally, Sect. 5 presents the final considerations and some directions for future work.

2 AR Technology within Maintenance Management In a Mixed Reality (MR) environment or application, real and virtual objects are combined and mixed to create a user experience incorporating both the real and the virtual world (Juraschek et al. 2018). Augmented Reality (AR) is part of MR and is characterized by predominantly sharing objects and the real environment. In other words, within the reality-virtuality-continuum (Milgram and Kishino 1994), AR applications are the closest to the real environment with the insertion of virtual objects while virtual reality (VR) is closer to immersive and virtual environments. Rapid development in hardware and software technology over the past decades has made AR a versatile and useful tool in many applications and fields (Safi et al. 2019). The industrial use of AR technology is growing and contributes to improve maintenance activities as well as the human-machine interface (Gallala et al. 2019). Among the main industrial AR application areas are production assistance, remote maintenance, training, product design, and others (Quandt et al. 2018). AR applications for maintenance management are growing in the literature in recent years. In a survey on the evolution of AR use in industry, Gallala et al. (2019) highlighted applications for assembly and disassembly and remote maintenance. For instance, in the aerospace sector, AR applications for maintenance are used to guide operator performance through a given task towards efficiency (Safi et al. 2019) while, in the military sector, they are used for maintenance training of military equipment (Wang et al. 2020). For better comprehension, Palmarini et al. (2018) conducted a systematic review to provide the state of the art of AR applications in maintenance which included 30 documents. The results showed that AR applications focused on four main maintenance activities: Assembly and disassembly (33%), Maintenance repair (26%), Inspection and diagnosis (26%), and training (15%). For that, the main hardware choices to implement them are using Head-Mounted Display (HMD) (30%) and Handheld Display (27%) such

110

R. F. da Silva and G. F. M. de Souza

as smartphones and tablets, Desktop PC (27%), and others (16%) such as projector, haptic and sensors. Regarding the way the AR interacts with the user, dynamic 2D/3D objects (40%), static 2D/3D objects (26%), text (26%), and audio (8%) are the interaction methods used for AR applications for maintenance. Finally, the AR applications for maintenance are mainly developed with marker-based tracking (52%) or model-based tracking (19%) (Palmarini et al. 2018). Although augmented reality is still under development for industrial applications, the literature reports promising benefits for the use of the technology in maintenance. AR can enhance human performance in carrying out technical maintenance tasks (Palmarini et al. 2018) and serves as a visual guide to assist users in inspections (Koh et al. 2020). It is found to decrease task completion time and the number of errors (Brown et al. 2021) and to simplify communication for remote maintenance assistance between on-site workers and remote experts (Aschauer et al. 2021). Even though several technological limitations with hardware and software that prevented AR to become a tool for industry in the past have been overcome (Masoni et al. 2017), it is worth mentioning that other aspects of AR technology can be barriers to application in maintenance. For instance, Brown et al. (2021) pointed out that greater difficulties for AR applications for maintenance stem from content creation. The selection of suitable concepts as well as hardware and software is also a challenge (Juraschek et al. 2018). Furthermore, their use for long periods is not recommended since they can cause eyestrain (Silvestri et al. 2020). Nevertheless, even with some reported concerns and difficulties, the potential benefits corroborate the current interest in AR technology for maintenance and the exploration of the use for maintenance management in hydroelectric power plants (HPPs) as the scope of this work.

3 Methodology As this paper aims to propose AR applications to support maintenance management in hydroelectric power plants, it is applied research that generates knowledge for practical solutions. Due to this exploratory objective, the research has a predominantly qualitative approach. Thus, a five-step method was structured to achieve its objective, as shown in Fig. 1.

Exploring Augmented Reality Applications

111

Fig. 1. Main steps of the research method

First, the AR technology was reviewed to gain a broad understanding of its concepts and potentialities. In parallel, the needs and opportunities for maintenance management were identified considering the operational context of a selected HPP. This information from both steps was combined to propose a set of AR applications for maintenance. Then, some proposed applications were demonstrated through AR prototypes that contributed to a better understanding of the AR technology in practice. Finally, the applications were discussed to highlight their potential benefits for maintenance management.

4 Results and Discussion The exploration of AR for applications in the maintenance management of HPPs followed the five steps foreseen in the research method. First, the review of AR technology in the literature provided the main concepts as well as potential uses for maintenance management, as presented in Sect. 2. Then, it was necessary to identify the needs and opportunities for maintenance management in the context of the hydroelectric power plants. In this second step, it was considered as a case study a Brazilian hydroelectric power plant composed of four Kaplan turbine generating units with a total installed capacity of around 200 MW. A technical visit to the plant, meetings with its technical personnel, and on-site data collection were carried out to support the identification of needs and opportunities for maintenance management. An understanding of the selected HPP and its maintenance management context was fundamental for the success of this step. Combining the results of both steps, it was possible to propose AR applications to support maintenance management in HPPs, as shown in Table 1. As can be seen in Table 1, different AR applications were proposed for two different maintenance processes. Although there are other processes in a maintenance management framework (da Silva and de Souza 2021), maintenance planning and maintenance execution condensed the main needs and opportunities for the use of AR technology in HPPs. In other words, these maintenance processes have the activities most likely to benefit from AR applications and to impact strategic objectives such as the improvement of the availability of power generation.

112

R. F. da Silva and G. F. M. de Souza Table 1. Proposals of AR application for maintenance management in HPPs

Maintenance process

AR purpose

AR application proposal Description

Maintenance planning

Information visualization

Enhanced maintenance work order

The use of AR to provide support content to enhance the maintenance work orders information

Training and education

Enhanced maintenance training

The use of AR to support the understanding of procedures, equipment, and maintenance tasks

Humanmachine interface

Enhanced equipment and systems interface

The use of AR to superimpose information on physical assets such as identification, specifications, and current condition

Remote assistance

Remote maintenance assistance

The use of AR to allow enhanced remote assistance in the maintenance activities of the HPP

Information visualization

Enhanced maintenance inspections

The use of AR to provide support content for inspections and diagnostics such as reference conditions and animations

Maintenance execution

For better understanding, two of the proposed AR applications related to information visualization were selected for demonstration through marker-based AR prototypes, as shown in Fig. 2 and Fig. 3. Then, as a third and last demonstration, the remote maintenance assistance was selected for AR prototyping with the aid of an interactive video calling system, as shown in Fig. 4.

Exploring Augmented Reality Applications

113

In the first AR prototype, the proposal for an enhanced maintenance work order aims to provide support content to the assigned maintenance technicians. For example, Fig. 2 represents a bearing replacement maintenance work order for a geared motor from a computerized maintenance management system. It was enhanced with the support of the UniteAR platform by adding a marker that triggers the AR mobile application and provides information visualization. In this case, it showed the model of the geared motor where maintenance should be carried out as well as the type of bearing and an indication of which bearing should be replaced.

Fig. 2. AR application prototype: enhanced maintenance work order

Fig. 3. AR application prototype: enhanced maintenance inspection

The use of AR to provide support content for inspections and diagnostics such as reference conditions and animations was demonstrated with a second prototype. Figure 3 represents an enhanced maintenance inspection in a compressed air pressure vessel of the speed governor system. It was also enhanced with the support of the UniteAR platform

114

R. F. da Silva and G. F. M. de Souza

by adding a maker in the vessel that triggers an AR mobile application and provides information visualization. In this case, it showed an interactive image highlighting the vessel manometer and the normal pressure limits for the system. Finally, the third prototype covered remote maintenance assistance using AR to allow experts from suppliers or other HPP from the same organization to assist local technicians in the maintenance execution remotely. For instance, Fig. 4 represents remote assistance where an external specialist guided the maintenance technician about the lubricating oil injection system in the generator’s combined bearing. Through an interactive video calling system, the maintenance technician can share a live camera or a photo of his/her point of view while the external specialist annotates it on the screen. Therefore, when the technician observes reality through the display, the inserted virtual content augments the real environment.

Fig. 4. AR application prototype: Remote maintenance assistance (a) Expert (b) User

Although not demonstrated, it is worth mentioning that the proposal for enhanced equipment and systems interface could also be demonstrated using a marker-based

Exploring Augmented Reality Applications

115

and mobile AR application as applied in the first two prototypes previously. Nevertheless, considering the operational context of an HPP, the use of handheld displays could decrease the immersion and feasibility of this proposal as the users would have to walk around the plant holding a mobile device in front of their field of vision. In this case, it is suggested to develop applications with the use of AR head-mounted displays as hardware combining maker-based and maker-less content. The enhanced maintenance training proposal, in turn, could also be demonstrated using the maker-based AR application prototype. As previously presented, the marker could trigger and track the content composed of 2D/3D dynamic objects, text, or audio about a maintenance procedure, equipment, or maintenance task. Thus, it has the potential to contribute to human reliability aspects in critical maintenance activities. However, although AR is used for maintenance training, this is the smallest part of maintenance applications (Palmarini et al. 2018) as they could be performed in Virtual Reality (VR) given greater immersion. Accordingly, all proposed AR applications presented in Table 1 can contribute to the digital transformation of maintenance management processes. The discussed prototypes were developed by the authors and presented to the selected HPP to demonstrate the capabilities of AR technology in its context and potential partnerships for future developments. It is also worth mentioning that these are not the only possible AR applications to support the maintenance of HPPs, but the most relevant considering the context of the selected case study. Furthermore, other AR applications for maintenance can be proposed considering different maintenance management processes such as performance evaluation and maintenance improvement (da Silva and de Souza 2021).

5 Conclusions Well-established maintenance management is essential to ensure the competitiveness of organizations. To achieve better results, it shall identify opportunities for improvement and address them, especially within the Industry 4.0 context. As AR has arisen as an enabling technology for handling increasingly complex maintenance processes (Gallala et al. 2019), there is an increasing interest in this technology and its applications for maintenance. Accordingly, this paper explored the use of Augmented Reality (AR) to support maintenance management in hydroelectric power plants by proposing several AR applications. For that, a hydroelectric power plant was selected as a case study to identify the needs and opportunities for the use of this technology in maintenance based on its operational context. As the main result, a set of proposals for AR applications were presented and discussed through AR prototypes to support maintenance management in HPPs. Although these results highlight the digital transformation potential of AR technology for maintenance, it is worth mentioning some limitations of this paper. Since it is predominantly exploratory, it was intended to provide a greater understanding of AR technology and to improve ideas by proposing possible applications for maintenance in HPPs. However, the AR proposals were not evaluated across important aspects such as cost-effectiveness, data security, setup time, ergonomics, accuracy, and reliability which

116

R. F. da Silva and G. F. M. de Souza

are important requirements to be considered for industrial applications reported in the literature (Quandt et al. 2018). As a direct suggestion for future work, there is the development of a method for evaluating the AR proposals for maintenance supported based on Multicriteria Decision Making (MCDM) methods. It could help organizations decide which is the best proposal considering relevant criteria of the AR technology and their preferences and physical assets’ objectives. In addition, other opportunities consider the development of selected AR applications for maintenance within the HPP using a software development kit (SDK), web framework, or game engines, and discuss the results achieved as well as the difficulties and recommendations. Finally, the findings of this paper are expected to contribute to the researchers and professionals in the field of physical asset management and maintenance management as Augmented Reality applications are innovative approaches to support maintenance management and provide advantages for organizations. The research methodology as well as the proposed AR applications can also be explored and replicated for other industrial sectors in a similar way.

References Aschauer, A., ReisnerKollmann, I., Wolfartsberger, J.: Creating an open-source augmented reality remote support tool for industry: challenges and learnings. Procedia Computer Science 180, 269–279 (2021) Bona, G.D., Cesarotti, V., Arcese, G., Gallo, T.: Implementation of Industry 4.0 technology: new opportunities and challenges for maintenance strategy. Procedia Comput. Sci. 180, 424–429 (2021) Brown, C., Hicks, J., Rinaudo, C.H., Burch, R.: The use of augmented reality and virtual reality in ergonomic applications for education, aviation, and maintenance. Ergon. Des. Quart. Hum. Fact. Appl. 106480462110034 (2021) Gallala, A., Hichri, B., Plapper, P.: Survey: the evolution of the usage of augmented reality in Industry 4.0. IOP Conf. Ser. Mat. Sci. Eng. 521(1), 012017 (2019) Juraschek, M., Büth, L., Posselt, G., Herrmann, C.: Mixed reality in learning factories. Procedia Manuf. 23(10), 153–158 (2018) Koh, Y.S., et al.: A review on Augmented Reality tracking methods for maintenance of robots. Jurnal Teknologi 83(1), 37–43 (2020) Masoni, R., et al.: Supporting remote maintenance in industry 4.0 through augmented reality. Procedia Manufacturing 11, 1296–1302 (2017) Milgram, P., Kishino, F.: A taxonomy of mixed reality visual displays. IEICE Trans. Inf. Syst. 12, 1321–1329 (1994) Palmarini, R., Erkoyuncu, J.A., Roy, R., Torabmostaedi, H.: A systematic review of augmented reality applications in maintenance. Robot. Comput.-Integr. Manuf. 49, 215–228 (2018) Quandt, M., Knoke, B., Gorldt, C., Freitag, M., Thoben, K.-D.: General requirements for industrial augmented reality applications. Procedia CIRP 72, 1130–1135 (2018) Safi, M., Chung, J., Pradhan, P.: Review of augmented reality in aerospace industry. Aircr. Eng. Aerosp. Technol. 91(9), 1187–1194 (2019) da Silva, R.F., de Souza, G.F.M.: Modeling a maintenance management framework for asset management based on ISO 55000 series guidelines. J. Qual. Maintenance Eng. (2021). https:// doi.org/10.1108/JQME-08-2020-0082

Exploring Augmented Reality Applications

117

Silvestri, L., Forcina, A., Introna, V., Santolamazza, A., Cesarotti, V.: “Maintenance transformation through Industry 4.0 technologies: a systematic literature review. Comput. Ind. 123, 103335 (2020) Wang, K.: Intelligent Predictive Maintenance ( IPdM ) system – Industry 4.0 scenario. In: Wang, K., Wang, Y., Strandhagen, J.O., Yu, T. (eds.) Advanced Manufacturing and Automation V, 1st edn., p. 432. WIT Press, Southampton (2016) Wang, W., Lei, S., Liu, H., Li, T., Qu, J., Qiu, A.: Augmented reality in maintenance training for military equipment. J. Phys: Conf. Ser. 1626(1), 012184 (2020)

Smart Water Dam Transformation in Industry 4.0 Gowrishankar Sabapathipillai(B) , Srijeyanthan Kuganesan, and Thanansan Kuganesan Tigernix Pty Ltd., Brisbane, QLD 4000 Level 14, 167 Eagle Street,, Australia {s.gowrishankar,k.srijeyanthan,k.thanansan}@tigernix.com Abstract. There are two most critical responsibilities of water dam infrastructure managers: ensuring the health and sustenance of the water dam infrastructure and reassuring the quality and quantity of dammed water in the reservoir is maintained favourably. Firstly, the paper discusses how the dam wall, gates and other infrastructural components of the catchment area can be maintained and constantly protected from unprecedented dam events using industry 4.0 technologies. The paper secondly identifies a list of industry 4.0 technologies that optimise the quality and maintain the precise quantity of stored water based on the water requirements of the community. Furthermore, it discovers how these tools can ensure the conservation of natural resources surrounding the water dam area. The reader will understand how the water dam structures can be advanced to avoid future dam threats, risks and degradation, plus how the water stored in the dam is protected from contamination and undersupply.

1 Introduction Water dam managers should familiarise themselves with recent digital technologies to optimise the activity of planning, engineering, developing and managing water resources across cities or states. One of the most integral responsibilities of quality water resource management is overseeing the quality of the catchment area. Catchment decision-makers are constantly impacted by uncontrollable climate changes, rainfall events, algae, decaying infrastructure etc.– and controllable - dam breaches, rising dam levels, water pollution etc.- Risk factors that lead to failures in spill gates, dam walls and more. Furthermore, with the uprising of water service demands due to increased population and gross domestic product (GDP) pro-capita, water infrastructure specialists are compelled to build new dams or renovate existing infrastructure and introduce smarter runoff and evaporation suppression technologies. To enable sustainable water supply and ensure the quality of dammed water, these experts should familiarise themselves with the concept of ‘Smart Water Dam Transformation with Industry 4.0’. Smart Water Management System is a scalable platform introduced by Tigernix Pty Ltd in 2021 that allows water asset specialists to manage 4 main pillars of the water system, namely the catchment, treatment, distribution and reticulation. Smart Catchment Management System is one of the modules integrated into this platform to allow dam managers, operators, investors and other decision-makers to get a comprehensive and updated version of all dam operations under a single scalable and interactive system. The conventional catchment system is powered with the latest modelling technologies to power up smart catchments, such as © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 118–129, 2023. https://doi.org/10.1007/978-3-031-25448-2_12

Smart Water Dam Transformation in Industry 4.0

a. b. c. d. e.

119

Optimal reservoir operating model to exactly calibrate optimal operating solutions. Accurate Dam inflow estimates and forecast Dam displacement prediction Flood simulation integration and gate operating strategies Non-invasive models to provide ground intelligence

Thus, the above smart features together make the normal catchment solution into a smart and integrated catchment system, like the Smart Catchment Management. After the advent of industry 4.0 technology, Smart Catchment Management was redefined and advanced by allowing water asset specialists to leverage the power of AI and other digital technologies to refine catchment infrastructure management using smart technologies like predictive analytics, prescriptive analytics, IIoT integration, 4D Digital Twin Environments, Satellite-based Remote Sensing Technology, GIS maps, real-time algal growth and water decontamination detectors etc. This paper will elaborate on two of the most critical responsibilities of modern water dam infrastructure managers: • ensuring the health and sustenance of the water dam infrastructure • ensuring that the quality and quantity of stored water in the reservoir are maintained favourably. The paper evaluates the impact of industry 4.0 technological advancements in ameliorating the meeting of these two responsibilities. The paper discusses how the dam wall, gates and other infrastructural components can be maintained and constantly protected from unprecedented dam events; it explains: • Ground motion and seepage monitoring to slow down dam deformation and terrain movement to reduce risks • Detect urban change in the areas that are served by the water dam • New analytical technologies powered by AI and ML models to predict the water catchment integrity, performance and sustenance for better maintenance approaches • How to activate real-time performance and risk identifiers before financial, operational and administrative conflicts appear via IIoT systems • The interactive and scalable visualisation tools like Digital Twin, Smart Dashboards and GIS Maps to gain a hands-on view of Dam Structures • Event-driven simulation platforms that help the dam operators create risk-mitigated maintenance plans for unique future dam events • How to automate scalable financial profiles that take fact-based predictions into account like water demands, dam usage, environmental conditions, demographic impacts and more The paper secondly explains industry 4.0 technologies that optimise the quality and maintain the precise quantity of stored water based on the water requirements of the community. It also describes how these tools can ensure the conservation of natural resources surrounding the water dam areas; it explains:

120

G. Sabapathipillai et al.

• New technologies that detect chemical imbalances in water by using historical parametric insights to train predictive models • Predictive technologies that isolate HAB(Harmful Algal Bloom)-prone areas on GIS Maps to ensure the safety of water users and the aquatic wildlife • Social and environmental impacts on water quality and quantity like drought, floods, landslides, bushfires, seasonal water demands, and extreme weather conditions. This will help in understanding how to seal water from being polluted after natural or human-made disasters. • What technologies allow for building pre-indicators of water contaminants and public health risks (like leaking septic tanks, urban pollution, infections, toxic metals and chemicals etc.) Thirdly, the paper concludes by elucidating to the reader how industry 4.0 technologies can ensure the preservation of the biodiversities surrounding the catchment area.

2 Issues Faced by Dam Managers Who Rely on Traditional Catchment Management Many asset-oriented events regularly impact the water service sector in countries on a daily basis due to an accumulation of social, environmental, and political influences. Water infrastructure utilisers must therefore gain a comprehensive insight by letting go of traditional asset technologies that allow asset managers to use inefficient and misleading time-based maintenance protocols to strengthen the health of water infrastructure. For example, water dam managers who use a time-based asset management strategy would follow the same agenda to maintain dam infrastructure in several locations. This retrofitting practice has various shortcomings because using a static method of managing dynamic degrading assets is ineffective. Time-based inspections may be inefficient; they are proven to monitor and identify known issues with dam civil structures as well as the electrical/mechanical assets. Two reasons why time-based asset management strategies are suboptimal in managing catchment infrastructure are explained below. 2.1 Dynamic Asset Degradation A substandard dam infrastructure management effort can cause a series of detrimental dam events: dam overtopping, flooding, slope instabilities, crack settling of the dam that causes dam bursts, fish passages, sinkholes in the dam, and more. However, catchment areas of the same city or state vastly differ from one another due to their particulars. A water dam can degrade due to a number of impacts: the strength, height, size, and materials of the dam infrastructure, the topography, the activity of the dam operations, human-made and natural disasters, vegetation, and fauna, the quality of the soil, etc. Thus, water dams’ degradation rates are never identical to one another. The dynamicity of the dam degradation can only be detected by a versatile technology solution that monitors all parameters around the clock. Even though it is of greater significance that a water asset manager should conduct to prevent dam disasters, the inconsistent degree

Smart Water Dam Transformation in Industry 4.0

121

of dam degradation has made it an arduous task. This proves that traditional time-based infrastructure maintenance alone cannot avoid expensive dam failure and mitigate dam asset risks. 2.2 Variational Water Quality and Quantity On the other hand, the quality and quantity of dammed water can also vary from one site to another due to the same influences. Poor water quality will create sanctuaries for invasive aquatic life that can smother the life of healthy and non-toxic aquatic ecosystems. The water quality and quantity of each dam can be diverse due to the varied impacts of maritime traffic, spillages, industrial activities, water dumping, deforestation, dynamic evaporation rates caused due to global warming, eutrophication, etc. Therefore, each dammed water has its own quality, quantity, and demand, which raises an important question: how can human labour alone test the quality and quantity parameters of large water bodies in an interminable loop? Thus, the need for industry 4.0 technology to optimise catchment water remains to be an advancement that can affect the water utility services’ bottom line (Fig. 1).

3 Transforming from Conventional Dam Management to an Industry 4.0-Driven Smart Catchment Environment

Fig. 1. How Tigernix Catchment Management Model works

122

G. Sabapathipillai et al.

Due to the obvious shortcomings of the traditional catchment management practices, the rise of smart and integrated digital technologies powered by the concept of industry 4.0 was inevitable. Smart catchment systems will address the issues of the catchment area that plague the dam areas, like dynamic asset deterioration, climate changes, and risk factors of spill systems, unprecedented algae-driven water contamination, which are the most critical issues of contemporary catchment specialists. This is why the use of industry 4.0 technology through the Smart Catchment System is best advised for catchment operators in the water industry. Such digital-based systems can deliver enhanced risk management profiles to optimise asset lifecycles and maintain water quality and quantity levels. Industry 4.0 technologies have allowed water asset managers to combine realtime, comprehensible and accessible water supply, quality and demand data with smart, automated insights generators of the Smart Catchment System. Modern water asset managers can therefore use this solution to utilise a collection of digital technologies to ensure that the catchment asset life cycles are healthy and that dammed water remains in the preferred states in terms of water quantity and quality. In general, the Smart Catchment System analyses three main input data from three main sources, as shown in the above diagram, namely Climate Change Analysis, Hydrological Models and Short Term Seasonal Prediction. To elaborate: Step 1: Climate Change Analysis • Examine and verify noteworthy climate signals • Future climate analysed and projected using the CMIP-5 • Climate data selected according to the regional performances and bias-corrected by means of observed rainfall data Step 2: Hydrological Modelling • Hydrological response of the basin for the present and future climate projected using AI models • This model is capable of estimating soil moisture, evaporation, groundwater, and inundation • Topographical data, soil, sand, vegetation data, evaporation data are used to train the model under real-world basin scenarios Step 3: Short-term Seasonal Prediction • The Standard Precipitation Index (SPI) and wavelet analysis conducted to predict the short-term seasonal prediction and as a result optimise the dam operation • This model is creating a virtual relationship with rainfall pattern and various real-time information gathered through sensors

Smart Water Dam Transformation in Industry 4.0

123

3.1 Ensuring the Dam Health and Sustenance The core industry 4.0 technology for ensuring dam health and sustenance is Artificial Intelligence (AI). AI unleashes new analytical technologies powered by ML models to predict the water catchment integrity, performance and sustenance for better maintenance approaches. Dam operators can utilise AI models to monitor odd occurrences in the catchment area, including the soil’s wetness, vegetation growth, asset degradation, environmental changes, and ground motion. AI is a powerful approach to wring insight out of disparate data, which may include free text as well as numeric data, in combination with several data feeds delivered by integrating disparate data sources: satellites, cloud drives, IoT sensors, applications, websites, databases and more it creates a unique and precise analytical picture of the dam over time. AI technology also allows the dam decision-makers to minimise safety risks of the labour, neighbouring communities and natural resources. It can implement predictive models that are matured by historical data and real-time data to simulate according to highly probable dam events to generate data-driven risk mitigation approaches to reduce accidents during dam operations. Moreover, modern dam operators can utilise industry 4.0 technology to activate realtime performance and risk identifiers before financial, operational and administrative conflicts appear via IIoT powered by AI. AI technology collects data from all data-rich points to assess the external parameters -temperature, evaporation, seepage, water flow, water pressure, etc.-that affects the dam infrastructure and the internal parameters like dam asset specifications- dam strength, resilience, sustenance, lifecycle stage, criticality, etc.-. It combines the internal and external parameters of the dam with historical data to give a complete perspective of how, when and why a dam infrastructure will fail and prescribe counter-reactive maintenance measures to avoid impending risks, failures and accidents. This will allow the dam managers to avoid expensive accidents by presupposing dam events and tailoring event-driven dam maintenance and risk mitigation programs to avoid ineffective operational, engineering, administrative and financial decisions (Figs. 2 and 3). Furthermore, industry 4.0 technologies allow catchment asset managers the luxury of detecting ground motion and seepage in all noteworthy ranges of the reservoir- even the hard-to-reach ground points. This disclosure slows down dam deformation and sustains the terrain movement reducing many risks that affect the dam infrastructure. IoT-driven sensors can be installed around the dam to share real-time information to automatically trigger alarms of impactful ground motions, giving the end-users a comprehensive profile to make insightful decisions at the central common centre (Sreekar et al. 2018). The dam operators can thereafter take precautions in ensuring the dam safety to either hold or release water to sustain the dam pressure and impact healthy levels. With the needful data collected to a centralised intelligence point by IoT, the dam smart catchment system can automate scalable financial profiles that consider fact-based predictions before generating financial reports. In the past, dam operators followed timebased maintenance approaches- regarding all dam infrastructure reactions and defects in the same manner- this led to inaccurate budgetary estimates. To overcome this shortcoming, industry 4.0-defined catchment solutions play an integral role in optimising the dam’s operational performance and reliability and extending its lifespan. It offers real-time visualisation, analytical perspectives and predicted insights throughout the dam

124

G. Sabapathipillai et al.

Fig. 2. Virtual dam aerial view via Tigernix Smart Catchment Solution

Fig. 3. Optimised valve point graph Tigernix Smart Catchment Solution

infrastructure’s lifecycle: planning, designing, engineering, maintaining and reconstructing. Industry 4.0 tools and features of the systems allows the dam designers, engineers and operators to collaborate in configuring the operations of the dam according to its predicted events to avoid impending failures or accidents such as overtopping, structural cracks, seepage failures, foundational defects, compression, engineering flaws and more. AI offers predictive analytical models that parallelly correspond to predicted issues with their significant maintenance decisions to control them with less damage. This way, the catchment managers can manage water demands, engineer catchment areas in the best locations, configure watergate operations, maintain dams and continuously improve catchment areas in a timely manner. AI and IoT technology allow financial decisionmakers of the catchment area to assess the real-time and predict the future water demands,

Smart Water Dam Transformation in Industry 4.0

125

dam usage, environmental conditions, demographic impacts and more before allocating budgets. Therefore, it also helps in creating comprehensive, realistic and reliable financial scopes before pencilling down estimated budgets and filling in fund requests. Additionally, IoT technologies reimagine the end-user perspectives of Smart Catchment Systems to ensure proactive decisions in sustaining dam health and integrity. The interactive and scalable visualisation tools like Digital Twin, Smart Dashboards and GIS Maps can be built based on the unique requirements of dam operators to gain a hands-on view of dam structures in real-time. 4D digital twin environments allow dam operators to use immersive devices like AR and VR to virtually visit dam sites and manipulate the time and events of catchment areas in the virtual world using simulation technology. These realistic insights give the dam operators a hands-on experience of how future dam events can affect the dam structures’ health and sustenance if not with proactive asset maintenance and watergate operations. This can use AR to manipulate virtual counterparts of the dam and right the wrongs based on critical dam events. Forbye, the user can also gain 24/7 updates via smart dashboards on how the water dam is maintained, managed and sustained despite the user’s distance from the actual dam site. Dashboardstyle screens allow the user to import and validate data by fully automated, scalable and easy-to-use visual representations of dams and other noteworthy parameters (Sreekar et al. 2018). Dam operators can also leverage the Smart Catchment System powered by industry 4.0 technology to narrate plots of dam events via GIS layered maps. GIS maps, combined with powerful remote sensing, play an important role in flood management in the catchment area by allowing simulation models and maps to forecast catchment events and map water evacuation agendas to ensure fewer risks (Ali 2021). Such damage assessment maps will allow the user to evade dam hazards in a timely, cost-efficient and riskless manner. In addition, event-driven simulation platforms also help dam operators create riskmitigated maintenance plans for unique future dam events. Simulation technology, as discussed before, can allow meaningful digital twin models, GIS maps and smart dashboards to compare and contrast the disparity between observed parameters (water seepage levels, water evaporation, temperature, water pressure etc.) and simulated parameters to gain meaningful insights into dam operations. Smart Catchment Areas should utilise the power of simulation of the Smart Catchment Solution into smart visualisation tools to assess a simulated time series to predict when and where the catchment area is susceptible to damage in future. This technology can help in building Dam Dynamics Models to provide perspectives based on the diverse interests of the catchment specialists, thus optimising the negotiation and large-scale management of water reservoirs on a greater scale (Fig. 4). Di Baldassarre et al. describe how urban change surrounding the catchment areas that are served by the water dam also plays an important role in catchment asset management. Dam operators should correlate the supply-demand cycle and understand the effect of the reservoir relying on the dependence on the water infrastructure’s health. Albeit the catchment area is capable of releasing an abundance of water to satisfy the rising water demand due to urban expansion, it also means the sustenance and reliance of the dam infrastructure are plummeting. Therefore, smart technologies can help water dam operators to understand urban change by creating high-hazard structures that are altered

126

G. Sabapathipillai et al.

Fig. 4. Virtual dam side view via Tigernix Smart Catchment Solution

based on urban changes. This will update the information centres with flood hazards, inundation ranges, environmental impacts, and other occurrences to regulate the hazard potential of the dam during drastic urban development measures (International Water Power and Dam Construction 2022). Such models are built by integrating GIS, Satellite and IoT technologies to study contemporaneous changes that can affect the good health and sustenance of the dam infrastructure and the catchment area at large (Fig. 5). 3.2 Ensure Water Quality and Quantity With regard to water quality management, industry 4.0 technologies are also advanced in detecting chemical imbalances in water by using historical parametric insights to train predictive models to react to significant changes in water quality. Industry 4.0 technologies allow pre-indicator models that visualise water contaminants and public health risks by building water quality assurance models via cyber-physical systems. Smart sensors can detect human-made water abuse like leaking septic tanks, urban pollution, infections, toxic metals and chemicals etc., to support water quality management by providing insights to reduce dumping, pollution, hazardous material and chemical discharges that spoils water and increases the complexity and budgets of untreated water management efforts. Beckett, Et al. Explain in their conference paper that water quality parameters must be collected at a higher frequency and analysed in real-time to scaffold the impurity traceability efforts of water operators that manage direct contact process water. They introduce four parallelly-implemented technologies that optimise water quality assurance: Cyber-Physical Systems, Data Analytics, Data Integrity, and Work 4.0. These

Smart Water Dam Transformation in Industry 4.0

127

Fig. 5. Tigernix Smart Catchment Solution Framework for water systems

technologies will grant technology assurance and water process assurance and optimise people’s engagement towards continuous improvement of water quality. Industry 4.0 technologies greatly impact advancing approaches to curing water from natural water-polluting hazards. One of the main natural threats to dammed water is the growth of HAB (Harmful Algal Bloom). Sediments and nutrients- nutrients and sediments can load via soil erosion, sewage, fertilisers, manures etc.- that flow to dammed water bodies encourages the growth of cyanobacteria, also known as blue-green algae. Water that HAB harms can be treated by reducing nutrient concentrations near water storage, water treatment efforts, biomanipulation and artificial stratification. However, if HAB growth can be detected before the occurrence, catchment operators can have the upper hand in preventing HAB growth and risk rooted in HAB-prone water zones. AI-driven predictive technologies can isolate HAB-prone areas on GIS Maps to ensure the safety of the water users and the aquatic wildlife. GIS technology visualises the Normalised Difference Chlorophyll Index (NDCI) developed by (Mishra and Mishra 2012) and incorporates all algae data (particularly from satellites like Sentinel-2) into a single, centralised location. They used the red edge band at 708 nm, a red band at 665 nm from the MERIS (Medium Resolution Imaging Spectrometer) red edge band centred at 708 nm plus a red band centred at 665 nm (nm)- these bands are similar to Sentinel-2, which has 39 nm compared to 20nm for MERIS; however, the red edge is 20 nm for both. This information can enlighten users with useful images on GIS maps to detect real-time HAB growth in catchment areas. On either hand, the data collected by these sources can be used to mature AI and ML models to predict the nature, pace and location of HAB growth by studying the parametric coordinates that are agreeable for HAB inhabitation.

128

G. Sabapathipillai et al.

It will automatically learn the favourable conditions for the domination of HABs like the concentration of the sunlight, temperature, level of nutrients, irradiance and more and predict HAB trends for faster water treatment efforts. The water quality and quantity of catchment areas can also be impacted by other natural causes other than algal growth, like drought, floods, landslides, bushfires, seasonal water demands, and extreme weather conditions. Incorporating industry 4.0 technology will help dam operators understand how to seal water from being polluted after various disasters. AI and other digital technologies can measure pore pressure, water level, filling, and drawdown conditions of dam events during natural disasters to prevent water quality and quantity level risks. Firstly, AI technology can use simulative models to predict dam disasters by studying probable dam events and identifying risk factors affecting dam integrity and dammed water. AI models can be trained to correlate the circulation patterns and moisture levels to narrate stories about the rainfall events that can affect the catchment area and interrelate them to the dam height, capacity, runoff levels, elevation and more to show the criticality of the dam water levels in a certain time. This way, during a flood, bushfire, landslide, or any other natural hazard, the dam operators can run simulations to prioritise dam maintenance to mitigate the risks of a disaster phenomenally. This way, dam managers can harness the insights of prioritising the sustainability of the dam by optimising critical infrastructure management to lower serious water levels and ensure the quality of water as a resource and level of water services during surging water demands despite uncontrollable natural occurrences that affect the catchment area.

4 Conservation of Natural Resources in Catchment Area Effective catchment management should ensure that all procedures consider the catchment’s terrestrial and aquatic biodiversity and the implications of any present and future actions. Catchment management may ensure the conservation and sustainable use of biodiversity in conjunction with other goals if approached in an integrated manner. For instance, if habitat needs are considered when choosing the places and species to be planted, tree planting for groundwater or riparian (streamside) management can help conserve biodiversity. The main benefit of integrated catchment management is that it encourages the sustainable and balanced management of natural resources. It acknowledges the relationship between environmental impacts and land and water use, the fact that decisions made in one area upstream will have a cumulative effect on other sites downstream, and the necessity of a comprehensive approach to the planning and coordination of land and water management. It is only possible to effectively address many of Australia’s long-term environmental degradation issues, such as dryland salinity, with an integrated strategy made possible through catchment management actions. Thus, catchment infrastructure managers must also take responsibility for conserving the dam area’s natural resources. Therefore, modern catchment managers blend industry 4.0 technologies to optimise infrastructure and protect the environment, sustain agricultural resources and manage natural resources by practising ecologically sustainable catchment development agendas. Therefore, integrated catchment management systems are optimised to comply with ESG compliances to restrict issues that can threaten natural resources

Smart Water Dam Transformation in Industry 4.0

129

in the catchment area. A Smart Catchment System can be built to ensure its reactivity to the environment by enabling a fact-based conscience to the use and impact of natural resources with less waste and byproducts that can harm the biodiversity around the catchment area. The amalgamation of smart sensors, AI-driven data analytics, cloud computing, location detection technology, IoT, satellite-driven EO data analytics, simulation models and more can ensure exemplary management structures to control and manage dam infrastructure, water quality and water quantity. Such technologies are proven to enhance natural environments to ensure the safety and perseverance of natural habitats and diverse ecosystems at large.

5 Conclusion The paper discusses how two of the most impactful challenges faced by water service sectors, generally, are facilitated by industry 4.0 technology application using a Smart Catchment solution, namely ensuring the health and sustenance of the water dam infrastructure and reassuring the quality and quantity of dammed water. It describes a series of technological advancements under these topics and explains how industry 4.0 technology can facilitate natural resource preservation in the catchment area.

References Ali, T.: How GIS mapping can help in flood management (2021 va). https://doi.org/10.13140/RG. 2.2.29515.92968 Beckett, R., Chapman, R., Berendsen, G., Dalrymple, J., Quispe-Chávez, N.: ‘Quality 4.0’ and Water Management Practices (2019) De Simone, S.V.: Data Management System for Dam Monitoring of Hydropower Projects. In: Bung, D., Tullis, B. (eds.) 7th IAHR International Symposium on Hydraulic Structures, Aachen, Germany, May 15–18 (2018). https://doi.org/10.15142/T3M634(978-0-692-13277-7) Di Baldassarre, G., et al.: Water shortages worsened by reservoir effects (2018). https://linkin ghub.elsevier.com/retrieve/pii/S0016328705000753 International Water Power and Dam Construction, 2022, August 15. Managing dam hazard in changing urban spaces - International Water Power (2022). https://www.waterpowermagazine. com/features/featuremanaging-dam-hazard-in-changing-urban-spaces-9928111/ Mishra, S., Mishra, D.R.: Normalized difference chlorophyll index: a novel model for remote estimation of chlorophyll-a concentration in turbid productive waters. Remote Sens. Environ. 117, 394–406 (2012). https://doi.org/10.1016/j.rse.2011.10.016 Sreekar, S., Bobba, P., Jain, P.: Water level monitoring and management of dams using IoT, pp. 1–5 (2018). https://doi.org/10.1109/IoT-SIU.2018.8519843

Can Industry 4.0 Keep Its Promises? A Literature-Based Comparison of Expectations and Experience Lasse Metso1(B) and Nils E. Thenent2 1 LUT University, Lappeenranta, Finland

[email protected]

2 Lufthansa Technik, Hamburg, Germany

[email protected]

Abstract. The purpose of this article is to show the main topics of Industry 4.0 and how expectations relate to experiences outlined in academic articles between 2012 and 2020. A quantitative keyword analysis is accompanied by a qualitative review of the top 10 keywords as well as expected benefits and experiences. Based on the top 10 keywords that accompany “Industry 4.0” a conceptual model is presented to show how these keywords relate to each other. Findings show that expected benefits of Industry 4.0 are efficiency gains, quicker ways to market, flexibility and significant cost savings in production processes. In contrast, the implementation in particular in small and medium enterprises is hampered by lacking expertise of new technologies and the required invest. While technology is available, companies lack strategy for its implementation. Companies that have successfully implemented Industry 4.0 benefit from efficiency, flexibility, quality and deliverability gains. It is also found that a focus on technology leaves aside other aspects such as implications on organizational culture and working conditions. This research is limited to journal and conference publications listed in the Scopus database. The use of specific search words and combinations of their synonyms and year further limits potential references. As such, some of the most cited articles about Industry 4.0 might be excluded. This article contributes to the discussion on Industry 4.0 through a condensed overview of the most prominent topics and by showing what promises of Industry 4.0 have materialized or not. As such, the value of this work is an orientation towards realistic expectations of Industry 4.0 in research and practice. Keywords: Industry 4.0 · IoT · Smart manufacturing · Experiences · Readiness

1 Introduction As an overarching theme Industry 4.0 integrates multiple concepts in the evolution of industrial production. Through the wide-reaching consequences of technological advancements nothing less than a fourth industrial revolution is anticipated. Industry 4.0 as a term originates in Germany in 2011 as a part of an economic policy outlook in © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 130–141, 2023. https://doi.org/10.1007/978-3-031-25448-2_13

Can Industry 4.0 Keep Its Promises?

131

high-tech strategies. Elements of Industry 4.0 include machine-based communication in highly automated manufacturing systems, commonly termed as the Internet of Things (IoT) and the integration of virtual and physical processes, so called Cyber Physical Systems (CPS) [1]. This work illustrates the proliferation of the term “Industry 4.0” in the Scopus database and shows its constituting key concepts followed by a comparison of expectations and experiences with Industry 4.0. This work illustrates the proliferation of the term “Industry 4.0” in the Scopus database and shows its constituting key concepts followed by a comparison of expectations and experiences with Industry 4.0. Figure 1 highlights the exponential increase of publications with the keyword “Industry 4.0” in the Scopus database from 2012 to 2019. Notably, 2020 shows a stagnation in the number of publications. Whether this reflects a trend or is related to the point in time this analysis was concluded in 2021 remains to be seen.

Fig. 1. The number of publications with the keyword “Industry 4.0” in the Scopus database from 2012 to 2020.

2 Methods This research reviews journal articles and conference contributions listed in the Scopus database that were published between 2012 and 2020. This is the first search string that was used KEY (“Industry 4.0”) AND (LIMIT-TO (PUBYEAR, 2020)). The resulting total number of publications with the keyword “Industry 4.0” for each year is presented in Fig. 1. The Scopus database shows 162 most used keywords and those were first listed and then analyzed. More detailed analysis leads to definitions and prospective benefits of Industry 4.0 outlined in articles from 2012 to 2014. These are compared to definitions and benefits as well as challenges and realized benefits as described in case studies in articles published in 2020. The search string was the same as in the previous search KEY (“Industry 4.0”) AND (LIMIT-TO(PUBYEAR, 2012)). For the deeper study of the current situation

132

L. Metso and N. E. Thenent

with Industry 4.0 articles from 2019 and 2020 were used. The used search string was (KEY (“Industry 4.0”) AND TITLE-ABS-KEY (readiness)) AND (LIMIT-TO (PUBYEAR, 2020) OR LIMIT-TO (PUBYEAR,2019)). The The result of the readiness search was 176 documents. Scanning the title and abstract and accessibility the number of relevant papers was limited to 73 in which 53 papers were included this study. Finally, 19 + 34 = 53 relevant papers were included to the study, of which 31 are referred to in this article. It was noticed that the Scopus keyword search gave more results than keywords listed in some papers. The reason is that indexed databases such as Scopus use both author-assigned and “Indexed keywords” in “Engineering controlled terms”. From the data extraction described above the resulting keywords were cleansed, e.g. “CPS” was renamed with “Cyber Physical System”. Some publications feature both expressions as keywords. So, this approach leads to a double count of “Cyber Physical Systems” and the numbers given in the following reflect the frequency of occurrence of a keyword and not (necessarily) the number of publications. An effect of this can be seen in Table 1 were the keyword “Cyber Physical Systems” has a relative occurrence of 110% with respect to “Industry 4.0”. The approach to derive the top 10 keywords is based on the relative frequency of each keyword. For each year the number of occurrences of each keyword was divided by the number of publications with the keyword “Industry 4.0” (for the latter see Fig. 1). Then, for the years 2012 to 2020 the average relative frequency of each keyword is calculated. In the following the top 10 keywords are discussed. The absolute occurrences of these top 10 keywords are presented in Fig. 2.

3 What Topics Are Discussed in Industry 4.0 This chapter presents an overview of the top 10 keywords, followed by their relative occurrence compared to “Industry 4.0” and a short description for each keyword. Figure 2 shows the total number of occurrences of the top 10 keywords from 2012 to 2020. As expected from the overall number of publications with the keyword “Industry 4.0”, as shown in Fig. 2, the is a general upwards trend across the curves of all top 10 keywords. One noteworthy observation is that up to the year 2017 the keyword “Cyber Physical Systems” (CPS) occurred most often. In 2012 CPS and “Industrial Revolution” are the only keywords out of the top 10. From 2017 the “Internet of Things” is featured most frequently. In the following years CPS decline and in 2020 falls below “Industrial Revolution”. For a better impression of the quantitative evolution of each top 10 keyword relative to the total number of Industry 4.0 publications (that is publications with the keyword “Industry 4.0”) the respective data is shown in Fig. 3. The logarithmic y-scale is % of “Industry 4.0” publications. The leftmost values show the average relative frequency and thus reflect the ranking of each keyword. The average of occurrence of CPS is 48%

Can Industry 4.0 Keep Its Promises?

133

and the occurrences of “Internet of Things” (IoT) and “Industrial Revolution” are 22% each. The average of “Embedded system” is 17% with descending tendency. This is even more pronounced for CPS that drops from 100% in the early years to less than 7% in 2020. In comparison, “Industrial Revolution” and IoT are relatively stable with the latter leading the ranking clearly in 2020. Considering the significant increase in Industry 4.0 publications and the increasing thematic breadth that comes along with it, it is not surprising that the relative frequency of most keywords decreases over the years. Nevertheless, it is remarkable that in 2019 and 2020 still more than 25% of all Industry 4.0 publications feature the keyword IoT.

Fig. 2. Count of occurrences of the top ten keywords in publications between 2012 and 2020 (note the logarithmic y-scale).

In the following subsections the top 10 keywords are briefly outlined based on the reviewed literature. At the end of this chapter a conceptual model of Industry 4.0 comprising these keywords and their relationships is presented. 3.1 Industrial Revolution Consistent with its designation, Industry 4.0 is identified as a significant industrial transformation comparable to the introduction of steam power, electricity and computerization [1]. While associated to technological advancements such as Cyber Physical Systems and the Internet of Things, industrial revolutions have wide ranging effects on social aspects such as work conditions and economic wealth [2]. 3.2 Manufacture “Manufacture” is an example of database assigned keywords that is not found among the author assigned keywords in the published manuscripts. Therefore, it is interpreted here as a domain specifying term. As such Manufacture represents the production of physical goods encompassing topics such as supply chain [3], factory planning [4], maintenance [5], additive manufacturing [6] and distributed manufacturing [7].

134

L. Metso and N. E. Thenent

Fig. 3. The relative occurrence of the top 10 keyword in publications with the keyword “Industry 4.0” (note the logarithmic y-scale).

3.3 Smart Manufacturing Smart Manufacturing is frequently used as a synonym for Industry 4.0 [5, 8, 15]. It stands for the most significant application of digitalization in the manufacturing industry [17]. Smart Manufacturing, is also described as a result or goal of Industry 4.0 implementation [11, 12]. It is associated to products that provide information about their production parameters, lifecycle and maintenance [13]. 3.4 Automation Automation is identified as an essential element of Industry 4.0 and one of the fields that itself evolves with technological advancements as for example in automated warehouses [20]. In production systems increased flexibility and adaptability are required for automation in an Industry 4.0 context [14], for example through an evolution from hierarchical automation structures to distributed services with Cyber Physical Systems [15]. 3.5 Smart Factory The “Smart Factory” is identified as a central element of Industry 4.0 and an enabler of value chain integration across companies [15]. Smart Factories are essentially the application of technologies such as CPS, the Internet of Things and Augmented Reality to integrate production and business processes in a way that supports people and machines in executing their tasks and provide management with meaningful insights [8, 11, 16, 25].

Can Industry 4.0 Keep Its Promises?

135

3.6 Cyber Physical Systems Cyber Physical Systems are interconnected embedded systems. They are considered as trigger for Industry 4.0 [17]. As an embedded system with the ability to communicate to other devices, CPS consists of a control unit, a microcontroller and communication interface. A microcontroller controls actuators and sensors measuring process data [18]. CPS are also entitled as intelligent machines with data storing and sharing ability [19]. Further characteristics of CPS include the merging of physical and virtual realities. CPS are commonly referred to together with the IoT and embedded systems [20, 21]. 3.7 Internet of Things The Internet of Things (IoT) is commonly understood as a network of interconnected devices that comprise sensors, actuators and the network infrastructure, as well as tools for data collection and analysis [22]. Among Embedded Systems, the IoT is seen as one of the enablers for CPS [21]. Furthermore, the IoT offers the foundation for automation, decentralized decision making through machine to machine interaction, remote controlling and diagnostics [23]. 3.8 Embedded Systems The combination of physical processes and computing is known as Embedded Systems in engineering. Embedded systems feature are able to self-organize locally and to control physical processes. In CPS Embedded Systems are connected into networks. [17]. Examples of embedded systems include smartphones, cars and household appliances [18]. Some authors identify Industry 4.0 with the evolution from Embedded Systems to CPS [20]. 3.9 Big Data The availability of real-time data from CPS through the internet of things linked with other sources from inside and outside the company, such as customer management or suppliers is referred to as Big Data in Industry 4.0 [5, 24]. Its analysis is seen as an enabler for algorithm-based decision-making and automated diagnosis [9, 18]. Big Data and production data as on particular example can be characterized by volume, variety and velocity [25]. In other words, Big Data is a quickly changing, large volume of structured, semi-structured and unstructured data that poses significant challenges for analysis [9, 24]. 3.10 Decision Making Decision Making is often associated to Big Data and its analysis [38]. It is also highlighted that decentralized decision making in automation is enabled by machine to machine interaction [23]. In the 86 publications reviewed explicit definitions of decision making in the context of Industry 4.0 were not found.

136

L. Metso and N. E. Thenent

3.11 Relationships Between the Top 10 Keywords Based on the compact descriptions above, Fig. 4 shows an interpretation of Industry 4.0 as relationships between the top 10 keywords. From top to bottom the domain of Industry 4.0 is marked around manufacturing. This is followed by the executing elements, namely the Smart Factory, automation and decision making that are expected to realize the goals associated to Industry 4.0. The execution cluster itself builds on such enabling technologies as CPS and the IoT.

Fig. 4. Industry 4.0 as an interpretation of the relationships between the top 10 keywords.

4 What is Expected of Industry 4.0 For the German industry improvements and important enhancements in productivity (manufacturing sectors by e90 billion to e150 billion), revenue growth (of about e30 billion a year), employment (6 percent increase during the next ten years) and investments (about e250 billion during the next ten years) are expected [1]. 4.1 Domain Cluster Within the Industry 4.0 by the German Federal Ministry of Education and Research among others, the following research and development areas were identified: Standardization, a broadband infrastructure for industry, work organization and design, regulatory framework, and resource efficiency [26]. In Smart manufacturing digital technologies are applied to improve supply-chain efficiency and profitability [11]. Manufacturing

Can Industry 4.0 Keep Its Promises?

137

Execution Systems (MES) producers should improve their systems to support the Digital Factory and the implementation of automation. This enables plug-in implementation of new sensors and devices in manufacturing lines. In 2014 fully automated operation was considered to be unrealistic in the near future. However, decision-making support, knowledge reusing, advanced planning and scheduling and machine simulations in cloud application were realistic. Virtual machine tools were used to verify and improve production efficiency [27]. 4.2 Execution Cluster The IoT enables new applications both in the industrial and other sectors like building automation. Internet connection and IoT changes the monitoring and control because standardization enables the use of different kind of field devices and sensors in the same network. Application for advanced data analysis and diagnoses can be added. Adding new field devices is easy because the network configuration is automated [20]. The target of Smart factories is to increase efficiency and productivity as well as to full-fill customers’ demands. Logistic processes connect a single factory into flexible supply chain network which is optimized through simulations [42]. The intelligence of a smart factory consists of big data analysis, simulations and algorithms in decision-making enabling corrective actions and prescriptive maintenance [5]. 4.3 Technology Cluster CPS improve collaboration enabling information sharing and sense-making. Information sharing can be easily organized while electronics and sensors are cheaper and can connect with technology of IoT. Simulation is an example of sense-making, because alternative scenarios can be assessed [21]. Markets and customers demand flexibility and CPS can be seen a solution to increase adaptability [14]. Cyber Physical Assembly Systems can support the management of complex assembly processes and high-level customization [17]. Through the IoT industrial applications are connected to sensors and other devices, enabling scalable systems [28]. Big data analysis and real-time decision-making will impact efficiency [13].

5 What is Experienced in Industry 4.0 The implementation of Industry 4.0 affects different areas such as employee skills, organizational culture, strategy, or organizational processes. Employees’ routine activities decrease because those are replaced by machines. Employees will develop digital processes and data analytic assignments. New technologies need companies to change strategies and processes and their overall organizational culture has to support the transformation into the new kind of business. Companies that implemented Industry 4.0 are in better competitive positions in the market [24]. Brozzi et al. [29] noticed that lager companies were better prepared to support the transition toward the Industry 4.0 than SMEs. In the evaluation of Industry 4.0 maturities

138

L. Metso and N. E. Thenent

in companies several aspects have to be taken into consideration, e.g., strategy, technological requirements, awareness and competences. Companies which are on high level in digitalization have also managed to implement Industry 4.0. SMEs do not have high turnover to invest into technology used in Industry 4.0 while large companies have the potential to invest in Industry 4.0 [30]. Industry 4.0 implementation in the production of automotive and aviation are on relatively high levels while other industrial sectors are in different situations, e.g. in the ceramics industry because of a lack of automation, monitoring systems and IT systems [30]. In many companies the focus of implementing Industry 4.0 is on technology. Companies are not scrutinizing what they want to achieve and how they can benefit from Industry 4.0. Both small and larger companies can have barriers to integrate new technology and create strategy supporting those goals [16]. However, SMEs have more challenges toward Industry 4.0. They have no clear vision to implement Industry 4.0, their staff have limited knowledge and skills about digital transformation, and SMEs lack decision-making to support high-level technology implementation [13, 29]. Several studies identified that companies frequently do not have a strategy for the implementation of Industry 4.0 [31] and on the operational level challenges exist too [10]. Data intensive industries are facing challenges in big data management such as data mining, classification and storage [12]. In some industry sectors essentially outdated technology is used, for example in ceramics and other refractory industry. This is seen as a barrier to the implementation of Industry 4.0 technology [30].

6 Summary and Conclusion The umbrella Industry 4.0 promises nothing less than a fourth industrial revolution. In the first chapters we have shown that generally the number of Industry 4.0 related publications has significantly increased from 2012 to 2020. Among the top 10 keywords of these publications in the Scopus database are technological subjects such as CPS and IoT. Other top 10 keywords include Smart Manufacturing, commonly understood as a synonym for Industry 4.0, and the Smart Factory that represents the implementation of Industry 4.0 principles in a production environment. All top 10 keywords are incorporated within a conceptual model to show how they relate to each other. The keywords are grouped in three clusters, namely Domain, Execution and Technology to show that the discussion on Industry 4.0 has a technological focus but with a wider scope towards strategic implications. On that level, the expectations in Industry 4.0 include increased productivity, reduced costs, quicker time to market and more flexibility. Accompanied by a lack of strategy, the implementation of Industry 4.0 is often technology focused. While both small and larger companies experience barriers to the integration of new technology, smaller companies struggle more with the required investment as well as the required skills. Other challenges concern the management of big data including data mining, classification and storage. In Table 1 expectations and experiences for the three clusters are highlighted. In conclusion it can be said that the implementation of Industry 4.0 shows that the conceptual benefits are not easily turned into practice. What makes the implementation of

Can Industry 4.0 Keep Its Promises?

139

Industry 4.0 concepts challenging is that its consequences affect multiple areas such as for example employee skills, organizational culture and strategy. These are, in contrast to technological aspects not well reflected among the top keywords in publications as well as in the implementation of Industry 4.0. Table 1. Cluster-based comparison of Industry 4.0 experiences and expectations. Cluster

Expectations

Experiences

Domain Cluster

Industry 4.0 offer a platform to implement new technology to support digitalization on manufacturing and those are linked to companies’ strategy

No strategy to implement Industry 4.0, benefits of Industry 4.0 are limited to costs and speed while flexibility, quality, deliverability and sustainability are ignored

Execution Cluster

Smooth supply chain network, decisions are based on monitored data

Collaboration problems data analytic and decision-support systems are not implemented

Technology Cluster CPS improve collaboration, information sharing, Big data and analytics

Integration is not working, lack of skilled staff

The research in this article is limited to journal and conference publications listed in the Scopus database. Therefore, only a fraction of Industry 4.0 related publications are reviewed. The use of specific search words and combinations of their synonyms and year further limits potential references. As such, some of the most cited articles about Industry 4.0 might be excluded. Hence, there is room for future work that takes into account more databases and includes a wider analysis of the themes and their relationships. In particular more focus on non-technological aspects, offers an opportunity improve the conceptual model of Industry 4.0.

References 1. Piccarozzi, M., Aquilani, B., Gatti, C.: Industry 4.0 in management studies: a systematic literature review. Sustainability 10, 3821 (2018) 2. Dombrowski, U., Wagner, T.: Mental strain as field of action in the 4th industrial revolution. Procedia CIRP 17, 100 (2014) 3. Borregan-Alvarado, J., Alvarez-Meaza, I., Cilleruelo-Carrasco, E., Garechana-Anacabe, G.: A Bibliometric analysis in industry 4.0 and advanced manufacturing: what about the sustainable supply chain? Sustainability 12, 7840 (2020) 4. Constantinescu, C.L., Francalanza, E., Matarazzo, D., Balkan, O.: Information support and interactive planning in the digital factory: approach and industry-driven. Evaluation 25, 269 (2014) 5. Silvestri, L., Forcina, A., Introna, V., Santolamazza, A., et al.: Maintenance transformation through Industry 4.0 technologies: a systematic literature review. Comput. Ind. 123, 103335 (2020)

140

L. Metso and N. E. Thenent

6. Hernandez Korner, M.E., Lambán, M.P., Albajez, J.A., Santolaria, J., et al.: Systematic literature review: integration of additive manufacturing and industry 4.0. Metals 10, 1061 (2020) 7. Schuh, G., Potente, T., Varandani, R., Schmitz, T.: Global footprint design based on genetic algorithms: an “industry 4.0” perspective. CIRP Ann. 63, 433 (2014) 8. Pirola, F., Cimini, C., Pinto, R.: Digital readiness assessment of Italian SMEs: a case-study research. J. Manuf. Technol. Manage. 31, 1045 (2020) 9. Pacchini, A.P.T., Lucato, W.C., Facchini, F., Mummolo, G.: The degree of readiness for the implementation of Industry 4.0. Comput. Ind. 113, 103125 (2019) 10. Schumacher, A., Nemeth, T., Sihn, W.: Roadmapping towards industrial digitalization based on an Industry 4.0 maturity model for manufacturing enterprises. Procedia CIRP 79, 409 (2019) 11. Narula, S., Prakash, S., Dwivedy, M., Talwar, V., et al.: Industry 4.0 adoption key factors: an empirical study on manufacturing industry. JAMR 17, 697 (2020) 12. Lu, Y.: Industry 4.0: a survey on technologies, applications and open research issues. J. Ind. Inf. Integr. 6, 1 (2017) 13. Chonsawat, N., Sopadang, A.: Defining SMEs’ 4.0 readiness indicators. Appl. Sci. 10, 8998 (2020) 14. Vogel-Heuser, B., Diedrich, C., Pantforder, D., Gohner, P.: Coupling heterogeneous production systems by a multi-agent based cyber-physical production system. In: Proceedings of the 12th IEEE International Conference on Industrial Informatics (INDIN), p. 713. IEEE (2014) 15. Forstner, L., Dümmler, M.: Integrierte Wertschöpfungsnetzwerke – Chancen und Potenziale durch Industrie 4.0. Elektrotech. Inftech. 131, 199 (2014) 16. Machado, C.G., Winroth, M., Carlsson, D., Almström, P., et al.: Industry 4.0 readiness in manufacturing companies: challenges and enablers towards increased digitalization. Procedia CIRP 81, 1113 (2019) 17. Dombrowski, U., Wagner, T., Riechel, C.: Concpt for a cyber physical assembly system. In: Proceedings of the IEEE International Symposium on Assembly and Manufacturing (ISAM), p. 293. IEEE (2013) 18. Jazdi, N.: Cyber physical systems in the context of Industry 4.0. In: Proceedings of the IEEE International Conference on Automation, Quality and Testing, Robotics, p. 1. IEEE (2014) 19. Makarov, O., Langmann, R., Frank, B.: Signal time deterministic for process control applications from the cloud. In: Proceedings of the 11th International Conference on Remote Engineering and Virtual Instrumentation (REV), p. 440. IEEE (2014) 20. Ungurean, I., Gaitan, N.-C., Gaitan, V.G.: An IoT architecture for things from industrial environment. In: Proceedings of the 10th International Conference on Communications (COMM), p. 1. IEEE (2014) 21. Schuh, G., Potente, T., Varandani, R., Hausberg, C., et al.: Collaboration moves productivity to the next level. Procedia CIRP 17, 3 (2014) 22. Wanasinghe, T.R., Gosine, R.G., James, L.A., Mann, G.K.I., et al.: The internet of things in the oil and gas industry: a systematic review. IEEE Internet Things J. 7, 8654 (2020) 23. Sari, T., Gules, H.K., Yigitol, B.: Awareness and readiness of Industry 4.0: the case of Turkish manufacturing industry. Adv. Prod. Eng. Manag. 15, 57 (2020) 24. Kohnová, L., Papula, J., Salajová, N.: Internal factors supporting business and technological transformation in the context of Industry 4.0. Bus. Theory Pract. 20, 137 (2019) 25. Stocker, Alexander, Brandl, Peter, Michalczuk, Rafael, Rosenberger, Manfred: Menschzentrierte IKT-Lösungen in einer Smart Factory. e & i Elektrotechnik und Informationstechnik 131(7), 207–211 (2014). https://doi.org/10.1007/s00502-014-0215-z 26. Jaschke, S.: Mobile learning applications for technical vocational and engineering education: the use of competence snippets in laboratory courses and industry 4.0. In: Proceedings of the International Conference on Interactive Collaborative Learning (ICL), p. 605. IEEE (2014)

Can Industry 4.0 Keep Its Promises?

141

27. Rehage, G., Bauer, F., Gausemeier, J., Jurke, B., Pruschek, P.: Intelligent manufacturing operations planning, scheduling and dispatching on the basis of virtual machine tools. In: Kovács, G.L., Kochan, D. (eds.) NEW PROLAMAT 2013. IAICT, vol. 411, pp. 391–400. Springer, Heidelberg (2013). https://doi.org/10.1007/978-3-642-41329-2_38 28. Imtiaz, J., Jasperneite, J.: Scalability of OPC-UA down to the chip level enables “Internet of Things”. In: Proceedings of the 11th IEEE International Conference on Industrial Informatics (INDIN), p. 500. IEEE (2013) 29. Brozzi, R., Riedl, M., Matta, D.: key readiness indicators to assess the digital level of manufacturing SMEs. Procedia CIRP 96, 201 (2021) 30. Kellner, T., Necas, M., Kanak, M., Kyncl, M., et al.: Assessment of readiness for industry 4.0 implementation in ceramic industry. Manuf. Technol. 20, 763 (2020) 31. Maisiri, W., van Dyk, L.: Industry 4.0 readiness assessment for South African indistries. South Afr. J. Ind. Eng. 30, 134 (2019)

Monitoring, Diagnostics and Prognostics for Smart Maintenance

Dynamic Maintenance Management Approach Based on Real Time Monitoring and Artificial Intelligence Using Digital Twins José Antonio Marcos-Alberca Carriazo1(B) and Adolfo Crespo Márquez2(B) 1 TALGO, Smart Maintenance Systems, Madrid Majadahonda, Spain

[email protected]

2 Seville University, Seville, Spain

[email protected]

Abstract. Improving maintenance and avoiding production downtime with predictive services has become a compulsory requirement for operators willing to optimise reliability while reduce their operational costs. To achieve these challenges, this paper explains how Talgo, with more than 80 years of experience in rolling, has developed a Dynamic Maintenance Management approach using digital platforms for real-time monitoring and predictive maintenance based on artificial intelligence and digital twins. The paper also explains the new digital platform for automatic train inspection based on machine vision. These platforms provide smart data supporting operators and maintainers activities with valuable information in real time. Because of this, the main operation KPIs such as: passenger experience, energy efficiency, safety, availability, reliability, and maintainability have been importantly improved. Higher trains availability due the savings in workshop inspections time and higher capacity in maintenance sites can be achieved easing rolling stock O&M.

1 Introduction Throughout its 80 years of history, Talgo has always been at the forefront of technology, incorporating new advances to improve the reliability and maintainability of its trains compared to the standard maintenance policies applied to conventional trains. Among these new technologies for optimising maintenance costs and improving train safety and reliability are the application of different processes and modern maintenance engineering methodologies, as well as the use of information technologies for remote maintenance, sophisticated predictive maintenance equipment and auxiliary machinery to improve the maintainability of rolling stock. All these technologies have been boosted by the advent of the Internet of things, and Industry 4.0, which has made cutting-edge technology much more affordable and achievable on an industrial level.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 145–153, 2023. https://doi.org/10.1007/978-3-031-25448-2_14

146

J. A. M.-A. Carriazo and A. Crespo Márquez

2 Smart Maintenance Based on RTM, RCM, CBM Nowadays, the maintenance of high-speed trains requires a high level of technification in maintenance processes and maintenance engineering techniques, to reach the required safety, quality and reliability standards (like the ones presented in Fig. 1), making these operations possible.

Fig. 1. Main challenges maintenance operations in high speed trains

The Smart maintenance approach that Talgo is developing is fundamentally based on the digitalisation of its maintenance processes and on real-time decision-making based on real-time data and information. This data and information are obtained by monitoring the condition of existing equipment on the train and modelling the health of the equipment itself. To perform these tasks in an orderly manner, the joint application of RCM (Reliability Cantered Maintenance) and CBM (Condition Based Maintenance) techniques is mandatory. In this way, the necessary resources are applied to avoid the appearance of the most critical failure modes in the most important train components for the business, in each specific operational context. Also, the train must be equipped, from its conceptual design, with an important number of sensors, capable of providing health indicators for possible critical systems. Currently, each train installs more than 10,000 sensors, which can transmit real-time information, more than 5 Gb per train per day, through the on-board communication system. This information is sent through a gateway system to a smart digital platform which monitors the information and analyse in real time this data using artificial intelligent, reporting the health status of the systems and their RUL (Remaining useful life). By analysing the evolution of RULs, it is possible to determine the schedule of systems repairs or replacement in the near term, preventing their failure.

3 Internet of Trains The state of development of new IoT technologies has enabled a great expansion and technification of industrial processes, resulting in a cutting-edge and affordable technology to be used. Talgo is committed to Industry 4.0, working together on new developments

Dynamic Maintenance Management Approach

147

based on the Google Cloud platform, so that all this technology can be used for the development of the Talgo IoT platform. So, the trend is that more and more systems and equipment are interconnected through electronic equipment receiving signals and data from the thousands of sensors that make up a train. Events, environmental variables, and train alarms are sent in streaming, in real time, and in a secure and encrypted way, so that once they are uploaded to the cloud, they are processed and filtered by platform allowed data processing engines. Once the data is processed in the Cloud, the data can be consulted, analysed, and sent through the different Apps that connect to the Cloud platform. Among the most relevant applications that Talgo has developed for data analysis and processing, which serve as a basis for continuous improvement and decision-making for predictive and/or condition-based maintenance, are the following: 3.1 Real-Time Monitoring System: TSMART This telediagnosis platform allows to know, in real time, the location and status of the main variables of the equipment, so that in the event of any failure, the system sends the information via cloud to the maintainer/operator to carry out the appropriate actions. In addition, the system always indicates the exact train position, to obtain the best possible information about the state of the train at the time of a failure. To send streaming data from the train (more than 10,000 variables/s, 5GB per train per day can be sent) requires a robust and efficient ETL system for ingesting, storing, and processing the data, using Google Cloud Platform. At the end of the day, this real-time monitoring together with the existing artificial intelligence in the different modules developed on the TSMART platform, makes the online visualization of the train’s main equipment health status possible. Modules of the platforms are presented in Fig. 2.

Fig. 2. Modules within the real time monitoring system. TSMART

148

J. A. M.-A. Carriazo and A. Crespo Márquez

3.2 Machine Learning and Predictive Maintenance Current high speed trains maintenance must have diagnostic tools that can indicate the progress and evolution of certain critical equipment’s health indicators, for the train’s maintainability and reliability (see some of the screens of the TSMART for predictive maintenance system in Fig. 3).

Fig. 3. TSMART. Predictive maintenance.

Talgo is committed to this type of technology which, through correct monitoring and analysis of the data sent by the train, allows the data to be processed in such a way that the time until the failure of different equipment or systems can be determined with an adequate degree of certainty, being decisive for the realisation of an adequate condition-based predictive maintenance. An example of a use case is the detection of bearing anomalies by means of machine learning algorithms based on the continuous monitoring of bearing temperature. In this case, a neural network has been obtained which, based on the monitoring of the temperatures of the remaining bearings on the same axle, together with the outside temperature and the speed of the train, can predict the temperature of any of the bearings. This temperature (calculated) differs by a certain value from the temperature (measured), a temperature discrepancy alarm is triggered, so that the evolution of the bearing failure can be determined with sufficient time to prevent it from breaking before it occurs. The target time for this is sufficient to be able to schedule the maintenance tasks for replacement within the scheduled workshop steps, while still within the RUL “remaining useful life” (Fig. 4).

Fig. 4. The concept of Remaining Useful life. RUL

Dynamic Maintenance Management Approach

149

4 Digital Twins Applied to Smart Maintenance This case study covers a DT developed for the maintenance of axle bearings in trains. A set of models were developed to understand and mimic the bearing temperatures when the trains and bearings were running under certain conditions, in specific travels where the railway and infrastructure changed as well as ambient temperature. The idea was to use this information to guide bearings maintenance, once it could be possible to detect bearing anomalies, to classify them and to offer an estimation of the remaining useful life of the bearing. A train axle bearing temperature depends on a set of factors, the operational regime, the type and dimensions of bearings, the antifrictional and hydrodynamic properties of the lubricant, the spaces between the bearing rollers and rings, the static and dynamical loads of the bearing, the train running speed, the duration of travel without stops, the ambient air temperature, and the road curves (Lunys et al. 2015) & (Mironov 2008). (See Fig. 5).

Fig. 5. Factors (physical model inputs) conditioning a train axle bearing temperature.

In a recently published article, Crespo et al. (Crespo Márquez et al., 2020a, b ) considered these temperatures to build the required predictive analytics for each bearing temperature. This is presented in that paper as an innovative approach (See Fig. 6).

Fig. 6. Crespo et al., approach to predict axle bearing temperatures.

Results in detection, diagnosis and RUL obtained with these models are presented now:

150

J. A. M.-A. Carriazo and A. Crespo Márquez

• Detection. One of the most relevant conclusions of the referred paper is the possibility to predict abnormal bearing temperature behavior with maximum precision and specificity, when selecting the Absolute Error descriptor with a threshold of 10 ºC in the rule, and regardless the speed of the train. • Diagnosis. A sorting algorithm was selected to attempt to separate bearings with internal deterioration from those with overtemperature caused by external causes, mainly train axle guidance problems. To that end, it was necessary: 1) to generate a specific ETL process to reduce a variable stress spectrum into a simpler, equivalent set of stresses; 2) to know the final diagnosis of all the bearings observed to have suffered overtemperature cycles. It is essential to have data on whether the bearing was replaced or not, and if it was replaced, whether the analysis performed by the quality department found it with internal deterioration or not. Bearings in the train that were not replaced, but which had overtemperature cycles recorded, were obviously classified as “non-deteriorating” bearings. Basically, most of these bearings went back to normal temperature conditions when the guidance problem was solved. All these records helped to better train the classification algorithm. • Prognosis/RUL. In this paper a statistical approach is followed to estimate the RUL (of any bearing of a train), once a positive (or anomaly detected for a failure mode) appears in a train axle bearing. A positive (according to the company existing Procedure for the Design and Implementation of CBM Plans) is defined as the occurrence of an absolute error (AE) of prediction greater than 10ºC between the actual bearing temperature and that predicted by the ANN designed for detection, when the train is running at more than 90 km/h (i.e., AE ≥ 10ºC, V ≥ 90 km/h) and for more than one minute. RUL is defined as a random variable that, estimated from the appearance of the first positive, offers a good prediction of the life of the element until its replacement due to over temperature or noise. This replacement is nowadays performed after the activation of the safety alert in the train monitoring and control system (TCMS) and/or because of a certain inspection (probably during a weekly train inspection in the workshop). It should be noted that this solution does not interfere with safety control system of the train which always acts as a safeguard and independently of the systems for predictive maintenance. Company’s objective through the analysis included in this part of the paper is to foresee the recommended time of bearing replacement, after its first positive, even without prior inspection, according to statistical estimates.

5 Dynamic Maintenance Scheduling and Management The main challenge to manage a fleet using the opportunities that data offers, as mentioned before, is calling from managing the condition of a single component to managing dynamically the fleet. This is translated to the allocation of the assets in the fleet to maintenance, operation, or staying idle, guaranteeing that the demand is fulfilled every day of the planning horizon. This has to be done respecting the limitations of the company and the existing systematic activities planned so a pragmatic and applicable plan is presented, otherwise, presenting a full predictive data-based strategy not considering existing regulations and safety policies is a thick as mince and naive approach when it comes to generating a real impact nowadays.

Dynamic Maintenance Management Approach

151

Fig. 7. Dynamic maintenance management approach.

It was essential the commitment of the Smart Maintenance team from Talgo to develop together the requirements of the business alongside the development of the data availability, the existing condition based maintenance programmes (which systems are being monitored) and the development of an optimisation model that truly provides an answer to the central planner dynamically for everyday operation allocating trains to operation, maintenance or idle every day or a certain planning horizon, respecting the existing systematic stoppages opportunistically, to maximise the usage and minimise the costs (Fig. 7).

6 Automatic Vehicle Inspection. TALVI This is a new technology aimed at automatic train inspection based on artificial vision and 3D scanners in the interests of continuous improvement and technological vanguard. This equipment is capable of inspecting trains at commercial service speeds of up to 300 km/h, with a precision and speed of analysis using machine learning that makes it highly reliable, safe and highly cost-effective in its use. TALVI platform integrates different equipment to perform automatic inspections of different parts of the train. The advantages offered by TALVI equipment for general track inspection (Fig. 8): • • • • • •

Improved operational safety Increased fleet availability Increased fleet reliability Cost savings in the scheduling and spare parts Carrying out inspections under real operating conditions. Repeatability and accurate tracking in measurement.

152

J. A. M.-A. Carriazo and A. Crespo Márquez

Fig. 8. TALVI. Automatic vehicle inspection

7 Conclusions In this paper, a dynamic maintenance management approach using digital platforms for real-time monitoring and predictive maintenance is explained. Artificial intelligence and digital twins, beside artificial vision and 3D scanning technologies, are used for compelling predictive maintenance. Health of the most critical train systems can be monitored, and maintenance actions properly released. Beside these technological challenges, improving dynamic management of maintenance activities according to the real-time assessment of failure mode risk is becoming a must.

References Canedo, A.: Industrial IoT lifecycle via digital twins. In: International Conference on Hardware/Software Codesign and System Synthesis, CODES+ISSS 2016 (2016). https://doi.org/ 10.1145/2968456.2974007 Crespo, A., et al.: Criticality Analysis for improving maintenance, felling and pruning cycles in power lines. IFAC-PapersOnLine (2018). https://doi.org/10.1016/j.ifacol.2018.08.262 Crespo Márquez, A., de la Fuente Carmona, A., Marcos, J.A., Navarro, J.: Designing CBM plans, based on predictive analytics and big data tools, for train wheel bearings. Comput. Ind. 122 (2020a). https://doi.org/10.1016/j.compind.2020.103292 Crespo Márquez, A., et al.: Exploiting EAMS, GIS and dispatching systems data for criticality analysis. In: Value Based and Intelligent Asset Management (2020b). https://doi.org/10.1007/ 978-3-030-20704-5_7 Durão, L.F.C.S., Haag, S., Anderl, R., Schützer, K., Zancul, E.: Digital twin requirements in the context of Industry 4.0. In: Chiabert, P., Bouras, A., Noël, F., Ríos, J. (eds.) PLM 2018. IAICT, vol. 540, pp. 204–214. Springer, Cham (2018). https://doi.org/10.1007/978-3-030-01614-2_19 Konstantinov, S., Ahmad, M., Ananthanarayan, K., Harrison, R.: The cyber-physical e-machine manufacturing system: virtual engineering for complete lifecycle support. Procedia CIRP (2017). https://doi.org/10.1016/j.procir.2017.02.035 Lunys, O., Dailydka, S., Bureika, G.: Investigation on features and tendencies of axle-box heating. Transp. Probl. 10(1) (2015). https://doi.org/10.21307/tp-2015-011

Dynamic Maintenance Management Approach

153

Marquez, A.C., Fernandez, J.F. G., Fernández, P.M.G., Lopez, A.G.: Maintenance management through intelligent asset management platforms (IAMP). Emerging factors, key impact areas and data models. Energies 13(15) (2020). https://doi.org/10.3390/en13153762

Heat Pumps Smart Asset Management Implementation Through Virtual Sensors Pedro Barandier(B) , Alexandre Miranda, and Antonio João Marques Cardoso CISE - Electromechatronic Systems Research Centre, Calçada Fonte Do Lameiro, 6201-001 Covilha, Portugal [email protected]

Abstract. This paper addresses an implementation plan for the smart asset management of a Heating, Ventilation, and Air Conditioning (HVAC) system, based on Air-to-Water Heat Pumps. Such plan aims to ensure a low cost and efficient asset management based on the use of Virtual Sensors (VS) and Fault Impact Models. An automated Fault Diagnostics (FD) approach ensures the improvement of energy efficiency in conjunction with a reduction on faults occurrence and service costs. Therefore, the main objective of this paper is to apply the know-how already acquired about this technology and thus, develop a profitable strategy of smart asset management implementation through an on-condition based maintenance.

1 Introduction In the last years, energy efficiency and renewable energies have been one of the main goals of European Union (EU) to ensure a sustainable growth. In order to achieve such goals, measures have been taken to reduce greenhouse gases emissions and energy consumption. Since space heating, cooling and water heating represent almost 80% of the energy used in households in EU, one of the solutions is the use of Heat Pump (HP) technology [1, 2]. This technology provides an efficient and sustainable solution for heating and cooling when compared to systems such as gas boilers or electrical heater [1–3]. However, faults may frequently occur over time and may result in a capacity degradation, which affects the runtime of HVAC equipment, and may lead to the reduction of energy efficiency and thermal comfort, and to compromise the equipment lifespan as well, causing thus higher costs of operation and maintenance of the asset [4–6]. Additionally, according to ASHRAE Handbook [7], several applications may also require special attention regarding their criticality, such as data centers, operating rooms, and even bioterium facilities, as depicted in [8]. Therefore, to prevent faults, automated FD systems may be used. Since an automated FD system can diagnosis faults and assesses their impacts, such scheme presents the potential to reduce energy and maintenance costs and enhance the equipment life as well. According to Li and Braun [9], an automated FD may provide a reduction approximately of 30% in the service costs and they also estimated that the payback period for a FD system was less than one year. To ensure a profitable asset management, an effective © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 154–161, 2023. https://doi.org/10.1007/978-3-031-25448-2_15

Heat Pumps Smart Asset Management Implementation Through Virtual Sensors

155

automated FD system should have low-cost, reliable sensors. This requirement may be fulfilled through the implementation of VS and Fault Impact Models [3, 10, 11]. VS do not require any extra hardware complexity because they are sensors that can replace expensive physical sensors (PS). Therefore, it is possible to acquire data with low cost sensors as inputs. A VS can be characterized, mainly, into two levels, basic and derived. A basic VS only utilizes real sensor data as inputs. A derived VS utilizes information from another VS [11]. The data acquired through VS may be used as inputs to detect a fault and to develop an analysis of its severity as well. This analysis is the Fault Impact model, which provides useful information for establishing thresholds for the system. Through these thresholds, equipment parameters such as health and economic status can be evaluated to diagnose the severity of the fault, and consequently, to provide a decision support regarding the requirement for service [5, 11].

2 Description of Facilities and Equipment A suitable Air-to-Water Heat Pump is available at the Guarda International Re-search Station on Renewable Energies (GIRS-RES), of the Electromechatronic Systems Research Centre (CISE), located in Guarda, Portugal. The facility presents an area of 75 m2 and its main function is the research of renewable energy systems. Guarda is in a region of Portugal where there is a significative weather variation over the year, with a maximum average temperature of 32 ºC and a minimum average temperature of −5 ºC. Based on that and on its building features, the facility presents a thermal load of about 6 kW for cooling or heating. 2.1 Equipment Used The equipment to be used consists of two units – an indoor unit, and an outdoor unit, which is the heat pump itself. The latter one is a package unit constituted, mainly, by a single-phase compressor, double expansion valves, a 4-way valve, condenser, evaporator, and an inverter control. The refrigerant fluid used by this equipment is the R410A. These technical features can be seen in Table 1. The other unit is an indoor one, composed of the internal unity which constitutes the control and hydraulic units, a Hot Water Tank, and a Heat Exchanger (HX), in this case, a fan coil unit, which may provide cooling or heating, according to the HP operating mode. Table 1. AWHP Technical Features. AW-7S Bosch Compress 6000 series model COP/EER

4.84/3.85

Sound Level (dB)

53 (continued)

156

P. Barandier et al. Table 1. (continued)

AW-7S Bosch Compress 6000 series model Compressor Type Expansion Device

Twin Rotary Electronic Expansion Valve (EEV)

Operation Limits Heating (ºC)

−20/35

Operation Limits Cooling (ºC)

15/45

Refrigerant Type

R410A

Refrigerant Charge (Kg)

1.75

Maximum Power (KW)

7

3 Methodology In order to ensure a good basis for the VS, a certain number of PS are needed. The use of thermocouples along the refrigeration circuit are the most economical option. Based on these temperature measurements, there are several data regarding the system thermodynamic parameters that can be estimated, such as the system pressures, enthalpies, refrigerant charges, refrigerant mass flows (RMF), flows in both HXs, and water flow in the internal HX. Once determined, these parameters may be used for fault diagnosis in the HP. There are some variations between the VS for fixed speed compressors and variable speed compressors, and since the equipment consists of the former kind, only the VS regarding this type of compressor will be considered. 3.1 Basic Virtual Sensors Kim and Lee [3] proposed a structure for a FD system based on VS and Fault Impact model, which is divided into four steps: preprocessor, fault detection, fault diagnosis and decision. This approach requires the input of 9 temperature measurements regarding the refrigerant, the air, and the water. Therefore, to enable an accurate data acquisition of the system, some temperature sensors will be used as physical sensors (PS). The parameters to be measured are shown in Table 2. Based on that, pressure measurements can be estimated through the acquired data by the thermocouples along the circuit and by the calculation of the pressure drops. Since the refrigerant is a two-phase mixture in the HXs under steady-state conditions, the evaporator and condenser pressures can be estimated through the refrigerant thermodynamic correlations for the properties. Meanwhile, when the refrigerant is in a subcooled liquid or superheated vapor state, their pressure cannot be directly determined by such correlations.

Heat Pumps Smart Asset Management Implementation Through Virtual Sensors

157

Table 2. Temperature Measured Parameters of the Fluids. Fluid

Parameter

Number

Refrigerant

Evaporating Temperature (ET)

1

Condensing Temperature (CT)

2

Suction Line (SL)

3

Liquid Line (LL)

4

Discharge Line (DL)

5

Air (evaporator in case of heating, and condenser in case of cooling)

Inlet Temperature (IT)

6

Outlet Temperature (OT)

7

Water (evaporator in case of cooling, and condenser in case of heating)

Inlet Temperature (IT)

8

Outlet Temperature (OT)

9

Therefore, it is necessary to establish suitable locations for the thermocouples to be installed and for the temperatures along the circuit to be reliably measured, and similarly, for the pressures. Therefore, as stated in [12], there should be thermocouples in the middle of the condenser, approximately 3 passes before the middle of the HX, and in the outlet of the condenser as well, in the inlet of evaporator, and in the inlet and outlet of the compressor, as it is shown in Fig. 1. There are 3 thermocouples in each side of the system since the HX may perform two different functions (evaporator and condenser) in reversible systems, according to the operation itself. Besides the locations of the thermocouples, in order to determine the pressures in the liquid, suction, and discharge lines, it is also important to consider the pressure drops (P) between the points where these parameters are measured. Therefore, it is necessary to calculate the P, which is given in Eq. 1 [12] P = k m ˙ 2Ref

Fig. 1. Thermocouples’ locations along the circuit.

(1)

158

P. Barandier et al.

where, according to [12], k is a proportionality constant and m ˙ is the mass flow rate that depends, mainly, on the compressor discharge and suction pressures (Pd and Ps). However, the compressor data is usually given as a function of the refrigerant saturation temperature parameters related to these pressures. Such temperature parameters can be termed as the discharge dew-point temperature (Td) and the suction dew-point temperature (Ts). It is important to clarify that these temperatures are different from the condensing and evaporating ones due to pressure variations. Thus, based on Td, Ts and on the compressor map, a VS regarding to m ˙ that will be better explained below can be estimated through Eq. 2. m ˙ Ref = compressormap(Td , Ts)

(2)

Therefore, according to Eqs. 1 and 2, the P through a HX may be estimated in terms of the actual and rated flow conditions as can be seen in Eq. 3. From that, the pressures in the system can be more accurately measured. P = P rated (

m ˙ ref m ˙ ref ,rated

)

2

(3)

According to [3, 11, 13], a compressor map is used to measure the RMF using compressor inlet and outlet data. Based on ARI Standard [14], Kim [3] states that a VS regarding m ˙ can be mathematically modeled as shown in Eq. 4   m ˙ comp = ρsuc × a0 + a1 Tc + a2 Tc + a3 Te2 + a4 Te2 + a5 Tc2 + a6 Te3 + a7 Te3 + a8 Te2 × Tc + a9 Tc2 × Te

(4) where ρsuc is the density at the suction of the compressor, the a’s are empirical coefficients which can be obtained in a regression analysis, Te is the evaporating temperature and Tc is the condensing temperature. This model has presented great accuracy for the determination of RMF rate. However, according to [3], it is not reliable when there is a valve leakage fault present in the compressor. For this reason, to provide some fault tolerance and to estimate the RMF rate as well, some derived VS regarding this parameter will be discussed in the next section. As stated in [11], another parameter that can be estimated based on the measured temperatures is the compressor power since it presents a direct correlation with the evaporating and condensing temperatures at the nominal compressor speed, the mathematical ˙ is the compressor power, the d ’s are empirical model can be seen in Eq. 5, where W coefficients and Te and Tc are the evaporating and condensing temperatures. ˙ = (d1 + d2 Te + d3 Tc + d4 Te2 + d5 Tc2 + d5 Te × Tc ) W

(5)

Through these installed thermocouples is also possible to estimate the refrigerant charge in the system. According to [15] and [16], a current practice to determine the charge level in the system is evacuating it and weighing the removed refrigerant. However, this method is very time consuming and expensive. Another approach is to use the superheat or subcooling parameters. Nevertheless, this approach is qualitative and can

Heat Pumps Smart Asset Management Implementation Through Virtual Sensors

159

only determine if the system is undercharged or overcharged, but not the system’s level of charge. Li and Braun [16] developed a quite effective method for systems with fixed speed compressors, except in situations of extreme undercharge or overcharge, using the temperatures measurements in the suction and liquid line. Despite the fact that this method could be applied to the equipment in the present study, since it presents a fixed speed compressor, this approach was improved by adding a correlation for refrigerant charge in terms of the evaporator inlet quality, according to [3, 11]. Therefore, a quite effective method to estimate the refrigerant charge is discussed in [3, 11] and can be seen in Eq. 6   mt − mt,rt mt,rt      1  sc × Tsh − Tsh,rt + K sc × Xev,in − Xev,in,rt K sc = [ Tsc − Tsc,rt − K sh x dsh Kch   (6) × Tdsh − Tdsh,rt where mt is the refrigerant actual total charge, mt,r is the nominal total refrigerant charge, K ch , K sc/sh , K sc/x , and K sc/dsh are constants that depend on the system attributes, such as the HXs geometries, and some nominal parameters. T sc , T sc,rt , T sh , T sh,rt , T dsh and T dsh,rt are the temperatures regarding the measured subcooling, nominal subcooling, measured superheat, nominal superheat, discharge superheat of the compressor and nominal discharge superheat of the compressor, respectively. X ev,in and X ev,in,rt are the evaporator inlet quality and the nominal evaporator inlet quality, respectively [3, 11, 16]. 3.2 Derived Virtual Sensors As previously stated, the VS for RMF rate based on temperatures is not very reliable when there is a valve leakage fault in the compressor. Based on that, another option to measure the RMF rates are the derived VS as described in [11]. One alternative proposed by [11] is to use an energy balance on the compressor to estimate the RMF rate as it is shown in Eq. 7 m ˙ energy =

˙ × (1 − αloss ) W hdis (Tdis , Pdis ) − hsuc (Tsuc , Psuc )

(7)

˙ is the compressor power consumption, αloss is the compressor heat loss ratio, where W and hdis (Tdis , Pdis ) and hsuc (Tsuc , Psuc ) are the discharge line and suction line refrigerant ˙ can be obtained by enthalpies. Discharge and suction pressures (Pdis and Psuc ) and W using other VS as previously stated. Despite presenting a very small value, lower than 5%, αloss can be estimated through Eq. 8 for FSC [3] αloss,est = c0 + c1 Pdis + c2 Psuc + c3 Tdis + c4 Tsuc

(8)

where the c’s are empirical coefficients, the Tdis and the Tsuc are the discharge and suction temperature. This VS presents a high accuracy even with valve leakage in the compressor. Therefore, when used with the previous VS to estimate RMF rate it can detect such fault in the system.

160

P. Barandier et al.

In order to diagnose HXs fouling, one of the most common HP faults, according to [11], there are two possible methods, which are HXs thermal capacity, which may provide later a VS for the HP’s COP, and the heat transfer fluid (air or water) flow in the HX. The HXs thermal capacity can be estimated based on other parameters measured by the already presented VS, as presents Eq. 9. ˙ high (Phigh , Thigh ) − hlow (Plow , Tlow )) QHX = m(h

(9)

˙ is the RMF rate measured by the previously where QHX is the HXs thermal capacity. m discussed VS, hhigh , Phigh and Thigh are the HXs high enthalpies, pressures, and temperatures, which represent the condenser inlet parameters and evaporator outlet parameters. hlow , Plow and Tlow are the HXs low enthalpies, pressures, and temperatures, which represent the condenser outlet parameters and evaporator inlet parameters. The flow in the HX is a function dependent on the thermal capacity and on the heat transfer fluid parameters. This flow can be estimated based on Eq. 10 [3, 11]   m(h ˙ high Phigh , Thigh − hlow (Plow , Tlow )) × vHX ,f (10) V˙ = (hf ,o + hf ,i ) where V˙ is the flow in the HX. hf ,o and hf ,i are the HXs fluids (air/water) outlet and inlet enthalpies. vHX ,f is the HXs fluid specific volume. Based on the VS regarding the thermal capacity and the compressor power, a VS regarding the HP’s efficiency can be obtained as it is shown in Eq. 11 COP =

QHX ˙ ˙ Wcomp + Wmechanicalcomponents

(11)

˙ comp is where COP is the coefficient of performance, QHX is the HXs thermal capacity, W ˙ mechanicalcomponents is the power regarding mechanical the measured compressor power, W components such as fans and water pumps.

4 Summary and Conclusion Considering the fact that PS usually present quite high costs, they also require maintenance, which increases even more their costs. Based on that, an automated HP FD approach through VS presents, in a technical and economical perspective, a promising approach to detect and diagnose faults in heat pumps systems since the discussed VS have presented not only high precision measurements, but lower implementation costs as well. Moreover, VS also present the possibility of an analysis of the precision of a given PS. Another great advantage of VS when compared to PS that corroborates its technical, economical, and environmental benefits is that since PS are quite invasive, their implementation, in the case of a HP pressure sensor, for example, may result in a leak of refrigerant of the circuit, which may be avoided by VS application. Conclusively, to ensure an efficient smart asset management, the VS present a consolidated basis, not only for analyzing the behavior of the equipment, but also to collect data regarding the equipment itself. These collected data can be used, in the future, to develop other useful FD tools, such as an equipment digital twin, for example.

Heat Pumps Smart Asset Management Implementation Through Virtual Sensors

161

Acknowledgments. This work was supported by the European Regional Development Fund (ERDF) through the Operational Programme for Competitiveness and Internationalization (COMPETE 2020), under Project POCI-01-0145-FEDER-029494, and by National Funds through the FCT - Portuguese Foundation for Science and Technology, under Projects PTDC/EEIEEE/29494/2017, UIDB/04131/2020 and UIDP/04131/2020.

References 1. Carroll, P., Chesser, M., Lyons, P.: Air Source Heat Pumps field studies: a systematic literature review. Renew. Sustain. Energy Rev. 134, 110275 (2020). https://doi.org/10.1016/j.rser.2020. 110275 2. Bellanco, I., Fuentes, E., Vallès, M., Salom, J.: A review of the fault behavior of heat pumps and measurements, detection and diagnosis methods including virtual sensors. J. Build. Eng. 39, 102254 (2021) 3. Kim, W., Lee, J.H.: Fault detection and diagnostics analysis of air conditioners using virtual sensors. Appl. Therm. Eng. 191, 116848 (2021). https://doi.org/10.1016/j.applthermaleng. 2021.116848 4. Kim, W., Braun, J.E.: Development, implementation, and evaluation of a fault detection and diagnostics system based on integrated virtual sensors and fault impact models. Energy Build. 228, 110368 (2020) 5. Kim, W., Katipamula, S.: A review of fault detection and diagnostics methods for building systems. Sci. Technol. Built Environ. 24(1), 3–21 (2018) 6. Breuker, M.S., Braun, J.E.: Common faults and their impacts for rooftop air conditioners. HVAC R Res. 4(3), 303–318 (1998). https://doi.org/10.1080/10789669.1998.10391406 7. Nicklas, S., et al.: ASHRAE Handbook HVAC Systams and Equipment. USA no. 28, p. 955 (2016) 8. Barandier, P., Marques Cardoso, A.J.: Asset management and energy improvements in a critical environment–the case of a University Bioterium. In: World Congress on Engineering Asset Management, pp. 364–373 (2021) 9. Li, H., Braun, J.E.: Economic evaluation of benefits associated with automated fault detection and diagnosis in rooftop air conditioners. ASHRAE Trans. 113 PART 2(2005), 200–210 (2007) 10. Cristaldi, L., Ferrero, A., Macchi, M., Mehrafshan, A., Arpaia, P.: Virtual sensors: a tool to improve reliability. In: 2020 IEEE International Workshop on Metrology for Industry 4.0 & IoT, pp. 142–145 (2020) 11. Kim, W.: Fault Detection and Diagnosis for Air Conditioners and Heat Pumps Based on Virtual Sensors (2013). https://docs.lib.purdue.edu/open_access_dissertations/153 12. Li, H., Braun, J.E.: Virtual refrigerant pressure sensors for use in monitoring and fault diagnosis of Vapor-Compression equipment. HVAC R Res. 15(3), 597–616 (2009). https://doi. org/10.1080/10789669.2009.10390853 13. Kim, W., Braun, J.E.: Development and evaluation of virtual refrigerant mass flow sensors for fault detection and diagnostics. Int. J. Refrig. 63, 184–198 (2016). https://doi.org/10.1016/j. ijrefrig.2015.11.005 14. A. R. I. S. 540, “Performance Rating of Positive Displacement Refrigerant Compressors and Compressor Units.” Air Conditioning and Refrigeration Institute Arlington, VA, USA, 2004 15. Kim, W., Braun, J.E.: Evaluation of a virtual refrigerant charge sensor. In: International Refrigeration and Air Conditioning Conference. West Lafayette IN, USA (2010) 16. Li, H., Braun, J.E.: Development, evaluation, and demonstration of a virtual refrigerant charge sensor. HVAC R Res. 15(1), 117–136 (2009). https://doi.org/10.1080/10789669.2009.103 90828

Driving Port Efficiency Through 5G-Enabled Condition Monitoring of Quay Cranes Adolfo Crespo del Castillo1(B) , Manuel Herrera1 , Manu Sasidharan1 , Jorge Merino1 , Ajith Kumar Parlikad1 , Loretta Liu2 , Richard Brooks3 , and Karen Poulter4 1 Department of Engineering, Institute for Manufacturing, University of Cambridge,

Cambridge CB3 0FS, UK [email protected] 2 Three UK, Green Park, 450 Longwater Avenue, Reading RG30 3UR, UK 3 Blue Mesh Solutions Limited, 9 Brackley, Weybridge KT13 0BJ, England 4 Port of Felixtowe, Tomline House The Dock, Felixstowe IP11 3SY, UK

Abstract. Ports play an important role in the worldwide economy as essential nodes in the global trading and supply chain network. The monitoring and maintenance of port assets are key to efficient terminal operations and need careful attention to ensure that their availability does not become a bottleneck. Recent advancements in digital technologies with their real-time potential and lack of reliance on human input for data collection, can improve the fault diagnosis and guide future investment or replacement of asset components. This paper presents the roadmap for applying condition-based maintenance and predictive analytics to improve the port efficiency by identifying operation critical assets, identifying the IoT sensors to monitor their failure symptoms, the algorithms to do so, and the whole data pipeline structure to support this process. The IoT sensors communicate using 5G technology by taking advantage of greater transmission speed and low latency; enablers for future applications in predictive asset management. Keywords: Smart port · Cranes · Condition monitoring · Asset management · IoT · 5G

1 Introduction Maritime transport is the backbone of the global economy. Over 80% of international trade is carried by sea (Zhang 2017) and any disruptions to port operations can cause local and global ripple effects in logistics and trade-dependent industries. The efficiency of the port operation often is influenced by the availability and condition of its critical operational assets such as quay cranes (QCs) (Lin et al. 2020) that load and unload the containers from the ships directly to the port platform. Previous works on improving port efficiency have focussed on the scheduling of vessels and workload allocation (Kaveshgar et al. 2012; Al Awar et al. 2016) and maintenance tasks while considering resource constraints (Li 2019; Zheng et al. 2019). It is also essential to have access to accurate and early fault detection and diagnosis of quay crane-related faults or disruptions before © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 162–171, 2023. https://doi.org/10.1007/978-3-031-25448-2_16

Driving Port Efficiency Through 5G

163

they impact port operations. Advancements in the Internet of Things (IoT) technology allow the collection of data and information from various components (e.g. vibration of crane motors, fatigue of steel gantries, knocking of couplings etc.) to inform asset management strategies; for example, where mitigation measures are required and when (Smythe 2013). Moreover, IoT deployments at ports are also aimed at improving energy efficiency, speed optimisation, and scheduling (Munim et al. 2020). The digitalisation of critical infrastructure systems such as transport, power, and water demonstrate the role of data and information in understanding the asset’s performance. It can also help the maintenance engineers to set condition-based maintenance (CBM) regimes through prognostics and health monitoring (PHM), anomaly detections and prognosis of the asset’s remaining useful life (RUL) (Guillén et al. 2016). The motivation for this paper is the ongoing project at the Port of Felixstowe (PoF) to monitor the QC condition using IoT and 5G technology and to identify pre-incident trigger conditions by employing artificial intelligence (AI) (Herrera Fernandez et al. 2022; Merino et al. 2022; Molavi et al. 2020). The PoF is the UK’s largest container port that handles more than 4 million containers from approximately 3000 ships each year. QCs have an operational life of 25 years, but the effective life is often not more than 20 years (Bartošek and Marek 2013). This paper presents a systematic approach by which a predictive maintenance programme can be adapted to enhance a port’s operational efficiency by harnessing the technological advancements in IoT, 5G and predictive data analytics.

2 Driving Port Efficiency Through 5G-Enabled Condition Monitoring of Quay Cranes A step-by-step guide to a 5G-enabled and IoT-based condition monitoring to aid the predictive maintenance of critical assets is given in Fig. 1. It also provides insights into the deployment of IoT sensors and 5G technology for the same. The following subsections will explain each step in detail.

164

A. Crespo del Castillo et al.

Fig. 1. A framework to drive port efficiency through 5G-enabled condition monitoring of quay cranes

2.1 Asset Identification and Hierarchisation in Systems According to ISO 17359:2018, the first stage to defining a predictive maintenance programme is to audit the assets while ascertaining the different levels where a maintenance action can be performed and identifying the associated failure modes. For the case in hand, the assets are broken down from the QC level to the components and subcomponents level depending on the detail required to identify failure precursors. 2.2 Failure Mode Analysis Associated with Asset Hierarchy and Physical Symptoms Associated Failure Modes and Effects Analysis (FMEA) was conducted to rank the criticality of failure modes of each QC component (spreader, hoist, gantry). FMEA considers the severity, frequency, and detectability of the incident of interest. In this case, for each component, the frequency can be determined by the study of disruption logs maintained by the PoF and the severity can be deduced from the average duration between when the fault occurred and when it was actioned (see Fig. 2). It was identified that the spreaders, hoists, and gantries contributed to most of the disruptions/failures in two years. It was also reported that 75% of the spreader-, 54% of the hoist- and 38% of the gantry-related incidents occurred at least more than twice during the analysis period. The symptoms associated with each of the failure modes were identified (see Table 1).

Driving Port Efficiency Through 5G

165

Table 1. Failure modes description and symptoms at asset hierarchy levels mapped against relevant sensor types Failure modes description

Sensor type

Asset hierarchy levels

Failure/outage

Symptoms

Main hoist engine bearings and gearbox

Uneven loading

Vibration indicating Vibration wear and tear of motor gearbox

Boom hoist gearbox

Lubrication

Vibration

Vibration

Hoist - trolley assembly -driver cabin

Overloading

Sound, vibration, orientation, and angular velocity

Acoustic, vibration, gyroscope

Gantry

Infrastructural failure

Deformation, pressure, Strain Gauge load, torque, position

Spreader - headblock

Shock Loading twist lock

Orientation, angular velocity

Gyroscope

Spreader - cable reel gearbox

Spreader up/down movement

The vibration of the gearbox, bearing failure, wear and tear

Vibration

Hoist - trolley drive input coupling

Knocking on couplings

Vibration, sound

Acoustic, vibration

Hoist - trolley assembly -rail Hoist - trolley assembly - sheaves

2.3 Define Monitoring Variables and Failure Mode State Once the detectable symptoms of each failure mode are defined, the next stage is translating these physical dimensions into monitorable variables defining the state of failure. For a certain failure mode, the vibration could evidence the state of the failure, but in other cases, the temperature and acoustics would also be necessary. Threshold levels are set for each monitorable variable to indicate the condition of the component. Such ranges of “normal” behaviour can be defined in consultation with the engineers, technicians, and the component manufacturer. 2.4 Sensor Identification to Capture Physical Dimensions The failure mode and symptoms identified will inform the type of sensors for capturing the physical parameters required to monitor the asset condition. The selected sensor candidates for vibration, accelerometers, gyroscopes, strain gauges, humidity, temperature, and acoustic sensors were selected for preliminary testing. Their specifications were compared to find the optimal for the final application. The locations of these sensors at the specific asset hierarchy levels were identified by working closely with the engineers and technicians responsible for the maintenance of QCs. The lessons learned from

166

A. Crespo del Castillo et al.

sourcing, installation, calibration, and communication of sensors in this deployment are detailed within Merino et al. (2022). 2.5 Algorithm Development for Anomaly Detection and Failure Prediction Data streams collected from each sensor are, naturally, time-series metering QC data that need to be converted into information about the occurrence of an anomaly. This places time-series data mining and time-frequency analysis among the best-suited candidate methods for approaching QC predictive analytics. For the first class of methods, using a symbolic aggregate approximation (SAX) representation benefits the analysis further of a data numerosity reduction, in addition to the straightforward implementation of timeseries clustering and classification. SAX lies in three main steps: a time series division into equal-length segments (Lin et al. 2007), averaging the values of each segment (Yeh et al. 2017), and approximating such average values by an alphabet symbol (Sun et al. 2017). Matrix profile is a more advanced technique than SAX for mining and pattern recognition in time series (Yeh et al. 2017). This is a time-series companion of the original data split into subsequences from which it represents the nearest neighbour distances. Matrix profile provides the capability for detecting subsequences having discord patterns as the first step in an anomaly detection procedure. Time-frequency analyses (Sun et al. 2017) are better suited for vibration analysis than the straight use of SAX and matrix profile. The reason lies in the high importance of features related to cycle behaviours in signals associated with a physical movement. A discrete Fourier transform (DFT) is a universal solution to analyse signal frequencies thanks to a time-series decomposition into its cycle components. A discrete wavelet transform (DWT) is an alternative to DFT (Wirsing 2020). A DWT decomposes the time series into a family of wavelets, or basis functions, describing the evolution of the signal frequencies. In this regard, wavelets are one of the main techniques for the further development of a joint time-frequency analysis (Zhang 2019). DWT has shown a better performance than a sliding DFT, or short-time Fourier transform (STFT), that iteratively computes the Fourier transform over a signal evolving on time (Mateo and Talavera 2018). In both cases, DWT and STFT, a general anomaly detection procedure includes a time-frequency feature extraction process along with such features comparison to their standard values. Regular and anomaly patterns found using the above-mentioned techniques create historic databases out of which it is possible to apply machine learning methods for their classification further. Such machine learning methods can be used for predictive analytics, inferring the asset condition that may trigger an anomaly in the future (and, ultimately, the type in which the anomaly is classified) (Wang et al. 2020). 2.6 Data Pipeline for Algorithms and Applications Design The most granular data available in the QC’s programmable logic controller (PLC) provides operational information of different components (e.g., power/energy used by the electrical engine of the hoist for each lift). Commonly, IoT sensors measure physical dimensions on a prescribed frequency (e.g., every second, minute, hour). The amount of data produced by the IoT sensors, and the data stored within the PLC is vast and needs to

Driving Port Efficiency Through 5G

167

be managed accordingly. Data is routed into the cloud server from both sources through a relay server with a DMZ to ensure security in the Port’s network (Fig. 2).

Fig. 2. High-level data flow diagram

Data from the sensors is streamed to the cloud via a publish/subscribe protocol (i.e., MQTT). PLC data is batched and uploaded to the Cloud and streamed similarly to sensor data. At the same time, it is pre-processed/transformed and cleansed to standardise the multifarious forms the data may come formatted into. Data can then be accessed in the form of queries from the integrated data storage for visualisation, analytics, and maintenance applications. Decision support systems for setting predictive maintenance plans can access and visualise not only data from the sources but also from the results of the analytics applications. 2.7 Sensor Installation on Physical Assets and 5G Testbed for Use Case Testing Sensor installation was managed between the crane engineers and the IoT engineers to combine the best fit between location for physical parameters and location to transmit data via WiFi. IoT devices power is supplied via 5 V USB C connectors and Raspberry Pi as Gateways with WiFi allowed data to backhaul locally. WiFi bridges were then used to carry the sensor signals into a 5G access point. Testing included considerations of quality of sensors and measured parameters, sampling frequency, data requirements, data management, device security, power consumption, data backhaul, location of sensors, location of gateways, data type, data labelling, and position labelling, maintenance of sensors when needed, and cost. The 5G System was deployed as a Non-Public or Private Network according to 3GPP. The system is based on SA Option 2 according to 3GPP R15. The System is deployed

168

A. Crespo del Castillo et al.

as full integrity of the PoF’s network and isolation from the public network, as a form of protection against malfunctions in the public network. The main components of the 5G System are the 5G ODUs, 5G NR RAN, and 5G On-remised SA Core as shown in Fig. 3. The 5G connection is non an additive but a core component of the novelty. 5G is about connecting in a more reliable and without lag way to manage things in real time, but it is also critical to send massive quantities of data. Therefore 5G application will allow to maximise speed but also the scale, as the coordination of multiple valuable assets’ data will be able to be extended to larger areas and assets. The reliability of the connection is also a critical point as this case requires high demanding data delivery requirements one placed online, due to the criticality of the QC and the port as a whole in the supply chain. This can be finally resumed enouncing that high complexity asset management applications with high demanding connectivity requirements as the one presented here, benefit from 5G by receiving massive quantities of data, faster and in a more reliable way; that will allow to apply the necessary powerful models that will provide companies with a better informed decision-making to manage their assets in real time.

Fig. 3. Standalone non-public network solution

2.8 Go On-Line The last step is to place the solution on production during the port’s normal operation. A data processing server was installed inline between the 5G network and the predictive analytics system to reduce the data processing requirement outside of the port’s IT network.

Driving Port Efficiency Through 5G

169

3 Concluding Discussion and Further Work The impact of condition monitoring on improving the efficiency of port operation is going to be significant in the upcoming years due to advances in IoT and communication. This paper provides a step-by-step guidance to drive 5G-enabled condition monitoring of valuable port assets based on a deployment to monitor QCs. 5G as the latest mobile technology is seen to be a major catalyst for driving entirely new operational models that cannot be addressed with existing 4G technology or successfully scaled commercially with fixed-line technologies. The 5G standards create opportunities for new types of technology, requiring not only a higher throughput, but also control over latency, security, and reliability. The root causes of QC related disruptions were identified at the component level. Sensing options to monitor the condition of the QC components were identified and tested for their suitability. Time-series data mining and time-frequency analysis were identified as suitable techniques for predictive analytics. The limitations of this work are mainly characterised by the IoT and data side. Ensuring a common time reference for all sensing data becomes one of the most important challenges, as it directly impacts the levels of data quality (Merino et al. 2022). Clocks in IoT sensors can drift because of power losses, communication irregularities, or because of the lack of a common time server (as most work in local networks). However the main limitation is the availability of data that represents failures of the asset. The algorithms applied are clustering and classifying discords but if the discord measured is not checked and identified at the source; that identified discord and the data related are meaningless. The biggest challenges from the data management side, were both the concurrent storage to generate flexible and manageable data lakes, together with the data life cycle, meaning the most efficient management of stored data to not collapse by overloading the data services (Merino et al. 2022). Future work will focus on the development of a decision support system based on anomaly detection algorithms to identify potential disruptions and alert the engineers and technicians for remediation measures. This will be achieved by improving the algorithms through the collection of more accurate failure data, relating the discords to actual failure modes. Another step ahead is the current development of an Asset Health Index driven by near real-time data to manage the lifecycle of the asset and the future prioritisation of capital investment in QC critical parts. Acknowledgements. This work was supported by the UK’s Department for Digital, Culture, Media, and Sport through the 5G Testbeds and Trials (5G Port of Felixstowe project).

References Al Awar, K., Alawani, M., Al Jaberi, S.: A multi-vessel quay crane scheduling problem. In: 7th International Conference on Information, Intelligence, Systems and Applications (IISA), pp. 1–5 (2016). https://doi.org/10.1109/IISA.2016.7785354 Bartošek, A., Marek, O.: Quay cranes in container terminals. Trans. Transp. Sci. 6(1) (2013). https://doi.org/10.2478/v10158-012-0027

170

A. Crespo del Castillo et al.

Guillén, A., Crespo, A., Gómez, J., Sanz, M.: A framework for effective management of condition based maintenance programs in the context of industrial development of E-maintenance strategies. Comput. Ind. 82, 170–185 (2016). ISSN 0166-3615, https://doi.org/10.1016/j.com pind.2016.07.003 Herrera Fernandez, M., Sasidharan, M., Merino, J., Parlikad, A.: Handling irregularly sampled IoT time series to inform infrastructure asset management (2022). https://doi.org/10.17863/ CAM.83951 ISO 17359:2018: Condition monitoring and diagnostics of machines. General guidelines (2018). www.iso.org/standard/71194.html Kaveshgar, N., Nathan, Huynh, N., Saeed, R.: An efficient genetic algorithm for solving the quay crane scheduling problem. Expert Syst. Appl. 39(18), 13108–13117 (2012). ISSN 0957-4174, https://doi.org/10.1016/j.eswa.2012.05.091 Li, H.: Integrated workload allocation and condition-based maintenance threshold optimisation. Doctoral thesis, University of Cambridge (2019) Lin, D., Lee, C., Guo, W.: Integrated preventive maintenance improvement and spare part ordering strategy for age deterioration components. In: 2020 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), pp. 1204–1209 (2020). https://doi.org/10. 1109/IEEM45057.2020.9309945 Lin, J., Keogh, E., Wei, L., Lonardi, S.: Experiencing SAX: a novel symbolic representation of time series. Data Min. Knowl. Discov. 15(2), 107–44 (2007) Mateo, C., Talavera, J.A.:. Short-time Fourier transform with the window size fixed in the frequency domain. Digit. Sig. Process. 77, 13–2 (2018) Merino, J., et al.: Lessons learned from an IoT deployment for condition monitoring at the Port of Felixstowe. In: Conference: 5th IFAC Workshop on Advanced Maintenance Engineering, Services and Technology, Bogotá (2022). https://doi.org/10.1016/j.ifacol.2022.09.210 Molavi, A., Lim, G.J., Race, B.: A framework for building a smart port and smart port index. Int. J. Sustain. Transp. 14(9), 686–700 (2020). https://doi.org/10.1080/15568318.2019.1610919 Munim, Z., Dushenko, M., Jaramillo Jimenez, V., Shakil, H., Imset, M.: Big data and artificial intelligence in the maritime industry: a bibliometric review and future research directions. Maritime Policy Manag. 47(5), 577–597 (2020). https://doi.org/10.1080/03088839.2020.178 8731 Smythe, T.C.: Assessing the impacts of Hurricane Sandy on the port of New York and New Jersey’s maritime responders and response infrastructure. Quick Response Report No. 238: Final Report to the University of Colorado Natural Hazards Centre (2013) Sun, Z., Zhou, L., Wang, W.: Learning time-frequency analysis in wireless sensor networks. IEEE Internet Things J. 5(5), 3388–3396 (2017) Wang, J., de Moraes, R.M., Bari, A.: A predictive analytics framework to anomaly detection. In: IEEE Sixth International Conference on Big Data Computing Service and Applications (BigDataService), pp. 104–108. IEEE (2020) Wirsing, K.: Time frequency analysis of wavelet and Fourier transform, wavelet theory. IntechOpen, 18 November 2020 Yeh, C., Kavantzas, N., Keogh, E.: Matrix profile VI: meaningful multidimensional motif discovery. IEEE International Conference on Data Mining (ICDM) 18 November 2017, pp. 565–574. IEEE (2017) Zhang, Y., Kim, C.W., Tee, K.F., Lam, J.: Optimal sustainable life cycle maintenance strategies for port infrastructures. J. Clean. Prod. [Internet] 142, 1693–709 (2018). https://doi.org/10.1016/j. jclpro.2016.11.120 Zhang, Z.: Spectral and time-frequency analysis. In: Hu, L., Zhang, Z. (eds.) EEG Signal Processing and Feature Extraction, pp. 89–116. Springer, Singapore (2019). https://doi.org/10.1007/ 978-981-13-9113-2_6

Driving Port Efficiency Through 5G

171

Zheng, F., Li, Y., Chu, F., Liu, M., Xu, Y.: Integrated berth allocation and quay crane assignment with maintenance activities. Int. J. Prod. Res. 57(11), 3478–3503 (2019). https://doi.org/10. 1080/00207543.2018.1539265

Remote Data Collection Motivational Drivers, Challenges, and Potential Solutions in Industrial SME Companies Teemu Mäkiaho(B) , Topias Kallio, Henri Vainio, Jouko Laitinen, and Kari Koskinen University of Tampere, Tampere, Finland {teemu.makiaho,topias.kallio,henri.vainio,jouko.laitinen, kari.koskinen}@tuni.fi

Abstract. Data collected from industrial machines enable companies to pursue new servitization offerings such as machine health condition monitoring as well as to monitor the machine behavior for research and development purposes. In this study, qualitative data was collected from four different industrial SME companies to create a better understanding of the motivational drivers industrial companies may have to create remote data collection systems. The results will discuss the initial motivational drivers for data collection setup construction as well as the challenges companies were facing along the way in designing and implementing their systems. Based on the empirical research results, the recognition of internal motivational drivers and data utilization targets should be clear before proceeding to further development of data collection infrastructure. Generally, the research results recognize various areas that an automation industry company needs to consider before planning and implementing an online data collection system, what challenges they may face as well as generalized proposed solutions are presented.

1 Introduction Machine system availability has a significant role in the overall profit making in the industrial environment and technology-advanced business models are emphasizing the importance of technology and data analytics in overall value creation (Schroderus et al. 2021). The profit margins can be increased using advanced condition monitoring techniques where machine condition can be accurately monitored and its current operation condition assessed based on the monitored parameters such as temperature, humidity, thermography, motor current, electrical inductance, electrical capacitance, acoustic signal, vibration, or pressure (Hashemian and Bean 2011). The broad use of sensors, data collection with real-time transfer capabilities, and data fusion technologies with advanced visualization may contribute to more effective results (Wang et al. 2018) when discussing machine system reliability monitoring and performance optimization. However, the infrastructure of machine system data collection is no longer limited to machine system parameters monitoring. The Industrial Internet of Things (IIoT) connected mass data can be composed by utilizing various data sources, such as condition and operation data (Ehret and Wirtz 2017), failure history, production process data, machine features © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 172–181, 2023. https://doi.org/10.1007/978-3-031-25448-2_17

Remote Data Collection Motivational Drivers, Challenges and Potential Solutions

173

(Bertsimas et al. 2019; Ehret and Wirtz 2017), operating conditions (Liu et al. 2019), repair history data (Kong et al. 2020), and operator attributes e.g. the attributes of the operator who uses the machine. Data collection for enhancing machine system maintenance or availability are only samples of motivation factors the SME’s may have to establish remote data collection capabilities. Increased competitiveness requires new business model approaches, therefore forcing companies to search for better ways to serve their customers as well as to make their products more cost-efficient and reliable. This qualitative research reveals the motivational drivers that the four industrial companies have towards the establishment of a data collection system as well as identifies the challenges industrial SME’s may encounter when designing or implementing a remote data collection system. By identifying the possible challenges in advance, companies will be more ready to create solutions and perform risk mitigation actions in advance (Kallio 2021). For the companies, the main driver of establishing remote data collection capability is to offer remote condition monitoring services for their customers. As a result of this study, the motivation for data collection planning and implementation is addressed. The last part of the results illustrates the challenges and identified potential solutions to how such obstacles could be alleviated or minimized in comparable companies or industries.

2 Methodology The research methodology is empirical qualitative research in the form of semi-structured interviews. Semi-structured interviews follow a list of predetermined questions under previously determined themes to form a general structure for the interview. Also, additional questions and talk are allowed during the interview in a controlled manner (Kallio 2021). This allows exploring rich background information about the studied topic (Saunders et al. 2019). Interviews for this study were conducted with four Finnish machine building small and medium size (SME) companies in the field of the automation industry. All the companies have their main operations in Finland, but all the companies are operating their businesses on a global scale. The companies were selected due to the following criteria: the size of the company corresponds to the small and mediumsized company definition from the machine building industry sector, and the company is currently developing or using remote data collection systems. One of the four companies is already performing data collection whereas the other three are in the design or implementation phase of the data collection system.

3 Motivation for data collection All the case companies share a common interest in developing remote data collection systems to create added value for their customers in the form of servitization. Grubic states that enabling novel servitization of industrial machine systems requires remote data collection (Grubic 2014). Companies are targeting to improve their data collection capabilities to enhance their data-based service offerings such as condition monitoring agreements. Common nominators of the industrial SME case companies, in addition to company size and industry sector criteria, are interest in servitization of machines,

174

T. Mäkiaho et al.

remote data collection, and condition monitoring to support maintenance activities and servitization. Another commonality between the companies is that the machines and machine systems are delivered around the world including no mass-produced products, and all the company products contain machine automation systems embedded in their main products with off-the-shelf industrial Personal Computers (PC’s) and Programmable Logic Controller (PLC) functions. Three of the four case companies can collect process-related control and operation parameters data on a local level, and one company has continuous remote access to their machine- and process-related data with remote diagnostics capability included. Based on the interviews, the case companies are driving towards servitization of their products while, at least partly, retaining ownership of the machines delivered to the customers. This creates a great incentive to keep machines in good condition. To reduce the risk associated with this type of availability-based contract, Grubic and Peppard illustrate that remote monitoring is crucial as it can result in higher availability and reliability of the machines (Grubic and Peppard 2016). From a technological perspective, built-in industrial PC’s enable data collection and partial processing capabilities from the machines with ease. However, in many applications, external sensors are needed to increase data collection capabilities about the machine’s health status or behavior that cannot be originated or are not available from the internal information systems. In many cases within the case companies, the initiatives for adding external sensors are often set by the customer demand of monitoring machine behavior to support their local maintenance planning and decision-making more precisely. Also, customer demand for remote data collection capability is recognized as being one of the top motivational drivers within the case companies. Based on the interviews, product marketing and selling are becoming substantially harder without any integrated data collection. On the other hand, the remote connection also enables companies to develop, tailor-made, and update their products continuously over their lifecycle which may also result in substantial benefits for the customer e.g. in the form of process optimization. Similar products are frequently operated in distinctive and dynamic surroundings and such information offers the product manufacturer enhanced capability to calibrate their machine operation to meet the customer demand and increase overall satisfaction in the customership. The motivational drivers for the companies to pursue remote data collection are listed below: • • • • • • • • • • •

Increased customer demand and satisfaction Predictive capabilities to avoid faults Remote diagnostics and technical support capability Continuous development of products. Learn about products and their deterioration during use in different environments Product servitization and additional services with data Optimal wear part changes Maintain and improve quality System control and tailored operational optimization for customers Troubleshooting and certain maintenance remotely Reduction of traveling for instance in warranty cases through remote fault finding and maintenance Safety improvements in condition monitoring

Remote Data Collection Motivational Drivers, Challenges and Potential Solutions

175

From the customer perspective, remote data collection and condition monitoring is deemed valuable, especially when considering the high price of the machine or the cost of maintenance, machine exposure to wear, and machine criticality to the overall production process. Consequently, remote condition monitoring may result in more stable and predictable operation, and availability increase of a machine (Grubic and Peppard 2016). Machine long-term operational data and the knowledge gained through the data can be further utilized for machine diagnostics, predictive and prescriptive maintenance capabilities to enhance machine uptime and reduce unnecessary operational stoppages (Wang et al. 2018). Additional servitization possibilities, machine optimization, continuous development of products, and better risk management controllability for instance in warranty cases are also recognized motivational factors for data collection within the case companies. Online diagnostics and machine accessibility also offer a sustainability perspective in terms of reduced need for traveling, where certain troubleshooting, system control, technical support, or maintenance can be performed from a distance. Safety issues are also improved when condition monitoring of devices requires fewer onsite measurements and inspections.

4 Selecting Data to Collect The selection of adequate data collection system components is highly dependent on the machine features and their connection to the operational environment. Certain usable data parameters can be obtained directly from the machine-integrated systems and sensors, but most importantly the collected data needs to support the goal of what type of information is requested to be conducted from the raw data. In addition to pure machine operational parameter-related data, other data sources such as machine condition (Ehret and Wirtz 2017), failure history, production process data (e.g. ERP and CMMS systems), machine features (Bertsimas et al. 2019; Ehret and Wirtz 2017), operating conditions (Liu et al. 2019), repair history data (Kong et al. 2020), and operator attributes e.g. the attributes of the operator who uses the machine also may be considered as data sources. Soft sensors i.e. virtual sensors are also deemed as an interesting source of data to replace some physical sensors, especially in applications with high physical stress, such as applications where high amplitude vibration or temperatures are present. Kabaday et al. describe that virtual sensor combine sensed physical sensor data and provide measurement results of abstract conditions that are not directly physically measurable (Kabadayi et al. 2006). According to the companies, by reducing the number of physical sensors, many sensor failures could be avoided thus resulting in better reliability and cost-effectiveness. Once planning sufficient data to collect, utilizing domain knowledge of the machine system operational features, individual subsystem or component level fault modes and effects, as well as the system interrelation to other process components is essential. In the case companies, the primary source of data acquisition was through integrated industrial PC’s or other integrated controllers due to already existing embedded capability. External sensors e.g. vibration sensors were used only if more accurate information about the machine or component condition was needed, or if the information supporting the use of the selected business model. Different data collection motivators may require different process steps for successful implementation. For instance, collecting data for research and development needs may

176

T. Mäkiaho et al.

differ from steps if the data collected is used in remote condition monitoring or process optimization. Simplified and generalized steps of initiating data collection planning and implementation for condition monitoring purposes are addressed as follows: 1. Identify business drivers and customer value 2. Identify machines or components to monitor based on their criticality, failure modes and root causes. Perform FTA or FMECA type of analyses when necessary. 3. Identify representative measurement parameters for root cause 4. Determine measuring and signal processing techniques and analyzing methods. Use primarily integrated technologies and add external sensors or data collectors if adding value. Evaluate the needed data properties 4V’s (Volume, Variety, Velocity, Veracity) (TechAmerica Foundation’s Federal Big Data Commission 2012). 5. Collect, save, analyze, and understand the data. Use various domain knowledge sources to find correlations from the data to create additional information or knowledge. 6. Decide if the data and technologies are sufficient, find development needs and iterate the process if necessary. 7. Implement a continuous data collection and monitoring system After the data has been collected, correct resources are needed to interpret and convert the received raw data into the desired form of information or knowledge. Contextualized technical understanding of machine system operation, fault behavior, normal component level operational feature and behavior knowledge, condition monitoring expertise, and interrelation to other processes to name a few are of the essence to understand when analyzing the collected data. Condition monitoring specialists, machine operators, maintenance technicians, and other professionals working with machines daily operational routines and challenges often hold the required domain knowledge for the correct data selection process and such domain knowledge is vital to utilize when selecting the correct data to collect. Identifying value-adding data before creating the data collection system proves to be challenging. Therefore, a strategy of collecting all the available data was seen as dominant within the companies leaving most of the challenges to find appropriate parameters for the data processing and visualization phase. However, initiating the data collection process quickly and improving the process over time -approach was seen as the most efficient way of starting data collection within the companies.

5 Implementation Challenges and Potential Solutions Generally, implementing IIoT based remote data collection system is considered challenging within the companies. The challenges are divided into two different categories as follows: • IIoT system and technology-related challenges, and • Resource and data-related challenges

Remote Data Collection Motivational Drivers, Challenges and Potential Solutions

177

Implementation of new technologies and systems requires resources. Due to their smaller scale, SME’s have more limited financial and knowledge-related resources available compared to larger companies. There are plenty of system technology providers in the market from sensors to communication and data storage providers, making the selection of correct partners and value-adding technologies demanding. Based on the study, it is recommended to utilize existing off-the-shelf products or embedded technologies as data sources yet remembering that all the IIoT system implementations require some level of customization. Despite this challenge, collaboration and teamworking with other similar companies, partners, or research institutes would significantly help companies in resource-related challenges. Designing the data collection system to meet possible future requirements such as updating communication protocols or gateways as well as scalability are identified challenges that must be accommodated in the system selection phase. Table 1 content consists of IIoT system and technological challenges with an identified potential solution both from literature and results from the interviews. Table 1. IIoT system and technological challenges with potential solutions IIoT system and technological challenges Challenges

Identified potential solutions

• Great variety of different IIoT products and • Discussions with providers solutions available • Specifying own objectives and needs • No standardized structure for IIoT platforms • Experimentation and test trials • Difficulty to get reliable and objective information to compare alternatives • Selection of IIoT platform and services (Contreras-Masse et al. 2020)

• Defining criteria for cloud-based IIoT platform selection (technological, economic, and social) and comparing different solutions based on needs (Contreras-Masse et al. 2020)

• Difficult to get specific costs of cloud and IIoT platforms

• Discussions with providers • Specifying own objectives and needs • Experimentation

• No ready off-the-shelf solutions available

• Understand your system requirements to prevent unnecessary customization

• Difficult to make a system with IIoT system • Find trustable, certified, and scalable characteristics: scalability, security, etc. subsystem platform partners from each of the needed areas. • Poor Internet connections and transmission restrictions in some locations

• Verifying that connections are enough • Use of local data storage • Use dual communication protocols e.g. LAN as primary and cellular network as a secondary connection method (continued)

178

T. Mäkiaho et al. Table 1. (continued)

IIoT system and technological challenges Challenges

Identified potential solutions

• Selection of proper technologies

• Having clear objectives • Data requirements are specified • Recognize internal skills and purchase the needed knowledge for instance by using external consultant services

• Determining required data sources, acquisition, transmission, and storage at the beginning of development

• Collaborate with peers, University co-operation, Training of staff, Consultants

• Specific component information can be hard • Collaboration with component to acquire manufacturers for instance joint development projects based on the collected data • Compatibility of device interfaces to allow data transfer

• Development in collaboration with system manufacturers

• Project-based deliveries make it hard to standardize data collection solutions, calculations, and interfaces

• Standardization of monitored products or components

• Verifying that remote data collection solutions work in customer’s facilities

• Train internal resources such as sales representatives to recognize potential issues and propose risk mitigation

Resource- and data-related challenges differ from the challenges related to technological infrastructure design and build-up, while there may be overlapping activities for instance when selecting proper data sources. Despite the selected motivation to remotely collect data, companies are struggling to recognize how to utilize the collected data and identify what knowledge and expertise are required for the purpose. Also, the case companies are aiming to start data collection rather quickly to start building the historical database of their machine behavior in a contextual environment. Another challenge may occur as SMEs begin to collect data from the machines that are in customers’ facilities. Customers may not be willing to share data from their factory for security or privacy reasons. If this is the case, it came up in interviews that it is helpful to negotiate and point out the benefits of data collection to convince a customer to accept it. Also, early-stage internal training on related technological or business strategy level change management would narrow the scope and reduce unnecessary overlapping of needed work steps, especially at the beginning of the data collection process. Collaboration with similar companies or partners is also considered beneficial for SME companies with limited resources. Resource and data-related challenges with identified potential solutions are presented in Table 2. The table content consists of challenges and identified potential solutions both from the literature and results from the interviews.

Remote Data Collection Motivational Drivers, Challenges and Potential Solutions

179

Table 2. Resource and data-related challenges with identified potential solutions Challenges

Identified potential solutions

Resource challenges • Hard to identify what knowledge and expertise is required and whether does company have it

• Collaboration with partners and needed external resources e.g. consultancy companies • Building internal experience and knowledge

• Limited financial and knowledge resources

• Limiting the scope of the project based on (Omri et al. 2020) • Use existing data collection systems • Limit data to be collected

• Lot of important knowledge focuses on only • Teamwork and collaboration inside the one or few professional company and with partners • Enhance internal documentation processes Data challenges • Identifying relevant data and its properties

• Detecting knowledge needs from a business perspective and selecting a database on them • Based on analytical needs derived in preparatory activities e.g. criticality analyses • Collect data, and learn relevant parameters by studying and analyzing with contextual knowledge

• Missing data: context and historical data

• Use external data sources with PLC data • Knowledge of maintenance personnel • Gradually build a historical database

• Getting access to data from the machine in customer facilities

• Negotiate and show benefits of data collection to customer • Include data collection and accessibility as a prerequisite to sales contracts • Obtain external security certification for your data collection system • Create joint data usage ownership agreements and benefits maps

• Small volume or poor quality of data can result in poor analysis (Omri et al. 2020)

• Collect more data with better properties (Omri et al. 2019) • Think about the 4 V’s of data: volume, variety, velocity, and veracity (TechAmerica Foundation’s Federal Big Data Commission 2012)

180

T. Mäkiaho et al.

6 Conclusions Different business opportunities and cost savings through remote data collection are more obtainable due to recent development and reduced prices of information and communication technologies. Short and long-term data collection combined with condition monitoring and data analytics knowledge will enhance service-based contract capabilities from single machine operation monitoring to accurate machine availability, predictability, and overall fleet management prognostics. The results of this study highlight the initial motivational factors for setting up a data collection system and challenges encountered in the design and implementation process within SME case companies. The motivational factors presented in this paper are collected from companies working in rather homogeneous automation-related business environments and cannot be generalized to all business sectors. However, the listed motivators could act as an initiator for an industrial company processing possibility, why a remote data collection system could be established for their own business. Following, the research results represent related challenges industrial SME’s may discover when intending to move toward data collection-based service contracts such as remote condition monitoring. Challenges are complemented with identified potential solutions which could help industrial players to consider recognized areas early in the process and therefore mitigate design and implementation-related risks. Utilization of the data and determining its value can take time. Establishing a remote condition monitoring system also requires a rather significant resource allocation from human and capital perspectives. The value potential and benefit may also be difficult to showcase in advance for the internal organization as well as to potential customers. Once the system is operational, it is advised to monitor and collect improvement-related information from the system to demonstrate the received benefits.

References Bertsimas, D.J., Supervisor, T., Jaillet, P., Jackson, D.C., Science, C.: Predictive and prescriptive methods in operations research and machine learning: an optimization approach (2019) Contreras-Masse, R., Ochoa-Zezzatti, A., García, V., Pérez-Dominguez, L., Elizondo-Cortés, M.: Implementing a novel use of multicriteria decision analysis to select IIOT platforms for smart manufacturing. Symmetry (Basel) 12 (2020). https://doi.org/10.3390/sym12030368 Ehret, M., Wirtz, J.: Unlocking value from machines: business models and the industrial internet of things. J. Market. Manage. 33, 111–130 (2017). https://doi.org/10.1080/0267257X.2016.124 8041 Grubic, T.: Servitization and remote monitoring technology: a literature review and research agenda. J. Manuf. Technol. Manage. 25, 100–124 (2014). https://doi.org/10.1108/JMTM-052012-0056 Grubic, T., Peppard, J.: Servitized manufacturing firms competing through remote monitoring technology An exploratory study. J. Manuf. Technol. Manage. 27, 154–184 (2016). https://doi. org/10.1108/JMTM-05-2014-0061 Hashemian, H.M., Bean, W.C.: State-of-the-Art Predictive Maintenance Techniques. IEEE Trans. Instrum. Meas. 60, 3480–3492 (2011) Kabadayi, S., Pridgen, A., Julien, C.: Virtual sensors: abstracting data from physical sensors. In: Proceedings - WoWMoM 2006: 2006 International Symposium on a World of Wireless, Mobile and Multimedia Networks, pp. 587–592 (2006). https://doi.org/10.1109/WOWMOM.2006.115

Remote Data Collection Motivational Drivers, Challenges and Potential Solutions

181

Kallio, T.: Remote data collection for condition monitoring in machine building companies. Tampere University (2021) Kong, Q., Lu, R., Yin, F.: Privacy-preserving continuous data collection for predictive maintenance in vehicular fog-cloud. IEEE Trans. Intell. Transp. Syst. 1–11 (2020) Liu, B., Lin, J., Zhang, L., Kumar, U.: A Dynamic prescriptive maintenance model considering system aging and degradation. IEEE Access 7, 13 (2019). https://doi.org/10.1109/ACCESS. 2019.2928587 Omri, N., Al Masry, Z., Giampiccolo, S., Mairot, N., Zerhouni, N.: Data management requirements for PHM implementation in SMEs. In: Proceedings - 2019 Prognostics and System Health Management Conference, PHM-Paris 2019 232–238 (2019). https://doi.org/10.1109/ PHM-Paris.2019.00046 Omri, N., AlMasry, Z., Mairot, N., Giampiccolo, S., Zerhouni, N.: Industrial data management strategy towards an SME-oriented PHM. J. Manuf. Syst. 56, 23–36 (2020). https://doi.org/10. 1016/j.jmsy.2020.04.002 Saunders, M., Lewis, P., Thornhill, A.: Research Methods for Business Students Ebook, 8th edn. Pearson Education, Limited (2019) Schroderus, J., Allan, L., Menon, K., Kärkkäinen, H.: ScienceDirect Towards a Pay-Per-X Maturity Model for Equipment Manufacturing Companies (2021) TechAmerica Foundation’s Federal Big Data Commission: Demystifying Big Data: A Practical Guide to Transforming the Business of Government Listing of Leadership and Commissioners Global Executive Vice President and General Manager. UNICOM Government, pp. 1–40 (2012). Wang, J., Ma, Y., Zhang, L., Gao, R.X., Wu, D.: Deep learning for smart manufacturing: Methods and applications. J. Manuf. Syst. 48, 144–156 (2018). https://doi.org/10.1016/j.jmsy.2018. 01.003

The Effect of Knowledge Based Feature Extraction on Failure Detection of Control Surface Failures of Fighter Aircraft Tauno Toikka(B) , Jouko Laitinen, and Kari T. Koskinen Tampere University, Kalevantie 4, 33100 Tampere, Finland {tauno.toikka,jouko.laitinen,kari.koskinen}@tuni.fi

Abstract. While the area of maintenance is developing from scheduled maintenance toward the condition based maintenance also the failure detection that utilizes system operational data becomes more important. The failure detection from system data can be done in many manners but a process of feature extraction is present more or less almost when the system data is high dimensional, that is the case also with fighter aircraft systems. In this study we examine an effect of system knowledge based feature extraction on the further performance of an algorithmic tasks of failure detection on the operational flight data of a fighter aircraft. The failures for validating the results are several flight control surface failures from the flight data. The feature extraction has been done by using system knowledge of fighter aircraft experts. The failure detection algorithms are comprehensive set of algorithms from the field of anomaly detection, novelty detection, one class classification and unsupervised machine learning. This study demonstrates that some specific failure detection algorithms are more robust for feature extraction and can perform well even with low level of feature extraction when detecting the flight control surface failures. This result can be further used for selecting algorithms for failure detection tasks for other subsystems of aircraft in cases when the system knowledge and expertise are lacking and thus the feature extraction that can be done is little.

1 Introduction Feature extraction is a necessary preliminary step when performing failure detection algorithmically from high dimensional data. The need of feature extraction with high dimensional data is a consequence of the universal phenomenon called the curse of dimensionality [1]. One of the most efficient way of avoid the curse of dimensionality is to perform feature extraction as a preliminary step of data analysis. Besides lowering the effect of the curse of dimensionality, another benefit of doing feature extraction is making the data analysis computationally less heavy. The curse of dimensionality is present in all data analysis tasks where the data is high dimensional. In practice the curse of dimensionality is because the fact that having an increase in the features space dimensionality will increase the relative sparseness of samples in the feature space. The”curse” is due to the exponential increase in sparseness relative to the increase of feature space. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 182–194, 2023. https://doi.org/10.1007/978-3-031-25448-2_18

The Effect of Knowledge Based Feature Extraction

183

The effect of feature extraction has been studied also by others. In Yang et al. [2] feature extraction was done on the data of esophageal X-ray images. The feature extraction process in [2] was a sequential forward selection algorithm (SFS) and principal component analysis (PCA). It was stated in the results that”The step of feature selection not only reduces the dimension of the input vector, but also improves classification performance.”. The feature extraction approach studied in [2] was algorithmic not knowledge based, but the findings will in some level agree with the findings of this study. System knowledge based feature extraction is an approach of feature extraction where feature selection and feature combination is done based on the prior knowledge of the system in hand. The feature extraction process lowers the dimensionality of the data and in that way dilutes the problem of the curse of dimensionality. Occasionally a system might be extremely complex, and an expert knowledge for a data analyst is not available, thus making the system knowledge based feature extraction almost impossible to perform. In these cases, the data analyst is forced to proceed directly on algorithmic approaches. Currently there is not much knowledge available about which algorithms are robust with poorly extracted data and thus would be preferable in this type of situation. The aim of this study is to discover the algorithms, from the fields of anomaly detection, novelty detection, one class classification and unsupervised machine learning that are suitable for failure detection from high dimensional system operational data, and which are robust with low level of feature extraction. In this study the operational data of the fighter aircraft has been utilized to construct a several versions of a data with different level of knowledge based feature extraction. A set of selected algorithms suitable for detecting failures of control surface has been tested with the data. The validation of the results has been done by comparing the algorithm output to the carefully selected preliminary periods of flight control surface seizure failures from real life flight cases. As a result, it has been found out that kernel based algorithms are more robust with low level of feature extraction compared to other algorithms. This finding is useful in cases where there is a little system knowledge present for performing feature extraction and simultaneously there is a lack of data describing potential failures for measuring performances of algorithms.

2 Methods for Failure Detection The aim of the failure detection is to detect failures beforehand. Thus in practice the target of failure detection is not to detect failures but rather potential failures. The potential failure is a state of a system that is functioning but process of significant degradation is present in such level that the near future failure is evident. When doing online monitoring the line between a healthy functioning system and a system with potential failure is vague, on the other here where a data of a failed system is analyzed afterwards, the state of potential failure is unambiguous. Potential failure here means the time period that ends to the failure and starts from reasonable time before failure while the system data is indicating abnormal system state. The data available for this study is from one single fighter aircraft (see Table 1). This specific aircraft has been chosen as a data source of this study since it has suffered three separate cases of flight control surface failures during its flight history and thus it has data samples presenting each of the following three data classes:

184

T. Toikka et al.

1. Normally functioning system 2. Potential Failure in functioning system 3. Failed system The functioning system is defined here as the system that does its task without causing any operating problems, warnings in existing monitoring system or suspicions in pilot. The main characteristics of the data used in this study is described in Table 1. Table 1. General description of the original data of this study. System

Fighter aircraft

Number of fights

40

Number of failure flights

3

Failure type

Flight control surface seizure failure

Feature space size

>104

Recorded samples

>107

Nature of recording

Time series with constant frequency

2.1 Knowledge Based Feature Extraction Process Feature extraction is a process of generating relevant parameters for data analysis algorithm from high dimensional original data. The process is usually parameter selection but can be also a process of generating new parameters. Knowledge based feature extraction is a process of doing parameter selection and generation based solely on the system knowledge and expertise. This is in contrast to the algorithmic feature extraction process where relevant parameters are found by using dimensionality reduction algorithms like PCA. The size of the feature space of the original data (described in Table 1) was several thousands of recorded parameters and thus not directly applicable for any data analyzing algorithm due to curse of dimensionality [1]. For this reason, the data was extracted from the original data by the following way. Step 1: (D26) By the expert of fighter aircraft systems in issue the dataset feature space was extracted to 26 parameters, by selecting parameters which may in any imaginable way be related to the control surface failures, while not being strict with irrelevant parameters for demonstrative reasons. Those parameters are listed in Table 2 in column D26. This dataset will be referred here as D26. Step 2: (D13) The D26 was further extracted to 13 parameters by selecting environmental parameters effecting to the control surface operation and operational features directly related to the control surface operation. The parameters are listed in Table 2 in column D13 and the dataset will be referred here as D13. Step 3: (D6) The D13 was further extracted to 6 parameters by selecting only operational features of control surfaces. The parameters are listed in Table 2 in column D6 and the dataset will be referred here as D6.

The Effect of Knowledge Based Feature Extraction

185

Table 2. The parameter selection steps in feature extraction process. Parameters labelled with * are included. Measured parameter

D26

Time

*

Left power leaver angle

*

Right power leaver angle

*

Left engine inlet temperature

*

Right engine inlet temperature

*

Left compression pressure

*

Right compression pressure

*

Left drain air temperature

*

Right drain air temperature

*

D13

Left low rotor speed

*

Right low rotor speed

*

Left high rotor speed

*

Right high rotor speed

*

Dynamic pressure (I)

*

*

Dynamic pressure (II)

*

*

Static pressure

*

*

Ambient temperature

*

*

Barometer corrected pressure

*

*

Pressure altitude

*

*

Air speed

*

*

D6

Leading edge flap position command

*

*

*

Leading edge flap differentiating command

*

*

*

Left inner leading edge flap position

*

*

*

Right inner leading edge flap position

*

*

*

Left outer leading edge flap position

*

*

*

Right outer leading edge flap position

*

*

*

Step 4: (D4) The D6 was further extracted by creating 4 new parameters. Those parameters were error positions compared to command signals of all four leading edge flaps. This dataset will be referred here as D4. Step 5: (D2) Finally the D4 was further extracted by creating 2 new parameters that were differences in error positions between the inner and outer leading edge flaps of the same side. This dataset will be referred here as D2. The characteristics of the datasets used for this study are summarized in the Table 3. The process of the feature extraction here is more or less heuristic, but as will Fig. 1

186

T. Toikka et al.

Table 3. Descriptions of the test datasets of this study. The number as a suffix of D in naming refers to the feature space dimensionality of the dataset. Dataset name Extracting procedure D26

Extracted from the original dataset of Table 1 by a parameter selection

D13

Extracted from the D26 by selecting environmental and operational parameters related to control surface operation

D6

Extracted from the D13 by selecting only operational parameter related to control surface operation

D4

Extracted from the D6 by combining parameters

D2

Extracted from the D4 by combining parameters

demonstrate it has led to a proper solution. On the other hand, in the scope of this study the actual feature extraction process is irrelevant, since whatever it is, the Fig. 1 proves that it works, and now we can now start to analyse how it effects on the performance of algorithms, that is the aim of this study. The result of D2 is illustrated in Fig. 1 after applying the related eigenvalues. The resulting two parameters of the dataset are somewhat abstract but the resulting figure demonstrates the fact that data of potential failure of the system is obviously out classifiable from the normal data.

Fig. 1. Two dimensional presentation of the fighter aircraft flight data describing functional flight control surface operation. Two parameters are achieved by doing system knowledge based feature extraction on the high dimensional original data of Table 1

The Effect of Knowledge Based Feature Extraction

187

2.2 Failure Detection Algorithms In this study a set of selected algorithms, estimated to be potentially capable for failure detection, have been used to estimate the effect of feature extraction rate on the performance of the algorithms. The set of algorithms used in this study are: Density-based spatial clustering of applications with noise (DBSCAN) is a distance based algorithm capable of detecting close backed clusters [7]. During the detection the data sample excluded from the clusters can be considered as an anomalies or novelties in the data. The anomalies in the system data may indicate potential failure. Euclidean distance (E-dist) method measures Euclidean distance between the sample of the examination against to an all training data [4]. The distances exceeding a threshold value are labeled as outliers. An ongoing stream of outlier may indicate a potential failure. Gaussian mixture model (GMM) estimates the underlying probability distribution of the data as a combination of component distributions [9]. Data samples that have low probability according to probability density of the whole data are considered as outliers. A constant stream of outliers may indicate a potential failure. Kernel Density Estimate (KDE) is a nonparametric density estimator formed by placing kernel on each data point and summing local contributions of each data points [9]. Data points appearing in regions of low densities are considered as outliers. A continuing stream of outliers may indicate a potential failure. Kernel Principal component analysis (K-PCA) extends traditional principal component analysis (PCA) by mapping an input data on higher dimensional space compared by the original dataspace by using kernel function [9, 11]. A novelty of the test data is measured by the reconstruction error of a test point with respect to subspace of principal components. A constant stream of novel data may indicate a potential failure. k-means is a clustering method for finding a present number of k clusters from a data. Cluster membership of the data sample is determined by some distance metric (usually Euclidian distance) [4, 9]. If test data has high distance to all k cluster centers it may be considered as an outlier. Stream of outliers may indicate novel state in the system that may further indicate a potential failure. Nearest Neighbour (NN) is a distance based method. The underlying assumption of the method is that normal data point is close to its predefined number of nearest neighbours while the abnormal data sample is far from its nearest neighbours according some selected distance measure (usually Euclidian distance) [4, 9]. A stream of abnormal data samples may indicate a potential failure. One-Class Gaussian Process (GPOC) is a probability density estimation method that models the probability of random process [12]. The data samples of the test data are assumed to be random Gaussian variables. Outliers are the variables with low probability. A stream of outliers can indicate a potential failure. One-class Support Vector Machine (SVM) is a modification of SVM that is a kernel based method that finds a maximum margin hyperplane to separate two classes. In contrast in One-class SVM normal data samples and noise are separated [4, 13–15]. If a stream of test data samples is classified on the side of the noise of the hyperplane the stream may indicate a potential failure. Parzen window (PW) is a special version of Kernel Density Estimate where the variance of the kernel (i.e. window size) is predetermined [4, 9]. Test data samples in

188

T. Toikka et al.

regions of low probability density are considered as outliers. A stream of outliers may indicate a potential failure. Principal Component Analysis (PCA) finds a highly correlated components of a multidimensional data which are called as principal components [9]. If a test data correlates highly with the components of lower end principal components, then the test data may be considered as anomalous. An anomalous data may indicate potential failure. Self-organizing maps (SOM) is a special type of neural network that can be trained as an unsupervised manner [9]. In SOM the distances of the neurons can be measured by the Euclidian distance manner. Test data samples that inhibit the neurons with great distance to other neurons may be considered as an anomalous data samples. A stream of anomalous data samples may indicate a potential failure. Detailed descriptions about the algorithms used in this study can be found from the references listed in the Table 4. The table also refers to acquisition sources of the algorithms. Table 4. ND algorithms used in this study. Algorithm name

Referred here as

Described in

Available at

Density-Based Spatial Clustering of Applications With Noise

DBSCAN

[7]

[8]

Euclidean Distance

E-dist

[4]

[6]

Gaussian Mixture Model

GMM

[9]

[6, 10]

Kernel Density Estimate

KDE

[9]

[6]

Kernel Principal Component Analysis

K-PCA

[9, 11]

[6]

k-means

k-means

[4, 9]

[6, 10]

Nearest Neighbour

NN

[4, 9]

[6, 10]

One-Class Gaussian Process

GPOC

[12]

[6]

One-class Support Vector Machine

SVM (A)

[4, 13]

[6]

SVM (B)

[4, 14]

[6]

SVM (C)

[15]

[16]

Parzen window

PW

[4, 9]

[6]

Principal Component Analysis

PCA

[9]

[6, 10]

Self-Organizing Maps

SOM

[9]

[6, 10]

3 Results of Failure Detection with Varying Feature Extraction The performances of the algorithms used in this study are estimated based on the following metrics:

The Effect of Knowledge Based Feature Extraction

189

• Accuracy: The rate of data samples classified correctly when the assumption is that PFs should be classified as novel data/anomalies/outliers and normal data should be classified as a normal data. • Sensitivity: Rate of classifying PFs as a novel data/anomalies/outliers samples. Property of 1 - Sensitivity describes the rate of missing the relevant information (that is PFs). Table 5: Classification accuracy, specificity, and sensitivity of algorithms for various datasets having a various level of feature extraction done. • Specificity: Rate of normal data samples seen as normal. Property of 1 - Specificity describes false alarm rate. The metrics of accuracy, sensitivity and specificity was chosen because they are generally widely used in literature and they were achievable for all of the algorithms of this study. The results have been summarized in Tables 5. Table 5. Classification accuracy specificity and sensitivity of algorithms for various datasets having a various level feature extraction done. Accuracy [%]

Specificity [%]

Sensitivity [%]

Algorithm D26 D13 D6 D4 D2 D26 D13 D6 D4 D2 D26 D13 D6 D4 D2 DBSCAN

4

11

13 13

97

E-dist

50

50

50 50

64 100 100 100 100 100

0

7

51

55

92 100 100

GMM

51

60

60 62

97

3

20

45

62

KDE

87

92

69 72 100

86

95

54

60 100

89

88 83

K-PCA

94

66

50 50 100

99

99 100 100 100

89

34

0

0 100

k-means

50

50

50 50

57 100 100 100 100 100

0

0

0

0

NN

54

50

50 50 100 100 100 100 100 100

8

0

0

0 100

GPOC

91

98

95 94 100

81

97

96

96 100 100

99 95

92 100

SVM (A)

96

97

94 99

99

97

94

96

98

97

95 100 91 100 100

SVM (B)

97

99

95 98

99

96

97

97

96

97

98 100 92

99 100

SVM (C)

90

93

75 88

97

87

92

56

82 100

94

95 95

95

77

89 80

0

0

4

3 100

0

0

94 100 100 76

PW

85

87

66 70

98

93

84

51

80

PCA

62

89

57 57

57

23

78

97

98 100 100 100 17

96

SOM

92

91

51 52

98

90

90

97

97

96

93

92

5

27

62 100 84 100 14

93

60 100 15

14

8 100

From Table 5 it can be seen the improving effect of feature extraction on accuracy in case of algorithms, the accuracy improves when advancing from column D26 towards the column D2. The effect is specially clearly visible in case of DBSCAN algorithm:

190

T. Toikka et al.

The accuracy is 4% with D26, it improves to 11% with D13, it further improves to 13% with D6, it remains with D6 and it further improves to 97% with D2. The row of DBSCAN in Table 5 also shows that the DBSCAN algorithm was in this specific case highly dependent of feature extraction since it does not perform good until the extraction level of D2. The poor performance of DBSCAN algorithm with the data of low feature extraction can be explained by that the data with low feature extraction is high dimensional. In high dimensional space the relative distances between the data points are higher compared to the low dimensional space. Due to the high distances between the samples the DBSCAN algorithm probably did not form enough clusters but rather classified almost everything as a noise. In contrast to DBSCAN algorithm, that was highly dependent on feature extraction we may examine SVM algorithms from the Table 5. All SVM algorithms have great accuracy even with the data having low level of feature extraction, indicating that SVM’s are much less dependent of feature extraction than other algorithms. This may be explained by the kernel function and kernel trick involved in SVM algorithm that maps the data of examination to high dimensional space for classification task. SVM also outperforms other kernel based algorithms like KDE and K-PCA probably because the optimization approach of solution finding, that is a central part in SVM algorithm. Specificity is an important measure in context of failure detection, since low sensitivity implies high false alarm rate. Failure detection system are for human use and humans are not tolerable with high level of false alarm rate since it means for running in vain. Systems with high false alarm rate are probably ignored and forgotten soon in practice. In Table 5 some algorithms like E-dist, k-means, NN, and all SVM’s did perform well measured in specificity. This result need to be but in to the prospect with accuracy since there is no use of specificity of the algorithm if it is not accurate. E-dist, k-means and NN have poor accuracy and thus they are not good for failure detection in this case even thou the specificity is good. All SVM’s have good accuracy and good specificity with all levels of feature extraction and thus they seem to fit well for failure detection in this case. Sensitivity is not that relevant measure compared to accuracy and specificity in context of failure detection, since by classifying everything as a potential failure will lead to 100% sensitivity. This can easily be achieved by constructing a detection system that does simply nothing but have a warning light always on. On the other hand, those algorithms that have 0% of sensitivity are not capable of detecting potential failures at all. The conclusion is that sensitivity must be up but the measure alone does not tell much. Sensitivities of the algorithms are listed in Table 5. The mean accuracy, sensitivity, specificity and times related to each test datasets are presented in Fig. 2. The mean accuracy, sensitivity, specificity and times related to each algorithms used here are presented in Fig. 3. In the Fig. 4 it is illustrated how the performance of some algorithms are strongly affected by the FE and some are not.

The Effect of Knowledge Based Feature Extraction

191

Fig. 2. Mean values for accuracy, sensitivity, specificity and times related to each test dataset.

Fig. 3. Mean values for accuracy, sensitivity, specificity and times related to each algorithm used in this study.

In Table 5 values appear below 50% The value 50% can be considered as an absolute minimum requirement for a realistic novelty detector, since the random detector (i.e. flip a coin) would produce average of 50% for accuracy, sensitivity and specificity. The implementation of some algorithms does require a prior decision about the use of its free parameters. For example, DBSCAN requires a prior decision about the minimum quantity of neighbours and the neighbour radius. On the other hand, One-class SVM requires a selection of a kernel function, kernel parameters and the share of the training data that will be classified as novel/anomalous during training. In this study the

192

T. Toikka et al.

Fig. 4. The accuracy, sensitivity and specificity for all the algorithms with the datasets of D26 (blue bars), D6 (green bars) and D2 (yellow bars).

parameter selection was not optimized for any algorithm, but instead the default values implemented by the algorithm original programmer was used. The general view about the feature extraction process in the novelty detection and the machine learning literature is that the feature extraction should be performed before the actual implementation of an algorithm. This view seems to be in some level true also based on the results of this study, since the highest accuracy was achieved with the D2 (furthermost extracted) data in case of all algorithms (see Table 5 and Fig. 2). The highest sensitivity and specificity in average were also achieved with the given D2 (see Fig. 2). The general view about the improving effect of feature extraction on classification performance was confirmed here, but there also seem to exist a counter effect, which can be observed from Fig. 2. Based on the results in Fig. 2, the second highest results in accuracy was achieved with D13. Thus the feature extraction from D13 to D6 and further to D4 did lower the performance of classification algorithms. The actual feature extraction process from D26 to D13 and further to D6 was the feature selection. Thus it seems that while the prior knowledge was added to the data via feature extraction, some of the information of the data was lost and specially in this case, the negative effect of the information loss was greater than the positive effect of the feature extraction. The interesting finding here is that there was a great difference between the algorithms when measuring the feature extraction effect on the algorithm performance. For example, by observing the accuracy in Fig. 4, it can be noticed that, SOM and K-PCA are suffering for information loss from D26 to D6 but further benefited by the feature extraction from D6 to D2. The general trend is that feature extraction here has an improving effect on potential failure detection performance, but on the other hand the performance some algorithms, for example SVMs, KDE and GPOC, has not been significantly affected by the feature extraction.

The Effect of Knowledge Based Feature Extraction

193

SVM seemed to be performing well with our data compared to other algorithms, based on our performance metrics (accuracy, sensitivity and specificity) in case of all three versions of SVMs. This finding is in line with the findings of Clifton L. A. study about Multi-channel novelty detection [4] where it was stated that”The one-class SVM classifiers outperform density models”. When comparing the performances of the algorithms of this study, it is worth for note that some of the algorithms may by chance perform better with this specific set of data used in this study. On the other hand, by observing Fig. 4 it seems that kernel based algorithms seem to be able to encapsulate essential characteristics from the data without system knowledge based feature extraction. The fact that kernel based algorithms did perform well with the dataset D13 (see Table 5) proves that they have been able to encapsulate some essential information concerning the environmental conditions which we were not able to discover based on our system prior knowledge. On the other hand, most of the kernel based algorithms did perform almost as equally effectively with D26 as D13 which proves that they have not been distracted by the data that is presumably non phenomenon related. The success of kernel based algorithms can be explained by their kernel function. This function maps the input data to a high dimensional feature space. In this space the class separation of data instances between novel data and normal data becomes more apparent compared to a class separation task in the original feature space. The success of SVM algorithms over other kernel based algorithms here may be explained by the internal optimization approach of the SVM algorithms. This study is in line with the general view of the improving effect of feature extraction when using classification algorithm, but on the other hand here have been found out that feature extraction may generate the counter effect when Feature Extraction process is only a feature selection due to loss of information of data. It has been also found out here that the kernel based algorithms are less sensitive for feature extraction and can perform well with low level of feature extraction against non-kernel algorithms. Kernel based algorithms seem also be less sensitive to the presence of non-phenomenon related features in the data. These facts positions kernel based algorithms in advance against non-kernel algorithms in practical cases where system prior knowledge might be minor.

4 Conclusions Classification methods categorized as novelty/anomaly/outlier detection methods are providing a useful tool when detecting anomalous behaviour and potential faults in the system. Algorithms need a data but the data is usually high dimensional in many practical cases. High dimensional data will expose classification algorithms to curse of dimensionally. In order to reduce the effect of curse the expert knowledge of the system may be used to perform a feature extraction. In this study, the effect of system knowledge based feature extraction was studied with comprehensive set (total 14) of classification algorithms using the data from fighter aircraft. The main finding was, kernel based algorithms outperformed non-kernel algorithms and SVM outperformed other kernel based algorithms with the data having low level of feature extraction. It was also discovered that, in some cases the feature extraction did reduce the performance of classification algorithms.

194

T. Toikka et al.

The findings of this study will provide value for failure detection tasks where the failure data is not available for validating the performance of failure detection algorithm and the data analyst has a little of system knowledge available for performing feature extraction. This study did utilize a comprehensive set of classification algorithms suitable for failure detection algorithms. Limitations of this study are that each algorithm has been coded by different people and has been optimized in different level. Also the set does not cover all algorithms suitable for the task and certainly not all versions of the algorithms available. Thus similar study could be carried out with a different set of algorithms. The test data utilized in this study comes from one specific aircraft system and the data instances presenting the failures are representing one specific failure mode i.e. control surface seizure failure. In future in order to validate the result in a larger scale, additional studies are needed with data of different systems and with different failure modes.

References 1. Bishop, C.M.: Neural Networks for Pattern Recognition. Oxford University Press, Oxford (1995) 2. Yang, F., Hamit, M., Yan, C.B., Yao, J., Kutluk, A., Kong, X.M., Zhang, S.X.: Feature extraction and classification on esophageal x-ray images of xinjiang kazak nationality. J. Healthcare Eng. 2017 (2017) 3. McBain, J., Timusk, M.: Feature extraction for novelty detection as applied to fault detection in machinery. Pattern Recogn. Lett. 32(7), 1054–1061 (2011). http://www.sciencedirect.com/ science/article/pii/S0167865511000341 4. Clifton, L.A.: Multi-channel novelty detection and classifier ombination. Ph.D. dissertation, University of Manchester (2007) 5. Abu-Mostafa, Y.S., Magdon-Ismail, M., Lin, H.-T.: Learning from data: a short course. [United States]: AMLBook.com, cop. (2012) 6. Pimentel, M.A., Clifton, L., Clifton, D.A., Tarassenko, L.: Ndtool - a novelty detection toolbox. University of Oxford, February 2014. http://www.robots.ox.ac.uk/~davidc/publications NDtool.php 7. Wikipedia, “Dbscan — wikipedia, the free encyclopedia,” (2017). https://en.wikipedia.org/ w/index.php?title=DBSCAN&oldid=798606463 8. Dbscan clustering algorithm. Yarpiz (2015). https://se.mathworks.com/matlabcentral/fileex change/52905-dbscan-clustering-algorithm 9. Nabney, I.: NETLAB: Algorithms for Pattern Recognition. Springer, London (2002) 10. Nabney, I.: Netlab, December 2002. https://se.mathworks.com/matlabcentral/fileexchange/ 2654-netlab 11. Hoffmann, H.: Kernel pca for novelty detection. Pattern Recognition 40(3), 863–874 (2007). http://www.sciencedirect.com/science/article/pii/S0031320306003414 12. Kemmler, M., Rodner, E., Denzler, J.: One-class classification with gaussian processes. In: Kimmel, R., Klette, R., Sugimoto, A. (eds.) ACCV 2010. LNCS, vol. 6493, pp. 489–500. Springer, Heidelberg (2011). https://doi.org/10.1007/978-3-642-19309-5_38 13. Tax, D.M., Duin, R.P.: Data domain description using support vectors. In: ESANN, vol. 99, pp. 251–256 (1999) 14. Sch¨olkopf, B., Platt, J.C., Shawe-Taylor, J., Smola, A.J., Williamson, R.C.: Estimating the support of a high-dimensional distribution. Neural Comput. 13(7), 1443–1471 (2001) 15. Support vector machines for binary classification. MathWorks. https://se.mathworks.com/ help/stats/support-vector-machines-for-binary-classification.html 16. Products and services. MathWorks. https://se.mathworks.com/products.html?stid=gnps

Advanced Maintenance of Distribution Assets Through the Application of Predictive Techniques Using GE’S APM System: Real Case in a Spanish DSO Sergio Bustamante1(B) , Mario Manana1 , Alberto Arroyo1 , Antonio González2 , and Richard Maurice3 1 Electrical and Energy Engineering Department, Universidad de Cantabria, Santander, Spain

{bustamantes,mananam,arroyoa}@unican.es 2 EDP Redes España, Oviedo, Spain [email protected] 3 GE Digital, Massy, France [email protected]

Abstract. EDP Redes España applies advanced maintenance techniques such as prediction and assets monitoring, in order to increase the assets reliability, minimize operational expenditures (OPEX), and reduce the operational risks, as well as optimizing capital expenditures (CAPEX) in assets with high replacement cost. Predictive techniques, that integrate the asset’s health status, criticality, and aging models, can help to achieve the objectives, but the selection of the predictive maintenance system is not an easy task, given the limited experience that exists in the distribution business. The solution selected by EDP Redes España was the one provided by General Electric (GE), the Asset Performance Management (APM) system. APM software aims to assist EDP Redes España in integrating its extensive data ecosystem into a single system and optimize operational work processes. This paper presents the implementation of the APM system in the assets of EDP Redes España, which, as a reward for the great effort made, valuable results are beginning to be obtained for the advanced management of asset maintenance.

1 Introduction EDP Redes España is a Spanish distribution system operator (DSO) that owns 151 highvoltage (HV) substations, 118 medium-voltage (MV) substations, 17.8k transformation centres, and a network of 52,256 km, and distributes electricity to almost 1.4 M customers. Therefore, asset maintenance management is a complex task due to the large number of assets and the importance of each one of them within the network. EDP Redes España has been applying advanced maintenance techniques with the aim of increasing asset reliability, minimising operational expenditures (OPEX) and reducing operational risks, as well as optimising capital expenditures (CAPEX) in assets with high replacement costs. Predictive techniques that integrate asset health index, criticality, and ageing models help to achieve the objectives. Based on the 8-phase Maintenance © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 195–204, 2023. https://doi.org/10.1007/978-3-031-25448-2_19

196

S. Bustamante et al.

Management Model (Parra and Crespo 2015) shown in Fig. 1, the application of asset health index calculation techniques is part of Phase 7 – Asset life cycle analysis and replacement optimization.

Fig. 1. Maintenance management model (Parra and Crespo 2015).

Asset health index calculation techniques are tools that measure the condition of an asset and how close it is to the end of its useful life. In other words, the asset health index helps to know the current condition of a physical asset, allowing to: • Compare the health of equipment located in similar functional locations. • Study possible premature deterioration and to optimise operation and/or maintenance plans of the assets if necessary. • Understand the behaviour of assets from different manufacturers in specific functional locations. • Support decision-making processes for future asset investments or asset life extension. Predictive techniques, that integrate the asset’s health status, criticality and aging models, can help to achieve the objectives, but the selection of the predictive maintenance system was not an easy task, given the limited experience that exists in the distribution business. Several studies (Coullon and Rego 2014) (Wan 2017) were published and reviewed at the beginning of the project (end of 2018). These studies analyse digital transformation and present early stage case studies of asset management implementation. Another study (Vila and Pérez 2019) develops a model for more efficient asset management through the application of advanced analytical techniques. In this study, the most important steps for the development of the system are presented and explained. More recently, two companies have entered into an agreement to market a platform that brings together experience and expertise in electricity infrastructure management and

Advanced Maintenance of Distribution

197

the potential in artificial intelligence (Elewit & IBM, 2020). In addition, over time, several companies have developed APM systems focused on the electricity business, such as Hitachi (Hitachi 2021), ABB (ABB, 2021), or Siemens (Siemens 2021), or others that can be applied to this business such as SAP Hana or IBM Maximo. The solution selected by EDP Redes España was the one provided by General Electric (GE), the Asset Performance Management (APM) system (GE Digital, 2022). This paper presents the implementation of the APM system in the assets of EDP Redes España, which, as a reward for the great effort made, valuable results are beginning to be obtained for the advanced management of asset maintenance.

2 GE’s APM GE’s APM is a suite of integrated solutions that can be used independently or together to provide a comprehensive approach to help optimise asset performance, increasing asset reliability and availability, while optimising asset maintenance and replacement costs, and mitigating operational risks. The health module of APM applies asset health index calculation techniques and provides results for each asset. APM Health manages data from a wide variety of assets and systems, creating a comprehensive data repository. It also gives insight into asset status and condition, and provides early warning of potential failures. The health indicators calculated to assist in the decision-making process are: • • • • •

Asset health index (AHI): Health status of the asset. Estimated residual life (ERL): Estimated remaining useful life of the asset. Asset criticality index (ACI): Impact in case of failure of a given asset. Probability of failure (POF): Probability that a given asset will fail. Asset risk index (ARI): An estimate of the actual risk to which an asset is exposed if maintenance actions are not taken. ARI = ACI * POF. • Asset maintenance index (AMI): Maintenance status of the asset and identifies the maintenance actions to be performed in the short term. • Completeness index (CPLI): Completeness of the asset health model. The algorithms implemented in the system that allow the above health indicators to be calculated are called policies. As there are different assets within the distribution network, each asset type has a different policy defined in the system. These asset policies have two different approaches. In the first one (complex model), the degradation process of the asset is extensively described, analysed, and reported by main standardisation bodies (IEC, IEEE, CIGRE), in addition, there is a lot of data obtained from inspections, online or offline monitoring, and tests. So this approach is based on the condition of the assets. In the second approach (simple model), there are assets where there is insufficient data to define the parameters and their measurements that drive end-of-life or, when they exist, obtaining the necessary measurements is not cost-effective. So this approach is based mainly on the age of the assets. Figure 2 shows the differences in the calculation of health indicators between the approaches, and the assets that belong to each approach. From the results of the maintenance ranges and using the limits set in the policy, the level 2 condition parameters (CP2) are obtained. For the calculation of the level 1

198

S. Bustamante et al.

Fig. 2. APM policy approaches.

condition parameter (CP1), all active CP2s of the same family and their weights (wCP2i ) are used, and for the calculation of the AHI, all CP1s and their weights (wCP1i ) are used. The CPLI calculation relates the weights of the active CP2s (activewCP2i ) and the total weights of the CP2s. The equations used are as follows: n CP2i × wCP2i n CP1(%) = i=1 (1) i=1 wCP2i n CP1i × wCP1i n AHI (%) = i=1 (2) i=1 wCP1i n activewCP2i n CPLI (%) = i=1 × 100 (3) i=1 wCP2i As an example, the policies of two of the assets implemented in the system will be shown below. These assets are the power transformer and the distribution transformer. The function of both is to transform an AC voltage and current system into another AC voltage and current system, usually of different values. Their differences, in the case of the distribution network presented, are the voltage level at which they operate and their power rating. On the one hand, the voltage levels on the HV side of power transformers range from 30 to 400 kV (220 and 400 kV correspond to power transformers connected to transmission lines), with power ratings from 1 to 450 MVA. On the other hand, the voltage levels of distribution transformers on the HV side ranges from 6.6 to 30 kV, with power ratings from 25 to 1000 kVA. Another difference is the size of the assets, power transformers have a large size compared to distribution transformers. Therefore, their replacement costs are very different; Fig. 3 shows the approximate replacement costs of power and distribution transformers as a function of power rating.

Advanced Maintenance of Distribution

199

Fig. 3. Approximate replacement costs of distribution transformers (left) and power transformers (right).

The power transformer (Fig. 4), as one of the most complex and high replacement cost assets in the distribution network uses the complex model in its policy. It may contain a sub-asset, the on-load tap-changer (OLTC), which has its own defined policy. As a critical grid asset, the power transformer usually has continuous monitoring equipment (including oil temperature, gas concentration, and water content in the oil). The monitored data can be managed from GE Digital’s Predictive Analytics and can be used from the APM health module to update health indicators. In addition, within a single asset policy, there may be variants depending on the asset typology, so that all assets and their typologies are covered by the policies. The power transformer policy is shown in Fig. 5. In the case of power transformers, their policy has differences in five CP2s (in red in Fig. 5) from three different CP1s. The percentages next to each CP1 correspond to the weight used in for each of them. The percentages in the green box correspond to power transformers with OLTC, while the percentages in the grey box are those without OLTC. The differences in the limits of the three oil quality analysis CP2s depend on the voltage level of the power transformer as indicated in the IEC guide (IEC 2013). In this guide, there are three voltage levels for power transformers belonging to the DSO, so there are three possible variations of the policy due to these CP2s. The different limits for acetylene are based on whether or not it has OLTC and whether or not there is oil communication between the main tank and the OLTC compartment (Bustamante et al. 2020) (IEC 2015). Finally, the limits used in the bushing power factor (PF) depend on the bushing type (CIGRE 2019). These limit differences mean that there are twelve variants of the same policy, which are duplicated depending on whether it has OLTC. In contrast, distribution transformers have the simple policy model, as their large number makes it not cost-effective to obtain measurements. Figure 6 shows the distribution transformer policy. Unlike the power transformer policy, this one has fewer CP2s as they are obtained from the regulatory inspection. In addition, age is not used for the calculation of HI but for ERL and POF. Based on the results obtained from each CP2, short-term maintenance recommendations are generated. These maintenance recommendations have three alert levels, critical, urgent, and non-urgent. The AMI value takes the value of the active maintenance recommendation with the highest alert level.

200

S. Bustamante et al.

Fig. 4. Power transformer in service.

3 APM Implementation Process and Results Over 12k assets were selected for phase 1 implementation in the APM system. These assets were chosen because of their criticality within the distribution network, replacement costs, or because of the large number of them. For phase 2, 8.7k assets were selected, following the same procedure. Table 1 shows the assets selected in each phase and the percentage of assets of each type implemented in the system. The OLTC is an asset with its own policy and a sub-asset of the power transformer. The current target is to reach 100% of power transformers, OLTCs, HV underground cables, and NiCd and Pb battery-chargers by 2022. After the implementation of the assets in phases 1 and 2, health indicators were obtained for all assets. Figure 7 shows the AHI result as a function of age for six types of assets.

Advanced Maintenance of Distribution

201

Fig. 5. Power transformer policy.

All of them are above 50% except for two assets of type 6 and one of type 5. The two type 6 assets were inspected and revealed a fault, and the type 5 asset is a decommissioned asset so the low AHI is normal. Type 7 assets were not included in Fig. 7 because of the large number of assets. Figure 7 shows how the assets of the simple model (asset types 1 and 3) have AHI values close to 100%; this is because the inspections carried out on these assets are the regulatory ones, so there is little variation in their CP2 results. Figure 8 shows the number of assets by type and AMI value. As the AMI is generated from the highest alarm level of the active recommendations, Fig. 8 also shows the number of short-term recommendations with the highest alarm levels for each of the asset types. Based on the short-term maintenance actions generated by the system, the AHI results, and the criticality of the assets within the network, it is possible to plan and optimise the maintenance work on the network assets. As the system also keeps the history of maintenance actions, this allows prioritising the investment in those assets with many actions carried out and that are close to the end of their useful life, not only taking into account their age.

202

S. Bustamante et al.

Fig. 6. Outdoor distribution transformer in service and its policy.

Table 1. System implementation process. Phase

Assets

Assets deployed

Policy model

1

Power transformer + OLTC

64%

Complex

1

Distribution transformer

63%

Simple

1

HV underground cable

61%

Complex

2

MV underground cable

54%

Complex

2

NiCd battery-charger

54%

Complex

2

Pb battery-charger

63%

Simple

CPLI is a key indicator for optimising asset replacement or asset life extension, as a low CPLI does not accurately indicate the true condition of an asset. Therefore, when optimising asset replacement strategies, it is necessary to assess the health indicators of assets with a high CPLI.

Advanced Maintenance of Distribution

203

Fig. 7. AHI results by age and asset type.

Fig. 8. Number of assets by AMI value and asset type.

4 Conclusions With the implementation of a predictive maintenance model, EDP Redes España aims to meet several challenges, such as improving decision-making in the investment of existing assets, optimising maintenance strategies, anticipating asset failure, and cost-risk-benefit balance. The selection of the system was a complicated task given the limited experience in the distribution business. The collection and implementation of the system assets was even more complicated than the selection of the system, due to the large number of assets and data for each asset. Thanks to the great effort made, the results of the health indicators were obtained, as well as the short-term maintenance actions. This allows the useful life of the assets to be extended without decreasing their reliability, and helps in making decisions on asset investments.

204

S. Bustamante et al.

References ABB: ABB Ability™ Genix Asset Performance Management Suite (2021). https://new.abb.com/ industrial-software/asset-performance-management. Accessed 18 Oct 2022 Bustamante, S., et al.: Determination of transformer oil contamination from the OLTC gases in the power transformers of a distribution system operator. Appl. Sci. 10(24), 8897 (2020) CIGRE: Condition assessment of power transformers, Paris: WG A2.49. Technical Brochure No. 761 (2019) Coullon, J.-L., Rego, N.: Flexible analytics for management of grid assets. In: 2014 Saudi Arabia Smart Grid Conference (SASG), pp. 1–5 (2014) Elewit & IBM: Press release - Elewit and IBM launch a pioneering solution to accelerate the digitalisation of the management of electricity network assets (2020). https://www.elewit. ventures/sites/webretit/files/paragraph/2020/11/file/20201112_NP_Elewit_%20IBM_ENG_ 0.pdf. Accessed 18 Oct 2022 GE Digital: Asset Performance Management (APM) Software (2022). https://www.ge.com/dig ital/applications/asset-performance-management. Accessed 22 Apr 2022 Hitachi: Lumada APM - Asset performance management from the field to the boardroom (2021). https://www.hitachienergy.com/offering/solutions/asset-and-work-management/ lumada-apm. Accessed 18 Oct 2022 IEC: Mineral insulating oils in electrical equipment - Supervision and maintenance guidance, Geneva: IEC 60422:2013 (2013) IEC: Mineral oil-filled electrical equipment in service - Guidance on the interpretation of dissolved and free gases analysis, Geneva: IEC 60599:2015 (2015) Parra, C., Crespo, A.: Ingeniería de Mantenimiento y Fiabilidad Aplicada en la Gestión de Activos. Desarrollo y aplicación práctica de un Modelo de Gestión del Mantenimiento (MGM), 2nd edn. INGEMAN, Escuela Superior de Ingenieros Industriales, Sevilla (2015) Siemens: Asset Performance Management 4.0: Predict with confidence within the Digital Twin (2021). https://assets.siemens-energy.com/siemens/assets/api/uuid:9877421f-1180-4ffa-9451f0fe67354329/technical-paper-apm-4-0-e-.pdf. Accessed 18 Oct 2022 Vila, C., Pérez, F.: How Machine Learning can support Utilities in their current challenges: Iberdrola experience. Córdoba, Advanced Research Workshop on Transformers (ARWtr2019) Wan, S.: Asset performance management for power grids. Energy Procedia 143, 611–616 (2017)

Challenges on an Asset Health Index Calculation E. Candón1(B) , Adolfo Crespo Márquez1 , A. Guillén1 , and U. Leturiondo2 1 Department of Industrial Management, University of Seville, Seville, Spain

{ecandon,adolfo,ajguillen}@us.es

2 Control and Monitoring Area, Ikerlan Technology Research Centre, Basque Research and

Technology Alliance (BRTA), Pº J.M. Arizmendiarrieta, 2, 20500 Arrasate/Mondragón, Spain [email protected]

Abstract. The new era of Industry 4.0 highlights the challenge, and at the same time the problem, which involves an adequate data capture and a correct data processing to achieve the success of this revolution. The basis for a correct treatment of the data is based on an adequate capture of the data. In many companies this is a very complex problem, since the challenge is not the excellent capture of the data from the start-up of the analysed asset, but a minimum amount of data that allows us to process it properly or even to be able to estimate this from a solid and reliable base of information. The lack of information will mean a deviation in the processing of the data to be carried out, but it will be possible to compare assets of the same type that have the same problem in capturing and processing the data. For example, it will be possible to make a comparison of the health index of several transformers located in different electrical substations and under different operation regimes. If the data capture of the operation and maintenance variables are equally deficient and the same estimates are made, the health of these equipment can be compared. Throughout this paper, the example of calculating the Health Index of different pumps will be developed in which the start-up of these goes back to times prior to the date of capture of the operation and maintenance data. Due to this lack of information, it will be necessary to start from the estimation of different fundamental variables for the processing of the data to be calculated.

1 Introduction An Asset Health Index (AHI) is an asset score, which is designed, in some way, to reflect or characterize the asset’s condition and thus, its performance in terms of fulfilling the role established by the organization [1]. An AHI represents a practical method to quantify the general health of a complex asset. For simple assessments, Condition Based Maintenance (CBM) technologies can precisely estimate the status of a specific asset with defined and specifics failure modes. However, most of these assets are composed of multiple subsystems, and each subsystem can be characterized by multiple modes of degradation and failure. From a pure theoretical perspective, every failure mode of every item that composes a system can be modelled and estimated. In some cases, it may be considered that an asset has reached the end of its useful life, when several subsystems have reached a state of deterioration that prevents the continuity of service required by © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 205–216, 2023. https://doi.org/10.1007/978-3-031-25448-2_20

206

E. Candón et al.

the business [2]. This calculation can be complex and cause a significant investment in time and resources. It is in the case of complex systems where the health index, based on the results of operational observations, field inspections and laboratory tests, produces a single objective and quantitative indicator. It may be used as a tool to manage assets, to identify capital investment needs and maintenance programs, allowing [3–5]: 1) Compare the health of equipment located in similar technical locations, to study possible premature deterioration and optimize operation plans and/or asset maintenance if necessary; 2) Communicate more accurately with manufacturers/builders, to understand the behaviour of assets of different manufacturers/builders in specific technical locations; and 3) Support decision-making processes in future investments in assets, or in extension of the life of these [6]. Thus, AHIs are widely used in supporting maintenance and replacement strategies based on asset condition and performance in some countries, to justify asset replacement schemes to the regulators [7–9]. A proper design of a health index should meet the following requirements [10]: • The index must be indicative of the suitability of the asset to provide continuity to the service and representative of the general health of the asset. • The index should contain objective and verifiable measures of the condition of the asset, instead of subjective observations. • The index must be understandable and easily interpreted. Several methods and models fulfilling these requirements have been reviewed, for instance, the ones by Kinetrics [3, 4], DNV GL [11], Terna [12] and GB DNO [7]. Although most of these models build a streamlined approach to introduce different influent factors to estimate the lifetime expectation/remaining useful life of an asset, several drawbacks are still present in their model formulation: i)

The AHI procedure seems not to be properly robust from the scientific perspective, as original models are built mostly by practitioners in specific sectors with very specific assets. ii) Many influent factors are evaluated based on assumptions that are never discussed (e.g., ranges of numerical values are given as scales for different factors while it is almost unclear what is the basis to define such ranges). iii) The procedure proposed is mainly presented in its development and never demonstrated completely with, at least, some case-based reasoning or at the minimum a complete industrial case which would enable a proper validation of the AHI model proposed. There are approaches in the relevant literature to identify asset health [13] used mainly in CBM applications based on dynamic health assessment, but the concept is different from the one used in this paper, now the health assessment allows comparison and decision-making among different assets. To overcome these weakness points, in this paper the methodology adopted to model the AHI is only loosely based on the OFGEM Network Asset Indices Methodology [7] (a similar approach as in the example previously presented in [14]). This method is selected

Challenges on an Asset Health Index Calculation

207

because it is considered simple for simulation model building purposes and very practical in its implementation, if a more robust scientific design of the model format is reached. More precisely, the method [7] requires: 1) The identification of the asset, which includes the category of the equipment under study, the current age, the expected life, the name of the manufacturer/builder, the model of the equipment and the location of the installation; 2) The operation and maintenance data recorded during a certain period of time; and 3) The condition of the equipment, that is, the results of the analyses performed on the equipment in site, results of readings of physical variables, results of visual inspections, etc. The health index model adopted in this paper contains values between 1 and 10, thus being able to compare health between different types of assets. There are other indices that go from 0 to 1 and others that range from 1 to 100. In any case, they all have the same functionality: normalize the health of different assets to be able to compare them with each other.

2 AHI Modelling Methodology The application procedure for calculating the health index is based on 6 consecutive steps, in which, starting from a design life associated with an equipment’s category, a current health index is reached. For this, a series of factors related to the location, operation and condition of the asset are considered. It is presented in the following Fig. 1, the model, with the 6 steps for calculating the health index of an asset. For a precise description of the methodology of the AHI the reader is addressed to [15]. In addition, for a precise description of the mathematical formulation of the model, the reader is addressed to [17]. A synthesis of formulation is as follows in Fig. 1.

208

E. Candón et al.

Fig. 1. Procedure to calculate the AHI.

3 Case of Study The motors of two motor pumps of a power generation plant have been selected as real examples on which the proposed methodology has been applied. The calculation of the health index will make it possible to know the current condition of the assets, making it possible to compare them objectively with each other. This indicator will make it possible to prioritise interventions, care and/or the renewal of the assets analysed. 3.1 Application of the Methodology Proposed As is shown in Fig. 1, in step 1 a design theoretical life for every asset depending on the equipment category is defined. Its design life can be adapted by the owner according to accumulated experience and the information provided by different manufacturers and builders. In this case of study, it has been considered that it can operate 24 h a day, every day of the year and for 10 years, so that an estimated normal life of 87,600 h is left. In step 2, the estimated owner life can then be adjusted according to the characteristics of the asset location and loading. In the installation where the analysis is carried out, the assets are located indoors, which means that distance to the coast, altitude above sea level, annual average of outside temperature, exposure to corrosive atmosphere or exposure to dust in suspension are factors that do not almost affect the deterioration of the equipment. Therefore, the location factor (FE ) is considered to have no influence, i.e. it is equal to 1. The load factor (FEL ) measures the load request that is made on the asset in that location, in front of the maximum admissible load. In this case, the

Challenges on an Asset Health Index Calculation

209

variable selected to calculate the load factor is the flow rate. The values for nominal and maximum allowable flow rate are available in the pump operating manual, resulting in a load factor of 81%. Equation 1 shows the calculation of the estimated life of these pumps: Estimated life = tEL =

tDL 87, 600 = 108, 148 h = FFL × FEL 1 × 0.81

(1)

A fundamental hypothesis of the methodology is that the irreversible degradation of an asset follows an exponential behaviour with respect to its age, and in step number 3, the aging rate (β) of the asset is determined by the natural logarithm of the quotient between the asset health index when new (Hnew ) and the asset health index when reaching its expected life (Hestimated life ). The equation for its calculation is the following, used in step 3: β=

HI new ln HI estimated life

Estimated life

=

ln 0,5 5,5 tEL

(2)

Then, in step 4, the initial health index (HIit ) is considered as a dimensionless number between 1 and 10, with an exponential behaviour with respect to the age “t” of the asset, which is characterized by the aging rate as follows: HIit = HI new · eβ·t

(3)

The health index (HI) is the result of adjusting the initial health index, using load, health, and reliability modifiers. In a first step, the initial health index (HIit ) of an asset is modified to obtain what we call the real initial health index (HIi Real) in Step 5, considering the load modifier registered for the current age (ML (t)), using the following equation: β·t

HI i Real(t) = HI new · e ML (t)

(4)

where the load modifier is the quotient between the load factor existing at an instant (FRL (t)) and the expected load factor (FEL ); ML (t) = FEL /FRL (t)

(5)

The load modifier is a health modifier of the asset, which is considered in this initial phase since it is very likely that in many assets the load recorded during each asset age will be significantly different to the one initially planned for the functional location. The introduction of HIiReal then allows the current asset degradation to be adjusted to compare with the anticipated degradation for the functional location. Finally, in step 6, the health index of the asset is determined by its operating conditions and reliability conditions at the time of the evaluation. To determine the health index, the following equation is used: HI (t) = HI i Real(t)e

(MH (t)+MR(t))

(6)

210

E. Candón et al.

where: MH(t): is the asset’s health modifier (condition and operation). MR(t): is the asset’s reliability modifier. For the evaluation of the health modifier (MH) that appears in this last equation, the different variables that can be measured and quantified for each asset sub-category are considered, and that, being independent in their impact on health, add information about it. From the large number of variables available in the plant information system, it is necessary to perform data mining to determine which variables to select as pump health modifiers. To do this, RapidMiner Studio software was used to pre-process the available database, eliminating missing data and outliers, creating a single database, and to analyse the correlation between variables, thus allowing the most representative variables to be selected to determine the final health modifier. In this case, the health modifiers are composed of the operating parameters of speed, flow rate, suction pressure, discharge pressure and suction temperature. These variables are obtained in real time from the PI System. For the reliability modifier, depending on the sub-category of asset, model and manufacturer, tables can be prepared with the value of this parameter. In this case, the reliability modifiers are made up of the unavailability of the pump and the number of major maintenance or overhauls that are carried out. Once the proxy variables for health and reliability modifiers have been determined, the challenge is to determine how they impact on the health of the asset. To do this, it will be necessary to determine, within a range of [1; 1.4], how each of these variables affects the health of the pump in a particular way, considering a value of 1 as having no effect on health and 1.4 as having a 40% negative effect on the health of the pump. Likewise, once these ranges have been determined for the different operating thresholds of each variable, they are dimensioned and the modifiers MH j (t) and MRk (t) are calculated, respectively, in a range [0;1], which multiplied by the weights of each modifier will give rise to the variable modifier (see Eq. 7 and Equation 8). In order to determine the effect of the modifiers, the participation of the organisation’s expert group is necessary, which, being familiar with the assets analysed, makes it possible to quantify how each variable affects the asset health. In this case, operational thresholds are established for each variable, and the corresponding modifiers are determined. Table 1 shows the results proposed for each variable. In addition, Table 2 shows the weights of each of the variables and the associated coefficient γi . The equations to obtain the value of the health modifier (MH) and the reliability modifier (MR) will be the following: MH (t) =

j=n 

γj · MH j (t)

j=1

With: j = 1… n is an index used for different health modifiers. γj : is the weight assigned to the health modifier j in the model. MHj (t): is the health modifier at time t, age of the asset.

(7)

Challenges on an Asset Health Index Calculation

211

Table 1. Health and reliability modifiers Pump variables

Below the admissible range

In the recommended range

Flow

0

0

1

Suction Pressure

1

0

0

Discharge Pressure

0

0

1

Suction 1 Temperature

0

1

BFP Speed

Inactivity Overhauls

Above the admissible range

Below the admissible range during t > = 30’

Below the In the admissible recommended range during range t < 30’

Above the admissible range during t < 30’

Above the admissible range during t > = 30’

0.5

0.25

0

0.5

1

From 0% to 50%

From 50% to 75%

From 75% to 100%

0

0.5

1

From 0 to 3

From 3 to 5

More than 5

0

0.33

1

Table 2. Relative weight and coefficient γ i Modifier

Relative weight

Coefficient γi

Flow

15.22%

0.046

Suction Pressure

15.22%

0.046

Discharge Pressure

14.13%

0.042

Suction Temperature

13.04%

0.039

BFP Speed

15.22%

0.046

Inactivity

13.04%

0.039

Overhauls

14.13%

0.042

And. MR(t) =

k=m 

γk · MRk (t)

k=1

With k = 1… m is an index used for different reliability modifiers.

(8)

212

E. Candón et al.

γk : is the weight assigned to the reliability modifier k in the model. MRk (t): is the reliability modifier at time t, age of the team. The determination of the weight assigned to each modifier must be done relative to the rest of the modifiers and assuming a maximum possible impact of the set of modifiers in their worst condition. This is achieved considering the following restrictions: The sum of the totality of the modifier weights will be equal to the so-called maximum impact rate γ (for this rate, all modifiers always take the value 1). j=n  j=1

γj +

k=m 

γk = γ

(9)

k=1

The value of γ is obtained by forcing the HIit to be equal to 10 (maximum limit of the HI of the asset in the model) upon reaching the estimated normal life (tEL ) of the asset. Therefore, γ = Ln(Ln(10)/Ln(5.5)) = 0.301 The effect that the modifiers of the asset’s health have in the calculation of the health index can be seen as an example in Fig. 2. Where the HIit is compared to the HIiReal (t) and the HI(t). This Figure considers the possibility that the asset’s health index improves with respect to the forecast (by reducing the load compared to the forecast). The figure includes the effect of the Overhaul of the pump on the mentioned indices.

Fig. 2. Effects of modifiers on the initial health index of assets

3.2 Results The case of studied have been applied on eight pumps of two different motor pumps (units A and B) and can therefore compare the state of health of all these pumps. The results obtained is shown in Fig. 3, representing the health index of each pump (y-axis) as a function of the operating hours since the last overhaul (x-axis). In addition, the diameter of the ball indicates the deterioration of each pump, deterioration being understood as the deviation between the initial health index (HIit ) and the final health index (HI). From the comparative health of the pumps in the Fig. 3 the following observations are worth noting:

Challenges on an Asset Health Index Calculation

213

• Pumps A2 and A4 accumulate the lowest number of operating hours and have the lowest HI, with the HI of pump A4 being higher despite accumulating 220 operating hours less than A2, due to the influence of the load factor and health modifiers. • The B2 and B4 pumps have a similar health index but the B4 pump accumulates approximately 3,400 h more than the B2 pump. However, the deterioration (HI-HIi) is higher for pump B2. • Pump B1 has the worst HI index and the worst accumulated deterioration, but not the one with the highest number of operating hours (pump A1). • Pump B4 is the pump with the best aging rate in the last sample analysed, and B1 the worst.

Fig. 3. Asset Health Index vs Operation Time of the pumps analysed

As an example, pump B1 is the most degraded pump from the rest of pumps analysed. This pump has accumulated approximately 59,902 operating hours since overhaul and a health index of 3.962. This pump is second only to pump A1 among all pumps in unit A and B in accumulating the most operating hours since overhaul, but shows greater deterioration than pump A1, with approximately 200 more operating hours than pump B1. This pump B1 is close to reaching the HI2 range between the values H = 4 and H = 6, which corresponds to the period of time when the first signs of wear begin to appear in the asset. In case any maintenance activity has to be prioritised among these pumps, pump B1 will take precedence over the rest, followed by pumps A1, A3 and B3 consecutively.

214

E. Candón et al.

Finally, is shown in Table 3 some recommendations for action for a simple interpretation of an AHI calculation. Table 3. Recommendations due to AHI interpretation. AHI

Condition

Requirements

0.5–4 Very good Normal maintenance 4–5.5 Good

Normal maintenance

5.5–7 Fair

Increase diagnostic testing, possible replacement depending on criticality

7–8

Poor

Start planning process to replace

8–10

Very poor

Immediately assess risk; replace or rebuild based on assessment

4 Conclusions A study has been carried out to calculate the asset health index of eight different assets of an industrial plants. To determine these calculations, it has been necessary to collect a large amount of location, operation and maintenance data of these assets, which after being processed allows estimating the current state of health of the assets analyzed. However, the methodology used has some weaknesses in terms of the mathematical formulation used and its consistency. For example, the calculation of the health index at instant t does not take into account the health index at instant t-1, only the condition that HI(t) is mandatorily greater than or equal to HI(t-1), except when an overhaul is performed. By setting this condition, Fig. 2 shows flat areas in the graph, since in the mathematical formulation HI(t) is lower than HI(t-1) due to a lower influence of the health and reliability modifiers. Despite the weaknesses of the methodology, the proposed objective is met. The objective of the application of the methodology has been to measure the current condition of the assets analyzed and to offer the possibility of being able to compare them, in an objective manner, against each other. As a result of the procedure, the organization has an indicator of the assets, to prioritize interventions, attention and renewal of significant equipment. Acknowledgement. This paper has been written within the framework of the projects INMA "Asset Digitalization for INtelligent MAintenace" (Grant PY20 RE014 AICIA, founded by Junta de Andalucía PAIDI 2020, Andalucía FEDER 2014-2020) and Geminhi (Digital model for Intelligent Maintenance based on Hybrid prognostics models), (Grant US-1381456, founded by Junta de Andalucía, Andalucía FEDER 2014-2020).

References 1. de la Fuente, A., Crespo, A., Sola, A., Guillén, A., Gómez, J., Amadi-Echendu, J.E.: Planning Major Overhaul and Equipment Renovation Based on Asset Criticality and Health Index. In:

Challenges on an Asset Health Index Calculation

2.

3.

4.

5.

6.

7.

8. 9.

10.

11.

12. 13.

14.

15.

16.

215

Crespo Márquez, A., Komljenovic, D., Amadi-Echendu, J. (eds.) WCEAM 2019. LNME, pp. 83–90. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-64228-0_8 Hjartarson, T., Otal, S.: Predicting future asset condition based on current health index and maintenance level. In: ESMO 2006 - 2006 IEEE 11th International Conference on Transmission & Distribution Construction, Operation and Live-Line Maintenance (2006) Naderian, A., Cress, S., Piercy, R., Wang, F., Service, J.: An approach to determine the health index of power transformers. In: Conference Record of the IEEE International Symposium on Electrical Insulation, July 192–196 (2008). https://doi.org/10.1109/ELINSL.2008.4570308 Naderian, A., Piercy, R., Cress, S., Service, J., Wang, F.: An approach to power transformer asset management using health index. IEEE Electrical Insulation Magazine 25(2): 2 (2009). https://doi.org/10.1109/MEI.2009.4802595 Azmi, A., Jasni, J., Azis, N., Kadir, M.Z.A.: Evolution of transformer health index in the form of mathematical equation. Renew. Sustain. Energy Rev. 76(March), 687–700 (2017). https:// doi.org/10.1016/j.rser.2017.03.094 Silvestri, L., Forcina, A., Introna, V., Santolamazza, A., Cesarotti, V.: Maintenance transformation through Industry 4.0 technologies: a systematic literature review. Comput. Ind. 123 (2020). ISSN 0166-3615, https://doi.org/10.1016/j.compind.2020.103335. GB DNO groups. DNO common network asset indices methodology (2017). Health & Criticality - Version 1.1. https://www.ofgem.gov.uk/system/files/docs/2017/05/dno_common_net work_asset_indices_methodology_v1.1.pdf Australian Local Government Association (2015). National State of the assets report. https:// alga.asn.au/national-state-of-the-assets-report-2015/ Federation of Canadian Municipalities (FCM) & other seven partner organizations (2020). The Canadian Infrastructure Report Card (CIRC) 2019. http://canadianinfrastructure.ca/dow nloads/canadian-infrastructure-report-card-2019.pdf Hjartarson, T., Otal, S.: Predicting future asset condition based on current health index and maintenance level. In: ESMO 2006 - 2006 IEEE 11th International Conference on Transmission & Distribution Construction, Operation and Live-Line Maintenance (2006) Vermeer, M., Wetzer, J., van der Wielen, P., de Haan, E., de Meulemeester, E.: Asset management decision support modeling, using a health and risk model. PowerTech, pp. 1–6. IEEE. Eindhoven (2015). https://doi.org/10.1109/PTC.2015.7232556 Scatiggio, F., Pompili, M.: Health index: the TERNA’s practical approach for transformers fleet management. In: IEEE Electrical Insulation Conference (EIC), pp. 178–182 (2013) de Calle, K.L., Ferreiro, S., Arnaiz, A., Sierra, B.: Dynamic condition monitoring method based on dimensionality reduction techniques for data-limited industrial environments. Comput. Ind. 112, 103114 (2019). ISSN 0166-3615, https://doi.org/10.1016/j.compind.2019. 07.004 CrespoMárquez, Adolfo, et al.: Defining Asset Health Indicators (AHI) to support complex assets maintenance and replacement strategies. a generic procedure to assess assets deterioration. In: CrespoMárquez, A., Macchi, M., Parlikad, A.K. (eds.) Value Based and Intelligent Asset Management, pp. 79–99. Springer, Cham (2020). https://doi.org/10.1007/978-3-03020704-5_4 Serra, J, et al.: A model for lifecycle cost calculation based on asset health index. Pag. 91-98. In: International Conference on Smart Infrastructure and Construction 2019 (ICSIC) (2019). ISBN 978-0-7277-6466-9. https://www.icevirtuallibrary.com/doi/https://doi.org/10. 1680/icsic.64669.091 De la Fuente, A. et al.: Strategic view of an assets health index for making long-term decisions in different industries. Chapter 126, Safety and Reliability – Safe Societies in a Changing World: Proceedings of ESREL 2018, June 17-21, 2018, Trondheim, Norway (1st ed.). CRC Press (2018). https://doi.org/10.1201/9781351174664.

216

E. Candón et al.

17. Crespo, A.: Asset health indexing and life cycle costing. In: Crespo, A. (ed.) Digital Maintenance Management. Guiding Digital Transformation in Maintenance. Springer, Cham (2022). https://doi.org/10.1007/978-3-030-97660-6_12. ISBN 978-3-030-97660-6

Asset Life Cycle Management

An Integrated Framework for Efficient Asset Life Cycle Costing in Case of Incomplete Historical Data Mohammad Baharshahi1(B) , Mostafa Yousofi Tezerjan2 , and Saeed Ramezani3 1 Department of Industrial Engineering, Iran University of Science and Technology, Tehran, Iran 2 Faculty Member of University of Applied Science and Technology, Karaj, Tehran, Iran

[email protected] 3 Faculty Member of Department of Industrial Engineering, IH University, Tehran, Iran

Abstract. One of the major strategic challenges confronting large organizations with the intention of reducing costs as well as increasing availability is determining the optimal replacement time and analyzing the life cycle cost (LCC) of heavyduty equipment. In the establishing LCC model for supporting human decisionmaking, it is inevitable to deal with uncertainty caused by vagueness intrinsic to human errors in stored data and contradiction or incompleteness in different data set repositories originating from various resources such as different divisions of an organization. Along with its reference methodology, BS EN ISO15663:2021, in this research project, a new model to produce life cycle costing prediction is established and sensitivity analysis provides the arguments for the key cost elements and then the capability of the proposed model in addressing incomplete historical data presented. Contextual research project of encounters inside the wellknown National Iranian Gas Transmission Company (NIGTC) is used to highlight and discuss the various aspects of the proposed model. This study involves case studies of four gas turbine fleet which is running in NIGTC including Siemens Gas Turbine (SGT600), Zorya DU80, Nevsky GTK01003, and Nuovo Pignone MS5002C (NP). The proposed model demonstrates that by making a trade-off between additional sources of data and filtering performance a justifiable optimal replacement time for each fleet could be estimated. Another interesting finding of this research project is that sensitivity analysis of key cost elements confirmed that overhaul maintenance cost has less impact on LCC than previously thought by experts.

1 Introduction It is vital to decide the best time to utilize and replace equipment and physical assets to lower the organization’s expenditures. A wide variety of life cycle cost models, both general and particular models, have been developed throughout the years. In the industrial sector, none of life cycle cost model is recognized as the standard model. The willingness of users, the nature of the problem, the existence of some alternative data set systems for the cost, and several types of equipment, devices, or systems are all causes for the

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 219–228, 2023. https://doi.org/10.1007/978-3-031-25448-2_21

220

M. Baharshahi et al.

lack of a standard model. Regardless of the varieties of models used in life cycle costing analysis, they should all be capable of giving clear and observable costs for equipment, systems, and subsystems (Iso 2021). In Turkey, decision-making is based on life-cycle cost analysis for solar energy transmission line scenarios. The net present value is used to calculate cumulative cash flow estimates (Acaro˘glu and Márquez 2022). The key uncertainties in calculating Life Cycle Costs for almost zero-energy buildings in Europe were identified and assessed. Using the differential and elementary effects methods, identify the benefits and shortcomings of sensitivity analysis. Eleven case studies show that uncertainties can cause variances of up to 37% around the median LCC, with an average of 26% (Pernetti et al. 2021). LCC may also be used to make purchasing decisions by weighing competing choices such as requirement, product quality, and post-purchase services. It can also be used to make decisions concerning plant performance, such as low total cost of product, high rate of survival, and reliability-centred design (Kamble 2022). A cost-optimal approach is developed based on the whole life-cycle cost of a gas turbine’s air intake system. The model indicated that increasing the number of filter units and increasing the capacity of the air intake system, the filter unit replacement cycle lengthens, and unit reliability improves, while the air intake system running maintenance cost decreases and the NPV remains low throughout the life cycle (Xueyu and Xin 2018). Some authors lay the foundation for a systematic research plan, management framework, and background knowledge evaluation that emphasize the connections and constraints within and between LCC and building management (BM). They developed conceptual frameworks in six stages with six goals in mind: establishing a starting point, mapping information sources, literature review, notion deconstruction and conceptual categorization, overview of the relevant background knowledge, and structuring a framework for LCC-informed decisions in BM (Salvado et al. 2018). In new research they have developed an indicator-based methodology called the Building Investment Index (BII), which enables real-time monitoring of the economic life-cycle performance of construction projects. They demonstrate how to utilize BII to monitor and enhance longterm investment plans for building systems, subsystems, and elements, as well as to future-proof choices (Salvado 2020). Replacement models include shortage and excess costs, which imply that replacing something too soon before it fails is a waste of time discussed by (Zhao 2016). The LCC is calculated based on the failure rates of the machine’s components. The issue is described as mixed integer programming, with the goal of minimizing total costs for machines of the same type with various age across a planning horizon of multiple periods (Seif and Rabbani 2014).

2 Gas Turbine LCC Analysis Principle The present literature on the definition of LCC analysis is primarily based on the British Standard EN ISO 15663:2021. As part of life cycle costing, systematic assessments and computations are carried out to compare competing choices using economic evaluation metrics. The following relationships can be used to illustrate the life cycle cost of

An Integrated Framework for Efficient Asset

221

equipment: LCC = Caq + Cow + Cdi

(1)

Caq represents the cost of equipment acquisition; Cow represents the cost of equipment ownership, which typically includes operation and maintenance expenses; and Cdi represents the cost of equipment disposal, which often includes waste costs and salvage value. The amount of Caq in the LCC of a gas turbine generally drops with time, while the proportion of Cow gradually grows. Cdi is generally constant, and the proportion is quite tiny, and can even be ignored in some circumstances, based on experience in the field. There is often no information on the value of disposal since a turbine of this type has not been sold for a prolonged period. In these situations, the cost must be established by a group of experts from various disciplines using methods like the Delphi method. A scrum framework to address this problem is to build a cost breakdown structure. The LCC approach relies heavily on cost concerns. After defining cost elements, the interrelationship between them must be accurately reflected in the system model. To avoid certain crucial cost aspects being overlooked, a method must be set up. 2.1 The Acquisition Cost of Gas Turbine The purchase price is main cost in acquiring of a gas turbine. Although asset purchase cost, installation, transport cost, insurance etc. are among those cost must be included in calculating acquisition of gas turbine but estimating of this cost in real practice may face with challenges. 2.2 The Ownership Cost of Gas Turbine In this study gas turbine ownership costs divides into three categories: a) opportunity cost of un-transmitted gas (Copp ), b) fuel consumption cost, and c) turbine maintenance cost. • opportunity cost of un-transmitted gas is the potential benefits that a unit of gas turbine misses out within unplanned shutdown. This cost calculated as Eq. 2 as below: Copp = gas sales rate × Lost transmission capacity

(2)

• fuel consumption cost is the price of gas consumed by a unit of gas turbine annually. It could be fairly estimated by Eq. 3: Cgas = gas sales rate × Annual turbine consumption

(3)

• Turbine maintenance cost (Cm ) is all cost associated for maintaining and overhauling gas turbine which include several items as follows. o Cost of preventive maintenance (electricity, instruments and mechanics and the cost of workforce, spare parts, number of days out of service of the equipment) o The cost of corrective maintenance (electricity, instruments and mechanics and the cost of manpower, spare parts, number of days out of service of the equipment) o The cost of performing repairs of various levels: A, B, C, D, E. in each level of repairs different job order performed.

222

M. Baharshahi et al.

2.3 The Disposal Cost of Gas Turbine The choice of depreciation method has high sensitivities. Because depreciation costs play a key role in tax savings. This savings can be maximized when the maximum amount of depreciation is concentrated in the early years of life. Therefore, financial division of organizations usually use double declining balance (DDB) depreciation method which is an accelerated method that multiplies gas turbine value by a depreciation rate, which is the most depreciation in the early years of gas turbine life.

3 Time Value of Costs The worth of money today is not equal to the value of money in the future. The “time value of money” is the term used to describe this idea. The degradation of money’s value through time, or inflation, and the cost of missed opportunities are the two elements that determine the temporal value of money. Opportunity cost is the advantage that might have been obtained from cash or current capital if it had been invested or used differently. Opportunity cost for borrowed funds is the fee for borrowing that money (e.g., the loan rate). Over time, inflation lessens the value or buying power of money. It is the end consequence of the cost of products and services gradually rising because of economic activity. Estimates of future expenditures can be established in current dollars and then converted back to present value using the correct calculations by subtracting inflation from all escalation and discount rates. It is possible to avoid estimating how inflation rates will behave in the future. For sake of brevity, calculation and formula of inflation didn‘t mentioned in this paper. Interested reader could refer to BS EN ISO15663:2021.

4 Combining Data Repository Certain critical context information will be missing due to a lack of knowledge and an incomplete data repository. Data originating from several divisions of an organization might occasionally be contradictory. As a result, past partially recorded data in organizations that are in some way contradictory will be combined. With such data repository, learning a global model for life cycle cost (LCC) becomes extremely needed for top manager of large organization. As a result, we’ve started looking for sophisticated LCC estimate strategies that aren’t limited to global models. Merging partial raw data originated from different data repository and turning them to integrated interpretable information is the key in the proposed method in this research. Incompleteness in data could be in following area: • Incomplete in historical data (i.e., incomplete cost data, incomplete operating data, un-known work orders, etc.) • Vagueness due to unsuitable of Cost Breakdown Structure (CBS) for LCC estimation purpose. • Un-integrated and contradict different data repository of organization

An Integrated Framework for Efficient Asset

223

• Different age of gas turbines in each fleet. Figure 1 schematically depicted how the proposed model combine total cost of gas turbines with different age.

Fig. 1. Combine total cost of gas turbines with different age

Full turbine life cycle data is not available, and the available data is from 10 years ago to now. The cost information of the previous years is not available, therefore, to calculate the costs of the turbine in different lifetimes, the turbines were sorted according to the life or hours of operation and a full/nearly full cycle was calculated for the life. As can be seen in Fig. 1, the combination of turbine data compensates for the lack of data. To calculate the right time for equipment replacement using life cycle costs, there is no need to calculate fixed costs, and they can be omitted for the simplicity of calculations and to reduce the data required at this stage of life cycle cost analysis. Therefore, according to the way of data recording, the data of the turbines were studied cumulatively so that it is possible to obtain more reliable data. Sometimes the data fields recorded in the information systems are not exactly the indicators that are required in the life cycle cost analysis, in these cases the indicators are calculated from the cost fields and data combination. If there are outliers, missing data, etc., data preprocessing and estimation techniques are used. After calculating the average values of operating hours and cost per operating hour for equipment in all periods, graphs of cost per operating hour should be drawn to average operating hours. Analyzing and examining this graph, the regression line and the scatter of its points show whether a good fit is obtained from it or not. Generally, due to the presence of outlier data caused by factors such as breakdowns due to lack of proper maintenance, lack of use of equipment, etc., the graph shows an error. Diagrams should be drawn in the following three modes: • Draw a graph with all the data • Draw a graph by removing outliers • Draw a graph with aggregated data

224

M. Baharshahi et al.

5 Practical Implementation In order to collect cost data for gas turbine fleets of National Iranian Gas Transmission Company (NIGTC), it is necessary to extract and aggregate cost information from various organizational departments, including maintenance, finance, logistics and support. The life cycle costing model used in this project includes Siemens Gas Turbine (SGT600), Zorya DU80, Nevsky GTK01003, and Nuovo Pignone MS5002C (NP). Acquisition Cost: Fleets of NIGTC were purchased many years ago, so it is necessary to calculate the net present value (NPV) of turbines. Other shipping, insurance and installation costs are also unknown. However, the major cost of acquisition is related to the cost of purchasing this heavy-duty equipment. Ownership Costs: In terms of turbine repair costs, non-compliance with the cost breakdown structure (CBS) posed a significant difficulty. Due to their high value, these expenses are classified as capital expenditures (CAPEX). However, there are two fundamental issues with data collection in this area: • To implement fundamental of LCC, the costs must be calculated separately for each turbine, but financial costs for major repairs/overhaul are estimated not only at the turbine level, but also at the area level. Human mistakes in registering data in the financial system, on the other hand, and, of course, the inability to use financial data, are both problems in data gathering. • Stored cost data include combined overhaul charges for many stations; however, the difficulty is that these financial statements do not indicate which turbine was spent with its unique serial code. As a result, analyzing and categorizing these expenditures necessitates one-on-one examination, which is not always practical. The nominal gas transmission capacity for each compressor is 30 million cubic meters per day and 1.25 million cubic meters per hour. The average monthly operation of Siemens turbines No. 1 and 2 is equal to 730 h and the annual operation of these units is 7840 h. The rate of sale (export) of gas, which is 0.08 Euros per cubic meter and (80,000 Euros per million cubic meters) is also considered as the cost of missed opportunity gas transmission per cubic meter. Based on Eq. 2 opportunity cost of un-transmitted gas for a unit with 7 h of unplanned shutdown would be 700,000 Euros annually. The average annual fuel consumption of Siemens Turbine No. 1 is about 20 million cubic meters and for Turbine No. 2 is about 28 million cubic meters. Gas sales rates are also considered as fuel prices. According to Eq. 3 cost of fuel consumption for a unit with 66 million cubic meter fuel consumption would be 4.8 million Euros. Ownership cost of SGT600 shown in Fig. 2. Disposal Cost: It is possible to estimate the acquiring value of a turbine at the time of purchase. Since the current value of several types of turbines is accountable so calculating net present value, depreciation and disposal cost of it is possible. Based on expert experience disposal cost of scrapped turbine assumed 5% of new specific one. Cost of depreciation and net present value of SGT600 shown in Fig. 3 left and right respectively.

An Integrated Framework for Efficient Asset

225

Fig. 2. Ownership cost of gas turbine (SGT600)

Fig. 3. Cost of depreciation (left), net present value of turbine (right)

6 Discussion After all the calculations, the point where the sum of the uniform annual costs is the minimum is the optimal time to replace the turbine. The life cycle cost diagram is plotted using the scrap value, depreciation cost, and maintenance cost in terms of time (turbine operating hours) in Figs. 4, 5, 6 and 7.

Fig. 4. Total average annually cost of SIEMENS GT10B2

Fig. 5. Total average annually cost of ZORYA DU80

As shown in Fig. 4 total average annual cost (TAC) of SIEMENS GT10B2 decreases as operating hours passes, but over time, the intensity of the downward trend has diminished. By fitting a curve on this data, the minimum of this trend determined at 58,877 operating hours, which is indicate the optimal time for replacing of this type of turbine.

226

M. Baharshahi et al.

Fig. 6. Total average annually cost of NEVESKY GTK01003

Fig. 7. Total average annually cost of NP MS5002C

Figures 5, 6 and 7 show the TAC of ZORYA DU80, NEVESKY GTK01003 and NP MS5002C, respectively.

7 Sensitivity Analysis on Key Cost Driver Following the calculations, a sensitivity analysis is undertaken so that the employer may account for additional unknown aspects, such as the need to keep the equipment in the event of an emergency, a shortage of funds, or other issues, and adjust the calculations accordingly. Calculations and graphs of sensitivity analysis help to keep the efficiency of the model even in the conditions of changing parameters, and there is no need to redo the entire analysis process. For example, by predicting the change in gas prices due to the war between Russia and Ukraine, without the need to revise the entire model, it is possible to predict the results for new conditions based on the sensitivity analysis charts and make the right managerial decision quickly and accurately. Gas price is the key cost driver in LCC of a gas turbine so sensitivity analysis of it presented in Table 1. Table 1. Sensitivity analysis of cost element of ownership cost by changing gas price Gas price Opportunity Fuel Overhaul Manpower (euro/mm3 ) cost of consumption cost cost un-transmitted cost gas 0

0%

0%

53%

47%

8

0%

1%

53%

47%

80

2%

6%

49%

43%

800

10%

35%

29%

26%

8,000

20%

69%

6%

5%

80,000

22%

76%

1%

1%

As shown in Table 1, It is obvious by increasing gas price from 0 to 80,000 euro per million cubic meter, opportunity cost and fuel consumption cost increase. Considering

An Integrated Framework for Efficient Asset

227

the fuel rate of 800 Euros, the share of opportunity cost of the Siemens fleet during the 10 years of normal operation is equal to 10% of the total maintenance and operation costs of that fleet. It‘s interesting that at fuel rate of 80,000 euro per million cubic meter– near to actual price of gas - the overhaul cost is just 1% of ownership cost. This part of overhaul cost is far away from expert guess about how much this cost element impact on LCC of a gas turbine.

8 Conclusion This research has developed a detailed life cycle cost analysis approach for evaluating gas turbine alternatives. The procedure’s steps include defining the system’s objective and needs, creating a cost breakdown structure, and creating mathematical models for each cost element. A series of electronic spreadsheets is used to run the models. Raw data of 4 type of gas turbine which is running in NIGTC including Siemens Gas Turbine (SGT600), Zorya DU80, Nevsky GTK01003, and Nuovo Pignone MS5002C (NP) acquired, preprocessed and analyzed. In order to address of incompleteness in data, an innovative approach propose. Data of different aged turbine in each fleet by a comprehensive LCC model combined and the optimal replacement time of all type of gas turbine determined. Result of this research work revealed the share of each cost element in life cycle of turbine. Although overhaul of turbines has a key role in operating planning but it‘s costs have less impact on LCC from which experts thought before. Acknowledgment. The authors express their gratitude to NIGTC for their assistance, particularly to distinguished corporate engineers, Mr. Rahmatnejad and Mr. Ghashghaipour.

References Acaro˘glu, H., Márquez, F.P.G.: A life-cycle cost analysis of high voltage direct current utilization for solar energy systems: the case study in Turkey. J. Clean. Prod. 360, 132128 (2022). https:// doi.org/10.1016/j.jclepro.2022.132128 Iso, B. S. E. N. 2021. Petroleum and Natural Gas Industries — Life Cycle Costing (2021) N. Kamble, S., Rajiv, B.: Machine health monitoring with life cycle cost analysis by condition monitoring. Mater. Today Proc. 52, 893–97 (2022). doi: https://doi.org/10.1016/j.matpr.2021. 10.296 Pernetti, R., Garzia, F., Oberegger, U.F.: Sensitivity analysis as support for reliable life cycle cost evaluation applied to eleven nearly zero-energy buildings in Europe. Sustain. Cities Soc. 74, 103139 (2021). https://doi.org/10.1016/j.scs.2021.103139 Salvado, A., Almeida, N., Azevedo, Á.: Future-proofing and monitoring capital investments needs throughout the life cycle of building projects. Sustain. Cities Soc. 59, 102159 (2020). https:// doi.org/10.1016/j.scs.2020.102159 Salvado, F., Marques de Almeida, N., Vale e Azevedo, A.: Toward improved LCC-informed decisions in building management. Built Environ. Project and Asset Manage. 8(2), 114–33 (2018). doi: https://doi.org/10.1108/BEPAM-07-2017-0042

228

M. Baharshahi et al.

Seif, J., Rabbani, M.: Component based life cycle costing in replacement decisions. J. Qual. Maint. Eng. 20(4), 436–452 (2014). https://doi.org/10.1108/JQME-08-2013-0053 Xueyu, L., Xin, T.: Optimization of gas turbine filtration system based on life cycle cost (LCC) theory. 120(Ifeesm 2017), pp. 997–1006 (2018) Zhao, X., Chen, M., Nakagawa, T.: Replacement policies for a parallel system with shortage and excess costs. Reliab. Eng. Syst. Saf. 150, 89–95 (2016). https://doi.org/10.1016/j.ress.2016. 01.008

Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management Virginia Montiel(B) Atlantic Copper S.L.U, Condition Based Maintenance, Avda. Francisco Montenegro S/N, 21001 Huelva, Huelva, Spain [email protected]

Abstract. Atlantic Copper is the largest copper and sulphuric acid producer in Spain, with its production center being the Huelva Metallurgical Complex. As part of its continuous improvement plan and in pursuit of effective and efficient asset management to maximize the profitability of its assets, it is certificated in 2019 under the ISO55001 standard, being the first company in its sector to do so. The application and implementation of the ISO55001 standard led to the introduction of improvements and new tools for the management, control, and maintenance of the company’s physical assets, among which the “Asset Life Cycle Cost Analysis” stands out. This article shows the system applied by Atlantic Copper to determine the start of a “Life Cycle Cost Analysis”, as well as the method used to apply this tool. It includes a real case developed with this methodology, the results of which allowed it to be possible to identify the best option among the alternatives proposed for the management the remaining life of one of the Metallurgical Complex’s critical assets.

1 Introduction The metallurgical industry today faces great challenges to maintain its levels of competitiveness in a market in continuous transformation, so it is essential to know when it is the appropriate time to make a replacement, an overhaul or a change of technology in an equipment. The fact of not identifying this moment properly and not acting accordingly could cause high maintenance costs, a growth in the failure rate and, consequently, an increase in production losses caused by the non-availability of that asset. For this reason, Atlantic Copper´s asset management system has integrated the concept of life cycle cost and has designed the specific tool of “Life Cycle Cost Analysis”, from now on the acronym LCCA will be used. This tool has been created to facilitate the decision-making of possible alternatives for action through the objective, methodical and systematic assessment of different direct and indirect, fixed and variable costs associated with all the consecutive and/or interrelated stages throughout the existence of the asset, from its conception and basic engineering up to its disposal and dismantling. Understanding as possible alternatives those that have as their final objective the recovery of the expected performance of the asset. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 229–239, 2023. https://doi.org/10.1007/978-3-031-25448-2_22

230

V. Montiel

2 Economic Health Index of the Asset The useful life expectancy for an industrial plant is between the twenty and thirty years. During this period, it is expected that the equipment health degradation will remain within thresholds that allow an efficient operation and maintenance (Fig. 1). Several factors can contribute to shorten the useful life compared to the design values, so it is necessary to have an indicator to follow up the evolution of the health of the equipment and detect in advance the approach to the end of life.

Fig. 1. Pattern B-Wear out curve. Equipment failure patterns (Nowlan and Heap 1978)

For this reason, Atlantic Copper has created the EHI (Economic Health Index) which allows to proactively monitor the evolution of the health status of the company’s equipment through its translation to maintenance costs and the production losses generated by equipment failures and relating them in turn to the expected useful life ( Márquez et al. 2021). This type of indicator is used as an indicator of health status because due to the type of industry and process that Atlantic Copper has, and the experience accumulated over the years, it has been proven that when the associated costs of the asset increase significantly it is indicative that the degradation of the asset is very advance. This being the main indicator to determine the beginning of the application of the LCCA methodology and for decision-making in relation to the management alternatives of the remaining life of an equipment before a negative impact occurs for the company. This indicator only applies to those physical assets of the complex that annually comply with all the following conditions and requirements (Crespo et al. 2019): 1. Their location in the asset criticality matrix must comply with the following: • Refinery plant: Locations with a criticality rate equal to or greater than 120 and with more 0,5 failures per year. • Rest of plants: Locations with a criticality rate equal to or greater than 120, with more than 1 failure per year and whose production loss is greater than or equal to 50 t/failure (Production losses: lack of concentrate smelting in Flash Smelting Furnace). 2. Their failure rate has increased in the last three years. 3. Current value is greater than e 100.000 (Replacement acquisition) 4. These requirements are reviewed once a year. 2.1 Calculation of the Economic Health Index (EHI) It is an economic percentage indicator (Eq. 1) that relates the costs with the greatest impact throughout life cycle of an asset compared to the cost of replacement, in order

Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management

231

to be able to graphically evaluate the weight of economic activities with respect to the acquisition of new equipment (De La Fuente et al. 2018). EHI =

MC + IC + PC × 100 RC

(1)

where, MC: Annual preventive, corrective and extraordinary maintenance costs of the equipment, IC: Maintenance and operations investment, PC: Annual cost of production losses due to equipment failures, RC: Investment cost of replacement of equipment. The annual values of the EHIs will compose the economic health curve of the asset throughout its life cycle (Fig. 2). And as comparative reference parameter has been created the parameter called “Annualized life cycle” (Eq. 2; Fig. 2), this being the percentage that represents a year in the life cycle of the asset (UK DNO 2017).  ALC = 1 UL × 100 (2) where, UL: Expected useful life of the asset in years.

Fig. 2. Atlantic Copper’s Economic Health Curve and Health Status Levels by value “ALC”. GA Pr04 (A.Copper 2019)

To determine if it is necessary to start a LCCA, the asset must have been in the orange range of the health status level in the last two years or has reached the red range in the last year.

3 Cost of Life Cycle Analysis Methodology (LCCA) The modelling of the life cycle cost calculation is based on the guidelines of the UNE EN ISO 60300-3-3 standard “Application guide for the calculation of the life cycle cost”, and incorporates the concept of total cost of ownership (TCO) following up the following flowcharts (Fig. 3) (IEC 60300-3-3:2017 2017; Roda et al. 2020; Animah et al. 2018):

232

V. Montiel

Fig. 3. Flowchart to calculate the life cycle cost (IEC 60300-3-3:2017 2017)

In this way, it is possible to have a systematic and orderly approach to ensure that no parameter is forgotten. 3.1 Definitions of the Alternatives to Be Analysed In this first stage, the different alternatives that are going to be considered to recover the performance of the asset are established and characterized, and with which the life cycle cost of each of them will be analysed and compared. Below, you can see the alternatives that fit within this methodology: • Maintenance of the current situation of the asset, with the possibility of improving operation conditions, increasing maintenance frequencies, implementing new preventive/predictive maintenances. • Carrying out a partial replacement of the asset. Immediate replacement of some part of the asset for being close to or reaching the end of its useful life. • Overhaul of the asset • Replacement with a new asset similar to the current one. • Substitution by an asset with a different design. 3.2 Time Horizon of the Analysis Determining the time horizon is one of the most complex decisions of the LCCA. This is the period of years that will be considered for each of the alternatives considered within of the analysis. This period can be established based on the expected useful life of the assets or following other criteria that may be more appropriate to each case, such as the amortization period, or a specific budget, investment, or strategic cycle. 3.3 Selection of Cost Categories to Consider It is necessary to define which are the stages of the life cycle that will be contemplated in the analysis (Fig. 4). Within the established time horizon. In this sense, it should be noted that there may be different criteria when specifying the different stages and, above all, about their inclusion or not in their scope. The stages of the life cycle to consider can be: In each of the stages of the life cycle, various categories of costs can be considered to be included in the analysis (Table 1).

Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management

233

The stages of the life cycle to consider can be: 1. 2. 3. 4. 5. 6. 7.

Definition Acquisition Installation Operation Maintenance Energy Dismantling Fig. 4. Life cycle cost analysis (Crespo Márquez et al. 2012)

Table 1. Cost categories of the cycle life stages. GA-Pr04 (A. Copper 2019). 1 1.1 1.2 1.3 1.4 1.5 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 3 3.1 3.2 3.3 3.4 3.5 3.6 3.6

Definition Requirement studies Economic studies Basic engineering Measurements and tests Others Acquisition Suppliers search Bidding definition Evaluation of bids Purchase price Financial costs Renting and leasing Amortizations and deductions Installation Detail engineering Adaptation and training Civil work and installation Commissioning Test and certifications Insurance Others

4 4.1 4.2 4.3 4.4 5 5.1 5.2 5.3 5.4 6 6.1 7 7.1 7.2 7.3 7.4

Operation Equipment operation Safety Enviromental protection Insurance Maintenance Preventive Corrective Safety Enviromental protection Energy Energy consumption Useful life Engineering Decommissioning Safety Enviromental protection

3.4 Definition of the Cost Model The selection of life cycle stages and cost categories made in the previous phase, together with the cost-generating resources in each of these categories determine the definition of the cost model for the calculation. The cost-generating resources in each category can be as follows: • • • • • • •

RRPP: Dedication of own personnel resources. RREE: Dedication of external personnel resources. Services: Contracting of external services. Equipment: Acquisition of equipment, installation, and infrastructure Materials: Acquisition of raw materials, auxiliary and consumable materials. Energy: Energy consumption. Production losses: Production losses due to equipment stoppage or limitations to the planned progress.

Some categories of costs, such as installation costs, are incurred only once throughout the life cycle of the asset. However, other costs, such as preventive maintenance, are incurred periodically. The latter must be annualized for inclusion in the calculation, so that the accumulated for the years of the established analysis horizon is totalled. In case the cost of any of the categories are the same in the different alternatives considered, this category of costs can be excluded for the comparative analysis.

234

V. Montiel

3.5 Data Collection for Analysis The calculation of the life cycle cost of each alternative requires the collection of the quantitative data of the units of each of the resources considered in each cost category of the model. Some costs will be gathered from the different analysis carried out within the framework of the different management systems, the expected maintenance costs will be provided for maintenance and in other cases the cost by analogy method will be used, which is a cost estimate based on experience of similar products or technologies, using historical data, updated to reflect cost scaling and the effects of technological advances. 3.6 Calculation of Life Cycle Costs

Table 2. Summary table of the costs of asset the LC stages. GA-Pr04 (A. Copper 2019). LIFE CYCLE COST ANALYSIS Total € RRPP 1 Definition 2 Acquisition 3 Installation 4 Operation 5 Maintenance 6 Energy 7 Useful life end

RRHH

Services

Equipment Materials

Energy

Production

Once the data for each alternative considered have been filled in, the asset life cycle cost analysis form provides the cost at present value added of the different stages of the life cycle included in the analysis, as well as the aggregate cost of each of the types of resources assigned to each stage (Table 2). 3.7 Risk Estimation of Each Alternative The incorporation of a specific asset to the Metallurgical Complex will be associated with risks that will need to be considered for the assessment and comparison alternatives, and for their proper management if necessary. The type of risks to be considered will be as follows: • Safety: Equipment failures that could have an impact on safety • Environmental: Asset failures that could have an impact on the environment • Quality: Asset failures that could cause non-compliance with final product specifications. • Production: Asset failures that could cause production losses. • Obsolescence: Elements or technologies that may become outdated in the life of the asset. • Dependency: The asset creates an exclusive dependency with the supplier • Interchangeability: Risk of increasing spare parts by no incorporating standard elements • Operational: The asset requires the coordination and/or intervention of many operators

Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management

235

• Maintainability: maintaining the asset requires many operations or affects production The risk analysis will be carried out in terms of probability and impact, scoring them according to the different established criteria of occurrence and severity. The total value dimensionless obtained for each of the risks studied will be evaluated by applying the risk matrix impact-probability (Fig. 5). Risks above 160 points may not be acceptable, which could lead to discarding alternatives with risks at these levels, regardless of their life cycle cost. RISK MATRIX

Impact

1 Insignificant

4 Mild

10 Severe

5 Frequent 4 Usual

5

20

50

4

16

40

160

3 Occasional

3

12

30

120

2 Unlikely

2

8

20

80

1 Exceptional

1

4

10

40

Probability

40 Critical 200

Fig. 5. Risk Matrix Impact-Probability. GA-Pr04 (A.Copper 2019)

3.8 Comparative Evaluation of Alternatives The calculation of cost and the estimation of risks allow the comparative evaluation of each of the alternatives. The life cycle cost analysis form at present value provides the quantitative data and a graphical comparison of the aggregate cost of the life cycle stages, as well as the level of the different risks for each of the alternatives considered: In the cost evaluation, both the total cost of the life cycle of each asset can be compared, as well as the comparative cost of each of the phases of the life cycle. This methodology allows correcting the evaluation of annualized costs throughout the useful life, applying the discount rate provided to consider the future value of money. The cost evaluation must be complemented by the risk assessment. In particular, alternatives that present risks at critical levels should be considered since, although they may be positive economically, this level of risk may be unacceptable.

4 Illustrative Case Study In this case, the LCCA methodology carried out for one of the most critical assets of the Atlantic Copper Metallurgical Complex, which are the cranes of the refinery plant, will be explained. These are three overhead cranes from 1969 with which the movement of cathodes and anodes of the plant is carried out daily, these assets generate high maintenance costs and production losses annually due to the degradation of their structures and of their roller and translation systems, caused by the use and passing of time. To know if the LCCA methodology can be applied to the three assets, it must be ensured that they meet all the requirements indicated in Sect. 2. The three equipment are ideal candidates to start the LCCA since their criticality value is equal to or greater

236

V. Montiel

than 120 and have an annual failure rate greater than 0,5 (Fig. 6). In addition, this failure rate has increased during the last two years and, finally, their current replacement value is greater than e100,000.

Fig. 6. Criticality Matrix (A.Copper 2019)

With these premises met, we give way to the calculation and representation of their EHI (Eq. 1; Fig. 7) and whose annual values are represented graphically below in the right side of the figure for each of the cranes:

Fig. 7. EHI curves for the 3 overhead cranes. EHI Analysis Form (A.Copper 2019)

The last requirement to satisfy to start the development of the LCCA methodology is that in last year its EHI Index has reached the red zone or in the last two years it has remained in the orange zone. Requirements met by all three cranes. Crane No. 1 and No. 3 reached the red zone and crane No. 2 has remained in the orange zone for the past two years (Fig. 7). Note: As a support parameter, the comparison of the maintenance costs and cost of annual production losses with respect to annualized life cost of the equipment acquisition can also be obtained. In this case, it can be seen that all costs far exceed the annualized cost of purchasing the equipment (Fig. 7). 4.1 Application of LCCA Methodology The first stage of the methodology is the search and consideration of the possible alternatives to be assessed to recover the performance of the asset, with four the options chosen: • Alternative 1: Maintain the current situation with equipment over 50 years old.

Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management

237

• Alternative 2: Major reparation, repairing and/or replacing the rolling assembly of translations and the displacement system (rails and carriage), as well as part of structures (beams and headwalls). • Alternative 3: Overhaul to the 3 cranes replacing them with cranes with similar characteristics and changing their tenders. • Alternative 4: Change of technology, replacing the three manual cranes with two automated cranes. Once the alternatives have been defined, the categories and cost model follow for each of them. For each alternative, the relevant cost categories are selected at each stage of the life cycle, and each category is displayed in the relevant cost-generating resources, as shown in the following image (Fig. 8).

Fig. 8. Life Cycle Cost Analysis Form. GA-Re 06 (A.Copper 2019)

Within each cost category, the resources, which are applicable to them, are valued. The data completion for each resource allows the generation of the calculation of the life cycle cost for each alternative (Table 3). To complete the analysis, it is necessary to evaluate the risks of the asset according to the nine defined risk classes and assess them according to their occurrence and probability, transferring this value to the risk matrix to obtain the final value of each risk category (Table 3). Table 3. Summary tables of the costs of the alternatives considered for each of the stages of the life cycle and of the risks associated to the alternatives according to the type of risk. GA-Re 06 (A.Copper 2019). GA-Re 06 (A.Copper, 2019) Alternative 1 COST ANALYSYS OF THE LIFE CYCLE 1 2 3 4 5 6 7

Definition Acquisition Installation Operation Maintenance Energy Useful life end

Total

Major Repair 36.000 € 0€ 310.000 € 0€ 634.948 € 199.041 € -44.850 €

Alternative 2 New Manual Crane 36.000 € 706.179 € 113.977 € 0€ 293.286 € 199.041 € -44.850 €

Alternative 3

Alternative 4

New Manual New automatic Crane + Lift crane System 36.000 € 88.000 € 1.706.179 € 2.150.000 € 152.327 € 280.000 € 0€ 0€ 293.286 € 302.463 € 199.041 € 372.435 € -44.850 € -70.850 €

1.135.139 € 1.303.634 € 2.341.984 € 3.122.049 €

Alternative 1 Alternative 2 Alternative 3 Alternative 4 RISK ANALYSIS OF THE ASSET Safety Environmental Quality Production Obsolescence Dependency Interchangeability Operational Maintainability Total Risk :

Major Repair

New Manual Crane

New Manual Crane + Lift System

New Automated Crane

30 3 3 50 50 5 20 50 20

12 3 3 16 2 50 20 50 20

12 3 3 16 2 50 20 50 20

8 2 2 16 2 50 120 20 12

231

176

176

232

All these data are transferred to a bar chart for better visualization (Fig. 9).

238

V. Montiel

Fig. 9. Bar chart summary of the total cost of the analysed alternatives and of the total risk associated with each of alternative. GA-Re 06 (A.Copper 2019)

Thus, the economic valuation and risk of each alternative is obtained in order to be able to objectively assess which alternative is the most proper for the company. According to the results obtained, the following can be observed: • Alternative 1: It is the one with the lowest cost, but it is one of the alternatives that presents the most risks, especially in production, obsolescence, and safety. • Alternatives 1 and 2: These are the ones with the lowest risk, with the risk of both alternatives being very similar, but the alternative 2 is the one with the lowest life cycle cost. • Alternative 4: It is one of the alternatives that presents the greatest risk together with alternative 1 and the highest life cycle cost. In summary, of the alternatives studied, Alternative 2 “Acquisition of new manual cranes without lifting system” is the best option for the company.

5 Final Conclusions The achieving and oversight of the Economic Health Indicators (EHI) facilitates the convenience of starting of the Life Cycle Cost Analysis for the most critical and valuable assets for the company. Life Cycle Cost Analysis allows to compare different investment alternatives for an asset, mainly in case of acquisition, repairing, overhaul or replacement of critical assets with a significant investment volume, considering all relevant concepts of cost throughout the stages of life of the asset. This methodology generates a document that allows users to obtain an overview of the analysis and also in detail, allows them to review and understand the results obtained. Against the benefits of applying this methodology, there are a number of limitations that, to some extent, represent a counterweight and can make this type of analysis less effective as a support tool in decision-making:

Life Cycle Cost Analysis in Modern Heavy Metallurgical Asset Management

239

• Reliability of data: Not all the data required for this type of analysis are always available or with the degree of detail that is required, assuming them and therefore strongly conditioning the representativeness of the results obtained. • Subjective criteria: Sometimes the risk assessment process cannot be carried out objectively because some of the categories that are evaluated are not measurable and therefore are subject to the subjective criteria of the people who develop the methodology.

References Animah, I., Shafiee, M., Simms, N., Erkoyuncu, J. A., Maiti, J.: Selection of the most suitable life extension strategy for ageing offshore assets using a life-cycle cost-benefit analysis approach. J. Qual. Maintenance Eng. (2018) Atlantic Copper SLU, 2019. GA-Pr04 Life cycle cost analysis of the assets Atlantic Copper, 2019. Criticality Matrix Atlantic Copper SLU, 2019. Economic Health Index Analysis Form Atlantic Copper SLU, 2019. GA-Re 06 LCCA Form Crespo, A., Sola, A., Guillén, A., Gómez, J., Amadi-Echendu, J.E.: Planning major overhaul and equipment renovation based on asset criticality and health index. In: World Congress on Engineering Asset Management, pp. 83–90. Springer, Cham (2019) Crespo Márquez, A., Parra Márquez, C., Gómez Fernández, J.F., López Campos, M., GonzálezPrida Díaz, V.: Life cycle cost analysis. In Asset Management, pp. 81–99. Springer, Dordrecht (2012) De La Fuente, A., et al.: Strategic view of an assets health index for making long-term decisions in different industries. In Safety and Reliability–Safe Societies in a Changing World, pp. 1151– 1156. CRC Press (2018) IEC 60300-3-3:2017: Dependability management - Part 3-3: Application guide - Life cycle costing (2017) Márquez, A.C., Parajes, J.S., de la Fuente Carmona, A., Rosique, A.S.: Integrating complex asset health modelling techniques with continuous time simulation modelling: A practical tool for maintenance and capital investments analysis. Comput. Ind. 133, 103507 (2021) Nowlan, F.S., Heap, H.F.: Reliability-centered maintenance. United Air Lines Inc San Francisco Ca (1978) Roda, I., Macchi, M., Albanese, S.: Building a Total Cost of Ownership model to support manufacturing asset lifecycle management. Prod. Plann. Control 31(1), 19–37 (2020) UK DNO Common Network Asset Indices Methodology. Health and Criticality. Version 1.1. January 30th (2017)

Use Proposal of the Asset Health Index in the Public Health Sector. A Case Study in the Health Systems of the Republic of Costa Rica B. Picado Arguello(B) Engineering Mechanics and of Industrial Organization, University of Seville, Seville, Spain [email protected]

Abstract. Through time, the assets used to provide public health services in different countries and regions have become more relevant in terms of being considered key elements in the provision of quality service. Aspects such as high availability, less impact due to unexpected failures, and ease of recovery of the required asset function have become objectives of the managers of health centers and hospitals, intending to avoid serious effects on the level of social service they provide, such as an increase in waiting lists for treatments, long delivery times for response for test results and others. Maintenance management in this sector becomes a key aspect, especially the application of methodologies that allow improvement in decision-making in situations of uncertainty, such as the methodologies associated with the Asset Life Cycle Analysis that allow identifying the real state of its operation and the evaluation of alternatives for replacing assets. In the case of the public health system of the Republic of Costa Rica, managed by the Costa Rican Social Security Fund, hereinafter the CCSS, which is recognized worldwide as an example of a successful model in developing countries, there is an Institutional Management System (SIGMI). This article proposes a contribution to integrating into the SIGMI the use of the methodology called asset health index (HI(t)) that provides a greater degree of technical rigor and a quantitative approach that allows reinforcing the process of evaluating the state of operation of critical assets and the needs of replacement or technological update, beyond considerations such as the manufacturer’s estimated service life or the expected degradation at a certain period of the life cycle. Keywords: Public health · Maintenance strategies · Lifecycle · AHI

1 Introduction The adjustments in budgets, the needs that arise from unexpected events such as the situation of attention of the global pandemic and the necessary justification of the adequate use of public resources, in addition to the degree of dissatisfaction of users that has been detected in recent years such as 452 days in the waiting list for general surgery (OCDE © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 240–251, 2023. https://doi.org/10.1007/978-3-031-25448-2_23

Use Proposal of the Asset Health Index

241

2017) make it necessary to incorporate into the asset maintenance strategy implemented by the CCSS recognized methodologies to improve the maintenance management. Currently, each hospital defines its maintenance plan which is focused on predetermined maintenance based on functioning hours (Barboza Arguedas, 2013). Methodologies such as the Asset Health Index (AHI) consider the theoretical life cycle as a basic element that, added to the influence of other factors such as the (UK DNO 2017) technical location of the asset, location, frequency of use, and others associated with Operation and Maintenance, establishes an index that places the asset in a scenario of proximity or remoteness to the rearward of its useful life. All these factors are quantified and allow for the calculation of a new theoretical profile of the asset, which becomes a valuable input of information for decision-making. The Asset Health Index (HI1 ) corresponds to a dimensionless number between 1 and 10 where 1 identifies the condition of the equipment as new and the value 10 to the condition of the equipment at the end of its useful life. Its calculation based on data and technical characteristics of the equipment, location, and operating condition supposes an exponential behavior of the index against the age of the same, as shown in Fig. 1.

Health Index 10 9 8 7 6 5 4 3 2 1 0

IH5 IH4 IH3 IH2

IH1

0

5

10

15

20

25

30 35 Time

40

45

50

55

60

Fig. 1. Asset Health Index (HI) ranges. Source: DNO (2017).

The HI1 range, between the health values H = 0.5 and H = 4, is associated with the behavior of a new computer. The HI2 range, between H = 4 and H = 6, corresponds to a period in which the asset shows the first symptoms of deterioration, within this range the value corresponding to H = 5.5 is equivalent to the estimated normal life. For values greater than 5.5, the methodology establishes three intervals where the asset initiates a relevant degradation period corresponding to IH3 = 6, IH4 = 7, and IH5 = 8. From a range of 8, it is determined that the asset is at the end of its useful life. The organization currently has a procedure called “Evaluation and planning of the replacement of industrial equipment” where guidelines and considerations are defined to 1 HI Asset Health Index.

242

B. Picado Arguello

plan the replacement of this type of asset in all health centers managed by the organization through the preliminary evaluation of the state of operation concerning its estimated useful life, and where the following are considered as evaluation criteria: degree of compliance with the intended use (demand and quality of the services required), degree of compliance with safety and the environment, cost of maintenance, state of technology, availability of parts and spare parts, and percentage of useful life achieved. Specifically, this document is that it is proposed to incorporate the calculation of asset health index (HI) as a complement and methodology to comprehensively quantify the real and remaining useful life of each asset, which will allow the decision-making of replacement, replacement of equipment, prioritization of repair or technological update within a group of similar assets based on in quantitative technical support.

2 Scope This article describes a proposal for the application of the Asset Health Index on a group of Electric Generators, assets categorized as industrial equipment2 , whose function is to provide service or auxiliary support to other assets or systems, such as energy supply, compressed air, air conditioning, water treatment and the like. Electric generators are located within the hierarchy of assets used by the organization as assets of high criticality since their unavailability or low reliability can negatively affect the operation of other systems, whether medical or industrial equipment and in turn directly affect the care of patients and the preservation of their health. Electric generators are used in case of unavailability of the main electrical system supplying energy to both the life safety systems and the critical systems indispensable to keep the critical areas operating. They also contemplate preserving the operation of fire detection and alarm systems, access control, elevators, critical air conditioners, refrigerated drug protection systems, and diagnostic and treatment medical equipment, among others. For this study, a group of electric generators installed in different geographical locations was selected, to analyze if the climatological and operating conditions may affect their useful life. In addition to the extreme conditions of tropical climate, Costa Rica has different microclimates and other conditions such as effects on the electricity supply, emissions of volcanic ash, and electric discharges from storms, among others, which can also affect the asasset’sxpected life.

3 Application of the Asset Health Index (AHI) Calculation The calculation of the Asset Health Index (HI) is a methodology developed by the working group on network asset indices methodology (IAM, UK) that is displayed in the document DNO Common Network Asset Indices Methodology (2017), which seeks to quantify the state of operation of the asset and the proximity to the fulfillment of its useful 2 For maintenance and operation purposes, the organization has defined the following division

of assets: medical equipment, infrastructure (civil works) and industrial equipment.

Use Proposal of the Asset Health Index

243

life, providing a technical basis to inform the decision of replacement, prioritization of technological upgrades or major repair. By evaluating individual factors and the criteria of experts in each of the areas where the equipment is located, as well as a homologation of criteria, it is possible to use the results of operational observations, field inspections, and data capture, to simulate the real life cycle of an asset. This estimate allows knowing the current state of the asset based on its natural life cycle and the future state, in both cases providing valuable information for making investment and replacement decisions. The procedure for calculating the Asset Health Index (HI) consists of six stages as described in Fig. 2

Fig. 2. Procedure for the calculation of the Asset Health Index (HI(t)). Source: Ingeman (2022)

Below are the activities developed for the application of this methodology. • Step 1: Select the asset. Initial data collection and obtaining the estimated normal life (ENL) In this step, the identification of the asset is carried out and all the information associated with its technical location and any relevant characteristics are collected. The basic identification data of the asset is obtained from the technical manual of the equipment supplied by the manufacturer. To record this information, data a sheet is prepared with data such as make, model, series, technical characteristics, and considerations for operation and maintenance. For the selection of assets located in different geographical locations, a field visit was made to a sample of nine hospitals, where a preliminary characterization was obtained

244

B. Picado Arguello

and agreed with those responsible for the maintenance areas. Finally, four types of hospitals are defined whose location and data availability characteristics allow the methodology to be applied In the sample, two assets are considered for each hospital. Table 1 shows the characteristics of each selected hospital. Table 1. Characteristics of the selected health centers Name

Localization features

Hospital 1

Hospital C provincial category. Close to active volcanoes. Located in a central area of the country with relatively low temperatures such as areas of a similar location. The average age of 50%

Hospital 2

Hospital C provincial category. Close to coastal areas ( 30 ◦ C)

Hospital 3

Hospital C cantonal allegory. Located in the center of the country. Far from volcanoes and coastal areas. High % relative humidity (70%) in the rainy season. Moderate average temperature (25 °C)

Hospital 4

Hospital Provincial category. Located in the center of the country. The high volume of operations throughout the region was studied. Moderate temperature (25 ºC) and moderate average relative humidity (60%)

The estimated normal life (ENL) of the asset is obtained from the information provided by the different manufacturers and from the considerations of the area of maintenance and institutional equipment and is taken as a basis for the calculations to be carried out in the following steps. • Step 2: Evaluation of the impact of the location and load factors and obtaining the estimated life (EL) This step evaluates how the impact of the location and load factors associated with the location of the asset affect the estimated normal life (ENL) and determines the estimated life of the asset (EL). The location evaluation factors are determined under expert criteria of the group of maintenance engineers of the different hospitals and represent the geographical or climatological conditions that have a direct impact on the real health of the asset, as shown in Table 2. The definition of placement factor (FLT ) for each asset evaluated is defined as the maximum of the factors obtained by the following formula: FLT = max(FLT 1 , F LT 2 , FLT 3 , F LT 4 ) The load factor (FC ) assessment determines the load capacity under normal operating conditions or at the warranty point at a certain location about the maximum permissible load; to determine the load factor the following formula applies: FL =

Load capacity under normal conditions or warranty point Maximum permissible load

Use Proposal of the Asset Health Index

245

Table 2. Localization evaluation factors Factor “Proximity to volcanoes”

Factor “Proximity to the sea”

Distance in km Factor Value Distance in km Factor Value 0 km–1 km

1,6

0 km–10 km

1,6

11 km–5 km

1,2

10 km–25 km

1,2

5 km–10 km

1,1

25 km–50 km

1,1

10 km–20 km

1,05

50 km–80 km

1,05

>20 km

1

>80 km

1

“High humidity” factor

“Temperature” factor

% humidity

Factor Value Distance in km Factor Value

0%–30%

1

0 °C–20 °C

1

30%–50%

1,05

20 °C–30 °C

1,05

50%–70%

1,1

30 °C–35 °C

1,1

70%–90%

1,2

35 °C–40 °C

1,2

>90%

1,6

>40 °C

1,6

In determining the maximum permissible load factor, the data extracted from the equipment’s start-up and commissioning records and the manufacturer’s operation manual were taken into account. Table 3 shows the values obtained for each location factor (FLT) and load factor) in the electrical generators analyzed. For example, electric generator No.1 has a placement factor (FLT ) of 1.2 and a load factor (FL ) of 0.86. To determine the estimated life of the asset (EL) is the quotient that results from dividing the estimated normal life (ENL) between the product of the location (FLT ) and load (FL ) factors, as shown in the following formula: Estimated life(EL) =

Estimated normal life(ENL) FLT xFL

• Step 3. Calculation of the aging rate Since this methodology has a fundamental hypothesis that the aging of an asset shows an exponential behavior concerning its age, the aging rate becomes the parameter of the methodology that allows expressing its behavior mathematically. The aging rate is represented by β and is obtained by applying the natural logarithm of the quotient obtained from dividing the health associated with the new asset and the health it would have when it reaches its estimated life. To do this, the following formula is applied: β=

HI El ln HI new

EL

246

B. Picado Arguello Table 3. Summary of location and load factors

Equipment #

Location

Placement factor (FE ) Proximity factor to volcanoes

Proximity to the sea factor

High Humidity Factor

Extreme Temperature Factor

Load factor (FC )

E.G. Nº1

Hospital 1

1,2

1

1,1

1,05

0,86

E.G. Nº2

Hospital 1

1,2

1

1,1

1,05

1,10

E.G. Nº3

Hospital 1

1,2

1

1,1

1,05

1,00

E.G. Nº4

Hospital 2

1,05

1,6

1,2

1,1

0,85

E.G. Nº5

Hospital 2

1,05

1,6

1,2

1,1

0,74

E.G. Nº6

Hospital 3

1

1,1

1,1

1,05

1,00

E.G. Nº7

Hospital 3

1

1,1

1,1

1,05

0,88

E.G. Nº8

Hospital 4

1,05

1

1

1,5

0,95

E.G. Nº9

Hospital 4

1,05

1

1

1,5

0,90

where: • • • •

β = Aging rate EL = Estimated life = time calculated in step 2 HI new = 1 (Health value corresponding to a new asset) HIEL = HI estimated life = 5.5 (health value corresponding to a computer that has reached its estimated lifetime)

The result of the determination of the aging rate for the analyzed assets is shown in Table 4. • Step 4. Obtaining the Initial Health Index (HIi(t) ) The initial health index (HIi(t) ) represents the asset’s health index considering the aging rate for its current age in units of time. This value corresponds, According to the methodology, this value corresponds to a dimensionless number between 1 and 10, and reflects an exponential behavior concerning the team’s age (t) the following formula is applied: HI i(t) = HInew ∗ βt where • • • •

HIi(t) = Initial Health Index HI new = 1 β = aging rate t = current age of the asset in units of time

Use Proposal of the Asset Health Index

247

Table 4 shows values obtained from the application of the calculations made in the previous steps for each of the assets under study: Table 4. Estimation of life and rate of aging Equipment Location ALL FL Load FLT Location # Estimated factor Factor normal life (thousands h)

EL B Aging Estimated rate Life (thousand h)

HI i(t) Initial Health Index

E.G. Nº1

Hospital 10 1

0,86

1,2

9,707

0,000240 0,75

E.G. Nº2

Hospital 12 1

1,10

1,2

9,091

0,000200 1,50

E.G. Nº3

Hospital 10 1

1,00

1,2

8,333

0,000240 2,53

E.G. Nº4

Hospital 10 2

0,85

1,6

7,353

0,000240 3,40

E.G. Nº5

Hospital 10 2

0,74

1,6

8,473

0,000240 2,53

E.G. Nº6

Hospital 12 3

1,00

1,1

10,909

0,000200 1,93

E.G. Nº7

Hospital 10 3

0,88

1,1

10,331

0,000240 1,85

E.G. Nº8

Hospital 12 4

0,95

1,5

8,421

0,000200 2,10

E.G. Nº9

Hospital 10 4

0,90

1,5

7,407

0,000240 1,79

• Step 5. Obtaining the real health index (HIi Real(t)) This step adjusts the Initial Health Index obtained in step 4 considering a load adjustment factor. This is done because the methodology contemplates an adjustment to the initial load factor defined in step 2, to consider the differences or variations between the estimated normal load or the load at the guarantee point with the actual load records obtained during periods of operation of the equipment. This setting is identified as AC(t), and is calculated using the following formula: HI i Real(t) = HInew ∗ Ac (t)∗β∗t

248

B. Picado Arguello

To determine the load setting AC(t) the following formula is used: AC(t) • • • • •

FLr(t) = = FL

Actual load (t) Maximum possible load Load capacity under normal conditions or warranty point Maximum possible load

HI i Real(t) = Actual Health Index HI new = 1 AC(t) = load adjustment β = Rate of asset aging t = current age of the asset in units of time

• Step 6. Determination of the asset health index (HI(t)) In this last step, the health index of the active (HI(t)) is determined by adjusting this index with additional factors that condition its health, grouped into two categories: health modifiers and reliability modifiers. The obtaining of real health index of the asset (HI(t)) is obtained using the formula: HI (t) = [HIi Real(t)]e

(MS(t)+MF(t))

where: • • • •

HI(t) = Asset Health Index HI i Real(t) = Modified Initial Health Index MS (t) = Asset Health Modifier (condition and operation) MF (t) = Asset Reliability Modifier

It should be noted that the values of health and reliability modifiers normally take values within the range [0, 1]. The result of the multiplication in the final calculation of the HI(t) must be adjusted since it cannot be greater than 10 as stipulated by the model in its methodology and assumptions. Asset Health Modifier (Condition and Operation). MH(t) As modifiers of the health of the asset (Table 5) are considered those factors of condition and operation that affect the functioning of the asset; such as analysis and test results performed on the asset (thermography, oil analysis, vibration test, interpretation and analysis of operating data, such as voltage, operating temperature, and the like). For the specific case of electric generators, the response factor and the number of stars are used as a modifier. The value of the factor is determined under expert criteria and based on affectations detected through historical records.

Use Proposal of the Asset Health Index

249

Table 5. Asset health modifiers. Response factor and number of starts Response

Starts

Response factor (%)

Factor value

Starting factor (quantity)

Factor value

95%

1

>15

1,6

90%

1,2

10 < x < 12

1,4

80%

1,4

6 < x < 10

1,2

70%

1,6

1 year specific quality level with impact in training) and specific to the schedule and cost industry

4

Only the execution of the knowledge area can be outsourced but not the Management/Supervision activities

The knowledge area is complex (>1 year specific training) and specific to the industry sector

Significant impact on production or quality level with impact in schedule or cost

5

The knowledge area cannot be outsourced as it is strategic for the company. It can be patented

The knowledge area is complex (>1 year specific training) and specific to the company / product

Significant impact on production or quality level with impact in schedule and cost Loss of new business opportunities

Impact on production or quality level with impact in schedule or cost

The level values for each category are selected from the above ranking level definition, where they are drawn into the formula which calculates the "Overall Critically Score". Whichever has the highest score (Transfer/Outsourcing, Unique or Complex Knowledge) is multiplied by Operational Impact, so the Overall criticality score is calculated. The maximum value for asset criticality is set to 25 dimensionless units (notice that 25 = (MAX score 5) x 5, when substituting max values in the above mentioned formula). The criteria for prioritizing the criticality of assets is shown in Table 2.

Methods for the Criticality Assessment of Intangible Assets

361

Knowledge areas that fall anywhere within the Red category will be classed as "Core" knowledge. Knowledge areas that fall anywhere within the Orange category are considered risky and shall be managed at a lower priority than the Red category. Finally, knowledge areas that fall anywhere within the Green category are considered "Non-Core" and shall be evaluated accordingly for their future management. Table 2. Criticality matrix for knowledge Overall Cricality Score Criteria

5

5

10

15

20

25

4

4

8

12

16

20

3

3

6

9

12

15

2

2

4

6

8

10

1

(1-5) = Non Crical

Operaonal Impact

(15-25) = Crical (6-12) = Moderate Risk

1

2

3

4

5

1

2

3

4

5

Knowledge Level

4 Criticality Assessment Based on AHP 4.1 AHP Brief Description The Analytical Hierarchy Process (AHP) is a method proposed by prof. Saaty [7] useful for decision making. It consists of a mathematical model to evaluate alternatives when several criteria are met. In order to carry out the AHP method, it is necessary to provide a subjective evaluation regarding the relative importance of each of the criteria [8]. Among others, the AHP Method has the following advantages over other decision methods such as: • • • • •

It has a mathematical support. It allows breaking down problems by parts. It allows considering qualitative and quantitative elements through an evaluation scale. It may include the joint participation of an interest group. It allows correction in case of inconsistency when evaluating the criteria.

For the purposes of this contribution, it will be used to obtain the weighting of each competency and thus be able to obtain from the most critical competency to the least critical for each work process. In other words, the method will consider the following steps: i.

Definition of criteria to use and to compare from each other (for the next case study, the competences have been defined in advance).

362

V. González-Prida et al.

ii. Assessment of competences (criteria) in a matrix, in order to establish the relative criticality between them. This will be defined according to a punctuation where score “1” will be used if both competencies are equally critical and the score “9” will be used if they are extreme (one very critical versus the other). iii. Subsequently, in a judgment matrix, a priority vector is calculated and used to weight (compare) the elements of the matrix. Mathematically, the method demonstrates that the normalized eigenvector obtained from the matrix is the best evaluation approach of the analyzed criteria. The AHP technique allows the analyst to evaluate also the congruence of the judgments with the Inconsistency Rate (IR ). Before determining an inconsistency, it is necessary to estimate the Consistency Index (CI) of an n x n matrix of judgments, where CI and IR are defined by:

RI is a random index of one pairwise comparison matrix (tabulated in [7]). With this, judgments can be considered acceptable if IR is less than or equal to 0.1. In cases of inconsistency, the assessment process for the evaluated matrix has to be repeated. 4.2 Case Study As mentioned before, the case considers a business context where competences and skills have been defined in advance. That means, the organization has carried out the necessary surveys to its staff, analyzed its processes (strategic, operational and support processes), and has evaluated the sector/market along their trends. With all this, competences are identified and classified. A first classification can be observed in Table 3: Table 3. Some competences identified for the business goal and strategy Traditional Competences

Systems Engineering Electronic Architecture Calculation & Simulation

Prototyping Testing Systems Integration Product support

Disruptive Competences

Data management Artificial Intelligence Autonomous/robotics

Cybersecurity Model Based Engineering VR/AR/XR

Once identified the necessary competences, the next step is to analyze the relative importance of each one and their prioritization. That means, which one is more important than the other, analyzing its criticality for the organization. In the praxis, it is performed by expert panels together with leaders of each functional area, where Knowledge needs are assessed and based on:

Methods for the Criticality Assessment of Intangible Assets

363

• Current projects and opportunities demands (tactical) • Company product roadmaps and technology trends (strategic) Considering just some detected competences as an illustrative example, the judgment chart can be observed in Table 4. After obtaining the criteria matrix, it has to be transformed to a normalized matrix, and, once it is done, each competence has to be weighted, obtaining the result shown in Table 5. Applying the a.m. formulas (see Sect. 2) about Consistency Index (CI) and Inconsistency Rate (IR ), it is obtained a CI = 0,064698897 and IR = 0,049014316 < 0,1 thus the judgment can be considered consistent. With this, the ranking of competences according to its criticality can be observed in Table 6. Table 4. Competences judgment chart VR/AR/XR Calculation Cybersecurity Robotics Systems AI Electronic & Eng Arch Simulation VR/AR/XR

1

3

1/4

1/3

1/2

1/5 2

Calculation & 1/3 Simulation

1

1/6

1/5

1/4

1/7 1/2

Cybersecurity 4

6

1

2

3

1/2 5

Robotics

3

5

1/2

1

2

1/3 4

Systems Eng

2

5

1/3

1/2

1

1/4 4

AI

5

7

2

3

4

1

Electronic Arch

1/2

2

1/5

1/4

1/3

1/6 1

6

Table 5. Competences weighting VR/AR/XR Calculation Cybersecurity Robotics Systems AI & Eng. Simulation Weight 6,85%

3,15%

23,49%

15,69%

11,52%

Electronic Arch.

34,74% 4,55%

364

V. González-Prida et al. Table 6. Competences criticality ranking AI Cybersecurity Robotics Systems VR/AR/XR Electronic Calculation Eng. Arch. & Simulation

Criticality 1 ranking

2

3

4

5

6

7

5 Conclusions Engineering-related studies are commonly developed under the threshold of exact sciences or under stochastic models that help to analyze or predict certain phenomena. However, in this case, the object of study has been an element (knowledge) that, due to its intangible nature, has to be managed under the threshold of qualitative management models. In other words, the management of intangible assets such as knowledge requires the development of non-conventional models such as those that can be developed and used for control and prediction applied to engineering assets. In particular, specific models for knowledge management are necessarily qualitative or semi-quantitative models that provide objective information but without the level of precision and accuracy that other asset management techniques can demand, especially reliability analysis techniques or data analytics techniques for predictive maintenance. The proposed framework intents to be an aid for this management, organizing under a coherent manner a line of action and suggesting specific and realistic tools. Along that line of action, knowledge priority setting can be performed by a criticality analysis and, among the different possible techniques, one based on Risk Analysis has been depicted. In addition to this, another alternative is the used of AHP, which has been applied in a case study. Regarding such case, it is important to consider also that, once the criticality of factors (competences) has been assessed, the next step will be to evaluate whether those that are less critical are really necessary for the business goal or if they are aligned with the business strategy. With this, the company can get quick wins either by eliminating them, outsourcing them or spending less time on their development or maintenance. Future researches or contribution can provide more real life’s cases in order to keep on verifying and validating the stages established in the framework approach.

References 1. Gonzalez-Prida, V., Parra, C., Guillén, A., Candón, E., Martinez-Galán, P.: An intangible asset management proposal based on ISO 55001 and ISO 30401 for knowledge management. WCEAM 15th , Brazil, July 2021 2. Crespo, A.: The maintenance management framework: models and methods for complex systems maintenance. Springer Science & Business Media (2007). https://doi.org/10.1007/9781-84628-821-0. ISBN 9781846288210 3. ISO 55001:2015. Asset management — Management systems — Requirements. International Organization for Standardization

Methods for the Criticality Assessment of Intangible Assets

365

4. ISO 30401:2018. Knowledge management systems — Requirements. International Organization for Standardization 5. Khaira, A., Dwivedi, R.K.: A state of the art review of analytical hierarchy process. Mater. Today: Proc. 5(2), 4029–4035 (2018) 6. Jagtap, H.P., Bewoor, A.K.: Use of analytic hierarchy process methodology for criticality analysis of thermal power plant equipment. Mater. Today: Proc. 4 (2), 1927-1936 (2017). Part A, 1927–1936, ISSN 2214–7853 7. Saaty, T.L.: Decision making—the analytic hierarchy and network processes (AHP/ANP). J. Syst. Sci. Syst. Eng. 13(1), 1–35 (2004) 8. González-Prida, V., Viveros, P., Barberá, L., Márquez, A.C.: Dynamic analytic hierarchy process: AHP method adapted to a changing environment. J. Manufact. Technol. Manag., (2014)

What is Smart Maintenance in Manufacturing Industry? Antti Salonen(B) Mälardalen University, Eskilstuna, Sweden [email protected]

Abstract. The ongoing transformation of manufacturing industry into digitalized production, Industry 4.0, has put new perspectives on the maintenance of production systems. The technologies offer an array of new possibilities in optimization of maintenance and data driven decision making. On the other hand, these new technologies offer a lot of challenges in form of investment costs, need for new competences, and how to handle the equipment legacy, i.e. upgrading old equipment. Many researchers associate data driven decision making with intelligent sensors, cloud computing and cyber physical systems, but are these technologies the most cost-effective way of achieving data driven maintenance? The aim of this paper is to discuss how manufacturing industry should approach smart maintenance in order to improve the industry’s competitiveness, rather than spending money on technology that doesn’t contribute. The basis for the discussion will mainly be a literature study but additional empirical data may be included.

1 Introduction With the introduction of Industry 4.0, there’s an increasing interest for utilizing the new digital technologies in the maintenance domain. Previous research presents nine technological pillars of industry 4.0: 1) Industrial Internet of Things (IIoT), 2) Big Data and Analytics, 3) Horizontal and Vertical System Integration, 4) Simulation, 5) Cloud Computing, 6) Augmented Reality (AR), 7) Autonomous Robots, 8) Additive Manufacturing (AM), and 9) Cyber Security (Alcácer and Cruz-Machado 2019; Vaidya et al. 2018). 1.1 Smart Maintenance The technologies introduced through Industry 4.0 increase the demand for first class maintenance, but except for new demands, the technologies also introduce new possibilities in the maintenance domain (Silvestri et. al. 2020). Through IIoT and cloud computing, production equipment can be connected and continuously collect condition data, e.g. vibrations, pressure, and temperature (Amruthnath and Gupta 2018). Big data and Analytics offers new means of analysis, thereby supporting decision making and planning (Silvestri et al. 2020). Also, Artificial Intelligence (AI), especially Machine © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 366–374, 2023. https://doi.org/10.1007/978-3-031-25448-2_35

What is Smart Maintenance in Manufacturing Industry?

367

Learning is suggested for improving analytics of maintenance data in order to apply Predictive Maintenance (Carvalho et al. 2019). One technology that has received a lot of interest in the maintenance domain is Augmented Reality (Rauch et al. 2019; Mourtzis et. al. 2017; Palmarini et. al. 2018). Applications such as instructions for autonomous maintenance, and remote maintenance, i.e. local staff get support through AR from remote experts, are the more common (Masoni et. al. 2017; Webel et al. 2013; Fiorentino et al. 2014). However, the implementation of these new technologies is not unproblematic. Savolainen et al. (2020), found that maintenance organizations are dealing with organizational, as well as technical, and managerial challenges when implementing data driven decision making in maintenance. Through a literature study, Forcina et al. (2021) found that Industrial Internet of Things and Cloud Computing generally were mentioned as enablers for Industry 4.0 as well as Maintenance 4.0. However, Ashjaei and Bengtsson (2017) point out that cloud computing is problematic from the aspects of privacy, security, low latency, and high availability. Moubray (1997) assessed that condition monitoring only is feasible in 20% of failure modes, and further, only worth doing in half of those cases. Another misunderstanding with CBM is that it really doesn’t prevent components from deterioration, but rather prevent functional failures from happening, thus focusing on the P-F interval. Bengtsson and Lundström (2018), point out the importance of understanding the I-P interval, where the deterioration is initiated, in many cases, due to poor cleaning, poor lubrication or mis use of the equipment. Also, the utilization of AI in maintenance related analytics is challenging. Authors like Uddin Ahmed et al. (2021) showed some of the challenges of utilizing Natural language processing and Machine learning on free text fields in Computerized Maintenance Management Systems, CMMS. On the other hand, Stenström et al. (2015) concluded that the use of NLP improved the identification of failure causes. One important difference between these two studies is the number of failure records, being 1700 in Uddin Ahmed et al. (2021), while Stenström et al. (2015) analyzed close to 11 000 failure records. 1.2 Human Errors The emergence of Industry 4.0 changes the role of humans on the shop floor. One vision for Industry 4.0 is the operator less factory, fully automated in physical processing as well as decision making (Benesova and Tupa 2017). This operator less factory is most probably rather distant, but still it is clear that the operator’s role has changed, especially through increased automation and introduction of advanced ICT (Barroso and Wilson 2000). In most discrete item manufacturing industry, the level of automation is comparatively low. Therefore, human actions have a significantly high impact on the dependability of the production system. According to Nakajima (1989, p.99): “…breakdowns are the results of human factors – the erroneous assumptions and beliefs of engineers, maintenance personnel, and equipment operators”. This statement is supported by Salonen (2018), who in a case study found that about 20–50% of breakdowns in the studied companies were caused by human errors. Similar assessments were reported in a study on maintenance planning in automotive industry, where the maintenance planners

368

A. Salonen

saw human errors as the main obstacle for reduction of breakdowns in the production equipment (Salonen and Gopalakrishnan 2020). Still, according to Sheikhalishahi et al. (2015), the majority of studies of human errors in maintenance are focused on reliability centered industries, e.g. aviation, nuclear, and chemical process industries, while discrete item manufacturing is less studied. One exception is Böllhoff et al. (2016) who present a study of human error probability in cellular manufacturing. The study showed that the most common form of human errors (43%), was omissions, most frequently in the form of neglected cleaning. GonzálezPrida et al. (2022), discuss human reliability and define it as “the ability of a person (an operator) to fulfil a required function under given conditions for a given period of time” (p. 162). After a summary of various human error analysis methods, they conclude that the structure of the presented methods, often make them difficult to apply in other settings than the ones they were designed for. The aim of this paper is to discuss how manufacturing industry should approach smart maintenance in order to improve the industry’s competitiveness, rather than spending money on technology that doesn’t contribute.

2 State of Practice In this paper the empirical examples are mainly collected from previous research. Some additional examples are taken from the author’s 12 years’ experience in maintenance of industrial manufacturing equipment. Most examples are based on case studies, and thus, build an empirical base for the coming discussion. 2.1 Readiness in Manufacturing Industry Through a survey, Giliyana et al. (2022) studied what experiences and what attitudes 11 large manufacturing companies in Sweden had of “Smart maintenance”. The outcome showed that the companies, so far, have rather limited experience of smart maintenance technologies. Three of the companies have no experience at all, and a fourth company only had some smaller pilot studies within their operations. The respondents found it challenging to implement smart maintenance technologies cost-effectively. Another survey by Fusko et-al. (2018) showed that only 12.5% of the respondents considered themselves ready for transforming their maintenance into the digital technologies. Frost et al. (2019), propose a maintenance system design for Maintenance 4.0. The design emphasises the first stage setting the basic maintenance processes in order to build a strong foundation for the maintenance program at the company, see Fig. 1.

What is Smart Maintenance in Manufacturing Industry?

369

Fig. 1. A maintenance system design for maintenance 4.0 (Adapted from Frost et al. 2019)

By analysis of three years of failure records from two production cells, containing 11 machining centres, Salonen et al. (2020), found that about 40–56% of the failures were in components suitable for condition monitoring. However, 41% of the failures were in components with random lifetime distribution and no measurable deterioration, which means they cannot be prevented. Further, it should be noted that 20% of the recorded failures were due to human errors, and many of these were in components suitable for condition monitoring. This means that it is likely that the components would break down without any noted deterioration taking place. 2.2 Considering the I-P Interval While condition monitoring is an essential means of avoiding severe breakdowns, it doesn’t per se prevent failures from occurring. Hence, Plucknette (2010), points out that it should be the objective of every maintenance organization to work on maximizing the I-P interval as shown in Fig. 2 (i.e. the time from component installation until potential failure), and to do so by applying reliability centered maintenance (RCM) and/or other proactive maintenance techniques. Also, Salonen (2018) means that industry needs to apply a holistic view of dependability. “A good starting point for this, is to map the factors that limit the dependability, e.g. through root cause analysis, not only of failures and down-time, but of all deviations from the expected systems performance.” (Salonen 2018, p. 22). Bengtsson and Lundström (2018), present a case study where a thorough root cause analysis of bearings in a fan system led to change of lubrication procedures and thus drastically improved the I-P interval and hence the dependability of the system.

370

A. Salonen

Fig. 2. Example of an i-p interval of a ball bearing (Bengtsson and Lundström 2018)

Another example from industry is in an electronics manufacturing site, where bearings in the solder pumps of a wave soldering machine broke, on average, every 10 months. The maintenance staff realized that the bearings should have a substantially longer life length. Through a root cause analysis, the maintenance engineer found that the bearing type, used in the pumps had too small internal radial clearance to cope with the operating temperature of 270° Celsius. By changing bearing type to one with greater radial clearance, the life length increased dramatically. 2.3 Human Factors Most papers discussing the role of the operator and the maintainer in Industry 4.0, tend to focus on human-machine interaction. How these new technologies can be utilized for the reduction of human errors in operations, or in maintenance is not as commonly discussed.

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

Un-known External influence Soware failure Poor material/design End of lifeme Poor installaon (PM) Poor PM Poor cleaning Overload

A

B

C

Poor handling

Fig. 3. Root Cause Categories for breakdowns in three manufacturing sites (Salonen 2018)

What is Smart Maintenance in Manufacturing Industry?

371

By analysing the root causes of breakdowns in three manufacturing sites in a large company within automotive industry, Salonen (2018) found that between 50 and 63% of the breakdowns were due to the categories Poor handling, Overload, Poor cleaning, Poor preventive maintenance (Lack of PM), and Poor installation (Poorly performed PM), see Fig. 3. All these categories are directly related to human errors. The root cause category “End of lifetime” accounted for 1%, 7%, and 15% in the three production sites. This clearly indicates that most breakdowns in these plants are due to some form of abuse of ideal operating conditions. Further, Salonen (2018) found that poor autonomous maintenance is one substantial category of human errors in Swedish automotive manufacturing. One cause of this is insufficient training, and another one lack of time. The production leaders often underestimate the importance of autonomous maintenance and prioritize to maximize the production time. Also, according to one maintenance manager, an emerging problem is that there is a trend that fewer and fewer operators are supposed to do autonomous maintenance on larger and increasingly complex production systems. According to an informant, working as a maintenance specialist in one case company, a major problem is that manufacturing engineers are not always fully aware of the limits of the production equipment. In order to cut cycle times, they buy new types of tools that allow higher working speed in the machining process. The machine is then run beyond the recommended force limits for the ball screws, thus incurring an increased wear on the screws. It is also important to point out that several researchers highlight the importance of considering organizational maturity and the role of humans when implementing technologies, associated with Industry 4.0 (e.g. Baglee et al. 2016; Salzer 2017; Havle and Ücler 2018).

3 Discussion Industry 4.0 will have a large impact on manufacturing industry and the associated technologies present major opportunities in the maintenance domain. However, as argued by e.g. Frost et. al. (2019), there is a need for a stable foundation built on the basic maintenance activities to be able to utilize these new technologies to their full potential. The findings of Salonen (2018) indicate that this foundation is not present. This is also noted by Giliyana et al. (2022) as well as Fusko et al. (2018), through direct statements from industry. Based on the empirical findings presented earlier in this paper it seems clear that the more advanced maintenance techniques, e.g. prescriptive maintenance, will have a marginal influence on the reliability of discrete manufacturing systems. These techniques are off course vital in order to prevent critical failures and disturbances to occur, but for the major part of dependability loss, there is a need for establishing excellence in the processes for basic maintenance actions, such as cleaning, lubrication and exchange of consumption parts, e.g. filters. With 20–50% of breakdowns caused by human errors that in most cases are erratic and without deterioration intervals to be traced, manufacturing industry need to review its maintenance processes. Failures due to human errors are in most cases impossible to

372

A. Salonen

predict through technological solutions, at least without violating ethics. Non the less, sensors could aid in tracking loss of speed or increase of resistance in moving parts, due to poor cleaning or lubrication. This kind of monitoring could act as a leading performance indicator in order to improve the cleaning and lubrication execution at shop floor. One promising technique is Augmented Reality which has gained a lot of attention within maintenance. Utilizing AR for improved maintenance instructions, especially for autonomous maintenance provides a good means of avoiding human errors due to poor knowledge. Also, remote maintenance, utilizing interactive AR applications for equipment experts guiding skilled maintenance generalists (Mourtzis et al. 2017; Masoni et al. 2017; Palmarini et al. 2018). SMEs often can not motivate to have all kinds of expertise in-house, but this technology allow the local maintenance staff to perform expert maintenance with supervision of experts and thereby reducing the risk of human errors in advanced maintenance activities. The total elimination of human errors is probably impossible, or as stated by Reason (1995, pp.89):”Managing human risks will never be 100% effective. Human fallibility can be moderated, but it cannot be eliminated..”. However, Gonzáles-Prida et al. (2022, pp. 163), states that: “Although it is never impossible to eliminate human error, it is possible, through good maintenance management and an understanding of the issue that affect error, to move towards this goal and to control the likelihood of error.” When analyzing maintenance records, it is essential to check for deviations from reasonable life lengths and apply root cause analysis on identified deviations with noticeable economic impact. The examples presented earlier with fan bearings in a paint shop, and solder pump bearings in electronics manufacturing are good examples where condition monitoring without consideration of proper life length could incur unnecessary costs. From a factory perspective or even a company perspective, AI could assist in analyzing and comparing RCA documentation on order to identify common cause failures and thereby eliminate the common causes entirely. It is important to point out that this requires good RCA processes and documentation in place though. Regardless of which new technologies are introduced in industrial manufacturing, a lot of the basic maintenance actions will still be performed by staff. It is therefore essential to remember what Nakajima (1989) pointed out, that the staff also needs motivation, sufficient time, and skills to perform their tasks. The smart maintenance technologies offer a lot of improvement potential, but it is important to utilize them where they provide most return on investment. Rather than tracking component deterioration or predicting the theoretical end of lifetime, these techniques should be used in order to prolong the I-P interval (i.e. lifetime) by supporting the basic maintenance actions and by preventing human errors to occur. Smart maintenance should truly have a considerable impact on the production system’s dependability, otherwise it is just wasted applications of smart technologies.

References Alcácer, V., Cruz-Machado, V.: Scanning the industry 4.0: a literature review on technologies for manufacturing systems. engineering science and technology. Int. J. 22(3), 899–919 (2019) Amruthnath, N., Gupta, T.: A research study on unsupervised machine learning algorithms for early fault detection in predictive maintenance. In: IEEE: 5th International Conference on

What is Smart Maintenance in Manufacturing Industry?

373

Industrial Engineering and Applications (ICIEA). Singapore 26–28 April, pp. 355–361 (2018). https://doi.org/10.13140/RG.2.2.28822.24648 Ashjaei, M., Bengtsson, M.: Enhancing smart maintenance management using fog computing technology. In: 2017 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), pp. 1561–1565 (2017) Baglee, D., Jantunen, E., Sharma, P.: Identifying organisational requirements for the implementation of an advanced maintenance strategy in small to medium enterprises (SME). J. Maintenance Eng. 16–26 (2016) Barroso, M.P., Wilson, J.R.: Human error and disturbance occurrence in manufacturing systems (HEDOMS): a framework and a toolkit for practical analysis. Cogn. Technol. Work 2(2), 51–61 (2000) Benešová, A., Tupa, J.: Requirements for education and qualification of people in industry 4.0. Procedia Manufact. 11, 2195–2202 (2017) Bengtsson, M., Lundström, G.: On the importance of combining the new with the old–one important prerequisite for maintenance in Industry 4.0. Procedia Manufact. 25, 118–125 (2018) Böllhoff, J., Metternich, J., Frick, N., Kruczek, M.: Evaluation of the human error probability in cellular manufacturing. Procedia CIRP 55, 218–223 (2016) Carvalho, T.P., Soares, F.A., Vita, R., Francisco, R.D.P., Basto, J.P., Alcalá, S.G.: A systematic literature review of machine learning methods applied to predictive maintenance. Comput. Ind. Eng. 137, 106024 (2019) Fiorentino, M., Uva, A.E., Gattullo, M., Debernardis, S., Monno, G.: Augmented reality on large screen for interactive maintenance instructions. Comput. Ind. 65, 270–278 (2014) Forcina, A., Introna, V., Silvestri, A.: Enabling technology for maintenance in a smart factory: a literature review. Procedia Comput. Sci. 180, 430–435 (2021) Frost, T., Nöcker, J., Demetz, J., Schmidt, M.: The evolution of maintenance 4.0 – what should companies be focusing on now?. In: The Proceedings of 4th International Conference on Maintenance Engineering, Dubai UAE, pp. 123–132 (2019) Fusko, M., Rakyta, M., Krajcovic, M., Dulina, L., Gaso, M., Grznar, P.: Basics of designing maintenance processes in Industry 4.0. MM Sci. J. 3, 2252–2259 (2018) Giliyana, S., Salonen, A., Bengtsson, M.: Perspectives on smart maintenance technologies – a case study in large manufacturing companies. In: Proceedings of the 10th Swedish Production Symposium, pp. 255–266 (2022) González-Prida, V., Parra, C., Crespo, A., Kristjanpoller, F.A., Gunckel, P.V.: Reliability engineering techniques applied to the human failure analysis process. In: Cases on Optimizing the Asset Management Process, pp. 162–179 (2022) Havle, C., Üçler, Ç.: Enablers for industry 4.0. In: IEEE International Symposium on Multidisciplinary Studies and Innovative Technologies, pp. 1–6 (2018) Masoni, R., et al.: Supporting remote maintenance in Industry 4.0 through augmented reality. Procedia Manufact. 11, 1296–1302 (2017) Moubray, J.: Reliability-Centered Maintenance. Industrial Press Inc., New York (1997) Mourtzis, D., Vlachou, E., Zogopoulos, V., Fotini, X.: Integrated production and maintenance scheduling through machine monitoring and augmented reality: an industry 4.0 approach. In: Lödding, H., Riedel, R., Thoben, K.-D., von Cieminski, G., Kiritsis, D. (eds.) APMS 2017. IAICT, vol. 513, pp. 354–362. Springer, Cham (2017). https://doi.org/10.1007/978-3319-66923-6_42 Nakajima, S.: TPM – Development program – Implementing Total Productive Maintenance. Productivity Press, Cambridge (1989) Palmarini, R., Erkoyuncu, J.A., Roy, R., Torabmostaedi, H.: A systematic review of augmented reality applications in maintenance. Robot. Comput.-Integrat. Manufact. 49, 215–228 (2018) Plucknette, D.: The introduction of the I-P interval. (https://reliabilityweb.com/articles/ entry/the_introduction_of_the_i-p_interval). Accessed 08 Feb 2022

374

A. Salonen

Rauch, E., Linder, C., Dallasega, P.: Anthropocentric perspective of production before and within industry 4.0. Comput. Ind. Eng. 139, 105644 (2019) Salonen, A.: The need for a holistic view on dependable production systems. In: Proceedia Manufacturing, no. 25, pp. 17–22 (2018) Salonen, A., Bengtsson, M., Fridholm, V.: The possibilities of improving maintenance through CMMS data analysis. In: proceedings from the Swedish Production Symposium, SPS2020, pp. 249–260 (2020) Salonen, A., Gopalakrishnan, M.: The practices of preventive maintenance planning in discrete manufacturing industry. J. Qual. Maintenance Eng. (2020) Saltzer, M.: A blueprint for digitalisation of maintenance. In: Proceedings of 2nd International Conference on Maintenance Engineering, INCOME-II, pp. 384–391 (2017) Savolainen, P., Magnusson, J., Gopalakrishnan, M., Bekar, E.T., Skoogh, A.: Organisational constraints in data-driven maintenance: a case study in the automotive industry. IFACPapersOnLine 53(3), 95–100 (2020) Sheikhalishahi, M., Pintelon, L., Azadeh, A.: Human factors in maintenance: a review. J. Qual. Maintenance Eng. (2016) Silvestri, L., Forcina, A., Introna, V., Santolamazza, A., Cesarotti, V.: Maintenance transformation through Industry 4.0 technologies: a systematic literature review. Comput. Ind. 123, 1–16 (2020) Stenström, C., Al-Jumaili, M., Parida, A.: Natural language processing of maintenance records data. Int. J. COMADEM 18(2), 33–37 (2015) Uddin Ahmed, M., Bengtsson, M., Salonen, A., Funk, P.: Analysis of breakdown reports using natural language processing and machine learning. In: Proceedings from International Congress and Workshop on In-dustrial AI 2021, Luleå, Sweden (2021) Vaidya, S., Ambad, P., Bhosle, S.: Industry 4.0 – a glimpse. Procedia Manufact. 20, 233–238 (2018) Webel, S., Bockholt, U., Engelke, T., Gavish, N., Olbrich, M., Preusche, C.: An augmented reality training platform for assembly and maintenance skills. Robot. Auton. Syst. 61(4), 398–403 (2013)

Audit Models for Asset Management, Maintenance and Reliability Processes: A Case Study Applied to the Desalination Plant Pablo Duque1 , Carlos Parra1 , Felix Pizarro1 , Andrés Aránguiz1(B) , and Emanuel Vega2 1 Department of Mechanics, Technical University Federico Santa María, Viña del Mar, Chile

{pablo.duque,carlos.parram,felix.pizarro,andres.aranguiz}@usm.cl 2 School of Computer Engineering, Catholic University of Valparaíso, Valparaíso, Chile [email protected]

Abstract. The identification of improvements, shortcomings, and potential failures applied to maintenance has taken relevant attention from the scientific community in recent years. In order to carry out timely diagnosis, the employment of methods to properly measure the reliability of industrial processes has been a trend. In this work, AMORMS and AMS-ISO 55001 are applied to a seawater desalination plant in order to measure and generate a proper improvement plan. In this context, AMORMS is a model based on 8 phases which aims at asset management. Also, AMS-ISO 55001 which is based on the asset management norm ISO55001. The results achieved include the design and generation of actions to tackle the 20% more deficient categories needed to achieve a competitive industrial performance.

1 Introduction Copper production in Chile has fallen between 17 to 24% as a consequence of the current drought, forcing copper companies to diversify their water sources. The Chilean Copper Commission estimates that during 2020, 73% of the water from mining processes was recirculated and that 70% of this water came from continental sources. In addition, it has been proposed, as a National Mining Policy, to reduce in the employment of continental water related to industrial mining process, which should not exceed 10% of the water used in 2025 (Revista Minería Chilena 2022), thus, concluding in an increment in numbers of desalination plants. The development of such a plant has been a key issue in the improvements related to the copper industry. In this context, the necessity to identify, review, and optimize the asset management and maintenance processes regarding this type of plant has taken relevant attention in the last decade. In this work, we propose the employment of two audit processes, the AMORMS (Asset Management, Operational Reliability & Survey) and the AMS-ISO 55001 (Asset Management Survey ISO 55001), in order to identify gaps in maintenance management model to a desalination plant. Regarding the results achieved, the first model classifies the plant as having “average standard processes”, the second model illustrates the classification as “Processes with © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 375–384, 2023. https://doi.org/10.1007/978-3-031-25448-2_36

376

P. Duque et al.

very good practices”. Thus, we analyze and highlights the drawbacks and gaps detected. Also, action plans are designed, generated, and proposed aiming for short and medium term.

2 General Background In this section, we describe in details the employed models and the audited areas from the plant. 2.1 Audit Models The main objective behind a maintenance audit is to carry out a measurement of different processes and areas that make up the Maintenance Management. Once the opportunities for improvement have been identified, action plans will be generated (Parra Márquez and Crespo Márquez 2015). The main differences between the models to be used are: • AMORMS is an audit which focuses on the process by evaluating the 8 phases of the Maintenance Management Model proposed by Parra and Crespo (Parra Márquez and Crespo Márquez 2015). This audit includes some topics regarding the management model that has been proposed by the standard ISO 55001. Also, it gives special emphasis on the employment of support tools for the management process. The evaluated areas can be described as follows: (1) Definition of the maintenance objectives and KPI’s, (2) Asset priority and maintenance strategy definition, (3) Immediate intervention on high impact weak point, (4) Design of the preventive maintenance plans and resources, (5) Preventive plan, Schedule and resources optimization, (6) Maintenance execution assessment and control, (7) Asset life cycle analysis and replacement optimization, and (8) Continuous Improvement and new tech. • AMS-ISO 55001 focuses on auditing the processes that make up the asset management system during its life cycle in accordance within the definition indicated in the standards set of ISO 55000 (ISO 2014). This standard evaluates the existing gaps in the following requirements: (1) Context of the organization, (2) Leadership, (3) Planning, (4) Support, (5) Operation, (6) Performance evaluation, and (7) Improvement. 2.2 Background of the Audited Unit The audits will be applied to the maintenance area of a desalination plant related to a copper extraction mining company. Currently, this plant distributes a total of 95% of the water required by the company. The plant includes 5,000 assets and is required to have an availability of 95%.

3 Audit Results In this section the results achieved by the audits are illustrated and discussed. In this process, planning engineers, reliability engineers, Field Supervisor engineers, and Maintenance Superintendent have participated.

Audit Models for Asset Management, Maintenance and Reliability Processes

377

3.1 AMORMS Audit Result In Table 1 we illustrate the results achieved by AMORMS. The description of the table is as follows, column 1 depicts each phase reviewed. The column Average, display the score achieved after carrying out the measurement of such a phase. Lastly, column Process Description, illustrates the classification given by the audits based on the score achieved. We can observe that the value illustrated in the row. X (3.88), given by the 8 computed Average values, allows this plant to have a classification of “average standard processes”. Also, it is observed that 37.5% of the processes reach a classification of having “very good practices”, and the remaining 62.5% achieved an “average standard”. In order to identify the points to improve, the Pareto Principle will be employed at two levels: Firstly, the 20% of the subcategories with the lowest score will be considered. Secondly, for each of these subcategories, the 20% of the questions with the lowest score will be considered, and action plans will be generated for the latter. Table 1. AMORMS audit results Phases Audited

Average

Process description

1

Definition of the maintenance objectives and KPI’s

3.88

Standard average

2

Asset priority and maintenance strategy definition

4.07

Very good practices

3

Immediate intervention on high impact weak point

4.00

Very good practices

4

Design of the preventive maintenance plans

4.01

Very good practices

5

Preventive plan, Schedule and resources optimization

3.97

Standard average

6

Maintenance execution assessment and control

3.78

Standard average

7

Asset life cycle analysis

3.57

Standard average

8

Continuous Improvement and new tech

3.74

Standard average

X

3.88

Standard average

The subcategories identified within the lowest 20% achieved score are illustrated in Table 2. In this regard, the subcategories have an Average that allows them to be classified as processes with “average standard”.

378

P. Duque et al. Table 2. 20% of the Subcategories with lowest score according to the AMORMS Audit.

Subcategories of Phases of the Maintenance Management Model

Average

1.2

Asset Management Plan

3.59

6.3

Operations control processes

3.57

6.5

Workshop management

3.63

7.1

Asset Life Cycle Cost Management

3.15

7.2

Management of information in the Asset Life Cycle

3.56

8.3

Staff development programs

3.11

Table 3 and 4 presents the proposed action plans designed to reduce the gaps detected, which are aligned with the company’s maintenance strategies. Thus, in order to control the progress and compliance within the action plans, indicators will be designed by the experts based on the established objectives, goals, thresholds, and definitions (AENOR 2003). Also, from the analysis of Table 3 and 4, it can be seen that 6 questions depicted reached a score that categorizes them as an “average standard” process. Moreover, when reviewing the answers from the surveys, it can be concluded that there are tools to support management processes, which are used systematically and consciously. However, they are not known by all the members of the Maintenance Management, so re-instruction and diffusion of these processes is a must. Regarding question 8.3.5, this was classified as a “below average process”, which gives the recommendation of carrying out an update in the training program. The proposed plan should be focused on developing the technical skills, such as knowledge, skills, and aptitudes, which are specified for each job position and stated in the given description. Also, the addition of a periodic instruction based on the deficient processes is recommended. Table 3. Proposed action plans for AMORMS audit Subcategory Question – Score

Action

Indicator

Comprehensive Asset Management Plan 1.2.2 Is there a comprehensive plan designed to implement the various processes proposed by the asset management model? Score: 3.17

The comprehensive plan exists, which is why it is necessary to disseminate the existing one. The dissemination will be done through training, workshops and dashboard

Name: Re-instruction of the comprehensive asset management plan   total attendees = total maintainers ∗ 100% Goal: 80%; Range: 75 to 85% Responsible: Superintendent Reliability (continued)

Audit Models for Asset Management, Maintenance and Reliability Processes

379

Table 3. (continued) Subcategory Question – Score

Action

Indicator

Operations control processes 6.3.1 Is there a procedure detailing the operational processes? Score: 3.17

The procedures exist, but they must be updated and disseminated. Dissemination will be done through training and workshops

Name: Procedures Update   updated processes = obsolete processes ∗ 100%

Workshop management 6.5.3 Is there a standard contract model developed for all the services requested from the workshops? Score: 3.57

The standard contract model exists, but it is necessary to train staff in the scope of existing contracts. Dissemination will be done through training and workshops

Name: Re instruction scope contracts   total attendees = total maintainers ∗ 100%

Workshop management 6.5.5 Is there an auditing and benchmarking model certified under a local or international standard, which allows evaluating the Services offered by the workshops? Score: 3.57

It exists, which is why it is necessary to re-instruct and reinforce the existence of the evaluation process for transversal services in external repair shops. Dissemination will be done through training and workshops

Name: Re-instruction evaluation of services   total attendees = total maintainers ∗ 100%

Goal: 80%: Range: 75 to 85% Responsible: Operations Superintendent

Goal: 80%: Range: 75 to 85% Responsible: Contract Administrators

Goal: 80%; Range: 75–85% Responsible: Reliability Superintendent

Table 4. Proposed action plans for AMORMS audit (continued) Subcategory – Question - Score

Action

Indicator

Life Cycle Cost Analysis Processes 7.1.5 Is the information on the life cycle of the assets efficiently documented and are the results of the Life Cycle of the selected equipment audited? Score: 3.00

There is an Asset management model that incorporates the Life Cycle Cost Analysis processes, for which it is necessary to re-instruct the organization’s Asset Management model, in addition to disseminating it. The dissemination will be done through training, workshops and dashboard

Name: Re-instruction Asset management model   total attendees = total maintainers ∗ 100% Goal: 80%; Range: 75–85% Responsible: Reliability Superintendent

(continued)

380

P. Duque et al. Table 4. (continued)

Subcategory – Question - Score

Action

Indicator

Although the maintenance program includes training in the areas of modern maintenance techniques, reliability and asset management, it must be reviewed and adjusted according to the needs of each job In the short term, the development of training in the indicated areas is proposed. Topics associated with the re-instructions requested in the previous points should also be included

Name: Re-instruction maintenance techniques   total attendees = total maintainers ∗ 100%

Management of information in the Asset Life Cycle 7.2.1 Does the organization’s management regularly review the key factors of its asset management system to ensure its effectiveness, suitability Score: 3.33 Staff development process 8.3.5 Does the training program include training in the areas of modern maintenance techniques, reliability and asset management? Score: 2.86

Goal: 80%; Range: 75–85% Responsible: Superintendent of Personal Development and Superintendent of Execution

3.2 AMS-ISO 55001 Audit Result According to the results illustrated in Table 5, the 7 requirements defined by the AMSISO 55001 audit can be classified as “Process with very good practices”. This can be interpreted as the organization demonstrating being systematically and carefully optimizing their own asset management practice, which is consequence of being aligned with corporative objectives and their operating context.

Audit Models for Asset Management, Maintenance and Reliability Processes

381

Table 5. AMS-ISO 55001 audit results ISO 55001 requirements

Average

Process description

4

Organizational Context

3.95

With very good practices

5

Leadership

4.29

With very good practices

6

Planning

4.14

With very good practices

7

Support

4.34

With very good practices

8

Operation

4.42

With very good practices

9

Performance evaluation

4.22

With very good practices

10

Improve

4.44

With very good practices

As in the previous case illustrated in Subsect. 3.1, the 20% of the subcategories with the lowest scores are identified, all of which are classified as processes “with very good practices”. Table 6. 20% of the Subcategories with the lowest score, AMS-ISO 55001 Audit. ISO 55001 requirements subcategories

Average

4.1

Understand the organization and its context

3.9

4.2

Understand the needs and expectations of stakeholders

3.9

4.3

Determine the scope of the asset management system

4.0

4.4

Asset Management System

4.0

5.1

Leadership and commitment

4.1

In Table 6, we illustrate the results with the lowest scored Average. In this regard, we can observe a similar situation identified by the application of AMORMS, there is a lack of knowledge related to the activities associated within the measured requirements defined by the standard ISO 55001, and how they are connected to the activities performed by them. It is proposed that the diffusion and re-training needs to be carried out at the operational, tactical, and strategic level on a regular basis. The Table 7. Proposed action plans for AMS-ISO-55001 auditpresents the proposed action plans designed to reduce the gaps detected (Table 8).

382

P. Duque et al. Table 7. Proposed action plans for AMS-ISO-55001 audit

Subcategory - Question - Score Action

Indicator

Understanding of the Organization and its context 4.1.4 Has an internal and external analysis of the key business units been carried out? Score: 3.43

The analyzes exist. The internal and external analysis processes of the key business units (Business Risk) must be disseminated. The dissemination will be done through training, workshops and dashboard

Name: Re-instruction Procedures and scope ISO 55001   total attendees = total maintainers ∗ 100%

Understanding the needs and expectations of stakeholders 4.2.8 Have the requirements of the interested parties to Asset Management been identified? Score: 3.86

The analyzes exist. The requirements of the interested parties must be disseminated to the asset management defined by the company. The dissemination will be done through training, workshops and dashboard

Determining the scope of the asset management system 4.3.9 Has the scope of the Asset Management System been declared, in which the main assets for the system (portfolio) were identified? Score: 4.00

The statements exist. The Scope of the asset management system should be disseminated. The dissemination will be done through training, workshops and dashboard

Goal: 80%; Range: 75–85% Responsible: Reliability Superintendent

Table 8. Proposed action plans for AMS-ISO-55001 audit (continued) Subcategory – Question - Score

Action

Indicator

Asset Management System 4.4.10 Have the commitments been established for the implementation of asset management comprehensively in all management units? Score: 4.00

The commitments exist. Commitments for the implementation of asset management must be disseminated comprehensively in all management units. The dissemination will be done through training, workshops and dashboard

Name: Re-instruction Procedures and scope ISO 55001   totalattendees ∗ 100% = totalmaintainers Goal: 80%; Range: 75–85% Responsible: Reliability Superintendent

(continued)

Audit Models for Asset Management, Maintenance and Reliability Processes

383

Table 8. (continued) Subcategory – Question - Score

Action

Asset Management System 4.4.11 Have the processes for the Business Unit Asset Management System been defined and specified? Score: 4.00

The definitions exist. The processes defined and specified for the asset management system must be disseminated. The dissemination will be done through training, workshops and dashboard

Leadership and Commitment 5.1.15 Does the business unit have a strategic asset management plan (SAMP)? Score: 3.71

SAMP exists, it must be spread. The dissemination will be done through training, workshops and dashboard

Leadership and Commitment 5.1.16 Has the work team in charge of the Asset Management System and reporting to senior management been designated? Score: 3.57

Yes it exists. The existence of the work team in charge of the Asset Management System should be publicized. The dissemination will be done through training, workshops and dashboard

Indicator

4 Conclusions and Future Work In this work, two well-known audits, the AMORMS and AMS ISO 55001, are carried out on a desalination plant, resulting on identification of gaps in the internal process and generation of action plans to all the levels of the organization. Also, the audits assigned a classification based different metrics and scores reached, which illustrated that the maintenance unit evaluated has an “Average standard process” and “Processes with very good practices”, respectively. It was determined that the main problems behind the maintenance area are directly related to the lack of knowledge of processes, plans, and management models that exist in the area. In this context, it was proposed to give a higher priority to the diffusion of knowledge through workshops, training and the use of a dashboards. Regarding the future work, we aim to keep this line of work, thus, in accordance with the results achieved by the employment of AMORMS, the objective is to perform improvements regarding the “Personnel development process”. This process will be reviewed in detail and properly exploited in order to design a special training for all the technical areas in order to close the gaps and achieve world global performance all around the organization.

384

P. Duque et al.

References Aenor. UNE 66175:2003 Sistemas de Gestión de calidad. Guía para la implementación de sistemas de indicadores. s.l.:Aenor (2003) ISO. ISO 55000:2014 Gestión de activos — Aspectos generales, principios y terminología (2014). [Online] Available at: https://www.iso.org/obp/ui#iso:std:iso:55000:ed-1:v2:es Parra Márquez, C., Crespo Márquez, A.: Ingeniería de Mantenimeinto y Fiabilidad Aplicada en la Gestión de Activos, Segunda Asociación para el desarrollo de la Ingeniería de Mantenimiento, Sevilla (2015) Parra Márquez, C., Crespo Márquez, A.: Técnica de Auditoría para los procesos de Mantenimiento, Fiabilidad Operacional y Gestión de Activos (AMORS & AMS-ISO 55001). s.l.:s.n. (2021) Revista Minería Chilena. Minería Chilena (2022). [Online] Available at: https://www.mch.cl/ 2022/03/01/mineria-sustentable-conozca-las-metas-establecidas-en-la-politica-nacional-min era-2050/

Impact of Information Digitalization on Asset Availability - an Empirical Study Katja Gutsche(B) and Santina Schlögel Furtwangen University, Furtwangen, Germany {katja.gutsche,santina.schloegel}@hs-furtwangen.de

Abstract. Problem definition: Maintenance managers face different technological options coming with Industry 4.0. These technological advancements appear useful, however, research looking at the effects within maintenance coming along with an increase of digitalization is rare. As assuring availability is one the most important concerns of to the maintenance management, changes on availability have to be looked at carefully. The aim of this paper is to verify whether the digitalization of repair information can contribute to improving asset availability based on a series of empirical studies. Methodology: A framed field experiment in more than 10 German manufacturing plants with their own maintenance staff was set up where maintenance technicians were asked to fulfil a defined repair scenario with the use of different digital information tools. The available information technologies were paper instruction, video instruction, app instruction and Augmented Reality (AR) instruction. In this work, the effects on asset availability were analysed by looking at work duration, the amount of support needed, number of errors and mental workload of the participants. Results: The results show that digitalization of maintenance information is not clearly associated with an increase in performance or asset availability. To this point, AR as the most digitalized information tool could not outperform the more common digital tools video and app instruction. However, this study underlines once again that paper or pdf instruction are outdated and lead to lower repair performance and asset availability.

1 Introduction Asset availability is one of the key drivers of industrial asset value. Without availability, the expected return on asset cannot be achieved. Availability depends upon the combined characteristics of the reliability, recoverability, maintainability, and the maintenance support performance. Digitalization within the field of industry is expected to have a positive impact on asset availability. The use of digital technologies is meant to enhance the overall business result. The ways of how to apply digital technologies along the asset life cycle are numerous. A digital asset record, predictive maintenance application, spare parts analytics using AI are just some of the technological options. This paper focuses on such maintenance assistance tools which provide access to maintenance information on demand. The asset availability is expected to be improved by the use of more adaptive information providence. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 385–393, 2023. https://doi.org/10.1007/978-3-031-25448-2_37

386

K. Gutsche and S. Schlögel

2 Information Digitalization for Asset Availability 2.1 Availability Availability is the ability of an item to be in a state to perform a required function under given conditions at a given instant of time or over a given time interval, assuming that the required external resources are provided (IEC 60050 2015) (EN 2001) (EN 1534 2019). Equation (1) shows the general way of how to measure availability (A). A=

uptime total operating cycle time

(1)

Consequently, availability is determined by the failure frequency (reliability) and the length of down times (maintainability). This paper looks at the impact of information digitalization in reducing the down times and therefore improving asset maintainability. In general down times can be divided into the following time windows (Fig. 1): 1. Task assignment: Time to assign maintenance task to technician 2. Diagnosis time: Time taken to discover and diagnose an error 3. Response time: Time it takes for the service personnel to arrive at the fault location (including travel times) 4. Access time: length of time until the faulty component is accessible (switching off, dismantling panels, etc.) 5. Repair time: Duration of replacing and/or repair of the faulty component 6. Commissioning time: Duration until the failed system is fully commissioned (including calibration, assembly of panels, etc.)

Fig. 1. Tasks defining downtime (cp. Matyas 2018)

2.2 Digitalization within Asset and Maintenance Management Digitalization is the use of smart technologies to digitalize products, services and processes wherever suitable which as a consequence leads to feasible improvements for the industrial workforce (micro-level), departments and complete companies (meso-level) as well as entire economies (macro-level) (Bauernhansl et al. 2014). Within the field of asset management this use is meant to improve the level of accessibility, connectivity, intelligence, adaptivity, and autonomy of assets (Liu et al. 2021) and the assets’

Impact of Information Digitalization on Asset Availability

387

maintenance (Kans 2022). The way of information providence within down times is one of many ways how digitalization changes maintenance work routines aiming for improvements in output, efficiency, work design and safety (Oks et al. 2016) as well as maximizing device life and reliability (Kans 2022) and therefore supports an asset roadmap (Wilson 2013) towards Maintenance 4.0 (Candell et al. 2009). Figure 2 gives an overview of different digitalization measurements for an increase in asset availability through a decrease in downtime.

Fig. 2. Availability improvements through digitalization measurements

Using smart information tools is supposed to have this intended impact on asset availability by reducing down time. Digitalization as focused within this paper is assumed to have an effect on the time needed for the actual repair procedure (see Fig. 2, step 4). Potential indirect effects imposed by digitalized information like fostering continuous improvement and innovation or changes within internalizing tasks (Wuttke et al. 2022) or ergonomic benefits in fulfilling the tasks (Henderson and Feiner 2011) as these might all have a long-term effect on repair procedures and therefore on asset availability are not addressed within this paper. Repair times are especially but not exclusively dependent on. 1. 2. 3. 4.

Knowledge about failure causes Knowledge/ experience about repair steps Suitability of material and tools in place and repair requirements Easiness in fulfilling the repair, especially defined by accessibility, modularity, testability and modifiability of the asset (cp. ISO 25010, 2011)

Therefor the repair time is dependent on criteria defined by the assets (criteria 4) and on the other hand can be influenced by the design of the repair procedure and the use of supporting tools (criteria 1–3). The use of smart information tools as analyzed within this study comes with changes in the repair procedure.

388

K. Gutsche and S. Schlögel

3 Research This paper summarizes a series of empirical studies looking for changes on asset availability by the use of digitalized repair information. It has been looked at the intended positive effects of digitalization (2.2), especially looking at an increase of asset availability. The focus lies on reduction in repair times. 3.1 Approach As down time is very much correlated with the time need for performing the repair task, the work duration was measured. Also the number of errors as well as the amount of support needed have negative effects on down times (cp. Figure 1). Whereas duration, error rate and support directly have an impact on availability, the mental workload has an indirect one. Mental workload is an indicator for human well-being within the work setting (Schlick et al. 2018) (Johannsen 1993) and describes the technician’s strain level within the repair. The mental workload was measured by using the NASA TLX (Hart and Staveland 1988) (Hart 2006). The following hypothesis were set up in advance: • • • •

H1: The more digitalized the repair information the lower the duration. H2: The more digitalized the repair information the lower the needed support. H3: The more digitalized the repair information the lower the error rate. H4: The more digitalized the repair information the lower the mental workload.

3.2 Empirical Setting The findings of this paper are based on a broad empirical study, where actual technicians were subjects. Within the study 50 service technicians from three different industries in Germany (mechanical engineering, electronics and automotive) were asked to use four different kinds of information technologies – paper/ pdf, video, mobile application (using video, text and picture) and an Augmented Reality (AR) solution installed on the two generations of Microsoft Hololens® (see Fig. 3). AR is the form of Mixed Reality, where virtual content as computer generated virtual objects or environments is blended into the real world (Milgram 1999) (Barfield and Caudell 2001). The user therefore stays connected to the real world which is mandatory for maintenance.

Fig. 3. Studied information tools ordered by their digitalization level

Impact of Information Digitalization on Asset Availability

389

The four tools differ in their levels of innovation and diffusion rates. Mobile applications and AR-solutions are often part of an asset digitalization strategy. In addition, the AR-solution was tested on two different versions of the Hololens® analyzing the effect of technological advances on repair performance and consequently on asset availability. The participants were randomly sorted into four groups testing the different information products. There were 10 participants using paper, video and app, 20 participants used AR. Each group had to repair a chain saw by using the assigned information technology. They were asked to fulfill the task as quick as possible. The average age of all participants was 38,8 years, 2 out of 50 participants were female. The study took place between end of 2020 and beginning 2022. The study showed the following sample characteristics (Fig. 4).

Fig. 4. Age and gender distribution by group

3.3 Findings The hypothesis set in 3.1 were analysed and summed up. The four criteria measured were (1) duration = time needed by participant for repair, (2) needed support = counted advices given by instructor, (3) errors made = counted errors done by participant, (4) mental workload = rating given by participant based on NASA TLX. H1: The more Digitalized the Repair Information the Lower the Duration. The time needed for the repair differed statistically significantly for the different groups, F(4, 45) = 2.64, p < .05. Descriptive statistics show the results shown in Fig. 5 where (1 = paper; 2 = video; 3 = App; 4 = AR Hololens®1; 5 = AR Hololens®2). Interesting to see that AR causes longest repair times. The positive impact on asset availability by repair time decrease aimed for by AR as the most digitalized tool could not be confirmed. Consequently, H1 has to be rejected. H2: The more Digitalized the Repair Information the Lower the Needed Support. There was no statistically significant difference between the different information tools and the number of assistance required, F(4, 45) = 1.89, p = .129. Consequently, H2 has to be rejected.

390

K. Gutsche and S. Schlögel

Fig. 5. Differences in repair duration

H3: The more Digitalized the Repair Information the Lower the Error Rate. The number of errors differed statistically significantly for the different groups, F(4, 45) = 2.87, p < .05. On average, the subjects with the paper instructions made 1.7 more errors than the subjects with the AR instruction on Hololens®2. The Tukey post-hoc test showed a significant difference (p < .05) in the number of errors between the paper instruction and Hololens®2 instruction (−1.7, 95% CI[−0.09, 3.31]). Descriptive statistics show the results shown in Fig. 6 where (1 = paper; 2 = video; 3 = App; 4 = AR Hololens®1; 5 = AR Hololens®2). Digitalization helps decreasing the error rate – H3 is confirmed.

Fig. 6. Differences in error rate

H4: The more Digitalized the Repair Information the Lower the Mental Workload. Mental workload differed statistically significantly for the different groups, Welch test

Impact of Information Digitalization on Asset Availability

391

F(4, 22.24) = 3.08, p < .05. Descriptive statistics show the results shown in Fig. 7 where (1 = paper; 2 = video; 3 = App; 4 = AR Hololens®1; 5 = AR Hololens®2). The positive impact on asset availability through a decrease in the worker’s mental workload aimed for by AR as the most digitalized tool could not be confirmed. However, the usage of App instruction show a clear advance compared to analogous paper instruction. H4 can only partially be confirmed, as App is a tool on the digitalization roadmap but doesn’t represent the most advanced digital information technology.

Fig. 7. Differences in mental workload.

4 Summary and Outlook The results of the study asks for a differentiated application of smart repair information within maintenance management. Smart information as one digitalization measurement (cp. Figure 2) can be implemented on different technologies representing different levels of digitalization. Within asset maintenance the use of smart information cause a diverse impact on asset availability depending on if information is provided through paper/pdf, video, mobile application or AR. The empirical study presented within this paper looked at error rate, repair duration, support rate and mental workload to indicate effects on availability by the kind of information used. The study showed a clear reduce in the error rate by an increase of information digitalization (cp. Figure 3). Repair duration, support rate and mental load showed better results when using video and mobile application. Even advances within the AR-technology as implemented in the second generation of Hololens® 2 did not lead to the expected improvements on asset availability by decreasing down time through faster, less faulty and less stressful repair. The results emphasize on the need on setting up a study with the aimed users as participants instead of tech-students and instructors, as this showed different results in performance measurements (Gutsche and Droll 2020); (Tang et al. 2003); (Blattgerste et al. 2017).

392

K. Gutsche and S. Schlögel

Judging the current state of repair information tools by their effect on asset availability, video and mobile application perform best. Information digitalization to a certain extend supports the intended improvements (2.2). However, some limitations within the results may be find as. • there was a rather small sample size, • only short-term effects within a predefined repair scenario have been measured, especially effects on learning behavior and/or continuous improvement measures should be looked at, • the user experience with the four different technologies is very different, especially the use of the AR-technology was mostly unknown to the users, • no differentiation between different information requirements (e.g. expertise, age, task difficulty) was made (Havard et al. 2021); (Wuttke et al. 2022). Further studies must take these limitations into account.

References IEC 60050. International Electrotechnical Vocabulary (IEV) - Part 192: Dependability (2015) EN 13306. Maintenance terminology. CEN (2001) EN 15341:2019–11. Maintenance - Maintenance Key Performance Indicators (2019) ISO/IEC 25010. Systems and software engineering — Systems and software Quality Requirements and Evaluation (2011) Bauernhansl, T., ten Hompel, M., Vogel-Heuser, B. (eds.): Industrie 4.0 in Produktion, Automatisierung und Logistik. Springer, Wiesbaden (2014). https://doi.org/10.1007/978-3-658-046 82-8 Oks, S.J., Fritzsche, A., Lehmann, C.: The digitalisation of industry from a strategic perspective. In: Proceedings of the R and D Management Conference (RADMA), Cambridge, UK (2016) Liu, C., Zheng, P., Xu, X.: Digitalisation and servitisation of machine tools in the era of Industry 4.0: a review. Int. J. Prod. Res. 1–33 (2021) https://doi.org/10.1080/00207543.2021.1969462 Wilson, A.: Asset Maintenance Management: A Guide to Developing Strategy and Improving Performance. Industrial Press Inc., New York (2002) Kans, M.: The Journey Towards Successful Application of Maintenance 4.0 and Service Management 4.0. In: Pinto, J.O.P., Kimpara, M.L.M., Reis, R.R., Seecharan, T., Upadhyaya, B.R., Amadi-Echendu, J. (eds.) WCEAM 2021. LNME, pp. 137–147. Springer, Cham (2022). https:// doi.org/10.1007/978-3-030-96794-9_13 Candell, O., Karim, R., Söderholm, P.: eMaintenance—Information logistics for maintenance support. Robot. Comput. Integr. Manufact. 25(6), 937–944 (2009). https://doi.org/10.1016/j. rcim.2009.04.005 Henderson, S., Feiner, S.: Exploring the benefits of augmented reality documentation for maintenance and repair. IEEE Trans Vis Comput Graph. 17(10), 1355–1368 (2011) Matyas, K.: Instandhaltungslogistik: Qualität und Produktivität steigern. Hanser (2018) Hart, S.G., Staveland, L.E.: Development of NASA-TLX (Task Load Index): Result of empirical and theoretical research. In: P.A. Hancook and N. Meshkati (Eds.) Human Mental Workload. Amsterdam: North Holland Press (1988) Hart, S.G.: NASA-Task Load Index (NASA-TLX) - 20 Years Later.In: Proceedings of the Human Factors and Ergonomics Society 50th Annual Meeting, pp. 904–908. Santa Monica: HFES (2006)

Impact of Information Digitalization on Asset Availability

393

Milgram, P., Colquhoun, H.: Mixed Reality-Merging Real and Virtual Worlds. Springer, New York (1999) Barfield, W., Caudell, T.: Basic concepts in wearable computers and augmented reality. In: Barfield, W. and Caudell, T. (Eds.) Fundamentals of Wearable Computers and Augmented Reality. Lawrence Erlbaum Associates, Publishers, Mahwah, NJ (2001) Schlick, C., Bruder, R., Luczak, H.: Arbeitswissenschaft. Springer (2018).https://doi.org/10.1007/ 978-3-322-85388-2 Johannsen, G.: Mensch-Maschine-Systeme. Springer (1993) Gutsche, K., Droll, C.: Enabling or Stressing? – Smart Information Use Within Indus-trial Service Operation. In: Duffy, V. (ed.) Digital Human Modeling and Applications in Health, pp. 119–129. Springer, Safety, Ergonomics and Risk Management (2020) Tang, A., Owen, C., Biocca, F., Mou, W.: Comparative effectiveness of augmented reality in object assembly. In: Bellotti, V., Erickson, T., Cockton, G., Korhonen, P. (Eds.) Proceedings of the Conference on Human Factors in Computing Systems, CHI 2003, Ft. Lauderdale, USA, pp. 73–80 (2003) Blattgerste, J., Strenge, B., Renner, P. Pfeiffer, T. Essig, K.: Comparing conventional and augmented reality instructions for manual assembly tasks. In: PETRA ‘17 Proceedings of the 10th International Conference on PErvasive Technologies Related to Assistive Environments, pp. 75–82 (2017) Havard, V., Baudry, D., Jeanne, B., Louis, A., Savatier, X.: A use case study comparing augmented reality (AR) and electronic document-based maintenance instructions considering tasks complexity and operator competency level. Virtual Reality 25(4), 999–1014 (2021). https://doi.org/ 10.1007/s10055-020-00493-z Wuttke, D., Upadhyay, A., Siemsen, E., Wuttke-Linnemann, A.: Seeing the bigger picture? ramping up production with the use of augmented reality. Manuf. Serv. Oper. Manag. 24(4), 2349–2366 (2022)

Infrastructure Asset Management

Linking Organisation Objectives with Asset Information Requirements for Highway Infrastructure Projects Georgios Hadjidemetriou(B) , Nicola Moretti, James Heaton, Manu Sasidharan, Ajith Parlikad, and Jennifer Schooling Department of Engineering, University of Cambridge, Cambridge CB3 0FA, UK [email protected]

Abstract. The success of an organisation or a project in civil engineering relates with realising and utilising the full potential of their assets. This can only be accomplished once it has been established what information is needed to optimise performance and manage risk during asset life cycle. The Building Information Modelling approach comprehends a set of methods and tools for effective asset information management. However, Asset Information Requirements are often shaped without direct connection to, or sight of, the Organisational Objectives they aim to achieve. The “Line-of-Sight” methodology has been created to bridge this gap by creating Function Information Requirements. Presented herein is a thorough exploration of the benefits and limitations of applying Line-of-Sight to highway infrastructure projects. This highlights the need for infrastructure asset owners and managers to employ this methodology to optimise information management and decision-making. The proposed methodology has the potential to redefine asset management and realign it with organisational imperatives.

1 Background Maximising the success of any organisation is dependent on its ability to realise the full value of its assets (Carlucci and Schiuma 2006). This can only be achieved after establishing what information is required to optimise performance and manage risk throughout the whole life cycle of those assets (Succar and Poirier 2020; Hadjidemetriou et al. 2021). Nevertheless, asset management organisations are complex, often maintaining and operating many thousands of assets within systems of systems, spread over large geospatial areas across many decades, while being engaged with diverse stakeholders (Ouertani et al. 2008; Hadjidemetriou et al. 2020; Bin Wee et al. 2022). The high-level aspects of Organisational Objectives (OO) do not easily translate to the more granular level of Asset Information Requirements (AIR). Industry standards such as the UK Building Information Management (BIM) Framework, the suite of ISO 19650 BIM standards and ISO 55000 Asset Management standards provide definitions of Organisation Information Requirements (OIR), AIR and Asset Information Models (AIM) (ISO 55000 2014; ISO 19650 2020). The ISO 19650 series for “information management using building information modelling” states that OIR should encapsulate AIR. Despite © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 397–404, 2023. https://doi.org/10.1007/978-3-031-25448-2_38

398

G. Hadjidemetriou et al.

these standards provide what ‘shall’ be done, they are limited in providing any practice tools or frameworks to support their development. Researchers and practitioners have realised from the earliest developments of Information Management Systems (IMS) that developing information requirements for those systems is a complex and puzzling task that is often neglected and result in reduced operational performance (Wetherbe 1991). Furthermore, the adoption of IMS within Engineering Asset Management (EAM) is immature compared to other sectors. While the Requirements Engineering (RE) provides a framework for the development of information requirements within the software engineering discipline, the conventional techniques used within the framework have limited use within EAM due to its multifaceted nature (Kotonya and Sommerville 1998). Most notably, RE fails to address the wholelife management requirements of EAM. Building Information Modelling (BIM) is a crucial enabler for the adoption of an IMS within the construction and EAM sectors (Heaton et al. 2019b). While BIM has been actively implemented within the design and construction phases, its adoption in the operational phase is still limited (Pärn et al. 2017). This is partly due to the complexity of capturing use-phase-related OIR within the initial stages of the BIM information management processes. In spite of a few example methodologies in the literature assuming that OIR can be immediately used to generate the AIR, it can be witnessed that the jump from OIR to AIR is a challenging step for most organisations. This is because the OIR are the information requirements used to measure the performance of the OO, which are defined at the organisational level. The OO, and consequently, the OIS are often abstract and therefore poorly translate directly into AIR, which are used to inform decisions at the asset level. In fact, AIR are developed from a technical perspective, and capturing only partially (and often lacking) the high-level information for management processes such as financial management and risk management. The proposed “Line-of-Sight” methodology addresses these emerging challenges, enabling a structured and repeatable approach to the development of OIR and AIR. The Line-of-Sight allows the alignment of the different decision-making levels within an organisation, creating explicit links between the AIR (operative level) and OIR (strategic level). This reduces the ambiguity in the measurement of performances against OO and informs better decisions. The subsequent sections of the paper describe the methodology, its application on a highway infrastructure project, and the extracted conclusions, in that order.

2 Methodology The Line-of-Sight methodology has been initially introduced by Heaton et al. (2019a). Presented herein is the modification and evolution of the framework, after introduced to Asset Management Organisations, and applied in real-life case studies. The Line-ofSight is divided into three phases, namely Organisational level, Functional and Asset level, and Asset Information Model, as illustrated in Fig. 1. The key concepts adopted in the methodology are summarised in Table 1.

Linking Organisation Objectives with Asset Information Requirements

399

Fig. 1. Main steps of the line of sight methodology

Table 1. Key concepts used in the line of sight methodology implementation. Concept

Description

Organisational objectives (OO)

The objectives are defined by organisations to respond to their business needs

Organisational information requirements (OIR)

Information requirements needed to respond/measure an OO

Functional outputs (FO)

Functions needed to respond to a OO

Functional information requirements (FIR)

Information requirements needed to respond/measure a FO

Asset information requirements (AIR)

Information requirements used to describe the properties of the assets

Phase 1 consists of two steps. After completing Step 1 a single source for all Organisational Objectives (OO) is documented, and after Step 2 OIR are developed. An OO normally corresponds to multiple OIR, while an OIR rarely corresponds to multiple OO. If an organisation already has an identified and agreed set of OO and associated OIR (the Key Performance Indicators – KPI used to measure the OO), then these simply need to be verified at this stage. Phase 1 requires input from senior stakeholders. Phase 2 encompasses Steps 3 to 7, with the outcome of this phase being a structured and documented set of OIR, Functional Information Requirements (FIR) and AIR. Step 3 connects OIR with Functional Outputs (FO). A FO is the function provided by a combination of multiple asset systems (e.g. heating, ventilation). The FO corresponds with the functional classification of the objects forming a whole infrastructure, or a part of it (e.g., a bridge, a railway or a road section etc.). The FO can be defined through plain language questions, by non-experts or the senior management. During this step, the OO are translated into Functional Requirements (FR) that allow to bridge the gap between the high-level decision making (OO definition) and the asset level classification and information requirements definition, carried out at the asset level. Step 4 corresponds to the definition of the FIR, which is the main contribution to science of the method. FIR are a set of individual information requirements documented within a structured template, to fill the gap between OIR and AIR. Each FO corresponds to a number of FIR. Step 5 identifies the assets that support each FO. Step 6 identifies the AIRs, used to store the properties of the assets. Step 7 validates that the newly developed information requirements are fit for purpose. It also brings together the documentation of the information requirements within a single structured source with user and administrator

400

G. Hadjidemetriou et al.

management controls, enabling them to be used in the development of an AIM. Phase 2 chiefly require input from asset managers and other technical stakeholders. Figure 2 represents the relations among the main concepts of the Line of Sight methodology and the major outcomes (along with several examples) from completing Steps 1–7.

Fig. 2. Relations among the main concepts of the Line of Sight methodology.

Phase 3, which consists of Steps 8, 9 and 10, is the technical phase of the methodology, utilising the outcome of Phase 2 (i.e. information requirements) as a mean to develop an AIM. Stage 8 supports the development of a BIM model that correctly represents multiple organisational perspectives. This can be achieved through the classification of the BIM objects, if the model is already in places, or through the development of guidelines for the design stage. Step 9 focuses on the development of a relational database, handling AIR and connecting to existing asset management systems. Step 10 enables the Extraction, Transformation and Loading (ETL) of data between different software, used to access use and store the data enabling the Line of Sight methodology. Phase 3 is primarily the concern of those with a responsibility for data and information management, including BIM managers.

3 Case Study The Line-of-Sight methodology was validated through a set of case studies, on different types of infrastructural assets, such as water filtration stations, roads and buildings. In this section, the application to a National Highways project for the upgrade of an existing highway is presented. Highway infrastructure projects aim to achieve high level OO, as defined by senior and project managers. At the same time, highway infrastructure projects involve numerous types of assets and of information requirements. Thus, information needs to be flowed across different levels and communicated across different teams. Line-of-Sight methodology allows this flow, while generating discussions and ideas

Linking Organisation Objectives with Asset Information Requirements

401

as well as identifying challenges across different organisation levels. However, vast amount of information is developed during the application of Line-of-Sight to highway infrastructure projects and that is why a software is under development. The application of Phase 3 to a case study is out of the scope of this paper since it addresses the technical development of the digital model. Phases 1 and 2 were applied at the infrastructure level. Two workshops (one for each phase) were conducted with NH experts to tailor Line-of-Sight to a project on upgrading an existing highway. Figure 3 represents the steps of the methodology developed during the workshops and the outputs for each of them. Strategic drivers (i.e. published documents by asset owners) facilitated the initiation of Line-of-Sight Phase 1. NH published the Operational Metrics Manual, providing a comprehensive overview of the metrics selected for performance monitoring (National Highways 2020). OO (Step 1) are mentioned in that document as Key Performance Indicators (KPIs). In the same way, OIR are mentioned as Performance Indicators (PIs). The workshops were carried out online, using a collaborative platform, which represents the main shared work bench for testing the methodology. During the first workshop, senior stakeholders approved, removed and added new OO to agree on six final OO. For instance, the first OO was “improving safety for all”. In the same vein, OIR were agreed and connected with the corresponding OO (Step 2). Table 2 shows the template used to define OIR. It presents two of the OIR corresponding to the first OO. Asset managers and other technical stakeholders participated in the second workshop. The Functional Requirements (FR), corresponding to OIR, were identified in plain language (Stage 3). For instance, the FR of ensuring appropriate lighting is connected with the OIR of number of people killed per year. Then, FR were connected with FO that were already defined by the Uniclass2015 table EF (NBS 2015). Following the same example, the FR of ensuring appropriate lighting was linked with the FO of electrical power and lighting functions. For Stage 4, FIR, corresponding to each FO, were defined using a template that includes: the code and name of the FO, as defined by the Uniclass2015 table EF; the information requirement; category of the information requirement; and data type. For Stage 5, FO were connected with assets, as defined by the Highways England’s Asset Data Management Manual: Part 3 - Data Dictionary (Highways England 2022). Each asset has multiple AIR, normally defined by in-house classification systems (Stage 6). In this case, AIR were also defined by the same data dictionary. Finally, Stage 7 validated that the AIR are fit for purpose and can form part of any future technical development (Fig. 3).

402

G. Hadjidemetriou et al.

Fig. 3. Steps developed during the workshop and outputs.

Linking Organisation Objectives with Asset Information Requirements

403

Table 2. Example of developing OIR. OO

Critical success factor

Plain language question

OIR

Data type

Improving safety for all

Zero people killed

How many deaths per year?

N. of people killed/year

Integer

Zero people seriously injured

How many people N. of people are seriously injured seriously per year? injured/year

Integer

4 Conclusions The Line-of-Sight Methodology has been developed to address the challenge of developing AIR that support, and are informed by, OO. It achieves this through a set of tools to extract and connect OIR, FIR and AIR. The development of FIR bridges the gap in existing methodologies, and ensures that data gathered by organisations truly supports the achievement of OO. The methodology, as described in this paper, has been validated by practitioners working in organisations owning and managing assets, such as water filtration stations, roads and buildings. In the near future, the methodology will be applied to even more types of assets, to identify potential challenges and opportunities. The feedback received so far from the practitioners highlights the capabilities of the methodology to: show the relationships between information requirements at different levels; communicate information across teams and the supply chain; and generate discussions and ideas across different organisation levels. The proposed method has room for improvement and work is currently under way to enhance it. The same FO, assets and AIR can be named differently by different documents and organisations. Thus, thorough attention should be given on the standardised terms used within an organisation and a project. Additionally, it is complicated and almost impossible to manually manage the vast amount of information developed during the application of Line-of-Sight. Therefore, a software is under development, aiming to facilitate the process of defining and connecting information requirements. The personnel of asset management organisations will be able to add, delete and edit elements at the phase related to their role. For instance, Asset Managers will be able to work on Phase 2, but only view Phase 1, which is defined by Senior Management.

References Bin Wee, X., et al.: Simulation and criticality assessment of urban rail and interdependent infrastructure networks. Transp. Res. Rec. J. Transp. Res. Board, 036119812211035 (2022). https:// doi.org/10.1177/03611981221103594 Carlucci, D., Schiuma, G.: Knowledge asset value spiral: linking knowledge assets to company’s performance. Knowl. Process Manag. 13(1), 35–46 (2006). https://doi.org/10.1002/kpm.243 Hadjidemetriou, G.M., et al.: Comprehensive decision support system for managing asphalt pavements. J. Transp. Eng. Part B Pavements 146(3), 06020001 (2020). https://doi.org/10.1061/JPE ODX.0000189

404

G. Hadjidemetriou et al.

Hadjidemetriou, G.M., Herrera, M., Parlikad, A.K.: Condition and criticality-based predictive maintenance prioritisation for networks of bridges. Struct. Infrastruct. Eng., 1–16 (2021). https://doi.org/10.1080/15732479.2021.1897146 Heaton, J., Parlikad, A.K., Schooling, J.: A building information modelling approach to the alignment of organisational objectives to asset information requirements. Autom. Constr. 104, 14–26 (2019a) Heaton, J., Parlikad, A.K., Schooling, J.: Design and development of BIM models to support operations and maintenance. Comput. Ind. 111, 172–186 (2019b). https://doi.org/10.1016/j. compind.2019.08.001 ISO 19650: ISO 19650-5:2020-organization and digitization of information about buildings and civil engineering works, including building information modelling (BIM), ISO (2020). https://www.iso.org/cms/render/live/en/sites/isoorg/contents/data/standard/07/42/ 74206.html. Accessed 8 June 2022 ISO 55000: ISO 55000:2014-asset management-overview, principles and terminology, ISO (2014). https://www.iso.org/cms/render/live/en/sites/isoorg/contents/data/standard/05/50/ 55088.html. Accessed 8 June 2022 Kotonya, G., Sommerville, I.: Requirements Engineering: Processes and Techniques. Wiley, New York (1998) National Highways: Operational metrics manual (2020). https://nationalhighways.co.uk/media/ n4tak3gk/ris2-omm-final-shared-for-publication-14-july.pdf. Accessed 6 Sept 2022 Ouertani, M.Z., Parlikad, A.K., McFarlane, D.: Asset information management: research challenges. In: 2008 Second International Conference on Research Challenges in Information Science. 2008 Second International Conference on Research Challenges in Information Science (RCIS), Marrakech, pp. 361–370. IEEE (2008). https://doi.org/10.1109/RCIS.2008.4632126 Pärn, E.A., Edwards, D.J., Sing, M.C.: The building information modelling trajectory in facilities management: a review. Autom. Constr. 75, 45–55 (2017) Succar, B., Poirier, E.: Lifecycle information transformation and exchange for delivering and managing digital and physical assets. Autom. Constr. 112, 103090 (2020). https://doi.org/10. 1016/j.autcon.2020.103090 Wetherbe, J.C.: Executive information requirements: getting it right. Mis Q., 51–65 (1991)

A Methodology for Ensuring Strategic Alignment of Railway Infrastructure Asset Management Processes Irene Roda1(B) , Donatella Fochesato2 , Adalberto Polenghi1 , Margherita Luciano2 , Isabella Tordi2 , Lorenzo Di Pasquale2 , and Ivan Cavaiuolo2 1 Politecnico di Milano, Milano, Italy

[email protected] 2 Asset Management, Rete Ferroviaria Italiana S.P.A, Rome, Italy

Abstract. For railway infrastructure management companies, the effective management of the assets is essential for the achievement of business goals. Asset management (AM) translates the organizational objectives into technical and financial decisions, plans and activities, with the aim of realizing value from assets. Securing that AM plans and activities are carried out in accordance with the company business strategy is fundamental. To do so, in this paper, a methodology is developed together with the RFI (Rete Ferroviaria Italiana), the Italian railways infrastructure manager company, relying on the process maturity assessment model already built by the company. The methodology allows identifying the AM improvement actions with highest priority ensuring strategic alignment with company longterm vision. The methodology is already demonstrated through a PoC (Proof of Concept) and its implementation is underway.

1 Introduction Recently, Asset Management (AM) has been gained influence in organisations responsible for the management of railway networks, where cost and performance of the infrastructure are of national significance (UIC 2010). Generating value from assets is the goal of AM (El-Akruti et al. 2013; Márquez et al. 2019; Roda and Macchi 2018). The methods for generating value are dictated starting from the AM strategy, which is outlined the Strategic Asset Management Plan (SAMP). As indicated in the ISO5500X (ISO 55000 2014) standards, the SAMP should guide the AM System towards the development of AM plans. Nevertheless, very often, in complex organizations such as Infrastructure Management organisations (IMs), the decision-making process is still addressed in a siloed approach, each organizational function being driven by its own targets. To avoid this, on one side being able to map all the asset-related processes within the organization and having a maturity assessment model for continuous improvement is fundamental. At the same time, there is the need to ensure that the improvement actions defined based on the maturity assessment, are aligned with the organizational strategy, prioritizing the ones contributing most to value generation. This paper aims to present a methodology developed with the Italian IM company Rete Ferroviaria Italiana (RFI), starting from its © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 405–415, 2023. https://doi.org/10.1007/978-3-031-25448-2_39

406

I. Roda et al.

needs to integrate the already available AM process maturity assessment model with the SAMP, linking to the value framework.

2 Railway Infrastructure AM Current Frameworks and Developments The reference for practical implementation of AM in railway is the AM framework defined by the UIC, which identifies the key components of an AM system for IMs (UIC AMWG 2016). The framework, aligned with ISO55001, identifies the components through which IMs work, considering the objectives set based on the stakeholders’ requirements, the operation of the assets, and the strategic decision-making moments, which require deciding (Leadership and Commitment), planning (Strategic AM Plan), scheduling (AM Plans) and implementation (Implementation of AM Plan). AM maturity refers to the capability of an organization’s people, processes, technology, leadership and culture to derive and deliver value from its assets to meet the needs of the organization and its stakeholder is a sustainable manner (UIC AMWG 2016). A key part of achieving an appropriate level of AM maturity is to understand how pivotal processes are to delivering organizational and stakeholder value (UIC AMWG 2020), which is still difficult to be ensured in the long term. In fact, it is crucial to ensure that investments, operation projects and activities are being prioritizes correctly in alignment with the organizational objectives. For this, on one hand, a crucial role is played by the definition of the organization’s Value Framework which should be described in the SAMP, and which defines the organization value elements based on its stakeholders’ requirements, and how they are measured. For this, the UIC, by means of the AMWG - Asset Management Working Group and the support of the AM-DIP subgroup - Asset Management Developing Implementation Projects, defined an AM value framework for a railway IM, which can be taken as a reference for IMs. On the other hand, there is the need of a supporting methodology for assessing the strategic alignment of AM processes, helping prioritizing improvement actions not only based on the processes’ level of maturity but also based on how they contribute to value generation for the organization and its stakeholders. This paper is aimed at describing the methodology that was developed for addressing this issue.

3 Proposed Methodology The developed methodology entails four main steps, as shown in Fig. 1 using the IDEF0 formalism for Business Process Modelling, and they are the following: A1. Prioritize value elements, aimed to rank the value elements based on the company strategy. The prioritization method proposed is fuzzy Analytic Hierarchy Process (AHP) (Saaty 1990) used to collect opinions from the top managers of the organization about the relative importance of the different value elements. AHP results to be widely used in research, as suggested by (Kabir et al. 2014). The fuzzy approach is proposed since involved managers are intended to work into separate groups during the data collection through AHP. This facilitates a more comprehensive discussion about value elements

A Methodology for Ensuring Strategic Alignment

407

relevance for the strategy of the company, according to specific perspectives on different aspects of asset lifecycle. A2. Map processes and value elements, aimed to define a matrix relating each asset management process of the company with the value elements it contributes to. The association is performed based on the previous analysis of the stakeholders’ requirements specific for each process. A3. Prioritize processes, aimed at identified the pivotal processes which present highest opportunity for improvement, and which bring most value to the company. This is done by combining the ranking of value elements (A1) and the process-value elements map (A2) with the assessment of the Maturity level of each process. A4. Prioritize improvement actions, aimed to list those improvement actions having higher contribution to value creation based on the output of the previous step, addressing the pivotal processes. In brief, the methodology starts with the identification of the top-priority value elements according to long-term company strategy. In parallel, the value elements are allocated to one or more processes since each process contributes to reaching a specific element. By combining the two mentioned information and based on the assessment of the maturity level of each process, it is possible to identify those processes that are pivotal. A pivotal process is a process which impacts the top-priority value elements, and which presents most opportunities for improving its maturity level. Finally, given the identified improvement actions linked to each process based on the maturity assessment, it is possible to prioritize the actions related to pivotal process. This is a crucial information for budget allocation.

Fig. 1. Proposed fuzzy AHP-based methodology for ensuring strategic alignment.

408

I. Roda et al.

4 Application Case The methodology has been designed and developed during a collaborative project with the Asset Management direction of RFI. After a brief presentation of the company, the application and results are presented, including a discussion on the current limitations. 4.1 Company Overview and Propaedeutic Activities for Implementing the Methodology RFI (www.rfi.it/) is the company of the “Ferrovie dello Stato” Group that has the role of Manager of the National Railway Infrastructure, with the task of ensuring the maintenance and safe circulation of trains on the entire infrastructure, managing investments for the upgrading and development of railway lines and facilities. RFI promotes the integration of the Italian infrastructure into the European Railway Network, coordinating with the countries of the European Union on quality standards, actions and marketing strategies for services. Within RFI, the Asset Management (AM) direction aims at spreading the culture of best practice in AM. As a first relevant activity carried out by the RFI AM, which is relevant for the proposed methodology application, is the in-house development and application of a Self Assessment Maturity Model (SAMM). RFI’s SAMM is based on an extensive process mapping activity that was previously developed in the company, identifying all processes in the company and the existing relationships among each other’s. The SAMM is aimed at analysing the state of the AM system, linking to the AM processes, by measuring its maturity on a six-level scale (UIC AMWG 2016), and defining improvement actions on the analysed processes in relation to 5 criteria, namely process structure, process objective, internal organization, supporting IT systems, and procedure. In addition to the SAMM, which sets the ground for the methodology, a value framework was developed in the context of the project, which stems from the general one proposed by the UIC and customised according to the company business and its stakeholders needs. To get to the definition of the value framework, both a detailed analysis of the stakeholders needs, and of the company strategic objectives, in line with the Industrial plan 2022–2031 of RFI recently released, was carried out by RFI AM Direction. The realised value framework (see Fig. 2) entails 19 value elements, aggregated into three groups, i.e. physical asset requirements, company processes requirements, and delivered service requirements. The grouping of value elements is required as some value elements are different in nature, but also because the AHP should be favoured, hence no more than 7 criteria for pairwise comparisons at a time are suggested.

A Methodology for Ensuring Strategic Alignment

409

Fig. 2. Proposed value framework by AM direction and Politecnico di Milano.

Considering the importance of the sustainable development issue for the company, each of the 19 value elements was also mapped against the three sustainable development dimensions, i.e., environmental, economic, and societal. This provide a sustainabilitydriven insight for the decision-makers that are involved in the prioritization of the value elements in the first step of the methodology, in order for them to consider the strategy of the company together with the sustainability dimensions. The mapping is reported in Fig. 3.

Fig. 3. Proposed value framework mapped with sustainable development dimensions.

410

I. Roda et al.

The developed value framework was used in the context of the project to test the methodology, as cleared out in Sect. 4.2. 4.2 Application and Results The developed methodology was applied in the RFI company as a Prof of Concepts (PoC) and an overall schematic representation of the steps and its implementation mechanisms is reported in Fig. 4.

Fig. 4. Application of the methodology as PoC in RFI.

Step A1 is aimed at ranking of the value elements, ordered according to its importance. Importance is measured according to the centrality of the value element with respect to RFI’s long-term strategy and sustainability. Specifically, the group(s) of decision-makers are given a short interview in which they are asked to make a judgement of relative importance between one value element and another, taken two by two. The question that should guide this assessment is: “How much is element X more or less important than element Y in terms of the RFI strategy and its respective impact on the pillars of sustainable development?” By answering this question for each pair of value elements (X,Y), by means of a score (taking a rating scale as a reference), it is possible to create matrices that are then processed by the fuzzy AHP method which determines, automatically, the ranking of the value elements. In this application, to test the methodology applicability, the members of the AM direction of RFI were involved instead of the company top managers, for collecting information and applying the fuzzy AHP prioritization method. This is done considering a validation of the methodology before its adoption with RFI top management, to solve and mitigate possible shortfalls or doubts around the steps. Therefore, three groups of decision-makers were defined, each composed by around three people. During the discussion, each group filled in the pairwise comparison matrices of the fuzzy AHP, enabling coming up with a final rank of the value elements. The process is iterative as matrices consistencies were not reached at first application. As such, a recurrent process is implemented in MATLAB as supporting tool with a double option (see Fig. 5): i) automatically adjusting the weights so to come up with consistent matrices at first application or ii) provide information about

A Methodology for Ensuring Strategic Alignment

411

which pairwise comparison majorly create inconsistencies to solve it through discussion between managers to come up with an overall consensus.

Fig. 5. Iterative process for AHP matrices consistency.

The mapping of the value elements with the sustainable development dimensions, supported a sustainability-informed assignment of weights to the different value elements, together with a discussion about its relevance for the strategy of the organization.

Fig. 6. Step A1: value elements ranking.

412

I. Roda et al.

The obtained result is a rank of the value elements considered within the value framework of RFI, as reported in Fig. 6. The rank is in descending order, from the most relevant for the RFI long-term strategy to the least ones. Indeed, it is worth to underline that the main conceptual effort is related to how to classify historical elements of relevance for IMs, especially “Safety”. Safety, as a railway requirement, is mandatory and it underlines all the activities and processes within IMs. However, safety, as a value element, requires to be evaluated in light of its importance for long-term company strategy, which is a different perspective. As such, downgrading safety to a less important value element does not mean less attention to it, rather it is given for granted, while other value elements, like “Innovation and digitalisation” should guide the company strategy. Step A2 is based on the analysis of the stakeholders needs for each process, aiming to map each process with respect to the value elements which are mostly impacted by it. The result is a matrix where on the row all RFI processes are listed, and on the column all value elements. An “x” means that a specific process impacts on a specific value element, as depicted in Table 1. Table 1. Step A2: processes-value elements mapping.

Process 1

Value element 1

Value element 2

x

x

Process 2 Process 3 Process n

x

Value element 3

Value element n

x

x

x

x

x

x

The following step A3 relies on the outputs of A1 and A2, jointly with the results of the application of the SAMM, already done in the company, which formalises the knowledge of the maturity level of each process. Relying on this information, the objective is to prioritize processes according to increasing level of maturity (the lower the maturity, the higher the probability that the process does not fully contribute to the associated value element); and decreasing level of impact on business strategy (i.e. impact on value). In detail, the impact on value of each process is evaluated summing up the weights of the value elements (from A1) which are impacted by that specific process (from A2). Consequently, it is possible to realise a matrix, as in Fig. 5, where the four quadrant represents the priority of each process in terms of impact on value, namely: • High-priority processes are those processes whose impact on value is high, that is, they are core in the long-term company strategy to generate value for stakeholders, but their maturity level is low, meaning that their performance is not adequate with respect to their importance to strategy; as such, these processes are of high-priority and improvement actions are required. • Mid-priority processes are processes where the mix of impact on value and maturity level does not mean urgent actions. Namely, two situations may occur:

A Methodology for Ensuring Strategic Alignment

413

o High impact on value and high maturity level implies that the processes are core in generating value, yet their maturity level is high, hence they should be able to properly support company strategy; o Low impact on value and low maturity level corresponds to processes that properly balance their relevance with respect to the company long-term strategy with respect to their maturity level they have. • Low-priority processes are those processes whose impact on value is low and maturity level is high, hence they are well carried out. Therefore, no urgent action must be taken. The line separating quadrants are calculated as the mean of impacts on value and maturity levels over all processes. In this way, it is possible to have a relative reference, considering that a maturity level equal to 4 is set by the (UIC AMWG 2016) to be a practice coherent with the ISO 55001. In Fig. 7, the separation lines between quadrants are placed at 0.5 and 3 for value impact and maturity level, respectively, as example.

Fig. 7. Step A3: prioritisation matrix of process.

Therefore, high-priority processes must be given attention as they have a major impact on value generation, while low-priority processes, even though important as well to carry out company business, do not have direct impact on value generation. Consequently, budget allocation is made easier as there is a rank of processes requiring immediate actions and which do not. As a matter of visualisation, considering the huge number of processes within RFI and to provide a visual yet meaningful output, two matrices are created, one (Fig. 8) where the 10 processes with highest value impact are shown, while another one (Fig. 9) where the 10 processes with lowest maturity level are reported. Note that process identification number is not the same between the two figures. Also, information about which are the shown processes is not reported as it is private.

414

I. Roda et al.

Finally, based on the results of A3, in A4, the improvement actions to be prioritized given their alignment with the company strategy and addressing processes with opportunities for maturity level improvement, are indicated. The result is a list, organised in descending order by value impact, where each row contains the following information: RFI direction, relative processes, maturity levels, overall weights (i.e., priority) and improvement action/s.

Fig. 8. Step A3: prioritisation matrix of process in descending order of value impact.

Fig. 9. Step A3: prioritisation matrix of process in ascending order of maturity level.

A Methodology for Ensuring Strategic Alignment

415

5 Conclusions This research presents a methodology to ensure strategic alignment by prioritization of improvement actions considering the AM processes maturity level together with their impact on the value elements of the company. The methodology promotes value-based AM in railway IMs by centering decisions around the value framework, which acts either as a translation of company strategy into something tangible, and as a driver to identify processes that are pivotal for value creation. Indeed, to the best of authors’ knowledge, this methodology is the first-of-a-kind regarding value-based AM in railway infrastructure management. Future work will focus on validating the methodology with top management of RFI and on looking for further applications for further refinement and generalization. It must be noted that tackling decisions at strategic levels imply involving top management and addressing high uncertainty, for this reason, mediating the value elements prioritization is one of the critical and most delicate aspects of the methodology to ensure that the outcome of the process reflect the company strategy. Possible extension may include advanced methods to formalize dependence between value elements, as Analytic Network Process.

References El-Akruti, K.O., Dwight, R., Zhang, T.: The strategic role of engineering asset management. Int. J. Prod. Econ. Elsevier 146(1), 227–239 (2013) ISO 55000. Asset management — Overview, principles and terminology. BSI Standards Publication, International Organisation for Standardization (2014) Kabir, G., Sadiq, R., Tesfamariam, S.: A review of multi-criteria decision-making methods for infrastructure management. Struct. Infrastruct. Eng. 10(9), 1176–1210 (2014) Crespo Márquez, A., Macchi, M., Parlikad, A.K. (eds.): Value Based and Intelligent Asset Management. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-20704-5 Roda, I., Macchi, M.: A framework to embed Asset Management in production companies. Proc. Inst. Mech. Eng. Part O: J. Risk Reliab. 232(4), 368–378 (2018) Saaty, T.L.: How to make a decision: The analytic hierarchy process. Desicion Making Anal. Hierarchy Process Theory Appl. 48(1), 9–26 (1990) UIC. Guidelines for the Application of Asset Management in Railway Infrastructure Organisations (2010) UIC Asset Management Working Group. UIC Railway Application Guide, Practical Implementation of Asset Management through ISO 55001 (2016) UIC Asset Management Working Group. UIC SAMP Application Guide - ISO SAMP Application Guidelines for Railway Infrastructure Organisations (2020)

Hierarchy Definition for Digital Assets. Railway Application Mauricio Rodríguez Hernández(B) , Adolfo Crespo Márquez, Antonio Guillen López, and Eduardo Candon Fernandez Universidad de Sevilla, Sevilla, Spain [email protected], {adolfo,ajguillen,ecandon}@us.es

Abstract. Defining the existence of a digital asset, integrating multiple platforms that represent the element digitally and at the same time meeting the context and operational demand of railway infrastructure systems, represents an unresolved challenge for this industry. This study focuses on the search for common minimums, complementing the perspectives of the scientific community/research centers, with the real applications of models at European level. Finally, converges in the development of a scheme that collects the state of the art and praxis, providing a starting point for the development of scientific discussion and the search for future models that provide an effective solution to the problem. The integration of maintenance management models with architectures for the development of digital twins in industry 4.0, and the applied study of the railway industry itself, are part of the basis of study. Seeking to respect the principles already proposed for industry 4.0, the scheme presents new relationship factors, which will be prototyped in the industry, particularly railway infrastructures. Keywords: Railway maintenance · Asset management · Criticality analysis · As set hierarchy definition · Industry 4.0 · Digital twin

1 Introduction The principal motivation to develop this research is to give to the scientific community and the industrial world a frame to start the way to standardize the digitalization in the railways industry, for that the authors contribute with their experience in real applications integrated with the state of the art for the 4.0 industry. This paper tries to establish a hierarchical structuring scheme for (railway) assets from a systemic, holistic and digital perspective, taking into account 3 fundamental dimensions: The real-world dimension, reflected in physical assets (Zheng et al. 2021) the digital dimension (Schweichhart 2016), manifested in the multiple platforms that seek the generation of their own digital element; and the management dimension extracted from several models: MGM (Crespo 2007) ISO (55,000) and UIC standards developed in the industry. All of them together aiming at the same objective, the maximization of value, the development and simplification of processes that allow generating the expected results. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 416–427, 2023. https://doi.org/10.1007/978-3-031-25448-2_40

Hierarchy Definition for Digital Assets. Railway Application

417

The first challenge facing the industry is the choice of an appropriate Maintenance Management model, which allows the organization to efficiently manage its assets. For this it will be important to take into account from the beginning, both the outputs and the inputs of this (Crespo et al. 2018). In this sense, the inputs it receives and the way in which they are defined, will be decisive to ensure success in the implementation of any management model, in the face of the level of decision making (later “the last layer”) (Schweichhart 2016). It will be necessary then, to know the route to travel from beginning to end, before embarking on any technological or digital solution, which promises in itself to provide a solution to the “maintenance question”. Digitalization, on the other hand, which has taken on a life of its own in each of the OT and IT platforms, becomes a challenge, having to align all the solutions, in a way that they converge, adding value to the process and not, on the contrary, adding entropy. From the above, we know that there is still no transversal agreement regarding a general architecture that allows to meet the needs of industry 4.0 for networked systems, such as railway infrastructure systems. (UIC 2022).

2 Research Methodology. The methodology considered: The review of the scientific and technical literature in European research; the compilation of several consultations with infrastructure managers in Europe; and finally, the results of the research on practical experience in the development and implementation of a management and digitalization model in one of the European railways systems. The first phase has concentrated on a bibliographic/bibliometric review of scientific publications, taking as a reference the Scopus database, with the objective to identify the level and intensity of relationship between the 3 mains concepts on study: Hierarchy of Assets, Digital Twin, Railway. The search pattern that has been used is: “railway” AND “digital twin” OR “railway” AND “hierarchy”. The second phase of the research has focused on railway research centers and projects currently in force, seeking the existence of models in development or experimentation that can be considered as a basis and/or inputs for the model. This has been covered as: • Technological and railway regulation systems, on the subject of digitalization, hierarchy and maintenance. • Research in digitalization projects in European railways and their respective internal forums, allowing to rescue the real practices of the industry.

3 Synthesis Review 3.1 Technical Scientific Literature This review aims to provide a fundamental vision of the art state in which the discussion of these concepts within the scientific community is located, and the evolution over time of them. The results obtained, presented as number of publications over time are shown in Fig. 1-right and the bibliometric study developed through the VOS viewer platform,

418

M. R. Hernández et al.

allows to appreciate Fig. 2, both the intensity of the concepts, as well as their relationship and their connection with other concepts treated in the analyzed studies. Out of a total of 668 results obtained from SCOPUS.

Fig. 1. Evolution on SCOPUS of mains concepts

Fig. 2. Evolution, Intensity and Relationship of mains concepts

It is clear from the analysis that the digitalization, digital twins and 4.0 industry are still underdeveloped concepts in the railway context, and although the AHP is inserted and mainly linked to risk assessment processes, not with the approach of hierarchizing railway assets and/or their respective digital twins. It is possible to observe that in the last 5 years the subject has begun to emerge only from the hand of GIS, but this is limited to a very bounded context and does not provide a global view regarding the hierarchy of digital assets. In this sense, the bibliographic evidence, shows us that there is ample opportunity to develop models and open the scientific discussion around the subject, since neither the hierarchy nor the taxonomy of railway infrastructure assets itself seem to be regulated by any scientific model, but rather structures derived from praxis, in each of the railway operators.

Hierarchy Definition for Digital Assets. Railway Application

419

3.2 Research European Projects The approach from the scientific point of view as analyzed in 3.1 suggests that there is weak research, so it is of special attention to complement it with the applied visions, aiming to rescue the praxis from the industry itself, these 2 sources have been investigated: • Europeans Investigations: Shitt2rail (Rail, s.f.) Fig. 3: S2R Projects • Industry-Specific Research: UIC (UIC, s.f.).

Fig. 3. S2R Projects 2022

From the research 2 studies are specifically related, the first of them referring to the taxonomy of assets developed by “The Railway Innovation Hub, Spain” and the second to the study of Big data for Asset Management, developed by UIC. From the first reference, it is clear that in the last year it has focused efforts on the development of BIM Railway Classification System Manual. Focused mainly on BIM systems (Ali et al. 2022) and consequently on digital models for construction. From the second, the following key aspects for this research are rescued: 1. The first thing is that it is reaffirmed that it is possible to find multiple investigations that point to the digitalization of assets, but that most of them focus on rolling stock and its peripherals, leaving as a corollary that in the aspect of digitalization of railway infrastructure is limited to a few.

420

M. R. Hernández et al.

2. Second, it is argued that it is essential to adopt a systemic perspective for the management of physical assets (Macchi et al. 2012), which will certainly condition the way in which data is handled (big data management), because it is these same, coming from the individual asset (bottom up), which determine the strategies (strategic decision-making) that must be addressed systemically (top down). Figure 4.

Fig. 4. Top-down and bottom-up approaches for systemic and asset-centric, datadriven decisionmaking. (Roda 2022)

3. Third, the existence of a high-level taxonomy (Sedghi et al. 2021), characteristic of railway systems, is recognized, where it is possible to recognize 2 large groups of assets: network assets, discrete assets (Roda 2022).

4. Fourth, a methodology for the adoption of Big Data is recognized, where it is argued that one of the first activities will be the definition of relevant elements and the impact that business decisions may have on it. Where for each of them there will be the “data wish list” necessary to make those decisions better. We can indicate from this study that these 4 elements, as a whole, provide a first approach to the processing of data in the railway 4.0 industry, and together justify the need to have a standardized architecture, which allows the exploitation of data in a systematized way (Fig. 5).

Hierarchy Definition for Digital Assets. Railway Application

421

Fig. 5. Asset taxonomy for railways organization and current digitalization levels (Roda 2022)

3.3 Railway Administrators The third part of the research has been developed on the actual application of a maintenance management model for a railway operator in Europe. That it has considered the revision of its current data models and applications that allow its exploitation. In this context, because of an internal inquiry at European level, it has been possible to collect the empirical experience of several operators, which as a result demonstrates a high level of data variability in other railways. This experience allows us to verify that there are multiple models that have developed naturally over time, that is, they have been adaptations to a progressive growth of information, usually hyper specialized, which effectively solve their local problems, but wich do not have systemic capacities that allow the exploitation of data and the identification of assets in a digital and unique way. In the case study itself, we have found that such a level of personalization becomes an advantage from the perspective of a specialty and its own requirements, but that, on the other hand, it becomes a great disadvantage when it comes to exploiting the information, given the architecture of the system and the data. Problems such as taxonomy, multiple identification of the same asset, multiplicity of structures and/or hierarchies, multiplicity of information sources (ranging from databases to pdfs un-processable digitally), in short, the “big data” ends up transforming into “Frankenstein data”.

4 Conceptual Scheme for Development a Digital Ar Chitecture: Railway Application The work presented in this paper develops an adaptation of the current models in the scientific discussion about digital twins and 4.0 industry. Adding elements and their relationship in a multi-system context (Macchi et al. 2012), applied specifically to the context of the railway industry. This provides as a result a proposal that is based on the 3 groups of layers indicated at the introduction, the first that is related to the real physical assets, the second that groups all the capture, processing, modeling and integration of the data related to each real asset, and finally the third layer that consolidates decision making (Zheng et al. 2021), normally supported in multivariable decision models (Crespo et al. 2018), such as the criticality model, in the case of asset hierarchy (Fig. 6). As an approximation to a general data model (Candón et al. 2019) for the development of the railway industry 4.0, and as a result of the practical application in the case studies, a

422

M. R. Hernández et al.

Fig. 6. RAMI 4.0 Functional Architecture (Schweichhart 2016).

model is presented that proposes to correlate at least 5 elementary dimensions, these are: Hierarchical Structure of the assets which will establish their functional dependence for the required purpose of the set Functional Units of System, recognized as the minimum unit of process that generates value; Real Assets Layer, which uniquely identifies each physical asset; Digital Twin layer, which digests, recognizes and processes the data of each real Asset, providing the Digital identity and some result according to some given output; BDM Layer, Layer of business decisions from the Big data (Weik et al. 2022) of the system obeying the business rules and exploitation of information that the Stakeholders demand. 4.1 RDTA: Railway Digital Twin Architecture (Schematic) Based on the principles proposed in the digitalization model structures such as RAMI, CDT among others (Zheng et al. 2021), the scheme introduces a new concept that allows it’s adaptation to the railway industry, this takes charge of the way in which the minimum units of value are defined in a railway context, which. will be classified as both network units or discrete units (Leitner et al. 2017). We have called this concept a Functional System Unit (Fig. 7). Therefore, the integration of a real asset, duly structured in a hierarchy that places it at a certain level and a functional unit in which it provides a service, all on the layer of digitization rules (recognition of the same asset in multiple systems), will provide each asset with what we have called a digital identity, which can then be conveniently exploited in the last layer of business decisions. Those that may be to a lesser or greater degree digitalize them, according to the internal digitalization models (Weik et al. 2022) that each architecture implementation defines, being able to be these simple relationships between databases, up to complex Machine Learning models integrated in real time. FSU: Functional of System Unit, minimum unit within the value chain, where value is added to the process as such, autonomous unit where all the existing systems in the hierarchy coexist, which will be related, logically and functionally through

Hierarchy Definition for Digital Assets. Railway Application

423

Fig. 7. Railways Digital Twin Architecture

these UFS, allowing to establish a base relationship parallel to the hierarchy of systems/equipment/component, that allows to emulate and assess the function of each asset with it’s impact at the systemic level (Mohammadi and El-Diraby 2021), the above will complement the traditional structure/hierarchy of assets and will cross the layer of inventories, allowing to order from the base to the assets according to their taxonomy and their function provided at the same time. Recognizing, also, at this level two types of units, the “Network units” and the “Discrete Units” For example, a journey vs a bridge, a tunnel. (Carretero et al. 2003). HIERARCHY LEVELS: Defined as System, subsystem, component according to the railway taxonomy. LAYERS: Grouped into 3 macro levels, real level, digital level, decision level. 4.2 Proof of Concept: Railway Application The study on computer solutions for Asset Management EAM, BIM Systems, GIS Systems, allows us to recognize that it is feasible at the digital level to recognize an asset through a volume of control (real, digital or both), which may be as small or as large as the user defines. Step 1: Considering the criteria established in our RDT model, we define a digital asset within a control volume, which will have as an elementary reference geo spatiality, recognizing this as the most absolute feasible reference, each control volume will therefore have it’s 3 dimensions, expanding the current reference systems that are only

424

M. R. Hernández et al.

limited to flat referencing (v-pk). To make that, as a proposal proof of concept we consider all the assets existent into inventory (rail, sleepers, signally, welding, swishes, etc.) that belong to some functional section (for example from station A to station B), then propose the relationship with the element defined an BIM platform to identify and match the singularity elements defined on BIM model with the individual element defined on the inventory platform. Finally, we relate the 2 previous platforms and correspond data references with the GIS system, is very important to remark that the systems currently are not integrated. Figure 8. Step 2: categorization of functional units (network or discrete). To define if a unit is a network or discrete unit, we concentrate in the function “the proposal” of the UFS, in this sense we can define it like an example of network unit: the segment between 2 stations, including all kinds of assets that belong to this segment (permanent way, electrification system, signaling, telecommunications, etc.). For the other side if we find a bridge, a tunnel or switches and crossing elements, we are extracting them from the network unit and recognize like a discrete unit, with a proposal itself. Step 3: Hierarchical structuring of assets, defined by their taxonomy according to specialty. Is a principal effort to give a digital identity to the assets, mainly on the EAM systems, where we can give the parameters for the decision-making model.

Fig. 8. Example of Real word Railways Assets vs Digital word on BIM, EAM or GIS system (Image: Siemens Mobility)

Step 4: Allocation of the real assets into each UFS with it’s respective hierarchical structure associated, according to how many real assets you have, level of grouping by specialty connected online. Level of detail of each specialty according to each Linear or point element defined. In this point it is very important to define some rules to truncate or divide the asset, mainly the linear assets, because for example: not necessarily the limit or the end of one permanent way is the same limit or end of a telecommunication system. Then the final of the UFS will not be referenced to some unique linear reference system, but to multiples systems, integrated in a digital twin (Figs. 9, 10, 11).

Hierarchy Definition for Digital Assets. Railway Application

425

Energy

Track

V1 V2

Sign Inf. V1 + PK

10.400

10.900

10.400

10.900

11.000

11.100

11.000

12.100

11.100

14.100

12.100

14.100

V2 + PK

Fig. 9. Example of coexistence of complex multiple end-limit between specialities on a railway network

TruncaƟon Rules

U1: SecƟon A

U2: SecƟon B

U3: S&C Zone A

U4: SecƟon C

U5: StaƟon Zone

U6: SecƟon C

U7: Tunel Zone

U8: SecƟon C

U9: Bridge Zone

Swch 123

Division Rules

CDV A13 CDV 22 CDV 33 Point LAC 3 Sig Ph A1 Sig Ph A2

Assets may be classified into - Linear Assets - Punctual Assets The Unit FuncƟonal of System (Control Volumes) will be classified into: - Network UFS - Discrete UFS The locaƟons may indisƟnctly contain linear assets or punctual assets that will belong to a certain control volume based on their reference coordinate system (georeferenced)

Fig. 10. UFS Railways Multiply System Scheme

Fig. 11. Business, Network and Enterprise Units

Like a result of the application of the model we observe, first that if the organization aligns the digital definitions of each asset through one “management perspective”, according to one model, allowing all the uplevels for the exploitation of the information to decision-making layers. The homogenization and standardization of each unit with the same criteria, allow too to scale the model for all the company, gives the capability to use advanced techniques of data mining or business analytics to get the best result for the business, always aligned with one criticality perspective.

426

M. R. Hernández et al.

5 Conclusions The main contribution of the study is that it manages to establish at least 2 clear lines of research and development for the future structuring of digitalization models applied to the railway industry. The first line is the incorporation of the factor “management” (through a model) in the determination of a digital asset. Digitalization itself has only been focused on technical aspects, but not in management aspects that aim to maximize the result of the business. In this sense, the proposal is towards the integration of all those computer models under the line of a clear Management model (The MMM). The second line, is the model RDTC itself like a conceptual scheme, incorporating the systemic perspective in the asset management models, consequently, of their digital twins. The digital asset normally is defined individually, but not necessarily establishing a functional relationship, that determines the impact on the business. In this sense, it is proposed as an essential factor, the use of a systemic and functional criteria, which fixes the hierarchical dependence of the assets respect to their taxonomy, and their functional relationship (data source) from the equipment to the system (Button-up perspective), enabling decision-making (top down) with the apropriate information of all assets and their affectation to the whole. Acknowledgements. This paper has been written within the framework of the projects INMA “Asset Digitalization for INtelligent MAintenace” (Grant PY20 RE014 AICIA, founded by Junta de Andalucía PAIDI 2020, Andalucía FEDER 2014–2020) and Geminhi (Digital model for Intelligent Maintenance based on Hybrid prognostics models), (Grant US-1381456, founded by Junta de Andalucía, Andalucía FEDER 2014–2020).

References Ali, K.N., et al.: Citation: collaboration and risk in building information modelling (BIM): a systematic literature review (2022) https://doi.org/10.3390/buildings12050571 Candón, E., et al.: Implementing intelligent asset management systems (IAMS) within an industry 4.0 manufacturing environment. IFAC-PapersOnLine. 52(13), 2488–2493 (2019). https://doi. org/10.1016/J.IFACOL.2019.11.580 Carretero, J., et al.: Applying RCM in large scale systems: a case study with railway networks. Reliab. Eng. Syst. Saf. 82(3), 257– 273 (2003). https://doi.org/10.1016/S0951-8320(03)001 67-4 Crespo Márquez, A.: The maintenance management framework. Models and methods for complex systems maintenance. Springer, UK, ISBN 978-1-84628- 821-0 (2007). https://doi.org/10.1007/ 978-1-84628-821-0 Crespo Márquez, A., González-Prida Díaz, V., Gómez Fernández, J.F.: Advance Maintenance Modelling for Asset Management (2018). ISBN978–3–319–58044–9 International Union of Railways. Maximising the benefits of big data for asset management Recommendations (2022) Leitner, B., Môcová, L., Hromada, M.: A new approach to identification of critical elements in railway infrastructure. Procedia Eng. 187, 143–149 (2017). https://doi.org/10.1016/j.proeng. 2017.04.360

Hierarchy Definition for Digital Assets. Railway Application

427

Macchi, M., et al.: Maintenance management of railway infrastructures based on reliability analysis (2012). https://doi.org/10.1016/j.ress.2012.03.017 Mohammadi, A., El-Diraby, T.: Toward user-oriented asset management for urban railway systems. Sustain. Cities Soc. 70, 102903 (2021). https://doi.org/10.1016/J.SCS.2021.102903 Roda, I.: Maximing the benefits of big data for asset management. UIC Rail System Department (2022). ISBN 978–2–7461–3186–6 Sedghi, M., et al.: A taxonomy of railway track maintenance planning and scheduling: a review and research trends. Reliab. Eng. Syst. Saf. 215, 107827 (2021). https://doi.org/10.1016/J.RESS. 2021.107827 Schweichhart, K.: Reference Architectural Model Industrie 4.0 (Rami 4.0). An Introduction (2016). Available online: https://www.plattform-i40.deI40 UIC, A.W.G.: 220430_UIC E2E decision tools landscape and high-level roadmap.pdf. Paris (2022) UIC, S.F.: UIC webpage. [online] Available at: https://uic.org/research/Research-Projects. Accessed Mar 2022 Rail, E., S.F.: Europe’s Rail webpage. [online] Available at: https://rail-research.europa.eu . Accessed Mar 2022 Weik, N., et al.: DFT modeling approach for operational risk assessment of railway infrastructure. Int. J. Softw. Tools Technol. Trans. 24, 331–350 (2022). https://doi.org/10.1007/s10009-02200652-4 Zheng, X., Lu, J., Kiritsis, D.: The emergence of cognitive digital twin: vision, challenges and opportunities. Int. J. Produc. Res. 60(24), 7610–7632 (2021). https://doi.org/10.1080/002 07543.2021.2014591

Big Data Adoption in Strategic Decision-Making for Railway Infrastructure Asset Management Irene Roda1(B) , Adalberto Polenghi1(B) , and Vesa Männistö2(B) 1 Department of Management, Economics and Industrial Engineering,

Politecnico di Milano, Milan, Italy {irene.roda,adalberto.polenghi}@polimi.it 2 Finnish Transport Infrastructure Agency, Helsinki, Finland [email protected]

Abstract. In this study we investigate Big Data adoption in Asset Management (AM) decision-making at strategic level for Infrastructure Management (IM) in railway. In fact, the formulation of strategies and objectives for the asset portfolio management for IM organisations is a crucial and non-trivial issue, and proper information and data management is required. Several are the studies about Big Data analysis for AM that have been published in the last years, but they mainly aim to develop models and algorithms to support decisions at tactical and operational levels rather than focusing on strategic AM decisions. This paper presents the results of a project funded by the International Union of Railway (UIC) that involved representatives of nine IM organisations in Europe. First, an integrated framework which depicts the different AM strategic decisions of an IM organisation is provided. Then, insights on expected benefits, with focus on data source characteristics and stakeholders’ involvement are presented. Finally, opportunities and challenges of the adoption of Big data to support those decisions are identified. Moreover, recommendations and a roadmap towards Big data-based AM strategic decision-making are presented, which may help to define use cases for future R&D programmes.

1 Introduction In railways, infrastructure managers (IMs) have been given much attention to Asset Management (AM) in the last years. Indeed, for IM organisations, physical assets are core for providing service that can be measured through various outputs like customer satisfaction, safety, sustainability (economical, environmental, and societal) and availability. The formulation of strategies and objectives for asset portfolio management is challenging for IMs. In fact, there are different expectations among the various stakeholders and, consequently, different perceptions of asset value and the benefits that asset life cycle management can achieve. Moreover, strategic decisions impact the long-term, address ambiguous and complex issues, engage various departments, and involve a high level of organisational resources (McKenzie et al., 2011). Hence, gathering, analysing, and considering reliable data and information is critically important, and this is also remarked in the ISO 55001 standard. Big data is the new frontier for collecting and © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 428–438, 2023. https://doi.org/10.1007/978-3-031-25448-2_41

Big Data Adoption in Strategic Decision

429

analysing data and for turning it into usable information (Intezari et al., 2017). This study investigates Big data opportunities in railways for AM strategic decision-making processes to grant the balancing of costs, risks, and performances over the asset life cycle.

2 Methodology and State of the Art 2.1 Research Methodology The overall methodology this work is based upon, is composed by three phases, which are further detailed by specific research methodology, as depicted in Fig. 1, where core activities are highlighted in bold.

Output

Activities

Conceptualisation phase

Exploratory phase

Generalisation and reporting phase

Framework synthetisation

Interviews to experts in the field

Framework refinement and report preparation

Literature review on: • Big Data in railway • Big Data in assetintensive sectors • Available software tools • Relevant research programmes

•

Interview protocol definition Pilot interview implementation Interview protocol revision Experts interviews implementation

•

•

Results synthesis from conceptualisation and exploratory phases Workshop for results validation and roadmap definition Finalisation

•

•

Updated and integrated framework

• •

Finalised framework Roadmap

Proposed framework

• • •

•

Fig. 1. Methodology of the research project.

Firstly, in the conceptualization phase, a systematic literature review was carried out to identify and hypothesize the foundations on which expert judgments elicitation was used as qualitative method to leverage the experience and point of view of experts on the studied topic. Secondly, an exploratory phase was carried out so to grasp feedbacks from experts in the field. Semi-structured interviews were adopted for experts’ judgements elicitation, and two pilot interviews were carried out to better tune the questionnaire before implementing all the interviews. Finally, a focus group was organized involving the representatives of ten IMs organisations around Europe. Table 1 shows the panel of experts that were involved. Overall, the collected evidence allowed identifying the main data types and source which are useful for supporting AM strategic decisions. Moreover, opportunities and challenges of the adoption of Big Data were collected. In the following section the main findings are reported, based on the integration of the opinions of the interviewed experts and the literature review evidences.

430

I. Roda et al. Table 1. Panel of experts involved belonging to IMs organisations in Europe.

IM Organisation

ountry

Role of the experts in the organisation

A*

Italy

Head of AM

B*

Finland

Senior Advisor in AM

C

Germany

Head of Maintenance

D

United Kingdom

Head of Advanced Analytics

E

Belgium

Strategy Department

F

Austria

AM and Strategic Planning

G

Spain

Network Performance Deputy Director

H

Ireland

Head of AM

I

Sweden

Senior Advisor in AM

L

Switzerland

Senior Consultant Business Development

* engaged for pilot interviews in the exploratory phase

2.2 State of the Art The review of the literature was carried out using the research query: ((“Big Data” AND “Asset Management”) OR (“Big Data” AND “Railway Infrastructure”) in Scopus database. Considering the papers that were selected to be analysed after titles and abstract screening, the following evidence can be summarized. Several studies about Big data analysis for AM in railway have been published in the last years, but they mainly aim to develop models and algorithms to support decisions at tactical level, like managing train traffic or improving maintenance performance, as also confirmed by recent reviews and surveys (McMahon et al, 2020; Zhu L. et al., 2019; Ghofrani F., 2018). Overall, few are the articles discussing about strategic decisions supported by Big data analytics, and most of those that were identified are oriented towards evaluating the commercial performances of the railway network and how it satisfy the needs of the users of the infrastructure (Han et al., 2020; Moyano et al., 2018; Wei et al., 2018; Rotoli et al., 2016).

Fig. 2. Evidence from the literature review

Big Data Adoption in Strategic Decision

431

3 Big Data and AM Decisions in Railways 3.1 Modelling of Domain-Related Concepts The review of the literature as well as the experts judgment elicitation showed that the AM topic can be seen differently according to the considered viewpoints and, specifically, on the role within the organisation of the interviewees. Therefore, to harmonise the terminology and set a common background to which both interviewers and interviewees could refer to, a data model is realised, which is representative of the domain under analysis, i.e., Big data for strategic AM decision-making. The data model, shown in Fig. 3, is accompanied by a vocabulary (for the sake of brevity not shown here) that guaranteed consistency regarding terms and intended meaning so that questions regarding “data source” were properly understood and interpreted.

Fig. 3. Data modelling of Big data within AM decision-making.

The core concept is “data”, which are intended, in this specific case, at large: not only raw data from sensors or information systems, but also information, which are generally conceived as elaborated data. This was a mandatory simplification due to the very different roles each interviewee had. From “data”, the first connection is with “Big Data method” as a way to transform it into useful insights for one or more “decision” that refers to one or more “asset” which is in a specific lifecycle stage, i.e., Beginning of Life (BoL), Middle of Life (MoL) and End of Life (EoL). On the other side, “data” are generated by one or more “data source”, which, in this work, is intended as a hardware/software system. The “data source” could be owned by a one or more “stakeholder”, which may be an internal or external stakeholder to the IMs organisation. Based on the presented data model, the interviewing campaign was designed, and questions targeted specific aspects, from the data itself to the type of decision. Even though the proposed data model is not the main output of the project, it became a cornerstone that guarantees meaningful results, which are summarised in the following subsections.

432

I. Roda et al.

3.2 Strategic AM Decisions for Railway IM Organizations As a first outcome of the research project, the main AM strategic decisions for IMs organisation in railways have been identified and classified into two main categories (Fig. 4): • Lifecycle delivery strategic decisions: strategic decisions linked to the lifecycle of the assets divided into its three main stages: BoL, MoL, and EoL. • AM support decisions: transversal decisions through the lifecycle, helpful in ensuring an efficient and effective AM service delivery. • A first version of the framework was defined based on the literature review and was then integrated and refined based on the experts judgments elicitation.

Fig. 4. AM strategic decisions: AM lifecycle delivery and AM support decisions.

Each decision has been properly defined as reported in Table 2. Table 2. Description of AM strategic decisions for railway IM organizations Strategic decision

Description

Lifecycle delivery decisions Asset creation and acquisition

Definition of new asset development based on requirements, needs and functionalities that it must have or comply with

Maintenance strategy definition

Definition of the best inspection and maintenance policy mix

Operations strategies definition

Implementation of the programs, services, policies, or systems, and related procedures (continued)

Big Data Adoption in Strategic Decision

433

Table 2. (continued) Strategic decision

Description

Asset reconfiguration

Evaluation of the deviations of the technical-economic parameters

Maintenance policy redefinition

The redesign of the maintenance strategy that must take into account performance, risk and costs

EoL strategies definition

Definition of timing and type of EoL strategy

AM support decisions AM planning

Definition of the corporate objectives related to the asset portfolio and the means, tools and actions to achieve them

Finance and accounting

Documenting, analysing, summarising and reporting the transactions arising from business operations

Resource management

Planning, acquiring, scheduling and allocating of tangible and intangible resources

Asset information management

Definition of technology and software to be used to deliver asset information strategy, information flows and system interfaces

Asset management system monitoring Monitoring, measurement, analysis and evaluation of the efficiency and/or effectiveness of AM processes in achieving organisational objectives Risk management

Identification, evaluation and prioritisation of risks

Human resource management

Recruiting, selecting, training, development, compensation and appraising of employees

Legislation and technical standards

Specifications which lay down the characteristics of a product, process or service

The decisions and sub-decisions listed in Table 2 result from the analysis of technical and scientific documents and the integration of the evidence collected through the interviews. It must be noted that the strategic decisions taken by an IM organisation depend on a series of contextual factors: • • • •

Role of AM within the organisation. Dimension and available financial resources of the organisation. Managed infrastructure (single or multi-infrastructure). Country where the IM organisation operates.

Moreover, not all AM strategic decisions may be taken by all IMs organisations as other actors could cover them, depending on the contextual factors previously listed.

434

I. Roda et al.

3.3 Expected Benefits from Big Data for Strategic Decisions Another outcome of the research is the list of the expected benefits by exploiting Big data for AM strategic decisions, which have been categorized into three main groups: i) Asset performance: higher reliability, higher capacity; ii) AM performance: better resource planning, better system design, higher effectiveness, cost reduction; iii) Service offering: delay reduction, higher flexibility, improved sustainability, higher safety. Overall, Big data are expected to improve business performance of IMs organisation as they could be disruptive in how decisions could be judged. Especially, interviewees confirmed that BoL-related decisions may be majorly impacted by the use of Big data given their intrinsic characteristics of high-risk and long-term horizon. Nonetheless, within MoL, maintenance strategy redefinition is considered to be favoured by the use of the huge amount of collected data. Moreover, AM support decisions are expected to be positively impacted as they may become more data-driven rather experience-based. In fact, the data-driven decision-making approach promoted by Big Data favours more objective planning, habilitating lifecycle delivery decisions to provide value. As final reflection upon the benefits of Big data referred to the nature of Big data itself. The question that came to the mind of the experts was “are really all the 3 Vs of Big data necessary”? After the exploratory phase, the discussion on whether the 3 Vs together are needed for supporting AM strategic decision-making is still open, but some insights have been reached: • Regarding volume there is a widespread agreement that the largest the data to rely upon the better strategic decision could be supported; this is specifically true for asset-related data concerning failures, degradation, effectiveness of interventions and related cost, and many others. • Regarding variety, a consensus was reached as strategic decisions do include several aspects, hence data must come different sources. Indeed, the sustainability trend of these years is stepping up the challenge for decision-makers as economic, social, and environmental impacts of decisions must be assessed. As such, different datasets must be considered, also to maximise service offering to customers, both of passenger and freight trains. • Regarding velocity, instead, no agreement was reached during the interviews and the final focus group. On one side, to offer an up-to-date service for improving customer experience is of paramount importance to accomplish to the social dimension of sustainability; in turn, this makes real-time and updated data mandatory. On the other side, the open question is “for strategic decisions, does real-time data really make a difference?”; the objection refer to the almost static nature data used to strategic decisions. Operational decisions do need real-time data as asset performance must be managed continuously and related practices updated, enabled by high-velocity data from monitoring, strategic decisions do not, especially for BoL-related strategic decisions. 3.4 Data Types and Sources Considering the asset-centric and data-driven AM strategic decision-making process, bottom-up and top-down approaches should be put in place jointly. As emerged from

Big Data Adoption in Strategic Decision

435

the study, AM strategic decisions must consider asset-related data coming from single assets (bottom-up), which guarantees a thorough knowledge of the assets enabling informed decision-making at different aggregation levels by relying on systemic analysis. Moreover, when dealing with strategic decisions, not only asset-related data are needed, but external data either from other organisation department/s and from external stakeholders are recommended to be integrated as well. From a top-down perspective, based on the analysis carried out, strategic decisions can then address both the top level of the asset breakdown structure (systems/ systems of systems) and every single asset, guided by the systemic analysis of data (criticality analysis), allowing to prioritise assets and systems for budget allocation (Fig. 5).

Fig. 5. Bottom-up and top-down approaches for data-driven AM decision-making.

Among the main asset-related data the following were identified: lifecycle data referring to the technical characteristics of the assets in the asset register, inspections data, traffic data, sensors data, maintenance/safety reports data, and passengers/users data. The main external data are socio-economic data; environmental data; capacity data; energy consumption data; government data; stakeholder expectations data and financial data. So far, in the practice mostly asset-related data are collected and analysed rather than external data, and they are mostly numerical data used together with image data and textual data coming from track inspections and safety/maintenance reports. When referring to valuable data for AM strategic decisions, whether they are asset-related data or external data, it must be considered that they may be owned by the IM organisation itself or they may be collected from external stakeholders. Railway operators are among the relevant stakeholders who can provide additional data to the railway AM organisations. Other stakeholders identified as significant are government and local authorities and technologies and services providers. Referring to this aspect, strong data governance becomes crucial, meaning clear data accountability to define who is responsible and ensure ownership; also including data privacy and security to protect asset data and ensure sensitive information are safeguarded. As a summary, Table 3 reports IMs’ stakeholders in the railways sector and related data type that are useful for AM strategic decision-making.

436

I. Roda et al. Table 3. IMs’ stakeholders and related data.

Stakeholder

Data type

Railway operators

Track data Power consumption Demand data Traffic data

Government and local authorities

Demand forecasting Traffic data Inspection and safety reports Financial data

Technology and service providers

Lifecycle data

Construction industries

Infrastructure data

Electricity operators

Energy consumption

Local industries

Future transportation load

Others (Institutions, consultancy etc.)

Climate change Weather forecast Coastal erosion Satellite data

4 Towards Big Data Driven AM Strategic Decisions 4.1 Opportunities and Challenges of the Adoption of Big Data The foreseen opportunities related to the use of Big data for strategic AM decisions in the railway sector are several. On the one hand, Big data is recognised as a contributor for better planning of AM activities, enabling: crossing of multi-type data for improved decision-making; transparent and real-time data sharing; exploitation of cross-industry data; real-time data analysis and visualisation; defining objective decision-making criteria. Moreover, Big data are recognised as an opportunity for supporting the digital transformation and modernisation of railways industry and generating synergies among European IM organisations. Some of the opportunities are considered reachable in the shorter-term (less than 5 years), such as: crossing multi-type data, transparent data sharing, real-time data analysis and visualisation, and getting to objective decision-making criteria. Considering the main challenges for Big data exploitation in AM strategic decisions, both technical and organisational challenges have been identified. The main technical challenge is the information systems interoperability, given the large number of information systems used by the IM organisations. Regarding other technical challenges, it is relevant to highlight the high variety of data to be managed; cybersecurity related aspects and proper data governance to be put in place. Regarding the organisational challenges, the main ones that emerged are: organisational commitment to evidence-based decisions

Big Data Adoption in Strategic Decision

437

making; data culture; data transmission with external stakeholders and within the organisation; skills and competencies on Big data and the difficulty in quantifying benefits vs costs. 4.2 Recommendations and Roadmap Being aware of both opportunities and challenges of the adoption of Big data is essential to trace down the evolution path towards Big data-based AM strategic decision-making. On the technical side, the cornerstone is asset digitalisation: guaranteeing that asset are “smart” and that can provide the necessary data and information is fundamental to create the dataset/s for further analysis. At the same time, input data should be carefully managed. Cybersecurity is also a relevant aspect to be addressed. On the managerial side, the decisions should address the ultimate objective of AM which is value generation. Hence there is the need to be consistent with the value framework of the organisation that should be stated in the Strategic AM Plan (SAMP) document (UIC, 2020) and must set the course of any strategic decision to comply with company strategy. Clear mapping of the AM decision-making processes in the organisation is also necessary for setting up proper decision criteria to guarantee consistency with the value framework. Stakeholders must be more and more engaged in the decision-making process. Moreover, imprinting data culture in the organisation is crucial. Training must be set up to let people in the organisation have the right skills and competencies to understand the outputs of Big data analytics correctly and properly criticise the given inputs. Overall, the main actions to be followed in a roadmap towards adopting Big Data for AM strategic decision making in IM organisations are the following: (1) Define a list of prioritised value elements and the decisions that impact on them, (2) Identify for each decision the set of already available useful data as well as a data wish-list including data that may improve the judgement of such decisions, (3) Set up a set of Proof of Concepts (PoC) where data integration is pursued at first and where small-sized, fit-for-purpose data analytics are developed, (4) once the PoC are concluded, scale up the analytics towards the use of Big data in terms of volume and variety. On the cultural side, several are the actions that should be planned: (1) Commit from top and middle management to promote data-driven decision-making approach in the whole organisation, (2) Strengthen competences related to data and their analysis, (3) Engage employers into small projects and PoC, (4) Promote inter-department collaboration.

5 Conclusions The investigation performed provides guidelines to set the ground for appropriate use of Big data for Asset Management strategic decisions in IM organisations. What emerged from this study is that to date, Big data is under-utilised in railway strategic decision making while it could bring several benefits for improving both asset performances and service offering. As final output, the report provides a vision and a roadmap towards Big data-based AM strategic decision-making for IMs, which may also help to define use cases for R&D programmes.

438

I. Roda et al.

Acknowledgements. The research work was performed within the context of the project entitled “Maximising the benefits of Big data for Asset Management” sponsored by the UIC Asset Management Working Group (AMWG). More information can be found in the report: 5-22002E, Maximising the benefits of big data for asset management, March 2022, UIC. All output and all work carried out are exclusive property of UIC. The authors would like to thank the representatives of the organisations involved in the study who have provided important experiences, opinions and facts for the research.

References McKenzie, J., van Winkelen, C., Grewal, S.: Developing organisational decision-making capability: a knowledge manager’s guide. J. Knowl. Manag. 15(3), 403–421 (2011) Intezari, A., Gressel, S.: Information and reformation in KM systems: big data and strategic decision-making. J. Knowl. Manag. 21(1), 71–91 (2017). https://doi.org/10.1108/JKM-072015-0293 McMahon, P., Zhang, T., Dwight, R.: Requirements for big data adoption for railway asset management. IEEE Access 8, 15543–15564 (2020) Zhu, L., Yu, F.R., Wang, Y., Ning, B., Tang, T.: Big data analytics in intelligent transportation systems: a survey. IEEE Trans Intell Transp Syst 20, 383–398 (2019) Ghofrani, F., He, Q., Goverde, R.M.P., Liu, X.: Recent applications of big data analytics in railway transportation systems: a survey. Transp Res Part C Emerg Technol 90, 226–246 (2018) Han, B., Min Wan, Yu., Zhou, Y.S.: Evaluation of multimodal transport in china based on hesitation fuzzy multiattribute decision-making. Math. Probl. Eng. 2020, 1–9 (2020). https://doi.org/10. 1155/2020/1823068 Moyano, A., Moya-Gómez, B., Gutiérrez, J.: Access and egress times to high-speed rail stations: a spatiotemporal accessibility analysis. J. Transp. Geogr. 73, 84–93 (2018) Rotoli, F., Malavasi, G., Ricci, S.: Complex railway systems: capacity and utilisation of interconnected networks. Eur. Transp. Res. Rev. 8(4), 1–21 (2016). https://doi.org/10.1007/s12544016-0216-6 Wei, S., et al.: Open big data from ticketing website as a useful tool for characterizing spatial features of the Chinese high-speed rail system. J. Spatial Sci. 63(2), 265–277 (2018). https:// doi.org/10.1080/14498596.2018.1497561 Esveld, C., Esveld, C.: Modern railway track, vol. 385. Zaltbommel: MRT-productions (2001) UIC (2020). UIC SAMP Application Guide - UIC SAMP Application Guidelines for Railway Infrastructure Organisations

The Potential Value of Digital Twin in Rail and Road Infrastructure Asset Management João Vieira1(B) , Hugo Patrício2 , João Poças Martins3 , João Gomes Morgado2 , and Nuno Almeida4 1 IP/CERIS/IST, Lisbon, Portugal [email protected] 2 IP, Almada, Portugal {hugo.patricio,joao.gmorgado}@infraestruturasdeportugal.pt 3 CONSTRUCT-GEQUALTEC/FEUP, Porto, Portugal [email protected] 4 CERIS/IST, Almada, Portugal [email protected]

Abstract. Asset management aims at realizing value from assets, meeting stakeholders needs and expectations by balancing financial, environmental, and social costs, risks, quality of service and performance of assets. Asset managers need to know the assets they manage, so that organizational knowledge and decisionmaking are improved, and the global value generated by the asset portfolio is maximized. Therefore, asset management is a data-intensive activity, and asset managers need tools and processes to efficiently collect, assemble, manage, analyze, and use asset data. In this regard, Industry 4.0-driven approaches, such as the Digital Twin (DT), promise to contribute to asset management decision-making and realize value from asset data. However, organizations that may consider DT application face different interpretations on the concept and varying expectations regarding its potential impacts. Based on the contributions of diverse information sources, this work aims to identify opportunities and challenges arising from its use and discuss them in the context of road and rail infrastructure asset management.

1 Introduction Annual passenger traffic is expected to increase by 50% until 2030 and global freight volumes by 70% (SuM4All, 2017). The road and rail networks account for more than 63% of goods transport and almost 90% of passenger transport within the EU (EU, 2020). These networks face other equally relevant challenges, such as large maintenance backlogs (OECD, 2020), continuous aging of infrastructures (OECD, 2019), the need to increase their resilience, and increasingly demanding environmental targets. Investments in increasing the knowledge about road and rail infrastructures and in digital transformation are opportunities of high potential for disruption and value generation, both for organizations and their stakeholders. Although these solutions are currently spreading at an increasing pace (Tao et al., 2019; Lamb, 2019), organizations may face difficulties in perceiving their viability and the benefits generated to asset © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 439–447, 2023. https://doi.org/10.1007/978-3-031-25448-2_42

440

J. Vieira et al.

management activities. One such solution is the Digital Twin (DT). Like other approaches that integrate the so-called Industry 4.0, the DT has been attracting increased interest. However, organizations that may consider its application face different interpretations on the concept and varying expectations regarding its potential impacts. This work aims to identify opportunities and challenges arising from its use and discuss them in the context of road and rail infrastructures.

2 Digital Twin Although the concept of a DT is not yet stabilized (Callcut et al., 2021; Kritzinger et al., 2018), it is understood in the context of this paper that a DT is a digital representation of a physical asset or asset system and its operating environment, integrating a data connection to the physical asset or asset system, as well as other supporting tools and sources (such as physical models, data analysis, simulation, and predictive capabilities), used to synchronize the physical asset to the virtual twin and to generate insights aligned with a predefined purpose and, ultimately, to overcome obstacles and promote robust asset management decision-making processes (Vieira et al., 2022). Figure 1 built from the contributions of Harper et al. (2019), Stanford-Clark et al. (2019), Parrot and Warshaw (2017) and ARUP (2019), illustrates this definition in the form of a conceptual architecture for the DT, and describes the various phases of its action cycle.

Fig. 1. Digital Twin conceptual architecture

The proposed architecture is a template structure for a DT and may change depending on the context and purpose of each DT. The architecture is divided into 7 phases: creation (phase 0), communication (1 and 6), aggregation (2), analysis (3), perception (4), decision (5), and action (7). In short, data is collected (0) from the physical asset and its environment (e.g., temperature) through sensors and other available sources (e.g., inspections), and then it is transmitted (1) to the virtual space through communication networks and interfaces (e.g., IoT, wi-fi, 4G/5G, Ethernet). Next, data is cleaned and aggregated (2) with other sources (e.g., GIS, fault records, maintenance history), stored (e.g., in data lakes) and prepared for analysis. Data is analyzed (3) using a variety of

The Potential Value of Digital Twin in Rail

441

tools and approaches (e.g., monitoring, classification, prediction, simulation, etc.). The results are shown to the users through visualization interfaces (e.g., 3D models, GIS, 2D CAD model), producing perceptions (4) that support the decision-making process (5). After a decision is made, it is communicated (6) to the physical asset via automated (e.g., decoders) or manual processes and transformed into real actions (7), which can have more or less autonomy (e.g., action on an actuator vs. manual change by an operator). The technological development and complexity level in each phase should be adequate for the DT purpose.

3 Methodology The authors propose mapping the impacts a DT can generate through a literature review. Figure 2 illustrates the mapping process used in this work. The review is based on references obtained from a previous systematic literature review (Vieira et al., 2022) using the PRISMA methodology and concerning the DT in the road and rail context, and on other information sources (such as conference papers, web pages, reports/white papers and ISO standards), covering other sectors of activity (industry, oil and gas, etc.). The list includes 67 sources, consisting of 1 ISO Standard, 16 reports/white papers, 22 web pages and 28 scientific papers/theses.

Fig. 2. Methodology used for the mapping process

The impacts of DT are divided into two types: positive impacts (opportunities) and challenges (risks). The impacts are also divided into 3 levels of affected parties (according to the conceptual framework of value defined by Almeida et al., 2022): life cycle management of physical assets; organization; and stakeholders. The methodology of impact identification consists of taking the various textual descriptions obtained from the bibliographic references as input data and pointing out the existence of impacts (positive impacts or challenges) at the various levels of analysis (life cycle management of

442

J. Vieira et al.

physical assets, organization, and stakeholders), the dimensions of impacts, and the DT capabilities related to them.

4 Results and Discussion 4.1 Results Figure 3 illustrates the number of contributions identified between the information sources, the DT capabilities, and their positive impacts on asset life cycle management, the organization, and the stakeholders. The results show a balanced distribution of the contributions (519) by the information sources, except for ISO Standards, due to the small number of existing publications on this topic. Analytics and prediction (30%), simulation and modelling (23%), and monitoring (21%) are the DT capabilities with the highest flow of positive contributions (74% of all 505 contributions) to the life cycle management of physical assets. Integration and collaboration (12%), visualization (7%), connectivity (4%) and control (3%) have a smaller number of contributions. In terms of the life cycle management of physical assets, there are three groups of activities with distinct levels of contributions. Firstly, operations stand out as the life cycle activity with the most contributions, with about 38% of the total - roughly twice as much as the second largest activity (maintenance). In a second group, maintenance (21%), planning and acquisition (17%), and failure response (13%) have a considerable number of contributions (51% in total). There is a third group with fewer contributions that includes management of risks and opportunities (4%), resources (4%), disposal (2%), asset compliance (1%), and asset traceability ( 33,3 rush hour adds 10 min h delay. Delay is valued on average at e10 per hour

CO2 emissions

Reduction 150 g/kilometre CO2 valued at e 100/tonne

Sick leave rate

Cyclist need less sick 1 day leave (up to 50%). Conservative estimate 1 day per year Cost of sick leave e 400 per day (average FTE cost e 80,000 per year

3000*0,15*1/1000 = 0,5 tonne

Annual equivalent value e 333

e 50

e400

Safety incidents Fatality risk on bicycle 3000 km results in an -/- e120 about 10 times higher additional fatality every per km than car. 25000 years Fatality valued at Me 3 Net value

e663

found in establishing the added value of additional cyclists. Many countries and cities around the world are committed to reducing their carbon footprint (i.e. CO2 emissions) and have embraced cycling as a way to achieve this goal given that it reduces the emission per passenger kilometre to virtually zero. However, in absolute terms the impact is much less. Combined with the relatively low monetary equivalent value (price per tonne), it means that initiatives to promote cycling often struggle with funding. The budget claim thus typically is supported with additional qualitative arguments: reduced congestion, reducing the environmental impact (required space, air quality, noise) and improving the health of the population. What is often not mentioned and may be used to challenge the initiative is that cyclists are more vulnerable in traffic, especially in cities without proper cycling infrastructure. Even by only using a small number of indicators from the basis set a much broader view on the total societal value of additional cyclists5 can be developed. In Table 5 some numbers for the Dutch context are collected for a cyclist replacing 3000 car kilometres per year (200 days * 15 kms travel distance).

5 Such a marginal net value allows for the selection of an efficient portfolio of interventions: those

that cost less per additional cyclist. This is in contrast with the current practice of assigning a budget and only using qualitative (= unquantified) objectives.

458

Y. Wijnia et al.

The results of such a quantification can be a surprise to policy makers. First of all, the increased safety risk would offset the benefit in sustainability. Only quantifying CO2 thus would leave the policy vulnerable to opposition based on the safety impact. To improve safety additional investments would be needed, but the programs are already struggling for budget. Fortunately, CO2 reduction is by no means the most important benefit. Both the reduced production loss due to congestion and the reduced sick leave rate due improvements in the general health of cyclists are about an order of magnitude more important. Given the high net benefit it should be no problem to fund additional investments in the cycling infrastructure. Over all values it may even be amongst the best infrastructure investment options a city has. However to see this it is necessary to consider multiple aspects at the same time which can be difficult in highly compartmented organisations. A common value framework helps in crossing boundaries between de compartments.

8 Conclusion Infrastructure asset managers often struggle with quantifying the value impact their assets and investment programs have on their stakeholders. This complexity is often addressed by technical standards and compartmented organisations and budgets, Unfortunately this results in suboptimal decision, with low yielding or even net negative interventions. In this paper we presented a pragmatic approach for addressing stakeholder values in a common value framework. By clustering the stakeholders into groups with similar interest a more thorough analysis of these interest can be made, and peculiarities of specific context can be addressed. Aligning these interests with a more fundamental theoretical model for value allows for more awareness of the impacted values. The resulting reference model consists of 6 value domains, 18 values, 50 objectives and some 200 indicators. In practice, not all indicators are needed though. With a limited set of less than 20 indicators a 360° perspective can be maintained for a reasonable fraction of the decisions. Such a basic framework can help to cross borders and achieve a better understanding of the total value impact. This was demonstrated by the evaluation of the societal value of an additional cyclist. The basic framework currently is limited to negative impacts. To include indicators that could measure positive impacts further research is needed. Acknowledgments. The authors gratefully acknowledge the contributions to the first draft of the value framework from Alan Crilly, Michelle Baker and Sarah McHale from Atkins, a member of the SNC Lavalin Group.

References Ackermann, F., Eden, C.: Strategic management of stakeholders: theory and practice. Long Range Plan. 44, 179–196 (2011) BSI. PAS55–1 Asset Management. Part 1: Specification of the optimal management of physical infrastructure assets. London (2004a)

Addressing Stakeholders Needs in Infrastructure

459

BSI. PAS 55–2 Asset management Part 2: Guidelines for the application of PAS 55–1 (2004b) DECISIO. Waarderingskengetallen MKBA Fiets: state-of-the-art. Commissioned by Ministerie van Infrastructuur en Waterstaat (2017) Elkington, J.: Triple bottom-line reporting: looking for balance. Australian CPA, 69 (1999) GOUDAPPEL COFENG. SAMENWERKEN AAN MEER FIETS, Eindevaluatie van 6 jaar fietsstimulering in Zuid-Limburg 2012–2017. Commissioned by Programmabureau Maastricht Bereikbaar (2018) Herder, P.M., Wijnia, Y.: A systems view on infrastructure asset management. In: Van Der Lei, T., Herder, P., Wijnia, Y. (eds.) Asset Management. Springer Netherlands (2012) IIRC. International Integrated Reporting Framework. International Integrated Reporting Council (IIRC) (2021). https://integratedreporting.org/wp-content/uploads/2021/01/InternationalInteg ratedReportingFramework.pdf. Accessed 18 Apr 2021 ISO. ISO 55000 Asset management-overview, principles and terminology. Geneva (2014a) ISO. ISO 55001 Asset management-management systems-requirements. Geneva (2014b) ISO. ISO 55002. Asset management-management systems-guidelines for the application of ISO 55001. Geneva (2018) Klinke, A., Renn, O.: A new approach to risk evaluation and management: risk based, precaution based and discourse based strategies. Risk Anal. 22, 1071–1094 (2002) Maj, J.: Diversity management’s stakeholders and stakeholders management. In: Proceedings of the 9th International Management Conference, vol. 9, pp. 780–793 (2015) Merkhofer, M.W.: Decision sciences and social risk management: a comparative approach of costbenefit analysis, decision analysis and other formal decision-aiding approaches, Dordrecht, D. Reidel Publishing company (1987) NEN. NTA 8120:2009 Assetmanagement - Eisen aan een veiligheids-, kwaliteits- en capaciteitsmanagementsysteem voor het elektriciteits- en gasnetbeheer (2009) Project management institute. a guide to the project management body of knowledge (PMBOK guide), Newton Square, PA, Project Management Institute (2017) Tengs, T.O., et al.: Five-hundred life-saving interventions and their cost-effectiveness. Risk Anal. 15, 369–390 (1995) Thackara, A.D.: Terotechnology - what it is all about. Chart Mech Eng. 22, 88–90 (1975) UKWIR. Future asset planning – scenarios, frameworks and measures: Final report, London, UKWIR (2022) United Nations. Resolution adopted by the general assembly on work of the statistical commission pertaining to the 2030 agenda for sustainable development (A/RES/71/313), Annex (2017). https://unstats.un.org/sdgs/indicators/indicators-list/. Accessed 20 May 2022 Wijnia, Y.: Pragmatic performance management. In: PINTO, J.O.P., Kimpara, M.L.M., Reis, R.R., Seecharan, T., Upadhyaya, B.R. Amadi-Echendu, J., eds. 15th WCEAM Proceedings, 163-172. Springer, Cham (2022).https://doi.org/10.1007/978-3-030-96794-9_15 Wijnia, Y.C.: Processing risk. In: Asset Management: Exploring The Boundaries Of Risk Based Optimization Under Uncertainty For An Energy Infrastructure Asset Manager. PhD, Delft University of Technology (2016)

Asset Condition, Risk, Resilience, and Vulnerability Assessments

Assessment and Prioritization of Critical Assets for Updating Maintenance Plans in a Biomass Power Plant Daniel Gaspar1(B) , Odete Lopes1 , João Costa1 , and Elson Grilo2 1 IPV, Viseu, Portugal {danigaspar,odetel}@estgv.ipv.pt 2 CTBF, Coimbra, Portugal [email protected]

Abstract. Several power plants have more than 20 years, and strict and planned management must be carried out to keep this equipment in good condition. It is necessary to know these assets well, their importance and criticality in the factory’s operation and to have up-to-date an effective preventive maintenance plans. This article describes the work to prioritise critical assets in a biomass power plant by studying the critical equipment and determining which active assets are most mandatory and prioritised to make updating maintenance plans. For this work it was necessary to carry out a reliability study. Firstly, a list of assets organized by sectors/areas was made, followed by an analysis using the methods of a semistructured interview and a critical assessment, adapting the risk matrix method. The results allowed the application of a new maintenance plan for critical equipment, which will be more effective and will enable avoiding unscheduled stops, greater reliability and consequently greater availability.

1 Introduction The industrial world is always in constant improvement; the case of physical asset management is no exception; modifications and new updates also have to occur, so the “fourth industrial revolution” will affect the way maintenance and asset management will be conducted in next years. The introduction of intelligent machines will provide intelligent maintenance, focusing on predictive maintenance, anticipating failures, avoiding or reducing downtime, optimizing the use of assets, and increasing productivity and competitiveness. With technological innovations, better value creation is expected from industrial assets (Panegossi and Silva 2021). The management of physical assets, nowadays, is essential for the excellent state of conservation and functioning of the equipment, to avoid accidents and to reduce costs in the maintenance of assets. Maintenance is an activity that takes up a lot of time for companies and that sometimes also has high costs for organizations. Hence, a good management with maintenance plans actualized of the assets is an important tool and way to reduce costs. According to the 55000:2014 standard, asset management allows an organization to perceive and produce value from assets in achieving the organization’s objectives. What © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 463–473, 2023. https://doi.org/10.1007/978-3-031-25448-2_44

464

D. Gaspar et al.

constitutes value will depend on the goals, nature and purpose of the organization and the needs and expectations of stakeholders. Asset management supports the perception and production of value by balancing the financial, environmental and social costs, risk, service quality and performance associated with assets (ISO 2014).

2 Criticality Analysis and Maintenance Crespo et al. (2018) made a criticality analysis methodology to prioritize the assets within an industrial/infrastructure context, with the purpose of adjusting assets maintenance strategies to dynamic business needs over time (Crespo et al. 2018) presented the method of establishing the criticality of assets, similar to a Plan Do Check Act (PDCA) cycle. The method classifies each equipment with multidisciplinary valences, regarding criteria of Quality, Availability, Safety and Environment, Costs and Technological Complexity (Santos et al. 2019). For some authors, one of the key phases of the Reliability Centered Maintenance (RCM) is the Identification of Maintenance Significant Items (MSI). Tang et al. (2017) presented framework for identification of the MSI through combination of quantitative analysis with qualitative analysis (Tang et al. 2017). The management of physical assets has a significant influence on organizations. Still, in recent years it has become more important because companies want greater profits and, for that, need to reduce equipment and personnel costs (Maletiˇc et al. 2020; Alsyouf et al. 2021). In the management of physical assets, maintenance is of great importance, so as the best applications and maintenance methodologies for the organization, to increase the competitiveness of companies and the technology which is also fundamental. To have good management of physical assets, it is necessary to apply several methods to determine the state of the equipment. The study propose an assessment based on risk analysis to evaluate the assets and define which maintenance plans and operations need to be improved or updated. The evaluation also makes future decisions based on the research carried out. Decision-making in this type of assessment is fundamental because decisions can include: equipment improve, replacement, refurbishment and finally, which path should be followed. Including these doubts in the assessment demonstrates that it is not at all simple to decide on the analysis, as there are always associated risks (ISO 2018). One of the methods used in the risk analysis and referred in the ISO 31010:2019 standard is the Consequence/Probability Matrix method. The consequence/probability matrix combines the qualitative or semi-quantitative classification of the consequence to define a risk level. The format of the matrix and the definitions that apply to it depend on the context in which it is used, and an appropriate design for the circumstances must be used. The matrices, being very variable, allow us to choose the one that is most favourable and the one that will enable to obtain the best results (IEC 2019) (Table 1).

Assessment and Prioritization of Critical Assets

465

Failure Probability

Table 1. Risk matrix, adapt from (IEC 2019) 5

5

10

15

20

25

4

4

8

12

16

20

3

3

6

9

12

15

1a7

Low

2

2

4

6

12

10

8 a 15

Medium

1

1

2

3

4

5

16 a 25

High

1

2

3

4

5

Risk Scale:

Severity/Consequence

The use of an adaptation risk matrix method considered in this study result in a matrix with a scale of one to twenty-five which is divided into three levels: low risk (1 to 7), medium risk (8 to 15) and finally, the risk high (16 to 25) (IEC 2019).

3 The Biomass Power Plant The case study was carried out at the Biomass Thermoelectric Power Station, inaugurated in 1998, being the first power station in Portugal to produce electricity from forest biomass. This group is a European reference in the production of eucalyptus pulp and sustainable forest management (Altri 2022). The biomass used at the plant is residual forest biomass. It is composed of leftovers or forest residues from forest clearings or tree falling. This raw material has several types of biomass, which can then influence electricity production, but mainly eucalyptus forest biomass is used. The industrial process can be seen in the Fig. 1, where it is possible to see the main components in terms of equipment and the representation of the cycle that the biomass goes through until the production of energy. This diagram represents a regular operation of a plant, from the discharge of biomass in the park to the injection of electrical energy into the grid. In Fig. 1, it is also possible to see the main equipment of a biomass thermoelectric plant. Some of these main equipment are: the biomass transport system, the boiler (together with super heaters, economizers, and evaporators); also shows the refrigeration cycle and finally, the turbo-alternator where the electrical energy is generated. It is a closed cycle, where water losses are negligible. The description of the industrial process begins with the reception of the raw material, storage and transport of the biomass to the furnace. In this stage, the raw material is transported to the plant and stored in the park. Finally, when the biomass is prepared, it is placed on the conveyor that takes it to the furnace. The second stage is the boiler, where the process of transforming chemical energy into thermal energy is triggered, as it is where the burning of biomass takes place, and thus the whole process begins. From the biomass combustion process, heat is produced, which is used to vaporize the water in the boiler circuits and produce superheated steam. The exhaust gases are cleaned

466

D. Gaspar et al.

Fig. 1. Representative diagram of a biomass thermoelectric plant (Altri 2022)

through an electrofilter – which has the function of removing particles. In the third stage, the turbogenerator and substation represent the steam network and the part of electric energy injection in the network. The superheated steam is injected into a condensing steam turbine coupled to the electric alternator. The movement of this turbine triggers the electric alternator to produce electricity and inject it into the electrical distribution network. After the work production stage, the steam enters the condenser to return to the liquid state – water-water heat exchanger – to resume the thermal cycle. This (demineralized) water is reintroduced into the feed water tank and injected back into the boiler to vaporize.

4 Assessment of Critical Assets In the case of the study in question, the first step that was taken was the registration of critical assets with the help of maintenance experts, as they are the ones who deal with the assets every day and have a more profound knowledge about them. The critical assets were defined after a detailed analysis of each equipment and according to the criteria previously established, where the list of critical equipment of the plant was obtained. The list of critical assets determined was as follows (Table 2):

Assessment and Prioritization of Critical Assets

467

Table 2. List of critical assets Group

Critical asset

Electrical auxiliaries

Power transformer

Thermal cycle

Feedwater pumps

Alternator MV circuit breaker Condensate extraction pumps Recirculation pumps to the cooling tower Water collection pumps Air – smokes

Primary air fan Secondary air fan Induced draft fan

Steam generator

Grade vibration system

Ash and slag circuit

Redler 1 and Redler 2

The designation that the company uses for critical equipment is the KKS, which is a system commonly used to identify all the components of a power plant. This system is based on an international document from Siemens called “KKS - KraftwerkKennzeichen-System (Siemens 2010). The equipment has a code that allows all people working in and for the service external provider to identify the equipment (Table 3). Table 3. Identification of the KKS of critical assets Equipament name

KKS

Power transformer

BAT01

MT alternator circuit breaker

BBB02GS001

feed water pumps

LAC01AP001/LAC02AP001

Condensate extraction pumps

LCB01AP001/LCB02AP001

Recirculation pumps to the cooling tower

PAC01AP001/PAC02AP001

Water collection pumps

GAF01AP001/GAF02AP001

Primary air fan

HLB10AN001

Secondary air fan

HLB20AN001

Induced draft fan

HNB60AN001

Grade vibration system

HHC10AF001/HHC10AF002

Redler 1

ETG10AF001

Redler 2

ETG10AF002

468

D. Gaspar et al.

The second step in this case study was to analyze the list of critical assets. To obtain a more extensive set of equipment information, conducting a survey with maintenance experts was necessary. The applied tables calculate failure severity in terms of operation and safety, the downtime of failure in two dimensions: the maintenance time and the logistic time, and the probability of occurrence of failure. A scale from 1 to 5 with different attributions was used in the tables to quantify the different types of risk. The tables presented represent an average of the experts’ answers, and it is through them that it is possible to calculate the criticality level of each equipment (Tables 4 and 5). Table 4. Severity in the operation Critical asset

Failure severity In terms of plant operation (1–5)

Power transformer

5

MT alternator circuit breaker

5

Feed water pumps

3

Condensate extraction pumps

1

Recirculation pumps to the cooling 1 tower Water collection pumps

1

Primary air fan

5

Secondary air fan

5

Induced draft fan

5

Grade vibration system

3

Redler 1

5

Redler 2

5

Table 5. Severity in terms of safety Critical asset

Failure severity In terms of safety (1 – 5)

Power transformer

3

MT alternator circuit breaker

3

Feed water pumps

3

Condensate extraction pumps

3

Recirculation pumps to the cooling tower

2

Water collection pumps

2 (continued)

Assessment and Prioritization of Critical Assets

469

Table 5. (continued) Critical asset

Failure severity In terms of safety (1 – 5)

Primary air fan

2

Secondary air fan

2

Induced draft fan

3

Grade vibration system

3

Redler 1

2

Redler 2

3

In Table 6 presents the probability of downtime of a failure, which consists of the two variables: the average maintenance time of the equipment and the logistical time of waiting for the parts to the assets. Table 6. Probability of downtime of a failure (maintenance and logistics) Critical asset

Failure downtime (logistics spare parts)

Failure downtime (maintenance tasks)

Power transformer

5

4

MT alternator circuit breaker

4

3

Feed water pumps

4

4

Condensate extraction pumps

4

4

Recirculation pumps to the cooling tower

4

4

Water collection pumps

4

4

Primary air fan

4

3

Secondary air fan

3

3

Induced draft fan

3

3

Grade vibration system

3

3

Redler 1

3

3

Redler 2

3

3

Finally, the last parameter and the most important of all, since it calculates the probability of occurrence, in which we want to know the probability of a failure in the equipment. In Table 7 show the description of the scale for the probability of occurrence of failure (Table 8).

470

D. Gaspar et al. Table 7. Scale for the probability of occurrence of failure

5

Highly probable

Occurs many times throughout the year

4

Very likely

Occurs regularly throughout the year

3

Likely

Occurs a few times throughout the year

2

Less likely

Occurs in occasional situations

1

Unlikely

Never occurs

Table 8. Probability of occurrence of failure for critical asset Critical asset

Probability of occurrence of failure 1–5

Power transformer

2

MT alternator circuit breaker

2

Feed water pumps

2

Condensate extraction pumps

2

Recirculation pumps to the cooling tower

2

Water collection pumps

2

Primary air fan

3

Secondary air fan

3

Induced draft fan

2

Grade vibration system

3

Redler 1

3

Redler 2

3

The calculation of the criticality level according to the risk matrix is done by multiplying the failure probability (p) by failure severities (a, b), the probability of downtime of failure (c, d), where after we divide by four as it is an average. The respective calculation formula is as follows: Criticality level =

(a ∗ p) + (b ∗ p) + (c ∗ p) + (d ∗ p) 4

Variables: (a) – Failure severity in terms of plant operation. (b) – Severity of failure in terms of security. (c) – Failure downtime in terms of maintenance time. (d) – Failure downtime in terms of logistical time.

Assessment and Prioritization of Critical Assets

471

(p) – Probability of occurrence of failure. After making the calculations, Table 9 is obtained, which presents the criticality level of each equipment. Table 9. Results of calculating the risk level of each critical asset

Critical asset Power transformer MT alternator circuit breaker feed water pumps Condensate extraction pumps Recirculation pumps to the cooling tower Water collection pumps primary air fan Secondary air fan Induced draft fan Grade vibration system redler 1 redler 2

Criticality level score 1 - 25

8 7 7 6 5 6 9 9 7 9 9 9

Criticality matrix level

Low Low Low Low Low Low Medium Medium Low Medium Medium Medium

The Table 9 is the result of the calculation for critical equipment, and from its analysis, it is concluded that no equipment is highlighted that is at high level of criticality. There are five equipment at medium level, which makes this task more complex and challenging to choose to carry out the next step. But according to the calculations and the results obtained, the primary and secondary fans are at a medium level; however, there is a project and plan to replacement the fans in the next great intervention of the plant, for this reason is not profitable and productive to make a great effort to improve the maintenance plan focused on reliability for an asset that can be replaced by a new one. Redlers present a result that characterizes them with a medium risk level, but as they are very complex assets in terms of study and even mechanically, therefore, the application of the improve maintenance plan and the RCM process is also very difficult. Its number of failures compared to the grade vibration system is lower. The asset that has the highest number of failures in 2020 is the grade vibration system, which puts it in an excellent position to be first to be studied. Another relevant factor is that the grade vibration system is the asset with the highest number of failures from 2001 to 2020, which means that it is equipment that causes great concern to the plant and causes great inconvenience to the production of energy. In short, for the application of the RCM process, the choice, together with all the people involved in this project, was that the best would be to study and redo

472

D. Gaspar et al.

the maintenance plan for the grade vibration system since its elaboration presents more significant advantages, namely to avoid or reduce downtime and improve repair times.

5 Conclusions The management of physical assets nowadays is fundamental for the excellent state of conservation and functioning of the equipment, avoiding accidents, and reducing costs in the maintenance of assets. With this study, the company gained more knowledge about its equipment and a method to identify the critical equipment and the selection of the type of maintenance to avoid them. The initial work of surveying the equipment made it possible to know which are the most critical assets of the biomass plant, if they have a failure, could cause significant problems to the plant, such as prolonged stops or total loss of activity. After elaborating the list of critical equipment, it was possible to have a deeper knowledge of the equipment, such as its function in the energy production cycle, the main failures, the current state of the equipment, and the maintenance plans. For the analysis, tables were prepared that allowed the analysis of the severity at an operational and safety level, the failure downtime (in which the logistical time of parts delivery and the maintenance intervention time are counted) and, finally, the estimation of the probability of failure occurrence. The calculation of the level for critical equipment was done through the values obtained in the tables developed for the survey interview method. The criticality level rating was obtained using the risk matrix adapted to the specific situation of the plant. In short, with the survey to the expert and analysis and filling of tables, it was possible to collect a lot of information about the plant’s assets. With all the data collected, the level of criticality of each equipment was calculated using the risk matrix, which also gives valuable guidance and the sequence and priority to apply the RCM maintenance method to critical equipment of the plant. The limitations of this type of investigation are the time needed to collect the data, and the great dependence on experts. For future work, it is intended to apply the RCM to all selected equipment and also to develop the evaluation and prioritization methodology, to apply in other factories of the company.

References Alsyouf, I., Alsuwaidi, M., Hamdan, S., Shamsuzzaman, M.: Impact of ISO 55000 on organisational performance: evidence from certified UAE firms. Total Qual. Manag. Bus. Excell. 32(1–2), 134–152 (2021). https://doi.org/10.1080/14783363.2018.1537750 Altri: O Nosso Mundo—Altri (2022). https://altri.pt/pt/sobre-a-altri/o-nosso-mundo Crespo, A., et al.: Criticality Analysis for improving maintenance, felling and pruning cycles in power lines. IFAC-PapersOnLine 51(11), 211–216 (2018). https://doi.org/10.1016/j.ifacol. 2018.08.262 IEC: IEC/ISO 31010: 2019: Risk management—risk assessment techniques. IEC, Geneva, Switzerland (2019) ISO: ISO 55000: 2014: Asset management–overview, principles and terminology. ISO, Geneva, Switzerland (2014)

Assessment and Prioritization of Critical Assets

473

ISO: ISO 31000: 2018: Risk management—guidelines. ISO, Geneva, Switzerland (2018) Maletiˇc, D., Maletiˇc, M., Al-Najjar, B., Gomišˇcek, B.: An analysis of physical asset management core practices and their influence on operational performance. Sustainability 12(21), 9097 (2020). https://doi.org/10.3390/su12219097 Panegossi, A., Silva, E.: Pesquisas avançadas em engenharia de produção 02. A evolução da gestão de ativos Asset management evolution. Março 2021, 02, 30 (2021) Santos, T., Silva, F.J.G., Ramos, S.F., Campilho, R.D.S.G., Ferreira, L.P.: Asset priority setting for maintenance management in the food industry. Procedia Manuf. 38, 1623–1633 (2019). https:// doi.org/10.1016/j.promfg.2020.01.122 Siemens: KKS - Kraftwerk-Kennzeichen-System Identification System for Power Plants (2010) Tang, Y., Liu, Q., Jing, J., Yang, Y., Zou, Z.: A framework for identification of maintenance significant items in reliability centered maintenance. Energy 118, 1295–1303 (2017). https:// doi.org/10.1016/j.energy.2016.11.011

Application of Risk Management System for Intangible Assets in a Steel Company Manuel González(B) SMS Group GmbH, Mönchengladbach, Germany [email protected]

Abstract. Intangible assets (contracts, intellectual property, brand, data, corporate relationships, processes, etc.) are assets that cannot be seen or touched, but are capable of generating enormous value for companies. Investing in intangible assets leads to spending with a future return, and these assets are an essential element for organizations, especially in the face of a complex global horizon. It is essential to manage the risk of these assets in a structured and agreed manner. These risks cover vital aspects such as security-environmental, economic, financial, cultural, political, technical and market aspects. This case study presents a real annual risk assessment in a Steel Company following the basic steps of the ISO 31000 standard and supported by ISO 31010: risk classification, definition of the criticality matrix and criteria, risk identificationanalysis-assessment using the RAMP tool (Risk Analysis & Mitigation Plan) and its treatment. Keywords: Risk Management · Asset Management · ISO 31000 · ISO 55001

1 Introduction International standards such as the ISO 55000 series urge utilities to have a structured approach to AM (Asset Management) related processes and risk-based decision-making. Such a framework provides the platform for the AM system policy and methodologies. On the other hand, utilities are required to empower this platform with technical expertise and experience. The asset replacement policy and guidelines, therefore, should describe how to translate the corporate strategy into an asset replacement plan. ISO 55001 assumes that the organisation will be using the ISO 31000 risk management framework (or equivalent) which specifies all the steps in the risk management process. It therefore does not spell these out in the 55001 text. Risk is the potential that a chosen action or activity will lead to a loss, an undesirable event or outcome. Companies take risks in their everyday life. When we do any work or activity at work, home or in our personal life, such as driving to work, repairing a machine, engaging in a new venture or assignment or project, we accept a certain level of risk. Risk = CxL where: C is Consequence of Risk event © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 474–486, 2023. https://doi.org/10.1007/978-3-031-25448-2_45

(1)

Application of Risk Management System for Intangible Assets

475

L is Likelihood of Occurrence Risk management is the process of identifying, assessing and controlling threats to an organization’s capital and earnings. These threats, or risks, could stem from a wide variety of sources, including financial uncertainty, legal liabilities, strategic management errors, accidents and natural disasters. The purpose of risk management is to prevent, reduce, or control future impacts of unfavourable events as opposed to reacting to unwanted events after they have already occurred. The mitigation of every plausible risk may not be possible and is rather impractical due to resource limitations. Hence, effective risk management requires a process to determine which risks are actionable and can be mitigated, and which risks are nonactionable or residual and cannot be mitigated. These risks must be controlled instead (if identified early enough), watched, or transferred while being accepted by the appropriate authority. Risk management is (should be) a central part of any organization’s strategic management. It is the process whereby organizations methodically address the risks associated with their activities to enable the goal of achieving sustained benefit within each activity and across all activities of the organization. This System allows organizations and teams to increase the predictability of outcomes, both qualitatively and quantitatively. This principle is about reaching the appropriate level of organizational process maturity (the ability of an organization to apply a certain set of processes in a consistent manner) and the optimal level of performance. Excellence in risk management is not achieved by the strict and exhaustive application of related processes. Rather, excellence can be achieved by (a) balancing the benefits to be obtained with the associated cost and (b) tailoring the risk management processes to the characteristics of the organization and its portfolios, programs, and projects. Risk management is a critical part of Operational Excellence. Industrial organizations tend to be very conservative with respect to process and individual safety and minimize risks as much as possible. Safety risk tends to vary with the operating process, local environment, intensity of operation, equipment maintenance, utilization requirements and products. Thus, safety and risks associated with safety excellence often add an additional constraint on profitability. Developing a safety performance excellence strategy requires that the current safety risk within each section of the operating enterprise is effectively measured on an ongoing basis. Understanding the actual current safety risk might enable the plant operations personnel to push throughput higher during lower risk periods, which could result in significant improvements to profitability. This process is based in standard ISO 31000 Risk management guidelines and ISO 31010 Risk assessment techniques.

2 Risk Management System A risk analysis is a formal, systemized process to anticipate the future, including the means and likelihood of detecting incipient symptoms in time to prevent or minimize the occurrence and/or consequences of an actual event.

476

M. González

There are numerous specifications for risk analysis, management, and control published by ISO (ISO 31000 & 31010), American Petroleum Institute (API), American Nuclear Society, and others. The process is amplified by detailed instructive material. Every company must have a risk identification, analysis, and management/control procedure in place that is in full compliance with one or more internationally recognized standards for the specific industry. 2.1 Principles of Risk Management (According to ISO 31000) • Integrated. Risk management is an integral part of all organizational activities. • Structured and comprehensive. A structured and comprehensive approach to risk management contributes to consistent and comparable results. • Customized. The risk management framework and process are customized and proportionate to the organization’s external and internal context related to its objectives.

Fig. 1. Principles of Risk Management (according to ISO 31000)

Application of Risk Management System for Intangible Assets

477

• Inclusive. Appropriate and timely involvement of stakeholders enables their knowledge, views and perceptions to be considered. This results in improved awareness and informed risk management. • Dynamic. Risks can emerge, change or disappear as an organization’s external and internal context changes. Risk management anticipates, detects, acknowledges and responds to those changes and events in an appropriate and timely manner. • Best available information. The inputs to risk management are based on historical and current information, as well as on future expectations. Risk management explicitly takes into account any limitations and uncertainties associated with such information and expectations. Information should be timely, clear and available to relevant stakeholders. • Human and cultural factors. Human behavior and culture significantly influence all aspects of risk management at each level and stage. • Continual improvement. Risk management is continually improved through learning and experience.

2.2 Risk Management Steps • Scope, context and criteria. Defining the scope of the process, and understanding the external and internal context. • Risk assessment. Overall process of risk identification, risk analysis and risk evaluation. It should be conducted systematically, iteratively and collaboratively, drawing on the knowledge and views of stakeholders. – Risk identification. Find, recognize and describe risks that might help or prevent an organization achieving its objectives. – Risk analysis. Comprehend the nature of risk and its characteristics including, where appropriate, the level of risk. – Risk evaluation. Support decisions (do nothing further, consider risk treatment options, and undertake further analysis to better understand the risk, maintain existing controls, and reconsider objectives). • Risk treatment. Select and implement options for addressing risk. • Monitoring & review. Planning, gathering and analysing information, recording results and providing feedback. • Recording and reporting. Risk management process and its outcomes should be documented and reported.

478

M. González

Fig. 2. Risk Management Steps (according to ISO 31000)

2.3 Maturity Model for the Risk Management Process in Standard Companies Usually, the myths surrounding Risk Management are the following: • • • • • • • • •

Risks are up to the Risk Manager Any risk is bad Our project is perfect… it has no risks Proactive risk management brings no benefit The hope that things will not happen is not a strategy To analyse risks, it is necessary to use specific software It is not necessary to foresee contingencies, it must be managed properly If the procedures are followed, we eliminate the risks Forget the cost of mitigating measures

Application of Risk Management System for Intangible Assets

479

How standard companies really manage Risks? • Absence of corporate risk criteria and classification • Lack of integration of risk management with strategy planning and performance management • Carry out risk management exclusively when each corporate strategy is carried out, that is, do not analyse risks periodically • Not documenting incidents or documenting them erroneously • Avoid risks instead of managing them • Do not group activities that have the same risk • Not periodically reassessing risks • All departments claim to carry out risk management, but each one does it in a different way and does not inform the rest of the departments • Do not include human behaviors. Culture eats strategy for breakfast (Peter Drucker)

AS-IS Companies vs TO-BE AS-IS

TO-BE

1. Scope, context and criteria 6 5 7. Recording and reporng

4

2. Risk idenficaon

3 2 1 0 6. Monitoring & review

5. Risk treatment

3. Risk analysis

4. Risk evaluaon

Fig. 3. Maturity model for the Risk Management Process in standard companies

2.4 Risk Treatment Options Once risks have been identified and assessed, all techniques to manage the risk fall into one or more of these categories: • Avoid the risk by deciding not to start or continue with the activity that gives rise to the risk.

480

• • • • • •

M. González

Take or increase the risk in order to pursue an opportunity. Remove the risk source. Change the likelihood. Change the consequences. Share the risk. Retain the risk by informed decision.

3 Case Study The case study is done in a Steel Company following the basic steps of the ISO 31000 mentioned before and using the RAMP tool (Risk Analysis & Mitigation Plan) for intangible assets. This tool could be interesting to be done annually reviewed, approved and committed in any company. Is important to do it before ending each year because this will imply new investments to be included in next year strategic plan of the Company. 3.1 Risk Management for Intangible Assets For a better comprehension of this document, it is important to have a clear understanding about what is the difference of a tangible and intangible asset. • Asset. Item that has potential or actual value to an organization. • Tangible asset. Assets that have a physical form. They can be seen and touched, e.g. equipment, workshops, etc. • Intangible asset. Assets that have no physical form, e.g. business, financial, human capital, intellectual property, brands, knowledge, etc. As mentioned before, this case study is focused only on intangibles assets and their Risk Management/Assessment. 3.2 Risk Classification for Intangible Assets Risks for intangible assets can be classified in different ways: internal/external, performance/economics, etc. but this case study will use the following Risk Classification. 3.3 Criticality Criteria Each Steel Plant must have an agreed Global Criteria on Likelihood and Consequence. Even within the same company, the different production plants located in different countries may have different criteria due to the environment, political situation, operational strategy, etc. These criteria may be changed over the years due to new safety or environmental requirements, changes in the social and economic environment, etc.

Application of Risk Management System for Intangible Assets

481

Table 1. Company Risk Classification Risk Commercial/Finance

DefiniƟon Risks associated to commercial, financial, legal and contract related issues including financial transac ons that include company loans in risk of default Culture Risks associated to misalignment between an organiza on's values and leader ac ons, employee behaviors, or organiza onal systems Data Availability & Quality Risks associated to data gathering because of its availability or quality of exis ng data Economic/Market Risks associated to lose money because of the market- or economy-related factors, like poli cal uncertainty, the global slowdown, interest rate change HSE/Ecologic Risks associated to Health, Safety, Environment (physical and biological components) and Ecologic (living organisms) Organiza on/Stakeholder Risks associated to people who are affected by a organiza on, decision, treatment, strategy or process Others Risks associated to Others Poli cal Risks associated to business interests resul ng from poli cal instability or poli cal change as well as cultural issues Skillness & knowledge Risks associated to skillness or knowledge of people around a business Technical Risks associated to the evolu on of the design and the produc on of the system of interest affec ng the level of performance necessary to meet the stakeholder expecta ons and technical requirements

Likelihood. For the decision making of the level of likelihood, this analysis can find quantitative or qualitative risks. Qualitative risks need to be evaluated by personnel with experience in the business since it does not have tailor-made metrics. Consequence. In the case of consequence calculation, various criteria based on one or several factors can be used. In this case study, factors S (Health & Safety), E (Environment), Q (Quality), C (Cost) & P (Production) are used. Each of them has a weight factor (αS , αE , αQ , etc.) defined globally by the Steel Plant. The global Consequence will be calculated using the following expression: 2  Consequence = αS S + αE E + αQ Q + αC C + αP P

(2)

The Consequence level will be truncated to a maximum of 5 if it is higher than this value. This calculation system is based on the UK Def-Stan 0045 Reliability and maintainability data collection and classification standard from the UK Ministry of Defence standardization system (Tables 1, 2 and 3).

482

M. González Table 2. Steel Plant Risk Criteria LIKELIHOOD CRITERIA FOR QUANTITATIVE RISKS Level

1 Rare

2 Unlikely

3 Possible

4 Likely

5 Almost certain

Fa i l ures /yea r ≤ 0,1

Fa i l ures /yea r > 0,1

Fa i l ures /yea r > 0,3

Fa i l ures /yea r > 1,0

Fa i l ures /yea r > 10

1 Rare

2 Unlikely

LIKELIHOOD CRITERIA FOR QUALITATIVE RISKS Level

- Ima gi na bl e, but occurrence woul d - Ac vi es a re not very compl ex. requi re s evera l fa i l ures or a - Pers onnel i s experi enced wi th combi na on of s ome fa ctors . thi s ki nd of ma er: - The ri s k ha s occurred s ome mes Li kel i hood: [0%-20%] i n s i mi l a r projects or s i mi l a r compa ny. Li kel i hood: [20%-40%]

3 Possible

4 Likely

5 Almost certain

- Ac vi es a re of medi um or a l i l e hi gher compl exi ty. - Pers onnel i s not ful l y experi enced wi th thi s ki nd of ma er. - The ri s k ha s a nd ci rcums ta nces ha ve been known i n our compa ny.

- The a c vi es a re very compl ex. - Pers onnel i s qui te i nexperi enced wi th thi s ma er or newl y recrui ted. - The ri s k i s a l mos t i nevi ta bl e a nd regul a rl y ha ppens wi thi n our compa ny.

Li kel i hood: [40%-60%]

Li kel i hood: [60%-80%]

- The a c vi es a re extremel y compl ex. - Pers onnel i s not onl y qui te i nexperi enced wi th thi s ma er or newl y recrui ted but i t i s a l s o not qui te cl ea r who a nd i n whi ch wa y wi l l tra i n i t. - The ri s k i s a l mos t i nevi ta bl e a nd regul a rl y ha ppens wi thi n one bra nch. Li kel i hood: [80%-100%]

Ba s ed on the UK Def-Stan 0044 Reliability and maintainability data collecƟon and classificaƟon s ta ndard from the UK Mi nistry of Defence standardiza on

CONSEQUENCE CRITERIA Level

S (Health & Safety)

E (Environment)

Q (Quality)

C (Cost)

P (ProducƟon)

≥ 5 mi l l i on EUR once or ≥ 1 mi l l i on EUR/yea r

Los s Produc on > 4.000 t/fa i l ure

5 Very High

Mul pl e fa ta l i es , or chemi ca l rel ea s e, or cl us ters of ca ncer or termi na l i l l nes s , or i nvol ves s ta tutory i ns pec on or ma i ntena nce requi rements

4 Major

Si gni fica nt degra da on, or on-s i te One or two fa ta l i es , perma nent i mpa ct, or l oca l off-s i te i mpa cts , or Si gni fica nt a c vi es rega rdi ng redi s a bi l i es , or i s ol a ted ca ncers , or revers i bl e i mpa cts , or s peci fic a nd engi neeri ng, repl a cements etc. termi na l / di s a bl i ng i l l nes s tempora ry l ega l brea ches

≥ 1 mi l l i on EUR once or ≥ 200 thous a nds EUR/yea r

Los s Produc on > 2.000 t/fa i l ure

3 Moderate

Medi ca l trea tment or i njuri es wi th s ervi ce res tri c ons or l os s of me, or revers i bl e hea l th effects , or hea ri ng l os s

Short-term s i te i mpa ct, but correcta bl e or repa i ra bl e

Severe a djus tments , reengi neeri ng or repl a cements a re requi red

≥ 100 thous a nds EUR once or ≥ 100 thous a nds EUR/yea r

Los s Produc on > 1.000 t/fa i l ure

2 Low, less

Mi ni ma l i njuri es , or firs t a i d

Mi ni ma l mea s ura bl e tempora ry i mpa ct on the s i te

Few a djus tments need to be ma de to keep up func ona l i ty

< 100 thous a nds EUR once or < 20 thous a nds EUR/yea r

Los s Produc on > 500 t/fa i l ure

No envi ronmenta l i mpa ct

No qua l i ty i mpa ct

< 50 thous a nds EUR once or < 10 thous a nds EUR/yea r

Los s Produc on > 0 t/fa i l ure

20%

25%

1 Insignificant No i njury, no firs t a i d

Factor α :

Si gni fica nt degra da on i n s i tu, or i rrevers i bl e, i rrevers i bl e envi ronmenta l da ma ge or s eri ous l ega l brea ches

20%

Ma jor repl a cements , reva mps to ful fil contra ct requi rements

20%

15%

3.4 Criticality Matrix Criticality matrix is 5 x 5 using linear criteria (1, 2, 3, 4 & 5) for Likelihood value and non-linear criteria (1, 2, 4, 8 & 16) for Consequence value. Table 3. Relationship between Level & Value for Risk Matrix Likelihood Level according Value for Risk Criteria Matrix 1 1 2 2 3 3 4 4 5 5

Consequence Level according Value for Risk Criteria Matrix 1 1 2 2 3 4 4 8 5 16

Application of Risk Management System for Intangible Assets

483

Likelihood (probability of occurrence)

Elaboration of the final risk matrix would be as follows.

Value

5

1

2

4

8

16

Insignificant

Low, less

Moderate

Major

Very High

5

10

20

40

80

Almost certain

4

4

8

16

32

64

3

6

12

24

48

2

4

8

16

32

1

2

4

8

16

Likely

3 Possible

2 Unlikely

1 Rare

Risk values: - Low: - Medium: - High:

1-4 5-16 20-80

Comments: - Consequence:Non-linear 1) - Likelihood: Linear 1) UK Defense Standard. Defense Standard 02-45 (NES 45)-2000

Consequence (potenal impact)

Fig. 4. Company Risk Matrix

3.5 Results After multiple work sessions with managers from different departments, the RAMP (Risk Analysis and Mitigation Plan) worksheet was finalized, including all short, medium, and long-term risks. The evaluation of those risks was carried out following the methodology described above, although most of the values selected for the different risks were of a qualitative nature. That is the reason why the working group needs to incorporate personnel with high experience and knowledge of the business, as well as its context. For all the risks, one or several Mitigation Action Items were agreed with department’s managers nominating responsible person and completion date, as well as recalculating Expected Risk once recommended actions will be finished. The risks that involve a significant investment will be analysed in different sessions with Top Management to assess whether it is feasible or, on the contrary, the risks are assumed corporately (Figs. 1, 2, 3, 4, 5, 6, 7 and 8).

484

M. González

Fig. 5. RAMP Worksheet sample Category

Polical Commercial/Finance HSE/Ecologic Others Organizaon/Stakeholder Grand Total

Count of Risk Descripon 19 15 15 3 3 55

No of Migaon Acon Items 20

19

18 15

16

15

14 12 10 8 3

3

Organizaon/Stakeholder

4

Others

6 2

HSE/Ecologic

Commercial/Finance

Polical

0

Fig. 6. No of agreed Action Items in RAMP

Main target is to “move” risks with a high value in an area whose risk value is medium or low. The analysis of the considerable increase in inflation and energy cost is of special interest, since it represents practically 30% of the operating costs of the Steel Plant. 3.6 Priorization Criteria Once the Identification, Analysis and Evaluation phases have been completed, decisionmaking for actions resulting from the Risk Assessment must be done according to a prioritization criterion. It is important, not only to take into account the original and to final risk value, but also to the cost and implementation time of each action, as well as its specific importance because some of these decisions will have impact in investment strategy and production plan.

Application of Risk Management System for Intangible Assets

Risk value original

Expected Risk value aer migaon Acon Items

12 10

485

10

16

10

15 14

8

12

7 6

6

6 4 4

3 2

2

3

Number of Risks

Number of Risks

14

10

9

8

7

6

5

4

2

1

1

3

2

1

1

0

0 1

2

4

5

6

8

10

12

16

20

40

1

32

2

4

Ri s k Level

5

6

10

16

8

Ri s k Level

Fig. 7. Expected Risks values after implementation of agreed Action Items

Value

5 Almost certain

4 Likely

3 Possible

2 Unlikely

1 Rare

1

2

4

Expected Aim Risks 8

16

Insignificant

Low, less

Moderate

Major

Very High

6

10

4

1

0

0

3

0

3

0

0

2

6

0

0

0

2

0

10

0

1

7

0

0

0

Likelihood (probability of occurrence)

Likelihood (probability of occurrence)

Present Risks Value

5 Almost certain

4 Likely

3 Possible

2 Unlikely

1 Rare

1

2

4

8

16

Insignificant

Low, less

Moderate

Major

Very High

7

1

0

0

0

0

0

2

0

0

0

5

0

0

0

4

4

1

1

0

15

10

5

0

0

Consequence (potenal impact)

Consequence (potenal impact)

Fig. 8. Distribution of Risks in Risk Matrix after implementation of agreed Action Items

To do this, you can select between four criteria to prioritize actions: • Criterion #1. Criterion for initial risk value. • Criterion #2. Criterion for increasing the value of the risk (Initial-Final) Risk. • Criterion #3. Criterion for increasing the value of the risk (Initial-Final) and taking into account the cost and implementation time of each action. Risk Cost x Time

(3)

• Criterion #4. Criterion for increasing the value of the risk (Initial-Final) and taking into account the cost and implementation time of each action, as well as its specific importance.

Risk αC x Costx αT xTime In this case study, the most complete has been selected, that is, criterion #4.

(4)

486

M. González

4 Discussion and Conclusions Carrying out this analysis for intangible assets is absolutely useful for the Company for the following reasons: • The operational continuity and survival of the company may depend more on the management of these risks than on those related to failures in equipment and machinery. Especially in such a changing environment as the current one, risks such as inflation, energy costs and political conflicts can definitively stop production in the very short term. • Reputation and image of the company will be enhanced in the eyes of its customers and suppliers, since it will demonstrate that it has control over the most critical risks and will avoid unforeseen surprises in the short, medium and long term. • A substantial improvement in efficiency, performance and loss reduction is guaranteed. • Sustainability of the business will "earn points" for achieving a lower impact on issues related to Safety, Health and the Environment. • Company investment plan will be flattened since strategic investments (especially environmental ones) will be included in the budget of the Company well in advance, thus allowing the most relevant investments to be planned. In other words, the surprises of sudden investments for shareholders will be minimized. This will give signals also to regulators and stakeholders that intangible/tangible assets are in “good hands”. • The decision-making process is aligned with modern asset management standards, such as the ISO 55000 series, embracing and adopting risk-based decision-making. • Contingency budgets can be more accurately estimated and rely less on the professional guesstimates of the project team. • Increases the stability of business operations while also decreasing legal liability. • Helps establish the Company´s insurance needs in order to save on unnecessary premiums a lower case letter.

References 1. Crespo A.: The maintenance management framework. Models and methods for complex systems maintenance. Springer, London (2006). https://doi.org/10.1007/978-1-84628-821-0 2. Gulati, R.: Maintenance & Reliability Best Practices. Industrial Press Inc., USA (2013) 3. ISO 31000:2018. Risk management – Guidelines 4. ISO 31010:2019. Risk assessment techniques 5. Mitchell, J.S.: Operational Excellence Journey to Creating Sustainable Value. John Wiley & Sons Inc, USA (2015)

Case Studies on Condition Assessments of Infrastructure Assets Joe E. Amadi-Echendu(B) , Jedial O. Mvele, and Refiloe R. Lapshe Deparment of Engineering and Technology Management, University of Pretoria, Pretoria 0028, South Africa [email protected]

Abstract. A key challenge in managing discrete assets, asset systems, and systems of assets is to conduct holistic assessments that facilitate informed decisionmaking under volatile, uncertain, complex and ambiguous conditions. Typical infrastructure assets such as modern facilities in a primary school, an amusement park, or a water and sewerage treatment plant comprise interconnected components, equipment, machinery and cyber physical elements. As a follow up to a previous discourse, this paper discusses three additional case studies and the findings indicate that conventional approaches to assessing the condition of infrastructure assets are usually not comprehensive. The assessments tend to be discipline-based, narrow and biased, thus provoking questions as to how far-reaching decisions are made with regard to infrastructure assets. This paper reiterates the argument that multidisciplinary and multidimensional approaches are essential for conducting holistic condition assessments towards effective management of infrastructure assets given the capabilities proffered by 4IR technologies and digital age platforms in the Society 5.0 era.

1 Introduction A dictionary defines infrastructure as “the basic physical and organizational structures and facilities [and systems] needed for the operation of a society or enterprise.” The significance of infrastructure to humanity is not a matter for debate, however, reference [1] provides a useful discourse on the positive and negative effects of infrastructure and the interdependencies between infrastructure sectors, as well as a mapping of the influence of infrastructure on sustainable development (see Fig. 1). Modern infrastructure comprises man-made artifacts or engineered discrete assets, asset systems, and systems of assets. Infrastructure assets such as buildings, man-made or engineered components, equipment, machinery, physical gadgets and cyber physical elements are interdependent because of high levels of interconnectedness and intercommunicability made possible via 4IR technologies (especially Internet-of-Things (IoT)) and the digital age platforms. A key challenge in managing infrastructure assets is informed decision-making under volatile, uncertain, complex and ambiguous (VUCA) conditions that pervade the era of Society 5.0. The challenge is often complicated and exacerbated by the fact that both the condition and performance of infrastructure assets are affected and influenced by human © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 487–498, 2023. https://doi.org/10.1007/978-3-031-25448-2_46

488

J. E. Amadi-Echendu et al.

Fig. 1. Influence of infrastructure on sustainable development [Source: 1]

behavior and social actions, as well as the effects of stressors inherent in the natural environment and ecology [2]. In fact, the condition and performance of infrastructure assets are inextricably interwoven with human activities, societal trends and changes imposed by the natural environment and ecology. This paper is structured as follows. Some commentary on the impetus and techniques for condition assessment of infrastructure assets is provided in Sect. 2. Section 3 includes brief narratives of three additional case studies that augment two case studies previously discussed in [2]. Drawing from the case studies, Sect. 4 of the paper reiterates arguments for taking advantage of 4IR technologies and digital age platforms as well as multidisciplinary and multidimensional approaches to conduct holistic assessments of the condition of infrastructure assets to inform robust decision-making under the VUCA pervasion of Society 5.0 era.

2 Condition Assessment: Conventional Metrics Infrastructure facilitates every aspect of human endeavour, livelihood and way of life; infrastructure assets enable all activities performed in governance, commercial, industrial, and social enterprise. For instance, a typical road transportation infrastructure encompasses discrete assets (e.g., motorised vehicles), asset systems (movable vehicles combined with immovable assets like bridges, pavements, signages, tunnels, etc.), and systems of road assets (various classifications of road networks). For established infrastructure, changes in service delivery requirements, maintenance, refurbishment,

Case Studies on Condition Assessments of Infrastructure Assets

489

rehabilitation and upgrading tend to be the primary reasons for conducting condition assessments of the composite discrete assets, asset systems, and systems of assets. Condition assessments of infrastructure typically involve analyses and evaluation of data and information captured using techniques such as visual inspections, nondestructive testing, photographic and optical imaging, as well as from intelligent and smart sensing of a variety of parameters that generally represent physical phenomena. These condition assessment techniques are increasingly adopting 4IR technologies (such as IoT, distributed ledgers, i.e., blockchain, augmented, virtual and extended reality, as well as artificial intelligence algorithms embedded in e.g., cobots) across digital age platforms (e.g., RAMI 4.0) to provide multiple criteria decision support systems that facilitate effective management of infrastructure. Often these sophisticated techniques are narrowly applied to indicate the technical condition and/or financial performance of discrete assets, or focused on a particular discipline’s or stakeholder’s preference and subjective description of an asset system. Some condition assessment codes, gradings or ratings are formalised, for example, as building condition index or facility condition index [3], albeit that the formalisation may be jurisdiction specific, as well as subject to nomenclature and metrics agreed upon by the relevant stakeholders [4]. The authors in [5] opine that “there is a need to analyse the condition of the whole portfolio of infrastructure [i.e. system of assets] in a cost-effective way to provide input into general policy development and high-level budgeting.” On this basis, several countries publish infrastructure report cards {ref: [6]} by assigning letter grades to various categories of systems of assets based on assessments of physical condition and needed investments for improvement [5, 7] (see Fig. 2). Interestingly, the report cards tend to be conventionally championed by practitioners in professional associations dominated by the civil engineering and closely related disciplines. More often than not, the views and perceptions of some of the stakeholders to the system of assets are either muted or not adequately considered during, and in the reporting of the assessments. As hinted earlier, a majority of routine condition assessments are performed on discrete assets in order to. (i)

in the first instance, prioritise maintenance interventions to ensure that the asset can continue to operate within stipulated regimes in order to deliver the specified levels of service; (ii) in the second instance, to determine the scope of refurbishment and associated rehabilitation that may be necessary to continue to deploy the asset for particular purposes; and (iii) in the third instance, to implement upgrades typically necessitated by technology obsolescence or prompted by the feasibility of new technologies. In practice, all three instances may be simultaneously necessary, especially where the assessment is about the condition of an asset system or a system of assets. As a followon to the discourse referred to in [2], three more case studies are briefly described as follows.

490

J. E. Amadi-Echendu et al.

Fig. 2. Letter gradings of infrastructure report cards [5, 7]

3 Case Studies 3.1 Case 1: Amusement Park Thrill Rides Using an approach somewhat similar to that described in [8], this case study was limited to the collection of specified data. The condition assessment did not include data and information recorded from visual inspections, non-destructive testing, photographic and optical imaging, or from intelligent and smart sensors. Instead, the approach was extrapolated from a literal application of some normative claims articulated in ISO 55000. Curiously, the benefits of asset management as articulated in ISO 55000 include inter alia, ‘improved financial performance’, ‘informed asset investment decisions’, and ‘managed risk’; since ISO 55000 promotes asset management as a coordinated set of activities that “support the realization of value while balancing financial, environmental and social costs, risk, quality of service and performance related to assets.” Subsequently, the term “balancing cost, risk, and performance” has become a colloquial within engineering asset management parlance (e.g., [9]). Thus, this case study pre-supposed that a snapshot of data and information relating to cost, risk, and performance should provide some indication of the condition of an asset at the particular point in time when such data and information were collected.

Case Studies on Condition Assessments of Infrastructure Assets

491

The infrastructure in this case study is a commercial amusement park that features 18 thrill rides, restaurants, and shopping facilities. Given the paramount importance of the safety and comfort of patrons, and the reliability of the thrill rides, the data collected was further limited to information relating to cost, risk and performance of the asset system. The reason for the focus on the thrill rides is that this asset system constitutes the major attraction of the amusement park. The interval for the collection of data was limited to one month, and the data collected is summarised in Table 1. Table 1. Summary of cost, risk and performance data for thrill rides asset system Cost Yr1 Yr2 Yr3 Yr4

Risk

Performance

Maintenance Non-Maintenance Inspection Reactive Maintenance Preventative Maintenance Utilization Energy Value-add 10.063 150.107 3744 18 15 69.44 1144.27 157.29 10.744 154.563 3744 18 18 83.33 1009.13 161.92 3.173 94.448 1248 6 17 28.13 6829.93 (45.21) 8.304 116.157 3432 16 18 77.04 1111.57 115.93

For the case study facility, inadequate/inappropriate maintenance was regarded as the dominant source of risk. As a matter of intrigue, the cost of maintenance includes the costs of reactive and preventative jobs but excludes the cost of inspection activities, while non-maintenance cost includes energy, water, marketing, and other overheads. The trends revealed by the data in Table 1 provided a partial indication of the condition of the thrill rides, and the data was applied to support informed decisions made by the amusement park management particularly regarding budgeting and cost control. 3.2 Case 2: Waste Water Treatment Facility The second case study regards an assessment of a waste water treatment facility in order to establish whether the asset system was in the condition to meet the demand to treat increasing volumes of domestic and industrial waste. For this case study, visual inspection was the dominant method of assessment, and this was carried out by a team of technical persons using a specified guideline and rating procedure. The assessment was conducted over a period of five months including 14 weeks of data collection via visual inspections and retrieval of archived data plus 6 weeks of data collation and analyses. The grouping of the assets for the visual inspection exercise is illustrated in Fig. 3 together with the condition ratings, data confidence grades, and the spectrum rating encompassing risk of failure, safety for the operating personnel, and susceptibility to pollute the environment.

492

J. E. Amadi-Echendu et al.

Fig. 3. Condition assessment metrics for Case Study 2

A total of 7201 items that were assessed are depicted in the groupings in Fig. 4, together with the condition and risk ratings for each group. Taking into consideration that there are various stakeholders to the case study infrastructure, Table 2 is a 4-dimensional rating of the condition assessment results, viz: technical, functional/utility, economic/financial, and environmental/ecological. 3.3 Case 3: Primary and Secondary Schools Facilities The third case study concerns the condition assessment of the facilities in a number of primary and secondary schools within a region. This particular case study was prompted by a number of incidents, unfortunately, some resulted in fatalities. Curiously, due to unavailability of maintenance reports, the indication is that partial or holistic condition assessments are not conducted either routinely or at specified intervals at the primary and secondary schools within the region. Using the nomenclature adopted in this paper, the facilities represent a system of assets that facilitate teaching, learning and development of children in the primary and secondary schools concerned. Depending on the curricula offered and other factors, the

Case Studies on Condition Assessments of Infrastructure Assets

Fig. 4. Condition assessment results for Case Study 2

493

494

J. E. Amadi-Echendu et al. Table 2. A 4-dimensional rating of the Condition assessment in case study 2

Dimension

Performance indicator

Condition Rating (1–5) Comment

Technical

Availability for use Structural integrity

Marginal

3.7

The plant is available for use, although it is not in the best of conditions: non-operational pieces of equipment and the poor operability condition of other equipment

Functional

Operability

Marginal

3.7

The plant is operating, though not as efficiently as it should have been

Financial

Cost/benefit Financial strain

Marginal

3.7

The plant will impose a financial strain on its stakeholders for maintenance and refurbishment actions

Environmental Gas emissions river and Marginal soil pollution

3.7

The marginal state of the plant is also marked by the emission of Ammonia into the atmosphere, as well as the pollution of rivers when discharging water that was not treated at the best standard

facilities in a school may include, at least, ablution facilities, classrooms, libraries, computer rooms, playgrounds, offices, other building structures, and perimeter fences. The condition assessment was conducted by a very concerned individual over a one-year period and was further limited to visual observations and capturing of pictures of structural defects, degradation and deterioration of particular facilities across 10 schools. The impetus for the action by the concerned citizen was to provide concrete and credible evidence to the relevant governance department since there were no records made available regarding maintenance interventions or previous condition assessments at the schools that were under observation. Interestingly, the assessor adopted the condition rating shown in Table 3.

Case Studies on Condition Assessments of Infrastructure Assets

495

Table 3. Condition rating applied for Case Study 3

Not with standing that the assessment of the schools’ infrastructural facilities was limited to visual observations by a postgraduate researcher, however, the images captured and the condition rankings depicted in Fig. 5 have provided additional factual evidence to the relevant governance department to plan and implement appropriate actions to mitigate some of the unfortunate incidents that happened in the recent past.

496

J. E. Amadi-Echendu et al.

Fig. 5. Condition assessment results for Case Study 3

4 Discussion and Concluding Remarks In the first case study, the attempt was to deduce the condition of an asset system based on the trends in cost, risk, and performance data. Data quality factors, especially accuracy and validity are paramount and bedevil this ISO55000 normative approach to condition assessment. The point here is, for conformity, the colloquial ‘balancing of cost, risk, and performance’ should be unambiguously promoted within the ISO55000 family of standards by the provision of a specific set of guidelines on how the cost, risk, and performance of an asset should be respectively defined. The limited scope of the ISO 55000 normative approach to condition assessment may be extended by including data and information obtained through interviews of users, visual inspections, non-destructive testing, photographic and optical evaluation, as well as from analyses of data recorded from intelligent and smart sensors. The second case study briefly described the approach commonly practiced by engineers and technicians to assess condition of infrastructure, i.e., the so-called facilities management approach. In most instances, the process of condition assessment starts with visual inspection augmented by 4IR and digital age scanning technologies. Where necessary and feasible, non-destructive techniques are applied to obtain specific information, for instance, to identify and locate vulnerabilities inherent in discrete assets, asset systems, and the system of assets comprising the infrastructure under assessment. Although the third case study is very limited in many respects, however, it demonstrates how a knowledgeable person can readily conduct assessment of infrastructure aided by the capabilities available in present-day technologies. Given the increasing

Case Studies on Condition Assessments of Infrastructure Assets

497

importance of citizen activism, liberalism (i.e., individual rights and civil liberties), the sustainability paradigm, and the VUCA characterisation of the era of Society 5.0, this approach can prompt campaigns for multidisciplinary and multidimensional approaches to holistic assessments of the condition of infrastructure assets to inform: (i) robust decision-making by owners, custodians, stewards, and managers of infrastructure assets and other stakeholders; (ii) anticipatory learning (resiliency), policy-making and sustainable infrastructure development. Extrapolating from the three case studies discussed here, as well as the two case studies discussed in [2], three propositions are made as follows. Firstly, condition assessments should, as a matter of principle, include information that scientifically captures the opinions of the users of infrastructure assets. Secondly, as a matter of principle and pragmatism, condition assessments should be conducted by teams adequately constituted by relevant multidisciplinary skills and competences aided by available technologies. For example, and where feasible, cobots should be included in the multidisciplinary teams for the conduct of condition assessments. Finally, taking into consideration the views, perceptions and needs of a variety of stakeholders, condition assessment ratings of infrastructure assets should be interpreted in, at least, technical, functional/utility, economic/financial, and environmental/ecological dimensions. This aspect is discussed in good detail in [10].

References 1. Thacker, S., Adshead, D., Fay, M., et al.: Infrastructure for sustainable development. Nat. Sustain. 2, 324–331 (2019). https://doi.org/10.1038/s41893-019-0256-8 2. Amadi-Echendu, J., Botlholo, G., Raman, K.: Condition assessment of engineered assets in the era of Society 5.0. In: Pinto, J.O.P., Kimpara, M.L.M., Reis, R.R., Seecharan, T., Upadhyaya, B.R., Amadi-Echendu, J. (eds.) WCEAM 2021. LNME, pp. 335–343. Springer, Cham (2022). https://doi.org/10.1007/978-3-030-96794-9_30 3. Zubair, M., Anuar, T., Sarbini, N.N., Ibrahim, I.S.: Study on condition assessment metrics based facilities condition index and building condition index. IOP Conf. Ser. Mater. Sci. Eng. 1144, 012013 (2021). https://doi.org/10.1088/1757-899X/1144/1/012013 4. Mayo, G., Karanfa, P.: Facilities Condition Assessments - The How, When, and Why (30–35) FM JA18 F3, volume 34, no. 4, July/August 2018. Issue: Essentials in Facilities Management 5. Rust, F.C., Wall, K., Smit, M.A., Amod, S.: South African infrastructure condition – an opinion survey for the SAICE Infrastructure Report Card. J. S. Afr. Inst. Civ. Eng. 63(2), 35–46 (2021). ISSN 1021-2019 6. https://www.statista.com/statistics/264753/ranking-of-countries-according-to-the-generalquality-of-infrastructure/ 7. ASCE’s 2021 Infrastructure Report Card. https://infrastructurereportcard.org/ 8. Chattopadhyay, G.: Issues and challenges of balancing cost, performance and risk in heavyhaul rail asset management. In: 2016 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), pp. 521–525 (2016). https://doi.org/10.1109/IEEM. 2016.7797930

498

J. E. Amadi-Echendu et al.

9. Heldt, T.: Balancing cost, risk, and performance. Engineering for Public Works Institute of Public Works Engineering Australasia (2018). https://www.ipwea-qnt.com/journal 10. Amadi-Echendu, J.: Managing Engineered Assets. Springer, Cham (2021). https://doi.org/ 10.1007/978-3-030-76051-9. ISBN: 978-3-030-76051-9

Identification of Emerging Safety and Security Risks in Drone Operations at Work Sites Risto Tiusanen(B) , Eetu Heikkilä, Tero Välisalo, and Emrehan Öz VTT Technical Research Centre of Finland Ltd., Sustainable Products and Material, Visiokatu 4, P.O. Box 1300, 33101 Tampere, Finland [email protected]

Abstract. The use of Unmanned Aerial Vehicles (UAVs), drones, is strongly increasing in various purposes in industry and in different sectors of society in general. Digital transformation is a megatrend also in aviation and it will bring significant changes to manned and unmanned aviation. Drones are already widely used e.g., in surveillance and security control tasks and in rescue and search missions. The use of UAV technology and related systems in everyday work processes and tasks will change many of the traditional operations and services. Drone based services are going towards high autonomy and drones will soon operate over the populated areas. Drones and their operating systems and energy supply systems are new assets. The safety and security implications and risks of these systems are still not well known among the drone operators and people affected by nearby flying equipment. To solve these challenges and to manage new risks several international activities are ongoing. European Union Aviation Safety Agency (EASA) has defined new risk-based categories for drone operations: open, specific, and certified. A risk assessment method SORA (Specific Operations Risk Assessment) has been developed by EASA for the analysis and assessment of air and ground risks in specific category drone operations. The aim of our study is to create understanding of UAV related emerging risks at work sites in urban areas. In this paper we first review and discuss air and ground risks that are under consideration in SORA when applying the operational authorisation for specific category drone operations. We also look beyond the scope covered by the SORA assessment and identify other indirect risks or safety implications associated with drone operations when there are people working nearby. In addition to safety risks, we also introduce potential security threats and discuss their implications in drone operations.

1 Introduction The use of Unmanned Aerial Vehicles (UAVs), drones, is strongly increasing in various purposes in industry and in different sectors of society in general. Drones are already widely used e.g., in surveillance and security control tasks and in rescue and search missions (Yaacoub et al, 2020). The use of UAV technology and related systems in everyday work processes and tasks will change many of the traditional data collection, monitoring and logistic operations and services. In the long-term visions, even passenger transport in the future could be done using unmanned Urban Air Mobility (UAM) © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 499–507, 2023. https://doi.org/10.1007/978-3-031-25448-2_47

500

R. Tiusanen et al.

solutions (Anon., 2021). Safe, secure, and reliable UAV operation is essential for the development of new drone-based service business, thus trust on drone technology and drone operators is essential for public acceptance of this new technology. In unmanned aviation, ‘Drone’ is a term for an unmanned aerial vehicle (UAV), while an unmanned aircraft system (UAS) is a drone system including its supporting components and interfaces. Drones can be categorised in several ways. Usually, drones are defined as fixed wing and rotary wing drones. Fixed wing drones look like airplanes with high velocity and large payload, but they also require continuous forward momentum to stay in the air. Although, some fixed wing drones can take off and land vertically (VTOL) without moving forward. Rotary wing drones are smaller and slower, but they are capable to stay stationary and move in any direction in the air like helicopters. One more special type of drone is an engine powered airship, which floats with a buoyant gas lighter than air. (Anon., 2022a). The introduction of new technologies and changes in operating methods and processes bring new challenges to ensuring a safe operating environment and managing the life cycle of these new assets. Drones, their operating systems, and energy supply systems are new assets, and their safety and security implications and risks are still not well known among the drone operators and people affected by nearby flying equipment. A wide range of companies and operators are engaged in drone operations, from small one-person-enterprises to large-scale businesses. In Europe, European Union Aviation Safety Agency (EASA) has defined new riskbased categories for drone operations: open, specific, and certified (EASA, 2021). Commercial use of drones over urban areas are typically specific category operations. Therefore, the focus here is only on specific category drone operations. EASA has also developed a risk assessment method SORA (Specific Operations Risk Assessment) for the analysis and assessment of air and ground risks in specific category drone operations. When assessing the safety risks of drone operations for authorisation purposes, the focus is on immediate safety implications, such as a collision with another aircraft in the air or collision with people on the ground if the drone or its cargo lands uncontrollably. As drone applications increase, attention needs to be paid also to the possible indirect implications to safety and possible new security threats caused by drone operations. The aim of our study was to create understanding of UAV risks at work sites and how they can be assessed. Our research questions were: What is the regulatory basis for drone operations and how the related risks are considered in the operational authorisation? How to identify work site specific safety and security risks? In this paper we first briefly introduce the status of specific category drone operations and the regulatory basis for operational authorization by reviewing air and ground risks that are under consideration in SORA when applying the operational authorisation for specific category drone operations. We also look beyond the scope covered by the SORA method and identify other indirect risks or safety implications associated with drone operation at work sites and populated areas. Due to the wide variety of drone applications in industry and in urban environment - from video recording to parcel delivery - and as they are at different stages of development, there is still very limited safety data and published experiences of them. In this study we approach the issue through one drone application area - construction work sites - and discuss the results in more general context

Identification of Emerging Safety and Security Risks

501

of drone applications in limited work environments in the urban areas. In addition to safety concerns and safety risks, we also introduce potential security threats and other implications caused by everyday drone operations. The study has been conducted as part of the research done in Finland on drone safety and security aspects in a nationally funded research project ‘DroLo’ (https://www.dro lo.fi/). The research is being carried out together with industrial partners, technology developers, drone operators, authorities, and other stakeholders. Focus on our safety and security research is on ‘specific category’ drone applications performed with commercially available drones in an urban environment. The overall aim is building trust in safe drone operation in Finland and in EU by taking a broader perspective on the management of safety and security risks as part of overall system life cycle and asset management.

2 A Brief Overview to Current Status and Future Trends in Drone Applications In recent years, the use of drones has increased widely due to the evolution of technology. Digitalization is a megatrend also in aviation and it will bring significant changes to manned and unmanned aviation. Advanced communication technologies like 5G networks will replace current radio communication solutions and provide improved connectivity and advanced control functionalities for drones. Development of power supply and energy systems provide improved battery capacity and even new and energy sources like hydrogen and fuel cell systems for drones (Anon., 2022a). The variety of drone applications, drone services and use of multi-purpose professional drones are expected to grow strongly in the coming years especially in small-scale logistic, security and inspection services. Smaller and cheaper drones are easily accessible to everyone, and drones can be built from components. This has brought various new applications to the public and commercial use, such as weather monitoring, forest fire detection, traffic control, cargo transport and emergency search and rescue. The highest potential in the broader drone market is expected to be seen in the five major economic industries: agriculture; construction and mining; insurance; media and telecommunications; and law enforcement (Anon., 2022b). The variety of drone applications, drone services and use of multi-purpose professional drones are expected to grow strongly in the coming years especially in small-scale logistic, security and inspection services. The use of flying equipment above the working area in everyday work and use of autonomous drones will change many of the traditional manned operations and services but as always new technology brings in new challenges to work processes and asset management.

3 Legislation and Safety Requirements for Specific Category Drone Operations European Joint Airworthiness Authorities were established in 1970, later renamed as Joint Aviation Authorities (JAA), and since 2002 named as the European Aviation Safety Agency (EASA) (EASA, 2022). European unmanned aviation went through a big change

502

R. Tiusanen et al.

during 2021 when EASA’s drone regulation concept, published first time in 2015, came into force in the end of 2020 with a one-year transition period. EASA prepared and published European wide regulations and defined new categories: open, specific, and certified for professional drone usage. The new EU regulation for specific and certified categories came into force on 1.1.2022 (EASA, 2021). In Finland, the Finnish Transport and Communications Agency Traficom is the national authority in permit, licence, registration, approval, safety, and security matters. [https://www.traficom.fi/en/transport/avi ation]. European wide SESAR project (Single European Sky ATM Research) (https://www. sesarju.eu/U-space) have published the European ATM Master Plan, which defines next steps what will happen in European aviation technologies. UAS traffic management systems, so called ‘U-Spaces’, including air navigation system and framework are currently under development and will be soon available. The new U-Space systems are part of that (Anon., 2020). The change in the drone regulations was significant. Change from nationally specified rules for drone operations to the current operating authorisation practices and requirements for specific category operations including risk assessment done by using SORA methodology continues to confuse drone operators and other stakeholders in the unmanned aviation sector. Safe and secure solutions to manage manned and unmanned aviation in the future will be needed. Requirements for specific category operations are based on risk evaluation. Because of the complexity of UAS domain and different needs of different users, EASA has defined four possible options to obtain the authorisation for the drone operators (EASA, 2021): national or EASA standard scenario (STS), pre-defined risk assessment (PDRA), specific operations risk assessment (SORA) and light UAS operator certificate (LUC). The long-term vision expressed by EASA is that in the future authorisation for the most common specific category drone operations will be based on standard scenarios (STS), pre-defined risk assessment (PDRA) or LUC certificate. In this paper we will not go in more detail into these authorisation processes or the requirements. Instead, we focus on the SORA methodology and the identification of drone related risks in that approach. The above-mentioned four options for operational authorisation are described and discussed in more details in EASA (2021) and in a recently published study in ‘DroLo’ project by Öz (2022).

4 Identification of Safety Risks According to the SORA Method A specific risk assessment method SORA (Specific Operations Risk Assessment) has been developed by JARUS (Joint Authorities for Rulemaking on Unmanned Systems) according to EASA regulations (EASA, 2021) for the analysis and assessment of air and ground risks in specific category drone operations. The Specific category includes operation in residential and urban areas; therefore, assessment of safety, security and other risks and uncertainties is critical (JARUS, 2019). The assessment is a risk-based approach considering not only the UAS but the whole intended operation. The SORA process proposes a risk model that combines ground and air collision risks, and it categorises requirements for risk mitigation measures required in that drone operation. SORA is a specified process to assess the risks that the drone operation imposes

Identification of Emerging Safety and Security Risks

503

on people on the ground (ground risk) and on other manned aircraft (air risk). SORA assesses the intrinsic ground risk and air risk of the operation (risk without mitigation measures) and the final ground risk and air risk (considering mitigation measures). As a result, SORA divides the specific category in six specific assurance and integrity levels (SAIL). The SORA process is meant to be used iteratively to determine the SAIL and consecutively perform a risk assessment (Torens et al., 2020). The SORA methodology does not consider other potential risks that drone operation may pose to its operating environment or its flight path. A simplified schematic of the SORA process is shown in Fig. 1. The input for risk assessment and SAIL level determination is the concept of operations (CONOPS) document, which contains information on the operator, the planned operation, technical data of the UAS, mitigation strategies in case of a loss of control and information of the remote crew. Each SAIL entails different levels of robustness required for a set of operational safety objectives (OSO) established to ensure safe UAS operations. These objectives are meant to reduce the risk of an operation getting out of control. (Torens et al., 2020).

Fig. 1. A simplified overview of the SORA process derived from Torens et al. (2020)

5 Safety and Security Challenges in Drone Operations at Work Sites The use of UAV technology and related systems in everyday work processes and tasks will change many of the traditional work site operations and services. Drone based services are going towards higher autonomy and drones will soon operate over more populated work areas in urban areas. Drones and their operating systems and energy supply systems are new assets. The safety and security threats of these systems are still not well known among the drone

504

R. Tiusanen et al.

operators and people affected by nearby flying equipment. When drone services will start growing, it is obvious that new safety and security threats will emerge in drone flights and operations over the populated work sites and other areas. As new, partially, and fully automated drone concepts will be developed, it is important to understand the emerging risks and the uncertainties involved to be able to plan and execute safe and secure drone flights. As was discussed in the previous chapter, the SORA method used for the operational authorisation focuses on air and ground collision risks and their strategic and tactical mitigation measures. Due to the wide variety of specific category drone applications in industry domains and in urban environment in general (from remotely controlled video recording to autonomous baggage delivery) there is still very limited safety data or published experiences. Different applications are also at different stages of development. In our study, we approach the identification of emerging risks in drone operations by reviewing studies in one application area - construction work sites – in which a lot of research work has been done to get to know about the important safety and security aspects. We discuss the results in more general context of drone applications in work environments in the urban areas. 5.1 Drone Related Safety Risks at Construction Sites New safety risks are estimated to be associated with drone flights over the populated areas in cities and towns or work sites in which there are many people. The introduction of new technologies and changes in operating methods and processes also bring new challenges to ensuring a safe operating environment and managing the life cycle of systems. Construction sector has been fast-growing sector in UAV adoption (Jeelani & Gheisari, 2021). The use of drones for various purposes and the associated safety risks and other possible effects on work and working conditions have been studied widely in construction industry and especially at construction sites (Khalid et al., 2021). The research findings from the construction industry are used here as a reference to emphasize the need for identification and assessment of new drone related safety risks. Drone flying nearby can cause safety risks if it is disturbing people at work and distracting them from an ongoing task e.g., driving a machine or vehicle or working on high scaffolding on a construction site. According to Khalid et al. (2021) the rate of UAV integration into construction jobsites is increasing in the modern work environment utilising the latest technology. Several safety and security risks identified in construction work sites can be generalized in other operating environments. Figure 2 shows examples of direct and indirect risks caused by drone operations reflecting the research results from construction work sites.

Identification of Emerging Safety and Security Risks

505

Fig. 2. Examples of safety risks caused by drones at work sites

The current UAV safety regulations only aim to prevent physical accidents and do not account for indirect risks such as worker distraction, increase in stress, or anxiety about constant monitoring of work. Hence, there is an increasing need to develop a holistic understanding of the direct and indirect influences of UAVs at work sites. 5.2 Safety Challenges that are not Covered in SORA In our survey among industrial partners, technology developers, drone operators, authorities, and other stakeholders in ‘DroLo’ project in Finland we have identified several safety challenges that need to be considered when planning specific category drone operations. Some of them are listed here: Uncertainties related to wireless communication and connectivity: The existing dedicated radio solutions for drones in some operations will be replaced by advanced mobile network solutions like 5G networks in the future (Anon., 2022a). Reliable high speed communication connection is necessary to enable autonomous flying solutions that can fly and navigate long distances reliably and can be remotely controlled from control rooms whenever needed. Continuous internet access is also required for drones for future downstream air traffic control solutions. Gaps in operating know-how and safety management: The importance of organizational factors, safety and security management aspects is likely to increase as drone systems become more complex and connected and the operations are more networked, subcontracted, and performed globally. Extreme weather conditions: The weather conditions are an environmental challenge at work sites. For drones operating in the Nordic countries, weather conditions (wind, frost, rain, snow, ice…) and sudden changes in them can cause special problems, such as freezing rotors and shorten battery life and loss of environmental perception. Technical risks in drone power supply systems: Overheating, fire and even explosion risks related to drone batteries while charging has been raised for discussion. Use of hydrogen raises immediately concerns of fire risks and explosion risks on the ground. Although, hydrogen fuel cell solutions seem to be very promising additional power generation solutions for drones. They have also the advantage that they are a low carbon option. High reliability and lower noise can be also arguments for hydrogen usage. At their worst, these factors can lead to unsafe conditions such as collision accidents or worker distraction. These issues are not covered in the SORA methodology as they are not directly related to air or ground collision risks.

506

R. Tiusanen et al.

5.3 Security Risks in Drone Operations In addition to occupational and personal safety risks, security risks associated with drone operations have recently come to discussion. Security concerns in UAV operations at work sites have been mainly unauthorized video recording of persons or unauthorized trespassing to restricted areas (Jeelani & Gheisari, 2021). According to Tran (2021) cyber-security concern is one of the main problems that prevent the public acceptance of UAS applications. Today security and more precisely cyber-security of drone operations is understood to be associated with a wide variety of risks that need to be taken in account also in work sites. Drones can be used for illegal purposes such as illegal data collection or data transmission, or a drone can be hijacked and used for criminal activities. These are in principle the same as cyber security threats to critical information systems in industrial applications and in society in general. It is well known that the use of wireless communication connections makes information systems vulnerable to various attacks. But it is important to understand that cyber security vulnerabilities can be in all the main UAS elements: human operators, drones, gateways, wireless communications, cloud services and application SW layers (Yaacoub et al., 2020). Identification and assessment of security and cyber security risks is a challenge for drone operators because there is no regulatory basis or assessment methodology available for specific category operations. The evaluation of security and cyber-security risks is not included in the operational authorisation process. Tran (2021) emphasizes that for these reasons a lot more research and development efforts should be put to identify cyber-security threats, risk analysis methodology and risk evaluation criteria in drone operations in civilian domain. As an example of research results Tran (2021) introduces and describes an extended SORA methodology to integrate safety and cyber-security issues to the determination of SAIL levels for specific category drone operations.

6 Conclusions and Future Work The change in the drone regulations has been significant and it still confuses many stakeholders related to drone operations. The current SORA methodology considers air or ground collision accidents and does not consider indirect safety implications or security threats either. Based on the research results and operational experiences there is a need to develop a more holistic understanding of the direct and indirect influences of drone operations in urban areas when there are people exposed to the dangers and harms caused by drones. The results also highlight the importance of organizational aspects and information sharing on new requirements as well as new methods and tool that help companies to develop their capabilities to safe and secure drone operations at work sites especially in urban environment. The results from the studies in construction industry and experiences of drone usage at work sites can be used to identify and assess safety and security risks in other application areas and other types of work environments. Our results in DroLo project have so far raised several safety risk aspects that confirm and supplement the findings of the earlier studies. Our research work continues within the framework of practical flights

Identification of Emerging Safety and Security Risks

507

and studying organizational factors and safety and security management aspects. One research theme in the future could also be the LUC based drone operations. If an organisation aims to act as a LUC operator, an interesting question is what kind of safety and security management systems must then be developed and maintained. Acknowledgements. The work presented in this paper is part of the ‘DroLo’ project funded by Business Finland.

References Anon: European ATM master plan, version 2020, SESAR Joint Undertaking (2020). https://www. atmmasterplan.eu/depl/u-space Anon: Drone taxi market by range, propulsion, autonomy, passenger capacity, system, end-use & region - global forecast to 2030. Research and Markets, July 30, 2021 (2021). https://www. globenewswire.com/news-release/2021/07/30/2272076/28124/en/Drone-Taxi-Market-byRange-Propulsion-Autonomy-Passenger-Capacity-System-End-use-Region-Global-Forecastto-2030.html Anon: Unmanned aerial vehicle (2022a). https://en.wikipedia.org/wiki/Unmanned_aerial_vehicle Anon: Drone market outlook in 2022: industry growth trends, market stats and forecast (2022b). https://www.businessinsider.com/drone-industry-analysis-market-trends-growth-for ecasts?r=US&IR=T EASA: European union aviation safety agency (2022). https://www.easa.europa.eu/light EASA: Easy access rules for unmanned aircraft systems (regulation (EU) 2019/947 and regulation (EU) 2019/945) (2021). https://www.easa.europa.eu/document-library/easy-access-rules/easyaccess-rules-unmanned-aircraft-systems-regulation-eu JARUS: JARUS guidelines on specific operations risk assessment (SORA). JAR-DEL-WG6-D.04. Edition 2.0, Joint Authorities for Rulemaking of Unmanned Systems (JARUS) (2019). http:// jarus-rpas.org/sites/jarus-rpas.org/files/jar_doc_06_jarus_sora_v2.0.pdf Jeelani, I., Gheisari, M.: Safety challenges of UAV integration in construction: conceptual analysis and future research roadmap. Safety Science 144 (2021). Elsevier. https://doi.org/10.1016/j.ssci. 2021.105473 Khalid, M., Namian, M., Massarra, C.: The Dark Side of the Drones: A Review of Emerging Safety Implications in Construction. In: Leathem, T., Perrenoud, A.J., Collins, W., (editors) ASC 2021 57th Annual Associated Schools of Construction International Conference, vol. 2, pp. 18—27 (2021). https://doi.org/10.29007/x3vt Torens, C., Nikodem, F., Dauer, J.C., Schirmer, S., Dittrich, J.S.: Geofencing requirements for onboard safe operation monitoring. CEAS Aeronaut. J. 11(3), 767–779 (2020). https://doi.org/ 10.1007/s13272-020-00451-0 Tran, T.D.: Cybersecurity risk assessment for Unmanned Aircraft Systems. Doctoral Thesis. Université Grenoble Alpes, France. (2021). https://hal.archives-ouvertes.fr/tel-03200719v2/doc ument Yaacoub, J-P., Noura, H., Salman, O., Chehab, A.: Security analysis of drone systems: Attacks, limitations, and recommendations. Internet of Things 11 (2020), 100218. Elsevier (2020). https://doi.org/10.1016/j.iot.2020.100218 Öz, E.: Organisational Certification Process for Specific Category Drone Operations in Finland. Master’s Thesis. Aalto University, Scholl of Electrical Engineering, Finland (2022)

Asset and Risk Management Approach in the Context of Complexity in Industry 4.0/5.0 Systems Issa Diop1(B) , Georges Abdul-Nour1 , and Dragan Komljenovic2 1 Department of Industrial Engineering, University of Quebec in Trois-Rivieres (UQTR),

Trois-Rivières, Canada [email protected] 2 Hydro-Quebec Research Institute (IREQ), Varennes, Canada Abstract. Accurate systematic approaches and tools for managing risks in the context of industry 4.0 and its corollary industry 5.0 are lacking or less efficient, propagating unrealistic awareness of risk in various domains where risk management is needed. Traditional methods have their own limits and might not identify all aspects that influence system safety. When conventional industry challenges are combined with emerging risks along with new systemic and organizational risks as well as cognitive and motivational biases in human logic, there arises the necessity of building thorough Asset Management (AM) and Decision Support approaches. This should account both for conventional and emerging risk safety management. Therefore, innovative, and efficient approaches that can investigate issues from a broad systemic perspective to support AM practitioners to deal with those threats associated with the complexity of socio-technical systems are of interest. On these grounds, this paper focuses on identifying and analysing components of risk management approaches especially for new emerging safety risks within industry 4.0, as well as the rising of extreme, rare, and disruptive events that might generate fatal disturbance of the performance of organizations. We have opted for the relatively new methods that have been developed based on system theories, viz. The Functional Resonance Analysis Method (FRAM), the SystemTheoretic Accident Model and Processes (STAMP) and the global risk-informed decision-making approach (RIDM) in AM as the best suited approach for this research. Further research would validate their efficiency and practicality. Therefore, future research initiatives will be devoted to conduct case-studies in order to obtain more accurate data. Keywords: Asset management strategy · Enterprise Risk Management (ERM) · Emerging risks · Extreme-rare-and-disruptive-events · Resilience · Industry 4.0/5.0 · Risk-Informed Decision-Making Approach (RIDM) · Functional Resonance Analysis Method (FRAM) · System-Theoretic Accident Model and Processes (STAMP)

1 Introduction Escalating complexity of socio-technical systems along with emerging technologyrelated risks as well as “known-unknown” and “unknown-unknown” risks denote an © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 508–520, 2023. https://doi.org/10.1007/978-3-031-25448-2_48

Asset and Risk Management Approach in the Context of Complexity

509

outstanding challenge for conventional system safety approaches. The rising complexity of socio-technical systems inevitably leads to a rise in emerging risks (Leveson 2016). The effects of these risks in asset management should be studied considering the organization’s external and internal context involving human performance and socioeconomic as well as socio-cultural considerations. In such a strong complex environment of Asset Management (AM), extreme, rare, and disruptive events might arise because of uncertainties. Scientists recommend that modern organizations should be studied as Complex Adaptive Systems (CAS) using the techniques of complex systems theory (complexity theory) which was built to cope with complex systems (Checkland 1981; Komljenovic et al. 2016). Industry 4.0 and its corollary industry 5.0 inevitably result in a CAS. The idea of Complex system governance (CSG) might help coping with complexity in CAS (Katina et al. 2021). The concept of CSG involves a framework for the enhancement of system performance. For more details on the latter concept, the reader is referred to Katina et al. (2021). In the same vein, Abdul-Nour et al. (2021) propose a resilience management framework and decision-making under risk and uncertainty (see Fig. 1). This framework recommends using either (i) traditional risk management or (ii) management under uncertainty or resilience management) designed for CAS. In the same vein, ISO (2018b) (draft) “Guidance for managing emerging risks to enhance resilience”, as well as CEN (2013) “Managing Emerging Technology-related Risks” provide foresight and insights about the issue of new emerging risks. The latter might cause the biggest challenges to business continuity and resilience as well as Enterprise Risk Management (ERM) and Occupational Safety and Health (OS&H) constraints and requirements. Still, ISO (2018b) should be used as a complementary tool to ISO (2018a). This will allow to manage with confidence both known risks (ISO 31000) and emerging technology-related risks (ISO 31050). The major challenges for the most widely used conventional analysis techniques of safety risks are the rising complexity of socio-technical systems driven by industry 4.0 which inevitably leads to a rise in emerging risks. Examples of traditional analysis methods of safety risks are Failure Modes and Effects Analysis (FMEA), Fault Tree Analysis (FTA), Hazard and Operability Analysis (HAZOP), Event Tree Analysis (ETA), Bowtie analysis, etc. New tools are needed for the new problems. Nonetheless, it is worth emphasizing that traditional analysis techniques of safety risks should not be discredited but should be extended and enhanced. They perform best on mechanical elements or hardware. Though, they have serious limitations on for e.g., human operators, organizational and social considerations, software program-related aspects, etc. (Leveson 2016). On these arguments, both practitioners and scholars have been interested in relatively new advanced methods based on system theories. The most prominent are the FRAM (for e.g., Diop et al. 2022; Patriarca et al. 2020; Gattola et al. 2018) and the STAMP (Leveson 2016), as well as the RIDM (Gaha et al. 2021; Komljenovic et al. 2019; Dezfuli et al. 2010; Komljenovic et al. 2016; Zio and Pedroni 2012). Our researcher’s priority is to answer the research question arising from the need to bring together complementary approaches to risk management. On these grounds, the general objective of this research paper focus on developing a High-level Risk Management Framework combining FRAM, STAMP and RIDM for the assessment

510

I. Diop et al.

Fig. 1. Decision-making under risk, uncertainty, and resilience (Abdul-Nour et al. 2021)

of industry 4.0 and its corollary industry 5.0 related new emerging technological risks in socio-technical systems, as well as extreme, rare and disruptive events. The remainder of this paper is structured as follows: Sect. 2 summarizes the literature review in AM as well as industry 4.0/5.0 and FRAM, STAMP and RIDM. Section 3 describes the proposed approach for characterizing system safety risks in AM. Section 4 outlines the future case study. Finally, Sect. 5 concludes the study then provides new research directions as a starting point for upcoming targets for this research.

2 Literature Review 2.1 Asset Management Complexity and Uncertainty Associated with the Rising of Extreme, Rare and Disruptive Events Strategy for managing asset involves a variety of interacting and mutually dependent activities at different levels of the organization (such as strategic, organization-wide, project, product, process, etc.). This is supposed to be strongly associated with the organization’s strategic planning (IAM 2015; ISO 2014). Both practitioners and scholars will have to operate complex socio-technical systems along with decision-making processes at all stages of the organizational strategy. The process of managing these sociotechnical systems should align with different levels of organizational strategy (corporate,

Asset and Risk Management Approach in the Context of Complexity

511

business, and functional-level strategy) (Diop et al. 2019, 2021). The latter are characterized by unpredictability affecting the dimensions of resilience such as organizational, technical/technological, operational, social, economic, financial, reputational, and business model (Diop et al. 2021; Komljenovic et al. 2020). These systems are made up of a panoply of complex and uncertain technological objects including capital investment/definition of requirements/acquisition/installation/commissioning and decommissioning of assets (O&M)/shutdown and outage strategies/life cycle value realisation, etc. Furthermore, the context of aging assets obliges organizations to cope with dependability challenges. The latter are reliability, availability, maintainability of assets, coupled with OS&H constraints and requirements as well as ERM as mentioned by Komljenovic et al. (2016). Consequently, organizations have significant constraints as well as requirements to decrease equipment malfunctions or failures causing high-level expectations from maintenance. The Institute of Asset Management (IAM) has developed a conceptual AM model involving the six groups of themes primarily issued by the Global Forum on Maintenance and Asset Management (GFMAM). These themes are (i) strategy and planning, (ii) AM decision-making, (iii) lifecycle delivery, (iv) asset information, (v) organization and people, and (vi) risk and review (GFMAM 2014; IAM 2015). These are contained in the IAM “Asset Management - An Anatomy”, a framework made up of 39 subjects that detail the AM activities within an organization and aligned with the principles of ISO 55000 series of standards for evaluating AM maturity. The reader is referred to Diop et al. (2021) and their bibliographic references for more details on AM models for those unfamiliar with these models. Faced with the severe international competition and the volatility of global markets, as well as the deep global insecurity of all kinds combined with complexity in modern socio-technical systems, managing asset turns out to be challenging. Organizations deal with dreaded risks and uncertainties of all types that can affect organizational objectives, along with meaningful impacts on technical and technological systems and human operator activities. Most of these new kinds of risks are emerging enabling propitious conditions for the rising of extreme, rare, and disruptive events that might badly disturb the performance of organizations. For instance, asset decision-makers and stakeholders grapple with effects of the severe socio-economic inflation of prices and impacts on the global economy. For instance, the unstable global economic context combined with the highly insecure political context inflected by the recent conflict between Russia and Ukraine, along with the coronavirus disease pandemic (COVID-19) are compelling asset decision-makers to revise their economic AM models. This will permit them to cope with these challenges and uncertainty that might affect substantial business investment decisions and elevate costs of commodity as well as the price of doing business. Hence, the challenges would be strategic planning, operational excellence, supply-chain management, regulatory compliance, financial management, health, and safety, etc. In the electrical and nuclear power industry design and operation, such as power generation and transmission as well as distribution, AM and risk management play a pivotal role in the performance of assets. Electrical utilities management which are considered as capital-intensive assets need to get ready for complex emerging technology-related risks due to the rising in frequency and severity of extreme, rare, and disruptive events

512

I. Diop et al.

that might seriously disturb the performance of organizations. The World Energy Council (WEC) propose a “Dynamic Resilience Framework” which focuses on (i) extreme weather, (ii) cyber risks and (iii) geo-spatial analysis for managing those risks in order to contributes to creating capacity and capabilities (WEC 2022). For instance, the “Dynamic Resilience to extreme weather” stands as a blueprint for developing resilience to extreme weather issues. Examples are the Fort McMurray fire: 590,000 hectares damaged, 88,000 people displaced, 2,400 residences ruined, oil and gas operations threatened, 1% crash in GDP) (WEC 2022). The latter recommend improving the resilience to particular events and systemic changes by “situational awareness of the different types of risks preparedness for future developments”. 2.2 Industry 4.0/5.0 Challenges Looking back over the past few years, the concept of industry 4.0 has developed rapidly and became a worldwide adopted term in the technologically advanced countries. Industry 4.0 does not arise from a digital divide like the three previous revolutions, viz. (i) mechanization of production through the steam engine and water at the 18th century, (ii) mass production (Henry Ford) and creation of the assembly line through electricity at the 19th century, (iii) automation of production through information technology and electronics in the 20th century. The arrival of the new era of industry 4.0 influences organizations in various domains. It involves cutting-edge technologies which are capable to capture, optimize and deploy massive data (big data). Technologies such as internet of things (IoT), artificial intelligence (AI), cyber-physical systems (CPS), and cloud computing communicate, interact, and adjust continuously. Industry 4.0 has been shaping the future of organizations provoking overwhelming changes in the way of doing business. The shift to more and more digital systems will be inexorably escorted by a multitude of new challenges and emerging risks associated with OS&H constraints and requirements as well as ERM. For example, major cyberattacks, interconnectivity of digital technologies and interoperability of systems, as well as acquisition and storage of massive data, workforce acquisition, training and their retention in the workplace, etc. Decision-makers who fully comprehend these shifts and the benefits associated with numerical technologies will be best prepared to tackle the various challenges related to industry 4.0. For more details about this concept and its numerous technologies, the reader is referred to the paper by Diop et al. (2021) and their bibliographic references. The fifth industrial revolution (a.k.a. Industry 5.0) is an initiative from the European Commission (EC), the executive branch of the European Union (EU) (Breque et al. 2021). The EC announced the idea of industry 5.0 at the tenth anniversary of industry 4.0 introduction. According to the EC, this concept stands for a complement to the concept of industry 4.0 through supporting research and enablers of innovation. The latter is aimed to be used for the transition to a sustainable, human-centric, and resilient industry (Breque et al. 2021). It enables to position the comfort and safety of people at the centre of the manufacturing process, to realize societal objectives and social fairness beyond jobs and growth, in addition to deliver resilience of prosperity, respecting the boundaries of our planet. That is trying to capture the value of industry 4.0 tools while employing environmentally friendly processes at every stage in the production chain. Industry 5.0 entails three core values, namely (i) human-centric, (ii) sustainable and (iii)

Asset and Risk Management Approach in the Context of Complexity

513

resilient, complementing industry 4.0. In other words, industry 5.0 is considered to be value-driven while industry 4.0 is deemed to be technology-driven. These fundamental principles move the spotlight away from the shareholder value to the stakeholder value as well as strengthen the responsibility of industry to society. The EC has identified six enabling technologies in Industry 5.0, namely (Müller 2020): (i) Individualized humanmachine interaction, (ii) Bio-inspired technologies and smart materials, (iii) Digital Twins and simulation, (iv) Data transmission, storage, and analysis technologies, (v) Artificial Intelligence, (vi) Technologies for energy efficiency, renewables, storage, and autonomy. 2.3 Functional Resonance Analysis Method The Functional Resonance Analysis Method (FRAM) is a relatively new performance assessment method for accident investigation and risk assessment. The FRAM is consistent with the philosophy of the resilience engineering and reflects the “Safety II” concept rather than “Safety I” concept (Hollnagel 2012, 2014). The “Safety I” concept which is a conventional hazard analysis method, such as Failure Mode and Effects Analysis (FMEA) and Hazard and Operability (HAZOP), puts the spotlights on what might goes wrong (that is, how an element may fail). FMEA and HAZOP are bottom-up approaches for risk analysis (Sun et al. 2022). The “Safety II” concept focuses on what goes right (that is, identify the mandatory functions for the system to achieve its purpose). In other words, Hollnagel (2012) mentioned that this method concentrates on “the nature of everyday activities rather than on the nature of failures”. The FRAM concept was established for the benefit of “going behind human error and beyond the failure concept” by modelling the required functions for everyday performance to be successful. At the early stages in 2004, the FRAM idea was motivated by the limitations of deterministic and probabilistic approaches to understand complex systems’ comportment, based on the Stochastic Resonance Theory in Physics (Hollnagel 2004). These days, FRAM is adopted to model complex and dynamic socio-technical systems to capture not only why things sometimes end up going wrong but also succeed (Hollnagel 2012). Hence, the FRAM method supports decision-makers to assess activities in complex and dynamic socio-technical systems in term of the system’s functions as well as complex dependencies and interactions among functions. Therefore, the system’s functions and performance can be studied to understand where performance variability might arise before spreading all over the system. Sun et al. (2022) state that the socio-technical system must have appropriate resilience to withstand the disturbance and absorb the performance variability of its sub-systems and procedures. 2.4 System-Theoretic Accident Model and Processes Leveson (2016) proposes a quite new system thinking approach for accident causation namely the System-Theoretic Accident Model and Processes (STAMP). The latter considers factors such as human operators and organisational considerations along with the technical and technological aspects. The STAMP is a top-down system engineering approach which its theoretical foundation is based on overall systems theory, capable to assess highly complex systems better than the traditional analysis methods of safety

514

I. Diop et al.

risks. The STAMP process describes system safety and security as a “dynamic control problem” (i.e., considering component interactions, control, or enforcement of safety constraints for both component failures and component interactions) rather than a “failure problem or reliability problem”. In the STAMP process, accidents arise when the safety control system does not handle effectively defective interactions among system components (i.e., violation of these constraints or requirements.). Be aware that independent component failure accidents remain contained within the model. The STAMP causality model includes a top-down hazard assessment technique called the SystemTheoretic Process Analysis (STPA). The latter is a quite innovative hazard analysis method based on STAMP extended model of accident causation. The principal purpose of the STAMP-STPA is “to identify accident scenarios that encompass the entire accident process, not just the electromechanical components” (Leveson 2016). The STAMP-STPA method enables to control the comportment of both the components of the system and the system itself (taken as a whole) to make sure that safety requirements and constraints are implemented in the system in operation (Leveson 2016). Steps of the STAMP-STPA process are depicted in Fig. 2 as follows:

Fig. 2. Steps of STAMP-STPA process

2.5 Risk-Informed Decision-Making The concept of RIDM was developed by the US Nuclear Regulatory Commission (USNRC) and the National Aeronautics and Space Administration (NASA) in the 90s to cope with safety concerns that come with nuclear power and the aerospace industry. The International Atomic Energy Agency (IAEA) provides a generic framework for an integrated risk-informed decision-making (Lyubarskiy et al. 2011). For the intent of this study, the subsequent definition which is technology neutral is suggested: “Decisionmaking in which the decision maker considers all pertinent factors, including relevant uncertainties that have a potential impact on the resolution of the issue under consideration. These factors include both quantitative and qualitative factors that are weighted in the risk-informed decision-making process in accordance with the decision-maker’s

Asset and Risk Management Approach in the Context of Complexity

515

judgment and experience. The “risk” component constitutes an adequately weighted input among others, whose significance is situation specific. It is opposed to a riskbased approach where decision-making is solely based on the numerical results of a risk assessment” (Komljenovic et al. 2016).

3 The Proposed Approach for Characterizing System Safety Risks in Asset Management The proposed high-level risk management framework is a combination of the FRAM, the STAMP-STPA and the global RIDM as part of an overall asset management process. This model should be holistic and consider hazards occurring from the system dynamic to facilitate capturing the overall complexity of the socio-technical system. Figure 3 depicts a characterization of system safety investigation methods including the FRAM, the STAMP-STPA and the RIDM positioned in quadrant 2 for highly complex and difficult to control systems.

Fig. 3. Characterization of system safety investigation methods (Source: Hollnagel et al. (2008) - modified)

The proposed approach is three-fold as shown in Fig. 4: 1) To build a model using the FRAM process that can shows the coupling among functional modules described as the interaction and dependencies among functional modules. Therefore, we are capable to show the variability of upstream functional

516

I. Diop et al.

modules and their influences on other functional modules (downstream functional modules) by up-down coupling. In FRAM, risks might arise because of the variability of functional modules and their interactions as well as dependencies. This principle of functional resonance is in line with what Komljenovic et al. (2016) call a “combination of unusual circumstances should come together to produce an extreme or rare event”. These authors point out the growing complexity in modern socio-technical systems as the major causes of performance variabilities. 2) To build a model of the most variable functions from the FRAM model using the STAMP-STPA process that control the behaviour of both the components of the system and the system itself (taken as a whole) in order to make sure that safety requirements and constraints are implemented in the system in operation (Leveson 2016). 3) To use the outcomes from the FRAM model and the STAMP-STPA model, then outline the possibility to combine them into a single model with the RIDM model. The influence of the RIDM would support for long-term performance, and the sustainability of an organization in a constantly shifting and hardly predictable environment, then can consider the risks of extreme and rare events within the overall AM strategy and decision-making process.

Fig. 4. Depiction of the recommended approach

The global RIDM process in asset management (AM) is a novel decision-making methodology appropriate for large projects such as long-term performance and sustainability recommended by Komljenovic et al. (2019). Figure 5 depicts the global RIDM process. Step 1 set up the decision-making framework. It helps to adequately define the question, the context, the options to be studied and the decision to be made as well as the scientific and technical assessment techniques to be utilized. It should not be

Asset and Risk Management Approach in the Context of Complexity

517

neglected and can take a lot of time to achieve. Step 2 performs comprehensive qualitative and quantitative appraisals of engineering and risk, as well as current geopolitical and economical context. This phase is primarily conducted by dedicated subject matter experts by means of the suggested proper scientific and technical assessment methods, models and tools provided in Step 1. The outcomes will provide the decisions makers with relevant evidence-based information and insights to deliberate and make the final acceptable decision-making in Step 3. The latter is primarily achieved by the decision maker along with subject matter experts and stakeholders. Figure 6 describes in details aspects of the model in step 2 of the global RIDM process in AM which is made up of seven sub-models.

Fig. 5. Depiction of Global RIDM process in AM (Source: own representation based on Diop et al. (2021); Komljenovic et al. (2019))

Fig. 6. Depiction of aspects of the model in step 2 of the global RIDM process in AM (Source: own representation based on Diop et al. (2021); Komljenovic et al. (2019))

Furthermore, to perform generic analyses, we argue that it is required to develop a holistic AM strategy capable to consider key factors and components as well as complexity and risks. This requires integrating the seven sub-models and risk assessments outlined in the international standard ISO 31000 methodology (see Fig. 7 below).

518

I. Diop et al.

Fig. 7. Depiction of the seven sub-models and ISO 31000 standard

4 Future Case-Studies In fine, it would be interesting to see what future case-studies will reveal about the effectiveness and usefulness of the proposed high-level risk management framework. The overall structure of these case-studies would be devoted to investigating and analyzing the impact of new emerging safety risks within industry 4.0, as well as the combination of uncommon circumstances which might generate extreme, rare, and disruptive events. Firstly, we will perform a study using the FRAM process for system safety risk assessment. Secondly, the STAMP-STPA process will be combined with the above-mentioned FRAM process to identify and assess the hazards associated with the system dynamic. This will enable capturing the overall complexity of the socio-technical system and provide safety control actions in the system. Moreover, it will be outlined the contribution of the RIDM on this framework for long-term performance, and the sustainability of an organization in the overall AM strategy and decision-making.

5 Conclusion This research is aimed at providing an effective high-level risk management and decisionmaking framework for identifying, assessing, and managing those relatively new or unknown risks just a few years ago. In this respect, we have opted for a trio of concepts that we believe being the best method, viz. The FRAM, the STAMP and the RIDM in asset management. These techniques are much more powerful and useful than the traditional approaches to engineer the complex socio-technical systems. Further investigation would validate their efficiency and usefulness. Hence, upcoming research initiatives will be devoted to conduct case-studies in order to obtain more accurate data. This might well provide an understanding of the socio-technical system from the perspective of asset

Asset and Risk Management Approach in the Context of Complexity

519

and risk management in the context of industry 4.0/5.0 and extreme, rare, and disruptive events.

References Abdul-Nour, G., Gauthier, F., Diallo, I., Komljenovic, D., Vaillancourt, R., Côté, A.: Development of a resilience management framework adapted to complex asset systems: hydro-québec research chair on asset management. In: 14th World Congress on Engineering Asset Management, WCEAM 2019, pp. 126–136. Springer Science and Business Media Deutschland GmbH (2021) Breque, M., De Nul, L., Petridis, A.: Industry 5—towards a sustainable, human-centric and resilient European industry. Directorate-General for Research and Innovation. Publ. Off. Eur. Union (2021) Checkland, P.: Systems Thinking, Systems Practice. Wiley (1981) CEN, European Committee for Standardization: Managing Emerging Technology-related Risks, DIN CWA 16649 (DIN SPEC 91299):2013-10, Geneva (2013) Dezfuli, H., Stamatelatos, M., Maggio, G., Everett, C., Youngblood, R., Rutledge, P.: NASA risk-informed decision-making handbook (2010) Diop, I., Abdul-Nour, G., Komljenovic, D.: The functional resonance analysis method: a performance appraisal tool for risk assessment and accident investigation in complex and dynamic socio-technical systems. Am. J. Ind. Bus. Manag. 12(2), 195–230 (2022) Diop, I., Abdul-Nour, G., Komljenovic, D.: Overview of strategic approach to asset management and decision-making. Int. J. Eng. Res. Technol. 10(12) (2021) Gaha, M., Chabane, B., Komljenovic, D., Côté, A., Hébert, C., Blancke, O., et al.: Global methodology for electrical utilities maintenance assessment based on risk-informed decision making. Sustainability (Switzerland), 13(16) (2021).https://doi.org/10.3390/su13169091 GFMAM, Global-Forum-on-Maintenance and Asset-Management: The Asset Management Landscape (2014). Retrieved from Zürich: www.gfmam.org Hollnagel, E.: FRAM: The Functional Resonance Analysis Method: Modelling Complex SocioTechnical Systems. Ashgate Publishing Ltd. (2012) Hollnagel, E.: Safety-I and Safety-II. The Past and Future of Safety Management. Ashgate, England (2014) IAM, Institute-of-Asset-Management: Asset Management - An anatomy V3 (2015). Retrieved from https://theiam.org/media/1781/iam_anatomy_ver3_web.pdf ISO, International Organization for Standardization: ISO GUIDE 73:2009 Risk management— Vocabulary. In: Geneva, Switzerland: Technical Committee: ISO/TMBG Technical Management Board – groups, p. 15 (2009) ISO, International Organization for Standardization: ISO-55000: asset management— overview, principles and terminology. In: Geneva, Switzerland: International-Organizationfor-Standardization p. 19. Technical Committee: ISO/TC 251 Asset management (2014) ISO, International Organization for Standardization: ISO 31000:2018 Risk management—guidelines. In: International Organization for Standardization ISO Geneva, p. 16. Switzerland, Technical Committee: ISO/TC 262 Risk management (2018a) ISO, International Organization for Standardization: ISO 31050 Guidance for Managing Emerging Risks to Enhance Resilience. Work Group 8 (WG8) of the Technical Committee TC262. International Organization for Standardization ISO Geneva, Switzerland (2018b) Katina, P.F., Pyne, J.C., Keating, C.B., Komljenovic, D.: Complex system governance as a framework for asset management. Sustainability 13(15), 8502 (2021)

520

I. Diop et al.

Komljenovic, D., Abdul-Nour, G., Boudreau, J.-F.: Risk-informed decision-making in asset management as a complex adaptive system of systems. Int. J. Strat. Eng. Asset Manage. 3(3), 198–238 (2019) Komljenovic, D., Gaha, M., Abdul-Nour, G., Langheit, C., Bourgeois, M.: Risks of extreme and rare events in Asset Management. Saf. Sci. 88, 129–145 (2016) Komljenovic, D., Stojanovic, L., Malbasic, V., Lukic, A.: A resilience-based approach in managing the closure and abandonment of large mine tailing ponds. Int. J. Min. Sci. Technol. 30(5), 737–746 (2020) Leveson, N.G.: Engineering a Safer World: Systems Thinking Applied to Safety. The MIT Press (2016) Lyubarskiy, A., Kuzmina, I., El-Shanawany, M.: Advances in Risk Informed Decision Making– IAEA’s Approach. Paper presented at the Proceedings of the Nordic PSA Conference, Gottröra (2011) Müller, J.: Enabling Technologies for Industry 5.0. European Commission, pp. 8–10 (2020) Patriarca, R., Di Gravio, G., Woltjer, R., Costantino, F., Praetorius, G., Ferreira, P., et al.: Framing the FRAM: a literature review on the functional resonance analysis method. Saf. Sci. 129 (2020). https://doi.org/10.1016/j.ssci.2020.104827 Sun, L., Li, Y-F, Zio, E.: Comparison of the HAZOP, FMEA, FRAM, and STPA methods for the hazard analysis of automatic emergency brake systems. ASCE-ASME J. Risk Uncert Engrg. Sys. Part B Mech. Engrg. 8(3) (2022) WEC, The World Energy Council: Dynamic Resilience Framework (2022). Retrieved from https:// www.worldenergy.org/transition-toolkit/dynamic-resilience-framework Zio, E., Pedroni, N.: Overview of risk-informed decision-making processes: FonCSI (2012)

Risk Assessment Using FMEA to Identify Potential Risks of Positive Displacement Pump Failure in Aluminum Industry: A Case Study Hamid Ahmadi1(B) , Meysam Esmaeilzadeh Mofrad2 , and Abolfazl Sedghi2 1 Ferdowsi University of Mashhad, Mashhad, Iran

[email protected] 2 Iran Alumina Complex, Jajarm, Iran

Abstract. Positive Displacement Piston diaphragm pumps are widely used in various industries that deal with high viscosity fluids, including the production of aluminum. However, insufficient research has been done and most of the knowledge of maintenance and repair of such pumps is implicit. In this study, in order to identify, evaluate and provide appropriate measures to reduce and eliminate potential risks associated with the use, repair and maintenance of the piston-diaphragm pump, failure mode and effect analysis has been practised. The presented study carefully studied the documents and interviewed the experts to determine the main components of the pump and carefully reviewed the processes and then using the standard FMEA worksheet, has identified potential failure modes and their effects, and risk priority number. Then the identified risks were analysed and corrective actions were recommended and they were evaluated and prioritized by using Analytical Hierarchy Process method. In total, 42 cases of potential risk condition related to the pump were identified and corresponding actions proposed. Keywords: Filure mode and effect analysis · Analytical hierarchy process · Risk proirity number · Asset management

1 Introduction Risk assessment is the process of qualitative and quantitative analysis of risk potentials and the coefficient of realization of potential risks and identifying what hazards currently exist or may appear. (Alam Tabriz and Hamzehi 2011). Therefore, before starting the project, risks should be identified, quantified, and finally, an appropriate strategy should be adopted to prevent their occurrence (Barends et al. 2012). Risk management can be divided into six main stages, which include risk identification, allocating criteria, risk assessment, determining the acceptable level of risk, dedicating corrective measures to reduce risk and selecting and implementing corrective action. FMEA, is one of the most popular method for risk assessment, hence it is comprehensive, systematic, skill dependent and quantifiable. It is a systematic technique that, prior to the final implementation of each project, defines, identifies potential risks, causes and consequences, assesses the risk of their occurrence, and takes measures to eliminate © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 521–529, 2023. https://doi.org/10.1007/978-3-031-25448-2_49

522

H. Ahmadi et al.

or reduce them (Xia et al. 2011). As a result, the quality of a system can be improved by minimizing project risks, and be effective and increase environmental safety and increase economic efficiency (Nguyen et al. 2016). The main purpose of this method is to predict failure, analyse the results, reduce risk and prevent the occurrence of risks (Xia et al. 2011). This method was first developed by NASA in 1963 to meet their requirements for precision assurance capabilities and then used by Ford Motor Company in 1977 (Gilchrist 1993). In this method, first all the subsets of the system with their components must be listed and the failure states of each of these components must be explained as the input of this method. After performing this operation, the severity of failures should be ranked according to the rate of repetition and their catastrophic effects by the risk priority number, which is the product of the rate of deterioration, probability of occurrence and probability of detection (Puente et al. 2002). Each failure mode can be evaluated by three factors as severity, likelihood of occurrence, and the difficulty of detection of the failure mode. In a typical FMEA evaluation, a number between 1 and 10 (with 1 being the best and 10 being the worst case) is given for each of the three factors. However, in order to obtain a risk priority number (RPN), Severity (S), Occurrence (O), and detectability (D) must be multiplied. Then, the RPN value for each failure mode is ranked to find out the failures with higher risks (Kutlu and Ekmekcioglu 2012). Positive Displacement Piston diaphragm pumps are widely used in various industries including aluminum production and are arguably the most practical equipment for creating very high pressures in abrasive liquids. The pump subjected to this study is TZPM 1600 GEHO which has four main units, 21 sub-units and about 267 parts. This pump is one of the most important equipment of Iran Alumina Complex and failure of this pump for any reason will lead to stop of production process and approximately will cost $ 5740 per hour. The volume flowrate of the pump is 112–135 cubic meters per hour and its working pressure reaches 92– 145 bar and its motor power consumption is 485–746 kW. The power supply consists of a large bearing, crankshaft, connecting rod and connecting rod handle. The end of the connecting rod handle, by being placed in a bearing, has made it possible to rotate and move back and forth, thus transferring power to the cylinder-piston assembly. In Fig. 1, the schematic diagram of the pump is illustrated. Despite the importance of this category of pumps, a comprehensive and practical study of the analysis and evaluation of failures of these pumps has not been done in quantitative ways. Also, regarding the provision of solutions to improve performance and reduce the risk of equipment failure in the continuation of the FMEA analysis, most of the researches have limited themselves to just listing the proposed solutions and have not evaluated the solutions from different perspectives. In the present research, after carefully examining the equipment and identifying the risks with the highest RPN and providing a solution to reduce this number, measures from the perspective of efficiency, cost and required man-hours have been reviewed and prioritized with the help of decision-making methods.

Risk Assessment Using FMEA to Identify Potential Risks

523

Fig. 1. Pump diagram (WeirMinerals 2011)

2 Definitions and Methodology In order to perform a FMEA project, definition of system that is subject of analysis is mandatory. The steps to define the system is proposed by international standards and the steps are done in the following sequences. The taxonomy is a systematic classification of items into generic groups based on factors possibly common to several of the items (location, use, equipment subdivision, etc.). Our study is at the Equipment subdivision section. Levels 6 to 9 are related to the equipment unit (inventory) with the subdivision in lower indenture levels corresponding to a parent-child relationship. (ISO 14224 2016) (Fig. 2). For each equipment class, a boundary shall be defined indicating what reliability and maintenance data are to be collected. This may be given by using a text definition or a combination of both (ISO 14224 2016). Criticality assessment provides the means for quantifying how important a system function is relative to the identified mission. This system, adapted from the automotive industry, provides 10 categories of Criticality/ Severity. It is not the only method available. The categories can be expanded or contracted to produce a site-specific listing. (NASA Reliability-Centred Maintenance Guide for Facilities and Collateral Equipment 2008). The likelihood of occurrence ranking number has a relative meaning rather than an absolute value (Potential Failure Mode and Effects Analysis, fourth edition, 2008). For calculating probability, the number of part failure in a one-year period and number of operational hours of machine is needed. Detectability is the ability of the inspecting mechanism and/or design control to detect the potential cause or the subsequent failure mode. The proper inspection methods need to be chosen. This parameter, too, is given a numerical value between 1 and 10. This

524

H. Ahmadi et al.

Fig. 2. Boundary diagram of the pump (ISO 14224 2016)

ranking measure the risk that the failure will escape detection. A high detection number indicates that the chances are high that the failure will escape detection (Kiran 2016). The risk is measured based on three factors mentioned above. By multiplying the degrees of these three factors by each other, the risk priority score for each potential failure and its effects will be calculated. Those failures with higher RPN score, the cause should be investigated immediately. The risk priority score is the product of three numbers: severity (S), the probability of occurrence (O) and the probability of detection (D). RPN = Severity × Occurrance × Detection

(1)

In order to gather information and collect data about the machine parts and related failures, a team of experts was created from a wide range of expertise including mechanic and electronic engineers in charge of repair and maintenance, operators of machine and condition monitoring engineers. Group meetings were hold and all possible failures, cause and effect of failures, severity and detection of failures were discussed. Then every team member rated the severity and detection based on the standard evaluation tables for corresponding criteria and after discussing about their evaluation, the team agreed on final scores for each part. Also, the data about all kind of failures of machine and their repair time in past 5 years were collected in order to calculate probability of occurrence.

3 Results The 42 cases of potential risk condition related to the pump were identified and the RPN numbers varies from 4 to 504. Most of risks, have lower RPN number than the average

Risk Assessment Using FMEA to Identify Potential Risks

525 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

500 400 300 200 100 0

Fig. 3. Pareto graph

number hence to identify high priority risks, the Pareto principle was used twice and the areas to focus has been clarified for the FMEA team (Fig. 3). After recognizing high priority risks, FMEA team meet again to propose action required. By brainstorming, the team has come out with practical ideas to reduce RPN number. Clogging of the lubrication filter occurs because of the presence of contaminants in the lubricant, and due to the conditions of the working environment, it is almost inevitable that the contamination enters the oil. The result of filter clogging is lubricant pressure drop and lack of proper lubrication, which leads to damage to gearbox gears and disruption of power transmission to the pump. Despite the importance of this issue, there is no suitable tool for monitoring the oil pressure and the only oil-related item that is currently measured is the oil tank level, which makes it difficult to detect the filter clogging. Therefore, experts suggested installing a pressure gauge or transmitter in the lubrication path as a suitable solution in order to reduce the number of fault detection capabilities. This tool can be a precision instrument and inform the operator of clogging by sending an alarm. In the case of the solenoid valve, the problem of “no signal” that occurs due to failure in the internal circuit of the valve or connections, leads to the closing or opening of the propelling liquid path, which results in the diaphragm rupture due to pressure imbalance of its opposite sides. Discovering the exact point of failure and replacing it is a relatively time-consuming and complex process. Therefore, the tacit knowledge of experts should be converted into explicit knowledge by preparing instructions and documentation and be made available to other officials of this department, and with their training, the time of failure detection and repair will be reduced, which will lead to the reduction of the failure severity factor. This will be achieved with the cooperation of the training and precision instruments units. The piston leakage, which is due to the damage of its seals, is one of the common failures that leads to the exit of the propellant fluid from the cylinder and its leakage, as a

526

H. Ahmadi et al.

result of which the amount of propelling liquid is reduced and leads to the pressure drop behind the diaphragm and its rupture. The price of these seals is very small compared to the high cost that their failure imposes on the system, on the other hand, their failure is inevitable over time and statistics show that their average life is 400 days. Therefore, the solution proposed to reduce the possibility of this failure is to replace them during the annual overhaul of the pump, which will be done in coordination with the storage and repair units. Check valve passing is the most frequent pump failure and the second failure in terms of risk priority number. This failure has caused the return of the pumped fluid in the suction line, which in addition to greatly reducing the efficiency of the pump, if it continues, it will cause damage to the transmission lines and other parts. One reason for the high occurrence of this failure is the very high pressure during the discharge and impact closing of this valve, which damages the seat and leads to passing over time. To solve this problem, the FMEA team experts have proposed a plan by studying the check valve housing in similar pumps, which by installing an elbow, will restrain the impacts on the valve and prevent seal failure, and therefore the probability of occurrence will be reduced. Carrying out these reforms requires the cooperation of research and development, piping and repair units. On the other hand, detecting the occurrence of this failure in the early stages can prevent the failure of other valve components and leakage, but detecting passing in the early stages is a difficult task because the sound produced in the early stages is outside the range of human hearing. By examining different solutions, the use of ultrasonic technique was suggested in order to reduce the detection time. Carrying out this process requires the development of a condition monitoring plan and the cooperation of the corresponding units. Diaphragm rupture has the highest risk among pump failures and causes mixing of the propelling liquid and the pumped fluid, as a result of which, in addition to a severe drop in efficiency, the piston and solenoid valves will be damaged if the process continues. Diagnosing this failure is very difficult and the severity of its effect is high. By examining these two factors multiple times, the FMEA team did not find a suitable solution that has economic justification for these two parts. But due to the relatively low price of this part and considering that the average life of these parts is about 200 days, it was suggested to be replaced in every other overhaul, i.e. after 180 days. Also, this part is purchased from three suppliers, and statistical data showed that one of these brands has a significantly longer life than the other two brands. Therefore, it was suggested to use the capacity of this supplier in coordination with the commercial and warehouse units. The results are illustrated on Table 1 below: Final issue to tackle by FMEA team is evaluating recommended actions. The team used AHP (analytic hierarchy process) method in order to assessing the actions by three factors of efficiency, cost and man-hour and rank them by most priority and importance. The AHP is one of the most popular and widely employed multicriteria methods. In this technique, the processes of rating alternatives and aggregating to find the most relevant alternatives are integrated. The technique is employed for ranking a set of alternatives or for the selection of the best in a set of alternatives. The ranking/selection is done

Risk Assessment Using FMEA to Identify Potential Risks

527

Table 1. Recalculating RPN after proposed corrective actions Failure

Current

Recommended action

S

P

D

RPN

6

2

7

84

Install a flow transmitter with low alarm to discover possible plug

No Solenoid valve 7 Signal

3

6

Piston leak

5

6

Valve passing

6

9

Lubrication filter Plugged

Diaphragm rupture

8

7

Predicted S

P

D

RPN

6

2

4

48

126

Train inspectors to 5 discover problems and follow instruction in case of failures to prevent further damages

3

6

90

7

210

Replace piston sealing 5 and packings after 360 days of work

3

7

105

6

324

Change the housing design to more efficient type

6

7

6

252

Use ultrasonic analyser to discover passing at early stage and prevent total failure

6

9

5

270

9

504

Take Both Actions

8

7

5

210

Purchase higher quality diaphragm from another supplier available on the market

8

6

9

432

Replace diaphragm every other overhaul

8

4

9

288

Take both actions

8

3

9

216

with respect to an overall goal, which is broken down into a set of criteria. (Ramanathan 2004). The efficiency is calculated by comparing new RPN to current. The cost and manhour is the average amount of experts’ calculations and prediction. The outcome of AHP is presented by following Table 2 and the actions are sorted base on importance respectively:

528

H. Ahmadi et al. Table 2. Ranking of actions by AHP score

Rank

Recommended action

AHP score

1

Use ultrasonic analyser to discover passing at early stage and prevent total failure

0.222

2

Take both actions (purchase better material and replace part on overhaul) 0.142

3

Replace piston sealing and packings after 360 days of work

4

Replace diaphragm every other overhaul

0.121

5

Install a flow transmitter with low alarm to discover possible plug

0.113

6

Train inspectors to discover problems and follow instruction in case of failures to prevent further damages

0.088

7

Take both actions (change housing and use ultrasound monitoring)

0.085

8

Change the housing design to more efficient type

0.054

9

Purchase higher quality diaphragm from another supplier available on the market

0.047

0.128

4 Conclusion Since a large amount of the cost of aluminum production plants and similar industries is spent on the purchase and maintenance of equipment and parts, including equipment related to high pressure pumps such as Piston Diaphragm pumps, FMEA technique with its preventive approach can reduce these costs and help factories provide high quality products and satisfy standards. After precise analysis of the pump and collecting information about the performance of the equipment parts, by examining the relevant tables and diagrams, it was found that by applying nine recommended actions, which requires collaborations between different departments, the risks of the pump failure can be greatly reduced, which in turn leads to increased reliability and resilience of the system. Each action has different effectiveness (reducing RPN number), cost and man-hour needed, so the options have been prioritized by AHP decision making method. By using results of this study, managers will be able to revise the repair and maintenance plans to achieve company’s mid-term goals. Currently, due to the lack of information or their implicitness about the most frequent, severe and unknown failure risks of this kind of pumps, the appropriate maintenance and repair plan is not implemented, which results in spending extra money and time on unexpected and emergency repairs or unnecessary repairs and maintenance. But researches like this one, by collecting comprehensive information and converting tacit knowledge into practical, can lead to an increase in productivity by reducing unwanted breakdowns and improving the maintenance program, which is a big step in the direction of sustainable development.

Risk Assessment Using FMEA to Identify Potential Risks

529

References Alam Tabriz, A., Hamzehi, A.: Project risk evaluation and analysis using risk management based on PMBOK standard and RFMEA technique. Ind. Manag. Stud. 9(23), e1–19 (2011) Barends, D.M., Oldenhof, M.T.: Vredenbregt MJ, Nauta MJ. Risk analysis of analytical validations by probabilistic modification of FMEA. J. Pharm. Biomed. Anal. 62(64), e82-86 (2012) GEHO PD Slurry Pumps Installation, Operation and Maintenance Manual (2011) Gilchrist, S.H.: Modelling failure mode and effect analysis. Int. J. Qual. Reliabil. Manag. 10(5), e16–23 (1993) International Organization for Standardization. ISO 14224 (2006) Kiran, D.R.: Total Quality Management: Key Concepts and Case Studies, p. e380. Elsevier science(2016) Kutlu, A.C., Ekmekcioglu, M.: Fuzzy failure modes and effects analysis by using fuzzy TOPSISbased fuzzy AHP. Expert Syst. Appl. 39(1), e61–67 (2012) NASA Reliability-Centred Maintenance Guide for Facilities and Collateral Equipment (2008) Nguyen, T., Shu, H., Hsu, B.M.: Extended FMEA for sustainable manufacturing: an empirical study in the non-woven fabrics industry. J. Sustain. 8(9), e1–14 (2016) Puente, R., Pino, R., Priore, P., Fuente, D.D.L.: A decision support system for applying failure mode and effect analysis. Int. J. Qual. Reliabil. Manag. 19(2), e137–150 (2002) Ramanathan, R.: Multicriteria analysis of energy. Encyclopaedia Energy, e77–88 (2004) Xiao, N. Huang, H.Z., Li, Y., He, L., Jin, T.: Multiple failure modes analysis and weighted risk priority number evaluation in FMEA. Eng. Fail. Anal. 18(4), e1162–1170 (2011)

Asset Operations and Maintenance Strategies

Influence of the Income From the Use of an Asset on the Calculation of its Preventive Interval for a Planned Horizon. Use of Semi-Markov Processes and Degraded State Antonio Sánchez-Herguedas1(B) , Adolfo Crespo Márquez1 , and Francisco Rodrigo-Muñoz2 1 Department of Industrial Management, School of Engineering,

University of Seville, Seville, Spain {antoniosh,adolfo}@us.es 2 Department of Applied Mathematics II, School of Engineering, University of Seville, Seville, Spain [email protected]

Abstract. The calculation of the preventive interval requires a mathematical procedure that analyses the income and costs involved in the operation and maintenance process of an asset and considers the behaviour in the event of asset failure. For this purpose, a model is designed in which the asset can be in four states, with varying sojourn times and increasing failure rates. Semi-Markovian processes rep-resent these assumptions well. The proposed model provides the mathematical equation that accounts for the average accumulated returns over successive transitions between states and the equation that optimises the preventive interval for any planning horizon. The results allow identifying the importance of the income per operating hour in the calculation of the preventive interval.

1 Introduction Most industrial assets are subject to predetermined preventive maintenance tasks and corrective tasks when they fail. To optimise maintenance, the size of the preventive interval is the key. To calculate the size of the preventive interval T to be applied to an asset, it is necessary to perform a study where costs, income, and failures are considered. Usually, this study is not performed, and the asset owner chooses to follow the manufacturer’s maintenance plan. The asset that has been designed for the conditions set by the manufacturer is used by the owner in his conditions. The maintenance manager must try to optimise the preventive interval in these new conditions. To observe the influence of the asset’s uptime income on the size of the optimal preventive interval, a model is designed, where the degradation of the asset can affect the quality or quantity of items produced. It can also be applied to an asset, where the variation of its production demand affects the income from its sale. Or simply the change in market values can change the income of companies, as is the case, for example, with the values of metals or energy sources. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 533–543, 2023. https://doi.org/10.1007/978-3-031-25448-2_50

534

A. Sánchez-Herguedas et al.

A tool is presented to calculate the optimal preventive interval, when the income generated by the use of the asset is modified. The operation and maintenance process of the asset is adjusted to a mathematical model of four states: operational, corrective, preventive and degraded operational. From the initial instant, the model evolves over time, following a semi-Markovian process with an embedded Markov chain. The process is semi-Markovian because the sojourn times in each state are not subject to exponential functions Transitions between states occur according to the probabilities set in the Markov chain. At each transition the process accumulates costs or income as returns (negative or positive). The aim is to find the size of the preventive interval that maximises the accumulated returns at each transition. This allows finding optimal values for any time horizon. The model proposes a system of difference equations for the average accumulated return, solved by applying the z-transform. In its resolution, two types of solutions have had to be chosen, one for real roots and the other for complex roots that depend on the degradation. Subsequently, by derivation of the average accumulated return with respect to the preventive interval, the mathematical expression of the optimum preventive interval τo is reached. Many authors use semi-Markovian models to find the value of the optimal preventive interval, because they are well adapted to the performance characteristics of maintenance tasks. The model allows to analyse how the preventive interval is affected when the income of the degraded state is modified. The higher the input, the longer the interval, and vice versa. To explain the use of this tool, a case study has been presented.

2 Background In the literature we find numerous papers where the value of the preventive interval is sought using different scenarios and techniques. Several authors use semi-Markovian models to locate the value of the optimal preventive interval because it is well adapted to the execution characteristics of the maintenance tasks. Grabski (Grabski, 2014) explains how to build semi-Markovian models and discusses the different parameters and reliability characteristics. Lyubchenko et al. (Lyubchenko et al., 2018) present an approach for the evaluation of recommended preventive maintenance intervals of radio devices. Kumar and Varghese (Kumar and Varghese, 2018) use non-exponential failure and repair time distributions. They apply the golden section search technique to obtain the preventive interval that optimises availability. Yi et al. (Yi et al., 2018) develop a discrete-time semi-Markovian system with a state space consisting of three subsets: working, modifiable and faulty. They also use the z-transform. Wu et al. (Wu, Maya and Limnios, 2021) solve a continuous-time semi-Markovian process using algorithms from the discretetime case. They consider sojourn times in states that follow exponential and Weibull distributions. Other authors use Markov processes. Yang et al. (Yang, Li and Wang, 2021) formulate the integrated optimization problem as a Markov decision process framework. They simultaneously consider processing costs, maintenance costs, and completion rewards to find the optimal production policy. They maximise the long run expected average rewards over an infinite horizon. Zhao et al. (Zhao, Gao and Smidts, 2021) model the degradation process of a system using a hidden Markov model and develop two algorithms to infer the transition rates between states. Liu et al. (Liu, Huang and Deng,

Influence of the Income From the Use of an Asset

535

2022) study the desirability of performing the preventive maintenance task or continuing with the productive task, as a Markov decision process that minimises the expected system costs. Gu et al. (Gu, Guo and Jin, 2020) by developing a framework based on discrete-time Markov chain models, evaluate the system performance under the control limit policy in manufacturing systems consisting of multistate machines and intermediate buffers. They perform numerical analysis to demonstrate the impact of parameters such as maintenance duration. Finally, a mathematical tool has been used during the process of developing the equations, the z-transform. In the field of industrial maintenance, there is little documentation where this resolution technique is used, which is more often related to wave propagation and resonance (Vadalá et al., 2021) or acoustics (Mikhin, 2008). Analysing the articles where the preventive interval is calculated, most of them apply algorithms, in a few cases, a mathematical formula is offered whose inputs are data available to the maintenance manager: failure data or intervention costs. The exceptionality of this article lies in the fact that this formula includes the income from the use of the asset, and thanks to the degraded state, it allows the value of the optimal preventive interval to be analysed when the income is modified.

3 Material and Methods The objective is to find the mathematical expression of the preventive interval that optimises the accumulated economic return over time. Its value depends on the distribution of its failures, the costs of maintenance tasks and operating income and the possible penalties for its inactivity. The maintenance manager, without knowledge of modelling and calculation techniques, can apply the formula to plant assets. In this section, the data of a case to be analysed are exposed, and the semi-Markovian model is designed. 3.1 Real Case. Returns and Weibull Distribution Data A four-state model is developed: operational S1 , degraded operational S4 , corrective S2 , and preventive S3 , see Fig. 1. The model is applied to the case of the failure of an

Fig. 1. Representation of the states, transitions between states and accumulation of returns at each transition

536

A. Sánchez-Herguedas et al.

engine component of a mining truck. Sometimes during operation an event occurs (at  time τ ) during operation that affects the power delivered by the engine. The decrease in engine performance decreases the income that the truck brings per hour of operation to the business. Failures from 16 engines in an open pit mine have been collected over four years. The failure time data, ti which are collected in the Table 1 have been fitted to a Weibull distribution function of shape parameter α, scale parameter β, and location γ , using the Benard approximation (Sánchez-Herguedas, Mena-Nieto and Rodrigo-Muñoz, 2021). We obtained a Weibull of parameters: α = 3.33, β = 5368 and γ = 301. On the other hand, we estimate the distribution functions of repair time Fc (t) and preventive time Fp (t). Due to their duration, they are fitted to Normal means μ2 = 72 for the corrective and μ3 = 56 for the preventive task. Table 1. Failure time data, ti corresponding to 48 failures (hours). 6,635

4,087

4,225

3,964

6,118

3,775

4,377

6,851

2,823

6,684

6,915

1,733

5,645

4,471

4,890

5,887

5,358

4,305

6,536

6,622

4,232

3,661

7,861

5,714

7,421

4,585

4,566

4,358

6,672

4,132

3,209

4,979

6,927

4,562

2,283

7,616

6,618

3,620

3,478

2,668

7,641

3,879

5,775

7,030

4,226

3,829

5,415

6,390

The returns (income and costs) involved in the process are: • Ri income or cost per unit time that the system remains in state i (S1 , S2 , S3 , S4 ) • Rij , transition cost from state i to next state j. The values for these data are shown in Table 2. Table 2. Model input data. Failure data, times, average maintenance activities, costs, and income for each state Failure distribution function

Average repair time Average preventive time

Weibull (α, β, γ)

Normal

Normal μ3

α

β

γ

μ2

3.33

5,368.00

301.00

72

S1 Returns

S4 Returns

56 S2 Returns

S3 Returns

R1 (e/h) R12 (e) R13 (e) R4 (e/h) R42 (e) R43 (e) R2 (e/h) R21 (e) R3 (e/h) R31 (e) 5.0

-3,270

-1.0

4.0

-3,270

-1.0

-95.0

-360

-82.0

-360

Influence of the Income From the Use of an Asset

537

3.2 Equations To design the model, a homogeneous Markov chain with n states and transition probabilities at each step is considered. {Xn , n ≥ 0} With n states, and transition probabilities in each step pij = P(X1 = j|X0 = i). This chain constitutes the Markov chain embedded in a semi-Markovian process and determines the evolution of the process. The semiMarkovian process is characterised by the fact that the sojourn times differ in each state. The sojourn times and the transition between states have associated economic returns, which can be positive in the case of income or negative in the case of costs. The variable rij (m) contains the return from the state i to the state j in the transition m. At m successive transitions from the i state, the process accumulates returns, which added together with their respective signs constitute the accumulated return in m steps from the state i. Let be Ri (m) this random variable. At m transitions the system can establish many alternatives. For this reason, it is not possible to calculate the value of Ri (m) but its average value, the average accumulated return, can be calculated. vi (m) = E(Ri (m)). Following the process established in (Sánchez-Herguedas, Mena-Nieto, RodrigoMuñoz, et al., 2022) and (Sánchez-Herguedas, Crespo-Márquez and Rodrigo-Muñoz, 2022) the following system of difference equations is reached. V (m) = V (1) + PV (m − 1)

(1)

The length of time the asset is kept running is a random variable that we will call T and can take any positive value. Usually, this variable is called time to failure. If the system is in the operational state S1 . It will remain in this state until it fails and the system goes to the corrective state, S2 or after a time τ  , the system reaches the state. It is degraded state, S4 . The time T0 is the time the system is in the   operational  . It is defined as: the time-to-failure random variable, truncated in the interval 0, τ   T0 = min T , τ  . The time to repair is also a random variable called. Tc . But if the system goes to state S4 it remains there until the failure occurs or a time elapses τ − τ  and reaches the preventive state. The time Td is the time the system  is inthe degraded state. It is the time-to-failure random variable, truncated at interval τ , τ . It is defined    as Td = min{T , τ } − τ = min T − τ , τ − τ  . The time to the preventive task is also a random variable Tp . The distribution functions of T , T c , Tp will be denoted by F(·), Fc (·), Fp (·) and probability density functions f (·), fc (·), fp (·). The random variables T0 and Td have truncated distributions whose distribution functions are defined from the distribution function of T . The matrix of times spent in one state before moving to another is: ⎞ ⎛ 0 T0 0 τ  ⎜ Tc 0 0 0⎟ ⎟ ⎜ (2) ⎝ Tp 0 0 0⎠ 0 Td τ − τ  0

where the i, j-th element is the sojourn time in the state i before to jump state j. The average sojourn time in the operational state is:  τ  τ 1 1  A = E(T0 ) = tf (t)dt = τ − F(t)dt. F(τ  ) 0 F(τ  ) 0

538

A. Sánchez-Herguedas et al.

The average sojourn times in the states S2 and S3 , are the averages of the random variables Tc and Td , (B = μ2 and C = μ3 ):   ∞   ∞ B = E(Tc ) = tfc (t)dt, C = E Tp tfp (t)dt 0

0

The average of Td , i.e. the average time spent in state S4 is:  τ 1 (t − τ  )f (t)dt D = E(Td ) = F(τ ) − F(τ  ) τ  The average sojourn times matrix is the average of matrix sojourn times, Eq. 3: ⎡⎛

0 ⎢⎜ Tc ⎜ Q = E⎢ ⎣⎝ Tp 0

T0 0 0 Td τ

0 0 0 − τ

⎞⎤ ⎛ 0 A τ ⎥ ⎜B 0 0⎟ ⎟⎥ = ⎜ 0 ⎠⎦ ⎝ C 0 0 0 Dτ

0 0 0 − τ

⎞ τ 0⎟ ⎟. 0⎠

(3)

0

where E[·] means, mean value. As in all continuous-time random processes, there is no transition to the same state the system is in. It remains there until the next transition takes it to another state. In our system of four states, while the system is in state S1 , it can only  transition   to states S2 and S4 .  = F τ  = p . The probability is P T ≤ τ The probability of transitioning to state S 2 1      





of passing the state S4 is P T > τ = 1 − P T ≤ τ = 1 − F τ = 1 − p1 . go to states S2 and S3 . The probability of If the system is in state   S4 , it can only 2 −p1 , and to state S3 P T > τ |T > τ  = going to state S2 is P T < τ |T > τ  = p1−p 1    2 1 − P T < τ |T > τ = 1−p 1−p1 . In the states S2 and S3 , the system can only go to the state S1 , the transition probabilities both take the value 1. The transition probability matrix P can be written as: ⎛ ⎞ 0 p1 0 1 − p1 ⎜1 0 0 0 ⎟ ⎜ ⎟ (4) P=⎜ ⎟ 0 0 ⎠ ⎝1 0 2 −p1 1−p2 0 0 p1−p 1 1−p1 The returns matrix is composed based on the sojourn times given in the Eq. 3 and the returns expressed in Table 1. It is a matrix whose ij − th element is the sum of the return due to the permanence in the state i, and the one due to the transition to the j state, Eq. 5. ⎛

⎞ 0 AR1 + R12 0 R1 τ  + R14 ⎜ BR2 + R21 ⎟ 0 0 0 ⎟ R=⎜ ⎝ CR3 + R31 ⎠ 0 0 0    0 0 DR4 + R42 τ − τ R4 + R43

(5)

Influence of the Income From the Use of an Asset

539

4 Development of Mathematical Expressions In order to solve the Eq. 1, the z-transform will now be used. Recall that the z-transform of a sequence x(m), which is usually denoted Z[x(m)], is a complex variable function which is defined as the following Laurent series: Z[x(m)] =

∞ 

x(n)z −n , z ∈ C\{0}.

(6)

n=0

In this case a vector whose components are sequences is considered. The z-transform of a vector is another vector whose components are the z-transforms of the components. For convenience, the Eq. 1 is rewritten with the index increased by one unit, thus: V (m + 1) = V (1) + PV (m)

(7)

Multiplying by the inverse of the regular matrix I −z −1 P, dividing by z and reordering terms, the Eq. 8 is obtained: Z[V (m)] =

−1  −1 1  I − z −1 P V (1) + I − z −1 P V (0) z−1

(8)

At this point, it only remains to invert these z-transforms, to solve the Eq. 7. A similar process has been followed in (Sánchez Herguedas, Crespo Márquez and Rodrigo Muñoz, 2022) and (Sánchez-Herguedas, Mena-Nieto, Rodrigo-muñoz, et al., 2022). Throughout the development, a second-degree polynomial appears z 2 + z − p1 + 1. The roots of this polynomial can be real or complex depending on the value of p1 . Real if (4p1 − 3) is non-negative and complex conjugate if it is negative. This causes the model to have two types of solutions depending on the value of p 1. The vector V (1) is obtained from vi (1) = 4j=1 rij (1)pij withi = 1, 2, 3, 4 for which we need the probability and return matrices, P and R: ⎞   ⎞ ⎛ ⎛ (AR1 + R12 )p1 + R1 τ  + R14 (1 − p1 ) v1 (1) ⎟ ⎜ v2 (1) ⎟ ⎜ BR2 + R21 ⎟ ⎟=⎜ ⎜ (9) ⎜ ⎟ ⎝ v3 (1) ⎠ ⎝ CR3 + R31 ⎠    2 −p1 2 + τ − τ  R4 + R43 1−p (DR4 + R42 ) p1−p v4 (1) 1−p1 1 In the case of real roots, the first component of the average accumulated return is of the form: v1 (m) = A1 r1m−1 + A2 r2m−1 + A3 (m − 1) + A4 , m = 1, 2, 3, · · ·

(10)

In the case of complex roots, the first component of the average accumulated return is of the form: v1 (m) = 2r m−1 (Re(B1 ) cos(m − 1)ω − Im(B1 ) sin(m − 1)ω) + A3 (m − 1) + A4 (11) Deriving with respect to the preventive interval τ . For the case of complex roots:

540

A. Sánchez-Herguedas et al.

dv1 (m) = M1 (1 − F(τ )) + M2 f (τ ) dτ   a μ + b2 M1 = R4 (a2 cos(m − 1)ω + 2 sin(m − 1)ω) · r m−1 + (m − 1)c1 + d2 σ     a μ + b3        a μ + b2  R42 − R43 + 3 v2 (1) − v3 (1) M2 = r m−1 cos(m − 1)ω a2 R42 − R43 + a3 v2 (1) − v3 (1) + sin(m − 1)ω 2 σ σ       + R42 − R43 (m − 1)c1 + d2 + v2 (1) − v3 (1) (m − 1)c1 + d3

(12)

For the case of real roots, the same expression is obtained, but M 1yM 2 take different expressions. The same is true for the expressions of A1 , A2 , A3 , A4 , B1 for the average accumulated return. If the failure distribution function is a Weibull, by substituting them in Eq. 12, and equating it to zero, the non-zero common factor can be eliminated. α e−(τ −γ /β) . The optimal preventive interval is defined by:  1  β α M1 α−1 τo = − +γ · α M2

(13)

5 Analysis and Results We proceed to include the Table 2 data into the equations of the optimal preventive interval. The value of the time for which the transition to the degraded state occurs, τ  is an experimental value to be included in the model. To be able to make a comparison  exercise, different values of τ , between 1,000 h and 8,000 h, taken at intervals of 1,000, will be considered. Table 3 shows the values of the optimal preventive interval and the average accumulated return obtained for ten transitions, m = 10. The values for  τ ≥ 7, 000 are not meaningful since the preventive maintenance time would be reached before the time to degradation. Table 3. Values of the optimal preventive interval τo and average accumulated return v1 (m) for different time intervals until degradation τ  , at the transition m = 10 τ  (hours)

1,000

2,000

3,000

4,000

5,000

6,000

7,000

8,000

v1 (10) (e)

39,364

47,744

55,695

61412

66,996

74,656

78,585

78,166

τo (hours)

6,042

6,061

6,115

6,164

6,159

6,146

6,184

6,229

If the three-state model formulation (without degraded state) (Sánchez-Herguedas, Mena-Nieto and Rodrigo-Muñoz, 2022) is used with the same data, it is obtained that τo (R1 = 5) = 6, 617) and τo (R1 = 4) = 6, 040). These coincide with the results of the four-state model when R1 = R4 . The mathematical expression for the optimal preventive interval τo in the three-state case is: (τo − γ )α−1 =

βα · α R12 − R13 +

−R1 2m−1−(−1)m−1 (R2 B 2m+1+(−1)m−1

+ R21 − R3 C − R31 )

Influence of the Income From the Use of an Asset

541

It is also possible to know the behaviour of the τo as the transitions m are increased for a given τ  . Specifically, in the Fig. 2 the values of the optimal preventive interval τo values are plotted against the values of m, between 1 and 60, for the value of τ  = 4, 000 hours. This figure also shows the value of τo when there is no degraded state (3 states). As the value of τ  value increases, the curve would shift upwards. The upper limit of the displacement is set at the value of the τo when there is no degradation (three states), 6.617 h. The same is true for the lower limit coincides with the value of the input when there is no degradation, 6.040 h, and the input is the value of R4 .

Fig. 2. Representation of the optimal preventive interval τo , at each transition m. 3-state case and 4-state case with τ  = 4000

6 Conclusions This article presents a mathematical tool or formula to find the optimal preventive interval when income degradation occurs or income changes as a result of the market. The expression achieved for the preventive interval is directly dependent on the income in the degraded state R4 and the costs in the degraded state before reaching other states R42 and R43 . It also depends on the probability of system failure when the degradation occurs p1 and the costs of a failure v2 (1) and of preventive intervention v3 (1). Once the tool was obtained, it was applied to a case study, and the following observations were made: • The value of income R4 is a determining value in the economic optimisation of the preventive interval. If its value is less than R1 the preventive interval will decrease, if it is greater it will increase. • Once degradation is detected, the asset can continue to be used, although the preventive interval needs to be reduced to optimise maintenance. The same is true when income decreases for market reasons. In this case, an increase in revenue may also occur, which would increase the size of the optimal preventive interval.

542

A. Sánchez-Herguedas et al.

• In the execution of maintenance, a large number of transitions cannot be considered, the developed formula allows the calculation of the optimal preventive interval for any number of transitions (finite horizon), as it depends on m. The authors intend to continue the line of research, developing models with a larger number of degraded states and considering the degradation function as an element to determine transitions and sojourn times in the states. The ultimate goal would be to model the behaviour of an asset under condition-based maintenance.

References Grabski, F.: Semi-Markov Processes: Applications in System Reliability and Maintenance. Elsevier Inc., Waltham (2014). https://doi.org/10.1016/C2013-0-14260-2 Gu, X., Guo, W., Jin, X.: Performance evaluation for manufacturing systems under control-limit maintenance policy. J. Manuf. Syst. 55, 221–232 (2020). https://doi.org/10.1016/j.jmsy.2020. 03.003 Kumar, G., Varghese, J.P.: Optimum preventive maintenance policy for a mechanical system using Semi-Markov method and Golden section technique. In: 2018 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), Bangkok, pp. 232–236 (2018). https://doi.org/10.1109/IEEM.2018.8607343 Liu, B., Huang, H., Deng, Q.: On optimal condition based task termination policy for phased task systems. Reliability Engineering and System Safety. Elsevier Ltd, 221(108338) (2022). https:// doi.org/10.1016/j.ress.2022.108338 Lyubchenko, A., et al.: An approach for optimal maintenance planning of radio communication devices considering reliability and operational costs. In: Moscow Workshop on Electronic and Networking Technologies, MWENT 2018 - Proceedings, pp. 1–5 (2018). https://doi.org/10. 1109/MWENT.2018.8337301 Mikhin, D.: Analytic discrete transparent boundary conditions for high-order Padé parabolic equations. Wave Motion 45(7–8), 881–894 (2008). https://doi.org/10.1016/j.wavemoti.2008. 03.006 Sánchez-Herguedas, A., Mena-Nieto, A., Rodrigo-muñoz, F., et al.: Aplicación de enfoques Semimarkovianos a la mejora de políticas de mantenimiento predeterminado en activos industriales. In: Tomás de J. Mateo Sanguino. José Manuel Lozano Domínguez, Manuel Joaquín Redondo González, Iñaki Josep Fernández de Viana González, M. Á. R. R. (ed.) Actas de las IV Jornadas ScienCity 2021. Fomento de la Cultura Científica, Tecnológica y de Innovación en Ciudades Inteligentes. Huelva, pp. 23–26 (2022) Sánchez-Herguedas, A., Mena-Nieto, A., Rodrigo-Muñoz, F., et al.: Optimisation of maintenance policies based on right-censored failure data using a semi-Markovian approach. Sensors 22(4)(1432), 1–18 (2022). https://doi.org/10.3390/s22041432 Sánchez-Herguedas, A., Crespo-Márquez, A., Rodrigo-Muñoz, F.: Optimising the preventive maintenance interval using a Semi-Markov process, z-transform, and finite planning horizon. In: González-Prida, V., Márquez, C. A. P., Márquez, A.C., (eds) Cases on Optimizing the Asset Management Process. 2022nd edn. Hershey, PA: IGI Global, pp. 137–161 (2022). https://doi. org/10.4018/978-1-7998-7943-5.ch006 Sánchez-Herguedas, A., Mena-Nieto, A., Rodrigo-Muñoz, F.: A new analytical method to optimise the preventive maintenance interval by using a Semi-Markov process and z-transform with an application to marine diesel engines. Reliab. Eng. Syst. Saf. 207, 1–15 (2021). https://doi.org/ 10.1016/j.ress.2020.107394

Influence of the Income From the Use of an Asset

543

Sánchez-Herguedas, A., Mena-Nieto, Á., Rodrigo-Muñoz, F.: A method for obtaining the preventive maintenance interval in the absence of failure time data. Eksploatacja i Niezawodnosc – Maintenance and Reliability 24(3), 564–573 (2022). https://doi.org/10.17531/ein. 2022.3.17 Sánchez Herguedas, A., Crespo Márquez, A., Rodrigo Muñoz, F.: Optimizing preventive maintenance over a finite planning horizon in a semi-Markov framework. IMA J. Manag. Math. 33(1), 75–99 (2022). https://doi.org/10.1093/imaman/dpaa026 Vadalá, F., et al.: Free and forced wave propagation in beam lattice metamaterials with viscoelastic resonators. Int. J. Mech. Sci.193 (2020) (2021). https://doi.org/10.1016/j.ijmecsci.2020.106129 Wu, B., Maya, B.I.G., Limnios, N.: Using Semi-Markov chains to solve semi-markov processes. Methodol. Comput. Appl. Probab. 23, 1–13 (2020). https://doi.org/10.1007/s11009-020-098 20-y Yang, H., Li, W., Wang, B.: Joint optimization of preventive maintenance and production scheduling for multi-state production systems based on reinforcement learning. Reliability Engineering and System Safety. Elsevier Ltd, 214(107713) (2021). https://doi.org/10.1016/j.ress.2021. 107713 Yi, H., et al.: Stochastic properties and reliability measures of discrete-time semi-Markovian systems.’ Reliab. Eng. Syst. Saf. 176(2017), 162–173 (2018). https://doi.org/10.1016/j.ress. 2018.04.014 Zhao, Y., Gao, W., Smidts, C.: Sequential Bayesian inference of transition rates in the hidden Markov model for multi-state system degradation. Reliability Engineering and System Safety. Elsevier Ltd, 214(107662) (2021). https://doi.org/10.1016/j.ress.2021.107662

Factors Affecting the Quality of Network Services in Emerging Telecoms Operating Environment and Markets Charles Okeyia(B) and Nuno Marques de Almeida(B) Instituto Superior Tecnico (IST), Lisbon, Portugal {charles.okeyia,nunomarquesalmeida}@tecnico.ulisboa.pt Abstract. As an emerging market, the telecoms sector in Nigeria has undergone a considerable increase in teledensity and consumer base over a decade and is still on exponential growth. However, the consequence of this growth has been a continuous degradation of telecom operators’ network quality of service (QoS), which has impacted subscribers’ and customers’ needs, satisfaction, and expectations. Given the existing literature and research on asset management, this paper explores the roles of infrastructure and asset management activities in network service quality in the study context. Therefore, in exploring the QoS issues, the infrastructure and asset intermittent outages have been critical factors from a performance perspective. In addition, operating cost (OPEX) and risk were also factors from the management and operational domain. The paper uses a survey strategy, focusing on a single case study of Nigeria’s telecommunication operating environment, with a mixed method of quantitative and qualitative techniques and a systematic review of related literature and documents on telecommunications and asset management. The chosen participants are field operations technicians and managers from the network operators, managed service companies and regulatory agencies. Data were collected through a designed online structured questionnaire and semi-structured face-to-face interviews and analysed through gap analysis protocol. Despite significant growth in telecommunication in the study area, the results indicated an impact of infrastructure and asset outages on critical issues: QoS network performance, cost pressure, and risk. The key identified factors affecting the network’s quality from the infrastructure and assets perspective are poor diesel management with 47.50% lack of actual diesel consumption measurement and visibility, and the outages caused by poor maintenance and monitoring practices at 15.70%. Real-time fault escalation increase mean-time-to-repair (MTTR). These infrastructures and asset problems, in turn, affect the QoS, which is a critical component of network availability. Addressing the power outages caused by poor diesel management to reduce MTTR is essential in resolving the cooling system’s 15.30% failures. These findings recognised the challenges subscribers and network operators face in an emerging operating environment and markets due to deficiencies in public grid infrastructures and systematised maintenance culture. This explains why network operators should focus on deploying intelligent and predictive approaches toward managing their infrastructure and assets. Furthermore, this recommended intelligent and predictive system integrates with human-centric intervention in addressing QoS issues. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 544–560, 2023. https://doi.org/10.1007/978-3-031-25448-2_51

Factors Affecting the Quality of Network Services

545

1 Introduction Increasing teledensity and customer value creation in emerging telecom markets requires innovative operations management and processes. This includes more flexible and costeffective operations management due to the cost pressure on operations and maintenance and the effect of human and environmental challenges on asset management activities. Public electricity (power grid) is a crucial issue in Nigeria’s operating environment that impacts telco infrastructure asset management regarding reliability, performance and operation cost. Over 75% of the telecom infrastructure base stations are off public electricity. These infrastructures operate mainly on alternating current diesel generators (ACDG), direct-current diesel generators (DCDG), and green or clean energy solutions like solar and batteries. The performance of these infrastructures and asset impacts the telecom network values, such as network availability and quality of services QoS. Despite the planned preventive maintenance cycles for these passive infrastructures and assets regarding the manufacturer’s recommendations, these infrastructures and assets break down intermittently, causing poor network values that impact customers’ needs, expectations, value-added services and dissatisfaction. Customer dissatisfaction is typically ascribed to poor network availability and quality of service. These infrastructures and assets’ intermittent outages also cause increased operating costs (OPEX) and operating risks associated with operational and environmental activities. Given the importance of sustaining and addressing the issues with the network quality of service in the research domain, our purpose is to explore the factors affecting network quality of service in Nigeria as an emerging telecom operating environment and market. This understanding explains how the quality of network services greatly influences customers’ perceptions, needs and expectations. It is, therefore, required to identify the factors affecting the QoS. In achieving this exploration, the paper focused on the technical and operating environmental areas that have not been researched in this context, such as infrastructure, asset management, and maintenance practice. Previous studies

Fig. 1. Operations management challenges to infrastructure performance and QoS.

546

C. Okeyia and N. M. de Almeida

have overlooked these critical domains, focusing on marketing, sales, promotions, and regulatory and government interventions (Lewis 1983). Figure one illustrates the effect on QoS from infrastructure and assets operations management challenges from the perspective of the research context. Indicating the interconnect between diversified operations (performance), OPEX and challenging operating environment (risk). These factors were explored to explain how they affect QoS. In the case of performance, cost and risk, infrastructure and asset management in the operating environment is a complex operation that should address the technical and non-technical activities to identify the factors affecting QoS.

2 Literature Review Despite several pieces of literature that supported the importance of infrastructure and asset management, almost no published research addresses infrastructure and asset management and maintenance activities that impact QoS in Nigeria’s telecommunication context. This understanding justified the broad divergence in the description of network quality from a production perspective as services without defects (Brady and Cronin 2001) and the concepts that included internal resources and service (Duggal et al. 2013). However, from a customer’s critical perspective, Parasuraman et al. (1988) described quality as a global attitude or judgement relating to the superiority of the service. In contrast, Lewis et al. (1983) defined quality of service as customers’ expectations of the performance attained from the services offered. These descriptions could be linked to the paper’s concerns about subscribers’ needs and expectations of the overall added value services. Although, the performance of service could be quantified and examined based on the tangibility of the activities from the infrastructure, asset management, and maintenance practices that provide the service. From the perspective of physical environment and infrastructure, Almossawi (2012) noted that internal company policies, service challenges, customer satisfaction and organisation position are attributes that affect QoS. However, much of the existing work on similar emerging operating environments and markets focused on applied technologies for powering infrastructure and assets, such as hybrid solar, batteries, and other green systems (Oviroh and Jen (2018). Other research works aim at consumer dissatisfaction (Opata 2013), connectivity and the digital ecosystem (Adame 2021). Although, various research works agreed that no contemporary telecommunication markets could be developed and continued short of an efficient telecoms infrastructure and service (ITU 2019; Vu et al. 2020). These existing research works did not focus on infrastructure, asset management, or maintenance roles in sustaining network values such as QoS and network availability. Additionally, the infrastructure and assets performance, cost and risk were explored to identify the factors impacting the network quality of services in this situation. Nakajima (1988) maintained that infrastructure performances are the measurement and identification of outages of critical business characteristics such as availability, performance and quality. This explains the understanding that network availability is expressed as the uptime and downtime of infrastructure (Kehinde et al. 2017). QoS is influenced by a better experience and performance of network availability. This perspective is from

Factors Affecting the Quality of Network Services

547

the asset management systems, asset networks and portfolios (De Azevedo 2019). The paper focuses on addressing infrastructure and asset problems as it affects the network quality of services in an emerging telecom operating environment and market. Most of these approaches and tools simulate the dynamic process of infrastructure maintenance over time, usually the asset lifecycle and explore various techniques to detect the difficulty and ineffectiveness of the maintenance strategy and propose likely areas for enhancement. Shafiee (2015) noted that each instrument or approach is created to describe one or several characteristics of the infrastructure activities and span various planning perspectives. Although some tools include optimisation features, other means are restricted to characterisation, usually assessing the agreed key performance indicators (KPIs). Figure 2 visually describes the mechanism, inputs, constraints and outputs traditionally considered in a telecoms maintenance simulation tool.

Fig. 2. Maintenance tool (Rinaldi et al. 2016)

The automation maintenance approach is critical in the telecoms infrastructure maintenance process because of the various factors and activities – erratic load imbalance, low generator frequencies, diesel line blockage and low level, alternator failure, and operation cost. However, this automation maintenance approach aims at each infrastructure element independently because of the assumption that downtime, faults, and degradation situations are independent. The mechanisms described the guidelines for the simulation of the maintenance activities. The inputs and constraints include the task descriptions for the individual method and the whole infrastructure assets maintenance (Rinaldi et al. 2016). The outputs offer a technical and cost-effective task evaluation during the simulated cycle. Finally, the detailed assessments of the maintenance simulation tools are described, examined and categorised based on the objective, attributes, working standard and fundamental

548

C. Okeyia and N. M. de Almeida

methodology. However, several of these maintenance models are adaptable and can certainly be tailored toward the automation of asset maintenance practices. On the other hand, the emphasis of this automation maintenance approach could be on the optimal planned inspection. In this manner, prolonged or early interventions, avoidable infrastructure interruption and high mttr are decreased, and sudden downtime is prevented. Therefore, if the infrastructure asset elements can be monitored constantly through the automation devices, intervening for maintenance and inspection could be likely only at discrete periods that distinguish between periodic and aperiodic decision instants (Olde Keizer et al. 2017). The automation maintenance approach involves several tasks such as alarm installation, data and information classification, and asset management activities – high-temperature faults, cooling alarms, power alarms, low diesel levels processing and decision-making. This paper’s simulation instruments and intelligence analysis are drawn from artificial intelligence (AI) systems. The AI simulates human intelligence processes through human, and machine learning collaboration termed hybrid intelligence (Kamar 2016; Dellermann et al. 2019). This action is achieved through automation like sensors, meters, and vision systems that monitor real-time system performance and collect data to manage uncertainties in infrastructure activities—automating data and real-time escalation of predictive outage saves human time, cost and better decision-making. In the current network quality of service context, opportunistic maintenance could be an extra in network management because it aims to classify maintenance activities of several elements to lower maintenance costs. This perspective is related to Zhao et al. (2019) condition-based maintenance (CBM) approach, which typically results in greater availability and reduces maintenance costs since it aims to avoid unplanned outages and prevent avoidable preventive maintenance activities for an infrastructure. However, the advantage of CBM remains uncertain in multi-infrastructure systems such as the telecom setting, where opportunistic maintenance strategies can be used. CBM could be cost-effective (Zhao et al. 2019). In contrast, monitoring diesel consumption and delivery, which contributes to the higher impact, may be problematic for these approaches; thus, AI and human-centric collaboration can be applied through automation and systematised maintenance practice.

3 Research Method The research method for this paper is a case study with a survey strategy that includes a quantitative technique structured questionnaire, a qualitative approach and a systematic review of documents from network operation centre (NOC) reports and operations and maintenance policies used as the primary data collection. The justification for using a case study survey approach is to capture a variety of viewpoints and the opportunity of using mixed techniques. Yin (2014) concluded that a mixed-method of quantitative and qualitative case study method offers the opportunity to gain a greater understanding of a contemporary phenomenon, generate hypotheses and reduce the possibility of any bias (Kezar 2002), thereby reducing the researcher and participants’ positionality. Indeed, variables in Appendix A are likely scales from their descriptions and are employed in relevant asset management research. The five-point Likert scale uses intervals ranging

Factors Affecting the Quality of Network Services

549

from 1 = strongly disagree to 5 = strongly agree based on Newell and Goldsmith’s (2001) work on the question that evaluates perceptions of the variables. Additionally, Saunders et al. (2016) suggest that the survey strategy is one of the most common methods in management research that is generally used to answer the where who, how and what research questions. This understanding is based on deductive reasoning, which begins with a theory and attempts to agree or disagree through quantitative techniques, and inductive reasoning, that the assumptions are derived from a particular phenomenon through a qualitative approach. The study participants are the employees of the telecom organisations responsible for the managed services, network operators and regulatory agencies. The judgment and area sampling techniques were used to administer the survey. This technique involves identifying the population within the studied operating environment that conforms to the criteria of high density with a focus on the telecoms infrastructure and asset management. All participants had various background knowledge and experiences in telecommunications operations and maintenance operations, regulatory and stakeholder management. The online structured questionnaire was created and distributed to participants involved in infrastructure and asset maintenance design and implementation. These questionnaire questions reflect the research issue. For instance, one of the key questions was. In what ways does infrastructure and asset performance improve the quality of service in your network? This question focuses on a specific issue with detailed and remarkable answers. At the same time, the semi-structured qualitative method was administrated to the managers in charge of the operations strategies, decision-making and stakeholder management. The qualitative questions focus on the participant’s experience, knowledge and views concerning the research problems. For instance, How does the quality of services been affected by infrastructure and assets outages in your network? These questions were designed and structured to generate conclusive and quantifiable data. The data were then analysed to understand the infrastructure and asset management process and how decisions are made in the physical context where the outages occur and affect QoS. At the same time, the paper used the gap analysis to assess both significance (S) and agreement (A) of the various infrastructure and asset management and maintenance activities. Identifying the gaps is likely to see the factors impacting the QoS. The critical area associated with the factors is the assertion, where the distinction between significance and agreement is the highest. The paper also reviewed related published journals on telecommunications and asset management in emerging markets. This search includes keywords such as passive infrastructure, network availability, QoS, asset management and preventive and corrective maintenance. This document search was conducted in vital academic databases such as Google Scholar, ProQuest, Scopus, Science Direct and JSTOR.

4 Results As mentioned early in this paper, not adequate attention has been given in the literature to explore the impact of different organisational scopes on infrastructure and asset management, performance, cost and risk. Therefore, this paper explores the positive effect

550

C. Okeyia and N. M. de Almeida

of infrastructure and asset performance, cost and risk on QoS. Accordingly, various subjects were conceptualised to design the questionnaire, with each question examined on the five-point Likert scale of 1 – strongly disagree and 5 – strongly agree. The structured questionnaire was responded to by the managed services companies (70.50%), the network operators 25.90% and the regulatory agency (3.60%). From available information, the size of the companies involved in this study – managed service companies responsible for infrastructure and asset management and maintenance (36.50%) of the sample that made up of 75 employees or less, (55.50%) were the network operators employed 100 employees and above and the regulatory agency (9.00%). This population show adequate content validity based on Gable and Wolf’s (1993) suggestion that the number of professionals needed for content validity is between two and twenty. Therefore, the raw data from Appendix A show that the relationship of diesel mismanagement 47.5% has a significant positive impact on infrastructure and asset management performance. On the contrary, as shown in Appendix B, the Janitorial service has no significant effect on infrastructure and asses, which also does not impact QoS. The results show intermittent outages of infrastructure and assets impact network availability and QoS and increase OPEX. The reason has been frequent site visits by field technicians from the point of high mean-time-to-repairs (MTTR), penalties for not meeting the service level agreement (SLAs) and travel time, and off-course operational risks to personnel and the organisation. By risk, the paper focuses on the MTTR caused by the downtime from the infrastructure and asset failures and not only the risk caused by the improper planned maintenance of the infrastructures and assets. Therefore, the result was based on addressing the impact of infrastructure and asset performance, OPEX and risk on QoS. The factors concerning performance were created based on the participants’ response and their reviewed faults reports and documents. At the same time, the OPEX and risk were mainly from the available records and reports. 4.1 Summary of the Main Results The summary of the key findings was outlined accordingly. First, to understand how network values such as network availability and QoS are affected by infrastructure and asset outages and performance, OPEX and risk. Secondly, the data shows that intermittent outages of the infrastructure and assets affect the network quality of services despite the planned maintenance cycles and diesel management of the infrastructure and assets. Other factors affecting network service quality are ageing infrastructure, assets, and security threats. Table 1 present the summary of the main results of this paper. We found that 47% impact on the quality of the network services is because of infrastructure and assets failures caused to diesel mismanagement, resulting from incorrect diesel allocation, quality and quantity consumption, poor monitoring of diesel supplied to sites, faulty generator counters and no intelligent system to confirm actual supply and consumption. This concern about poor diesel management and the exact quantity consumed creates an increase in OPEX and outages that affect QoS and overall performance.

Factors Affecting the Quality of Network Services

551

Table 1. Main results Infrastructure performance tasks

Share (%)

Diesel management

47.50%

Planned preventive maintenance (Mtce)

15.70%

Cooling systems maintenance

15.30%

Battery backup maintenance

11.50%

Janitorial and tower maintenance

10.00%

Other critical factors were planned preventive maintenance 15.70% and cooling systems maintenance15.30% of activities, causing the frequent outages. In addition, the generator elements and engine oil are not replaced at the appropriate cycle according to the design specifications. Finally, we found an oversight and delay in battery backup mtce shows 11.50%, such as adding electrolytes or distilled water in the battery compartment and non-checking of loose-fitting cables, terminals, or plugs impact the batteries’ backup performance when an outage occurs. We also discovered that janitorial and tower maintenance represents a 10% factor, as indicated in Appendix B, which does not impact asset performance and are non-traffic affecting tasks. This means that they do not directly affect the quality of network service. Appendix C indicates outages that are traffic-affecting and impact QoS due to unreliable infrastructure and asset performance. The paper also found factors such as lacking real-time visibility of the technical and environmental activities that affect QoS, from the point of not predicting the outages before it occurs and resolving issues remotely before escalating to the field technicians to visit the site physically. Furthermore, not monitoring the maintenance activities remotely to ensure materials were appropriately replaced and actual diesel quantity delivered to the site were also factors that affected QoS through port network availability.

5 Discussion This paper supports the assumptions establishing intelligent and human collaboration by telecom operators in emerging operating environments and markets. This action will benefit them by executing appropriate infrastructure, asset operations, and maintenance practices through real-time applications that will address performance, cost and risk issues in their operations. Several simulation instruments and approaches optimise the maintenance strategies for asset management support. The existing infrastructure operations management is reactive, time-consuming, and not responsive or innovative enough to address network operations and maintenance challenges. Current operational and asset management activities rely greatly on human interventions, affecting OPEX. These routine and non-routine operations and maintenance costs and activities represent over 70% of the total OPEX cost.

552

C. Okeyia and N. M. de Almeida

The level of the existing automation in the study context is mainly low. For instance, the integration between various power redundancy procedures (public grids/diesel generators 1 or 2) is not swiftly and intelligent. Due to the more complex operating environments, such as across rivers, standalone or road coverage sites, that trigger network availability issues such that manual integration increases mean-time-to-repair (MTTR). This high mttr explains the problem with network availability which in turn affects QoS. Analysing and interpreting the evidence in Sect. 4 and Subsect. 4.1 indicated a leading gap between real-time escalation to outages that create high MTTR. Diesel mismanagement, which included spillage, undersupply, theft and adulteration, represented 47.50%. The no visibility of infrastructure and asset outages, faults and activities affected the poor diesel management and explained the impact of diesel outages on performance, cost and risk. Table 2 shows the gap analysis results of the average performance scores, significance scores and gaps. Table 2. Gap analysis of infrastructure and asset activities. Management and maintenance activities

Agreement (A) Significance (S) Gap

Diesel management (Daily and weekly consumption 3.2

4.8

0.6

Conduct monthly planned preventive maintenance

2.6

4.6

0.5

Cooling systems maintenance (Monthly)

2.8

3.6

0.8

Janitorial services (Monthly)

2

3.2

1.2

Spare management (Holding levels)

1.8

2

0.9

Tower maintenance

2

2

1

Total asset spot checks (Weekly and monthly)

3

3.6

0.6

From the gap analysis, the leading gap relates to janitorial services, which are not traffic affected, which means that it does not affect QoS. This is understandable based on the operating environment because field technicians do not see this activity impacting their performance, and often have been carried out by their host communities as support. However, the leading gap that affects QoS (performance is diesel management and planned preventive maintenance. The diesel mismanagement is a pivotal setback to the business and customers because of loss of revenue, penalties, and poor network values and subscribers’ experiences. Thus, the gap between the agreement and significance suggests the prospective concern for the business. Likewise, monitoring whether the spares are replaced or reused with the correct quantity and quality. These problems could be resolved by intelligent and predictive systems that would assist network operators in having visibility, enhancing performance, and reducing OPEX and risk. This is because the management and maintenance practice influences the infrastructure and asset performance, improving the QoS that addresses customer satisfaction and expectations.

Factors Affecting the Quality of Network Services

553

6 Conclusions and Recommendations The originality of this paper has shown that passive infrastructure, asset management, and maintenance are the key factors that impact network quality of services. Improved QoS cannot be achieved without the stability of the infrastructure and asset performance. Therefore, network operators should focus more on infrastructure and asset management activities towards addressing factors affecting QoS by integrating intelligence and a human-centric approach to optimise network values as a gateway for the future in sustaining better QoS. For network operators to address and achieve these QoS problems, a systematised management and maintenance practice that involves an intelligent and humancentric approach should be adopted and implemented to influence the infrastructure and asset performance, network availability and quality of services. Furthermore, the recommended management and maintenance system should be constantly revised using real-time fault data colle, prediction, amplification, and evaluation to improve maintenance decision-making based on existing situations and limitations. This is because quality management and efficient maintenance approaches are crucial predictors of quality of service. This perspective supports the significance of integrating an improved systematic – intelligent, and human-centric - maintenance management approach that is hybrid. Consequently, in exploring this identified problem of the intermittent outages of the infrastructure and assets, their operations and maintenance strategies need to meet the requirements of their potential network value. Therefore, innovative and knowledgedriven operation and maintenance processes should address the problem with operational reliability (performance), risk management, and operations and maintenance to ensure network availability, quality of services and infrastructure performance. Additionally, implementing intelligent and human-centric strategies in infrastructure and asset management and maintenance practices in this context helped network operators focus on addressing and stabilising network quality of service in the future. This conclusion explains that adopting intelligence and human-centric operational concepts entailed practical automation of the technical and non-technical procedures that predict real-time and remotely resolve issues. Besides, a multi-disciplinary method between intelligence and human-machine interface through monitoring and control was necessary to address the network quality of service caused by poor network availability due to infrastructure and asset failures caused by poor maintenance, delay in faults detections, resolution and human development. This paper is to be further developed by the principal author of this paper, focusing on Off-Grid Passive Telecommunication Infrastructure in Emerging Market: Efficient and Cost-Reduced in Asset Management Solutions for Sustainable Network Value. Further Reading. International Telecommunication Union (ITU) (2017), “Key ICT indicators for developed and developing countries and the world”. [Online] Available at: www.itu.int/en/ITU-D/Statistics/Pages/facts/default.aspx (Accessed 10 April 2022).

554

C. Okeyia and N. M. de Almeida

Appendix A Measurement Scales - Key Maintenance Performance (KPIs) Participants were asked to indicate how much significance is assigned to each of the following events or activities where 1 - strongly disagreed and 5 - strongly agree. Overall infrastructure performance: Power availability; Mean time to repair (MTTR); Planned vs Actual maintenance percentage and Infrastructure failure rates. Assessment of Network Quality of Services 1. Infrastructure Maintenance Practices Participants were asked to indicate how much significance is assigned to each of the following events or activities where 1 - strongly disagreed and 5 - strongly agree. IMP 1: We entrench maintenance into activities that could impact infrastructure performance.

1

2

3

4

5

Strongly disagree

Strongly agree

IMP 2: We consider remote management systems through monitoring tools.

1

2

3

4

5

Strongly disagree

Strongly agree

IMP 3: We examine maintenance practices, operations, and quality of services.

1

2

3

4

5

Strongly disagree

Strongly agree

IMP 4: We implement a quality assessment to minimise poor infrastructure maintenance practices.

1

2

3

4

5

Strongly disagree

Strongly agree

IMP 5: Maintenance process and strategy are incorporated components of infrastructure performance.

1 Strongly disagree

2

3

4

5 Strongly agree

IMP 6: The percentage of network availability has decreased during the past periods.

Factors Affecting the Quality of Network Services 1

2

3

4

555

5

Strongly disagree

Strongly agree

IMP 7: Mean-time-to-repair has increased during the past periods.

1

2

3

4

5

Strongly disagree

Strongly agree

IMP 8. In what ways does infrastructure and asset performance improve the quality of service in your network.

1

2

3

4

5

Strongly disagree

Strongly agree

2. Performance of Asset Management Activities Participants were asked to specify how much significance is assigned to each of the following events or activities where 1 – strongly disagreed, and 5 - strongly agree. PAM 1: We use asset records to improve asset understanding.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 2: We frequently assess the total effectiveness of asset management activities.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 3: We embark on benchmarking to support asset management activities.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 4: We monitor the key performance indicators to confirm the accomplishment of the organisation’s asset management objectives.

1 Strongly disagree

2

3

4

5 Strongly agree

PAM 5: We proactively pursue infrastructure and asset maintenance improvement of asset management activities.

556

C. Okeyia and N. M. de Almeida 1

2

3

4

5

Strongly disagree

Strongly agree

PAM 6: The organisation gathers and examines infrastructure and asset management data.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 7: We often examine the general effectiveness of infrastructure and asset management activities.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 8: We use remote management systems to support asset management activities.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 9: We examine the condition of critical assets.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 10: We assess organisations’ asset management performance.

1

2

3

4

5

Strongly disagree

Strongly agree

PAM 11: How does the quality of services been affected by infrastructure and assets breakdown in your network?

1 Strongly disagree

2

3

4

5 Strongly agree

PAM 12: How are asset management and maintenance affecting network availability and network quality of service?

Factors Affecting the Quality of Network Services 1

2

3

4

557

5

Strongly disagree

Strongly agree

3. Infrastructure Lifecycle Management Participants were asked to specify how much significance is assigned to each of the following events or activities where 1 - strongly disagreed, and 5 - strongly agree. ILM 1: We constantly modernise our infrastructure following business plans or evolving technology.

1

2

3

4

5

Strongly disagree

Strongly agree

ILM 2: We constantly enhance our infrastructures to increase network availability.

1

2

3

4

5

Strongly disagree

Strongly agree

ILM 3: We ensure the quality of the infrastructure’s process and implementation during the complete lifecycle stages.

1

2

3

4

5

Strongly disagree

Strongly agree

ILM 4: We ensure the implementation of maintenance practices within all lifecycle stages.

1

2

3

4

5

Strongly disagree

Strongly agree

ILM 5: We implement disposal of infrastructure in agreement with the environmentally friendly policy.

1

2

3

4

5

Strongly disagree

Strongly agree

ILM 6: We seek internal and external resources to acquire new knowledge and information about customers, technology and partners.

1 Strongly disagree

2

3

4

5 Strongly agree

558

C. Okeyia and N. M. de Almeida

ILM 7: The effectiveness of diesel consumption management has improved during the haulage cycles.

1

2

3

4

5

Strongly disagree

Strongly agree

ILM 8: Resource functioning such as public grid/electricity, hybrid and solar systems has increased in utilisations.

1

2

3

4

5

Strongly disagree

Strongly agree

Appendix B

Data from the Gap Analysis

Factors Affecting the Quality of Network Services

559

Appendix C

Downtime and Outages defects

References Adame, B.: The Ethiopian telecom industry: gaps and recommendations towards meaningful connectivity and a thriving digital ecosystem (2021). https://doi.org/10.1016/j.heliyon.2021. e08146. Accessed 23 Mar 2022 Almossawi, M.M.: Customer satisfaction in the mobile telecom industry in Bahrain: antecedents and consequences. Int. J. Mark. Stud. 4(6), 139–156 (2012) Barad, M.: Total quality in management. In: Warner, M. (ed.) International Encyclopaedia of Business and Management. Thomson Business Press, London (1996) Brady, M.K., Cronin, J.J.: Some new thoughts on conceptualising perceived service quality: a hierarchical approach. J. Mark. 65, 34–49 (2001) Dellermann, D., Ebel, P., Sollner, M., Leimeister, J.: Hybrid intelligence. Bus. Inf. Syst. Eng. 61, 1–7 (2019). https://doi.org/10.10007/s12599-019-00595-2. Accessed 20 Aug 2022 De Azevedo, C.: Asset Management Insights. Phases, Practices, and Value. Industrial Press, Inc. (2019) Duggal, E., Verna, H.: Service quality: construct comprehension and evolution over time. J. Serv. Res., 135 (2013) Gable, R.K., Wolf, J.W.: Instrument Development in the Effective Domain: Measuring Attitudes and Values in Corporate and School Settings. Kluwer Academic, Boston, Netherland (1993) ITU: Economic Impact of Broadband in LDCs, LLDCs and SIDS, an Empirical Study, Geneva, Switzerland (2019) Kamar, E.: Directions in hybrid intelligence: complementing AI systems with human intelligence. In: Proceedings of the International Joint Conference on Artificial Intelligence (2016) Kehinde, A., Lawan, S., Adunola, F., Isaac, A.: GSM quality of service performance in Abuja, Nigeria. Int. J. Comput. Sci. Eng. Appl. (IJCSEA) 7(3/4) (2017) Kezar, A.: Reconstructing static images of leadership: an application of positionality theory. J. Leadersh. Stud. 8(3), 94–109 (2002)

560

C. Okeyia and N. M. de Almeida

King, N.: Using interviews in qualitative research. In: Cassell, C., Symon, G. (eds.) Essential Guide to Qualitative Methods in Organizational Research, pp. 11–22. Sage, London (2004) Lewis, R., Booms, H.: The marketing aspects of service quality. In: Berry, L., Shostack, G., Upah, G. (eds.) Emerging perspectives on services marketing, pp. 99–107. American Marketing, Chicago (1983) Nakajima, S.: Introduction to TPM. Productivity Press, Cambridge (1988) Newell, S.J., Goldsmith, R.E.: The development of a scale to measure perceived corporate credibility. J. Bus. Res. 52(3), 235–247 (2001) Olde Keizer, M.C.A., Teunter, R.H., Veldman, J., Babai, M.Z.: Condition-based maintenance for systems with economic dependence and load sharing. Int. J. Prod. Econ., 319–27 (2018). https:// doi.org/10.1016/j.ijpe.2017.10.030 Opata, C.: The curious case of consumer dissatisfaction in Nigerian telecommunications sector (2013). https://ssrn.com/abstract=2356804, https://doi.org/10.2139/ssrn.2356804. Accessed 24 July 2022 Opata, C.B.: Sustainable development and rural access to telecommunications in Nigeria. In: Onwumechili, C., Ndolo, I.S. (ed.) Re-Imagining Development Communication in Africa, pp. 225–244. Lexington Books, Maryland (2013) Oviroh P.O., Jen, T.: The energy cost analysis of hybrid and diesel generators in powering selected base transceiver station locations in Nigeria. Energies 11, 1e20 (2018). https://doi.org/10.3390/ en11030687. Accessed 14 July 2022 Parasuraman, A., Zeithaml, V.A., Berry, L.L.: SERVQUAL: a multiple-item scale for measuring consumer perceptions of service quality. J. Retail. 64, 12–40 (1988) Peng, X., Scott, R., Prybutok, V., Sidorova, A.: Product quality vs service quality in the mobile industry: is there a dominant driver of customer intention to switch providers? Oper. Manag. Res. 7(3/4), 63–76 (2014) Rinaldi, G., Thies, P.R., Johanning, L., Walker, R.T.: A computational tool for the proactive management of offshore farms. In: Proceedings of the 2nd International Conference Offshore Renewable Energy, Glasgow, UK, 19–20, pp. 111–115 (2016) Saunders, M., Lewis, P., Thornhill, A.: Business Research for Business Students. Pearson Education, Essex (2016) Shafiee, M.: Maintenance logistics organisation for offshore wind energy: current progress and future perspectives. Renew. Energy 77, 182–193 (2015) Shafiee, M.: Maintenance strategy selection problem: an MCDM overview. J. Qual. Maint. Eng. 21, 378–402 (2015). https://doi.org/10.1108/JQME-09-2013-0063 Strauss, A., Corbin, J.: Basics of Qualitative Research, 2nd edn. Thousand Oaks, CA, Sage (1998) Vu, K., Hanafizadeh, P., Bohlin, E.: ICT as a driver of economic growth: a survey of the literature and directions for future research. Telecommun. Pol. 44(2), 101922 (2020) Wang, H.: A survey of maintenance policies of deteriorating systems. Eur. J. Oper. Res. (2002). https://doi.org/10.1016/S0377-2217(01)00197-7 Van der Wal, R.W.E., Pampallis, A., Bond, C.: Service quality in a cellular telecommunications company: a South African experience. Manag. Serv. Qual. Int. J. 12(5), 323–335 (2002) Yin, R.K.: Case Study Research Design and Methods, 5th edn., 282 p. Sage, Thousand Oaks (2014). https://doi.org/10.3138/cjpe.30.1.108 Zhao, H, Xu, F., Liang, B., Zhang, J., Song, P.: A condition-based opportunistic maintenance strategy for multi-component system. Struct. Health Monit. 18(1), 270–283 (2019). https://doi. org/10.1177/1475921717751871. Accessed 03 July 2022

Explaining Underlying Causes for the Degradation of Handover Information for Commercial Building Owners Janet Chang(B) , Jorge Merino Garcia, Xiang Xie, Nicola Moretti, and Ajith Parlikad University of Cambridge, Cambridge, UK {jc2019,jm2210,xx809,nm737,aknp2}@cam.ac.uk

Abstract. Evaluating possible causes of building handover information degradation is critical to asset owners because the handover information is the foundation for generating a thread of reliable asset information to minimise risks associated with operating commercial buildings. However, very little is known about possible explanations of information degradation, particularly during the operation phase of commercial buildings. This study, therefore, investigates the reasons for the quality of handover information deterioration by conducting semi-structured interviews with asset management professionals. Given that the handover information supports various asset management processes, fragmented and incoherent processes of managing different types of handover information during the operation phase are one of the leading causes of diminishing the quality of information. Moreover, leadership support is crucial to establishing robust information management processes with dedicated resources to sustain information quality. The contributions of this study are twofold. First, on a practical level, the identified causes of information degradation enable asset owners to inform decisions on improving existing information management processes and feasibly leverage emerging technologies like cloud computing to address the information management dilemma. Theoretically, this study contributes to the body of knowledge on information quality by identifying possible rationales for the quality degradation of handover information. Finally, this study yields evidence for seeking possible remedies and future research topics.

1 Introduction Quality handover information is pivotal for managing complex-built assets like commercial buildings. Reliable handover information provides guidance and instructions for the required routine maintenance at the operational level. For effective asset management, credible handover information is indispensable for developing a recurring investment plan because up to 85% of the total investment of a building occurs during the operation phase (Thabet and Lucas 2017a). At the strategic level, valid handover information helps establish long-term goals for development while optimising the use of buildings to minimise potential environmental impacts (Roberts et al. 2018). Additionally, the recent investigation of the Grenfell Tower fire showed that accurately updated and readily accessible handover information must be compulsory for managing risks involved in © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 561–570, 2023. https://doi.org/10.1007/978-3-031-25448-2_52

562

J. Chang et al.

building operations, notably fire-, life-, and safety-related issues (Hackitt 2018). Despite the vital role of handover information, the building industry has accepted ineffective ways of managing handover information as a norm. However, limited knowledge in this area has neither examined nor explained information degradation, its causes, and its sources. The recent adoption of BIM in the construction industry provides immense potential for generating quality handover information, but its ubiquitously cited benefits still require extensive implementations to support amorphous handover information management during the operation phase (Zadeh et al. 2017). Considering the negative impacts of poor-quality information on decision-making support, Zadeh et al. (2017) confirmed asset information deficiencies in BIM-based projects that possibly affect the quality of handover information. Moreover, Sadeghi et al. (2019) proposed a BIM-enable workflow to identify handover information requirements and its management, in addition to the stoic ongoing effort to gain BIM’s tangible information management benefits. Although various studies evaluated different methods of improving the quality of handover information at different phases of the lifecycle, an absence of research investigating possible causes of the quality of handover information decline during the operation phase continues to hinder informing effective asset management decisions. This study, therefore, aims to conduct an exploratory investigation to identify the probable reasonings for degrading the handover information, particularly during the operation phase. For this study, the pertaining building handover information includes but is not limited to: (1) as-built drawings, (2) a list of the installed products, (3) updated health and safety files, (4) O&M manuals, (5) warranties, and (6) testing and commissioning report. With semi-structured interviews with existing asset management professionals, this study systematically performed a two-way thematic analysis to identify the underlying reasons for information erosion. The two-way thematic analysis involves developing a list of first-order descriptive codes based on the participant’s frequently used terms and phrases during the interviews. Subsequently, the first-order categories are further filtered through an inductive coding process to generate theoretical second-order themes. As a result, this study illuminates a comprehensive understanding and deeper insights into the phenomenon of information quality decay for establishing appropriate strategies for controlling the quality of handover information. Further, this study elucidates future research opportunities in handover information management.

2 Literature Review This section discusses the relevant studies in managing handover information for postconstruction support. The literature review includes academic journals, conference papers, industry publications and standards, focusing on handover information and its desired information quality in the asset management domain. Asset management is the coordinated activity of an organisation to realise value from tangible and non-tangible assets (The International Organization for Standardization 2014). ISO 55000 further clarifies that asset management includes planning and associated implementation to balance the cost of ownership, optimisation of asset performance, and potential risks related to operating the asset while achieving the organisation’s business goals. Within the boundaries of this definition, successful asset management heavily

Explaining Underlying Causes for the Degradation

563

relies on credible information to make informed decisions surrounding assets. In this context, handover information includes a mixture of graphical (e.g., as-built drawings, photos, etc.) and non-graphical information (e.g., product data, warranty certificates, health and safety reports, etc.) about assets originating from the handover information from the construction phase (The Institute of Asset Management 2015; Mayo and Issa 2016). In addition, a blend of different types of handover information and operational data support various asset management activities; therefore, it is in the asset owner’s best interest to comprehensively manage accurately represented handover information. However, the challenges of managing the handover information arise due to the complex nature of the construction project generating asset information causing incompatibilities and inconsistency (Sacks et al. 2018). The Hackett report recently illuminated the significance of reliable handover information through the Building Safety Act in the UK. Understanding the vital position of handover information, East and Nisbet (2010) revealed that approximately thirty (30) per cent of inaccurate information in the handed-over information impedes decision-making processes. Additionally, Sadeghi et al. (2019) argued that information loss is the leading cause of the interior quality of information. In supporting this assertion, Eastman et al. (2011) remarked that a large volume of essential information is lost during the handover phase, consuming additional resources to augment the inferior quality of information to the usable levels. Moreover, Bayar et al. (2016) and Cavka and Poirier (2017) both contended that file conversion degrades the quality. Following their studies, Bayar et al. (2016) concluded that using multiple formats can generate inadequate and fragmented information, resulting in quality doubt in the handover information. However, the results of these studies vaguely discussed the significance of information quality degradation during the operation phase. A recent study discovered that the BIM approach projects suffer from design deficiencies, negatively impacting the quality of the handover information. Thabet and Lucas (2017b) explored adopting a BIM-based facility management approach to minimise data loss during the handover phase and enhance information quality. However, the proposed information-gathering model misrepresented multi-dimensional information flow during the project phase, which would result in the collection of inaccurate handover information. Finally, Bayar et al. (2016) argued that infrequent asset condition assessment, using multiple information formats, legacy data, and employee turnover causing loss of tacit knowledge also diminish handover information quality. The same authors, therefore, suggested that supplementary studies are needed to understand the possible causes of handover information degradation to inhibit probable quality decay.

3 Methodology This section describes the methodology adopted to probe the significance that triggers degrading the quality of handover information, particularly during the operation phase, in three parts: (1) sampling, (2) data collection, and (3) data analysis.

564

J. Chang et al.

3.1 Sampling Exploring the causes of handover information degradation requires a rich and realworld context that cannot be emulated in an artificial setting (Creswell 2014). This study, therefore, adopted the purposive sampling technique to recruit asset management professionals who regularly use the handover information for their job duties in similar asset management settings to obtain fruitful outcomes (Saunders et al. 2019; EasterbySmith et al. 2021). This study intentionally excluded participants with less than five (5) years of working experience because the study by Schmidt et al. (1988) discovered a strong relationship between the years of working experience and job knowledge, culminating in the fifth year. The main goal of exploiting this selection criterion is to obtain the best answers to the topic of study while gaining in-depth knowledge through each participant’s lived experience (Creswell 2014; Easterby-Smith et al. 2021). The selected participants comprised asset managers, senior management, building service managers and asset information managers, mainly in the following disciplines: (1) fire alarm systems and the related ancillary elements such as lifts, (2) electrical systems, (3) HVAC (Heating, ventilation, and air conditioning) systems including boilers, (4) space management, and (5) handover information management. Participants’ experience in the identified disciplines ranged from five (5) to twenty-three (23) years, and most have prior knowledge from their past professional experience. Table 1 exhibits the summary of participant rates. Table 1. Distribution of participant rates

3.2 Data Collection Considering the potential shortcomings of the survey, which may lack important details, this study chose semi-structured interviews over surveys to collect the relevant data to meet the study’s objectives. Interviews were deemed appropriate because lengthy, semi-structured interviews help capture embedded professional insights stored as tacit knowledge about reasons for the degradation of handover information during the operation phase (Pirila 2021). At the same time, interviews frequently encourage participants to share information voluntarily while producing and articulating unintended insights

Explaining Underlying Causes for the Degradation

565

and perceptions that emerge from discussions (Hoffmann 2007). Interviews were conducted until saturation was reached, meaning no additional new information surfaced from participants (Fusch and Ness 2015). This study reached saturation after interviewing twenty-five (25) participants from five organisations. This study assumed a strategy for categorising the interview questions to extract more specific data from each participant (Vermunt et al. 2019). The interview questions were predefined into five categories: (1) characteristics of assets, (2) processes of using the handover information, (3) the desired information quality attributes of different types of handover information, (4) use of asset management tools to manage handover information, (5) participant’s rationales for information degradation during the operation phase. The interviews consisted of open-ended questions, offering opportunities to gain participants’ first-hand experience and specific knowledge of their understanding of ‘why’ handover information degrades during the operation phase in organic ways. During the interview, the participants could freely use their previous professional experiences to explain certain events that they perceived relevant. At the same time, they have the option to decline to respond to any questions they deem sensitive. The average length of the interview process was approximately sixty (60) minutes. Each interview was recorded and transcribed with the participant’s permission. 3.3 Data Analysis In analysing the interview data, this study followed the recommended steps of qualitative thematic analysis suggested by Gioia, Corley and Hamilton (2013). The data analysis followed the steps of creating the structure of findings to analyse the results systematically. Firstly, the participant’s descriptive quotes and expressions describing apparent explanations for the information degradation were extracted from the interviews to generate 200 attributes. Then, the identified attributes were clustered to develop six groups of the first-order findings after comparing and analysing similarities and differences. Sequentially, the evaluation of the first-order classifications was further clustered into theoretical second-order themes, which served as the basis for the emergent theory on the underlying causes of the handover information decay during the operation phase. Finally, the emergent second-order themes were refined into second-order aggregate dimensions for discussion. Six (6) first-order categories were identified through the previously mentioned process: (1) fragmented handover information management processes, (2) irregular formats and non-conforming naming conventions, (3) poor quality information, (4) adoption of multiple technological solutions, (5) lack of resources for managing handover information, and (6) no value realisation of handover information. The first-order findings were further clustered into broader themes to improve the understanding of the reasonings for the quality decline of handover information during the operation phase. These themes include lack of data governance, consequences of using multiple technological solutions, and inadequate management support. The systematic process of summarising and evaluating emergent themes helped finalize rich textual interview data into compelling findings. Thus, the following section presents the results of the interviews by interpreting the collected data of the participants from the descriptive narrative formats to conclusive information that could enhance

566

J. Chang et al.

the understanding of various causes that trigger degrading the quality of handover information during the operation phase.

4 Results The following sections discuss the findings generated from the qualitative thematic analysis with supporting evidence from the interviews. Direct quotes were redacted and used anonymously to maintain confidentiality. As previously indicated, the results show three significant themes that degrade the quality of handover information: (1) lack of data governance, (2) consequences of using technological solutions, and (3) inadequate management support in managing handover information. Figure 1 illustrates the summary of the interview findings.

Fig. 1. The data structure of the thematic analysis

4.1 Lack of Data Governance The findings suggested that a lack of data governance produces inferior quality information inducing patchy handover management processes during the operation phase. According to the interviews, many formats are utilised to assemble a complete set of handover information, and converting diverse types of handover information into valuable formats increases the chances of losing helpful information. Various formats include but are not limited to PDFs, video clips, 3D models, CAD files, etc. While irregular information formats can vary from project to project, one participant illustrates how information can be lost during the conversion. For example, converting 3D models to PDFs could discard embedded information in objects, and reverting the PDF with the updated information to the 3D model will not automatically capture the updated information. Many underscored the importance of maintaining a consistent and accurate naming convention of buildings and sites for user safety and property insurance policy renewals,

Explaining Underlying Causes for the Degradation

567

along with the potential information loss. Finally, the participants argued that the accuracy of handover information alone is insufficient to support the complicated nature of commercial building operations. Completeness, timeliness, consistency, and availability are the desired quality dimensions in addition to accuracy. Accordingly, data governance establishing data policies and standards is necessary to manage the handover information that meets the preferred quality dimensions. 4.2 Consequences of Using Technological Solutions While adopting technological solutions is prevalent in managing various asset information, including handover information, several participants arguably discussed the consequences of employing technological solutions as part of the rationale for deteriorating the quality of information. Firstly, utilising new technological solutions without polemic clarity on the projected use and the associated processes often initiates the information quality decline. Secondly, expanding handover information types may require adopting additional software, potentially causing interoperability issues among the systems and incremental storage requirements. Thirdly, using outdated software and tools poses hindrances and obstacles in managing quality handover information because information types, formats and storage capacities are evolving according to technological advancement. At the same time, the recent BIM adoption produces non-traditional handover information that can be one of the reasons for the handover information quality decline as many organisations continue to manipulate legacy and non-traditional asset information to fit their needs. Finally, the proliferation of disruptive technologies such as BIM provides an opportunity to bridge the inefficient practice of managing handover information. Still, the participants expressed that a shortage of BIM experts and a lack of devoted technology support result from adopting technological solutions, contributing to creating quality concerns. 4.3 Inadequate Management Support The interviews suggested insufficient resources contribute to decaying the quality of information. The participants, however, argued their view of needing dedicated resources for managing the handover information. According to the participants, handover information still requires error-prone manual processes despite the digitalisation effort in the construction industry. Handover information is updated in various locations by multiple parties whose primary job responsibilities are not managing the handover information. As a result, inconsistent processes of managing the handover information are frequently formed, resulting in substandard quality of the handover information at the beginning of the operation phase of a building. Many participants suggested that centralised support would eliminate the possibility of mistakes by multiple hands. Secondly, the participants remarked that the timeline of receiving the handover information could also influence the quality of the handover information. According to the participants, it could take up to two years to receive handover information for a new building because the contractual due date of the handover information is unclearly stated in the contract. Meanwhile, minor projects are already in motion, requiring a coordination effort to incorporate minor project information into the final handover. Lastly, inconsistent and fragmented handover

568

J. Chang et al.

management processes and utilising multiple asset information management systems need committed assistance to prevent the quality decay of handover information. The interview revealed that up to twelve different asset information management systems are employed to manage the complex process of updating the handover information, even for a simple project. After analysing 200 attributes, the study discovered the fundamental causes of information degradation during the operation phase. The discussion of these findings will be included in the subsequent section.

5 Discussion The data structure of the interview findings uncovers two (2) fundamental causes of handover information degradation: (1) a lack of clarity on the intended use of handover information and (2) inadequate leadership support. This study uncovered a vague understanding of the projected use of handover information generate sub-standard information, causing flawed and incomplete information management processes. This finding alone suggested that a clear understanding of the utility of handover information is compulsory for establishing data governance, improving existing information management processes, and optimizing the size of asset information management systems. One way to clarify the usage of the information can be by leveraging the existing statutory requirements for commercial building operations to classify different kinds of handover information because of the constant use of the information. Similarly, this approach can be utilised further to create information requirements for future projects. The discussion of insufficient leadership support influencing the quality of handover information was explicitly indicated during the interviews. This suggested that top-down leadership support, one of the guiding principles of asset management, is obligatory for preventing quality deterioration, which is essential for alignment with asset management activities. Asset information management is a nonlinearly compounded process, requiring an optimal blend of technological solutions, appropriate processes, quality information and leadership support. However, the allocation of scarce resources stems from a lack of recognising of the value associated with managing handover information. Adopting technological solutions is inevitable; however, utilising such an approach does not resolve the concerns regarding information quality support. This study also revealed that emerging technologies require additional resources, such as experts, to take full advantage of the innovation.

6 Conclusion The study assessed probable causes for the degradation of handover information during the operation phase. This study produced six (6) categories of findings based on the 200 attributes collected from the interviews, identifying two ultimate reasons for the information deterioration. An unexpected finding is that a lack of clarity on the intended use of the handover information led to subsequent activities that triggered handover information degradation instead of key events like renovations. Although the quality of handover information is hinged upon the information gathering in the construction

Explaining Underlying Causes for the Degradation

569

phase, the study findings present opportunities for future studies on this topic. Using the findings as a basis, asset owners can re-evaluate the use of different types of handover information based on the compulsory requirements of commercial building operations to understand the role of different types of handover information instead of receiving a pool of generic asset information. With this view, the study findings can also be used to create trustworthy processes and data environments for asset management professionals to make plausible decisions based on reliable information. Acknowledgement. The research leading to these results has been fully funded by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 860555.

References Bayar, M., et al.: Optimizing handover of as-built data using BIM for highways. In: 1st International (UK) BIM Academic Forum Conference, Glasgow, Scotland (2016) Cavka, H., Staub-French, S., Poirier, E.: Developing owner information requirements for BIMenabled project delivery and asset management. Autom. Constr. 83, 169–183 (2017) Creswell, J.: Research Design: Qualitative, Quantitative, and Mixed Methods Approaches, 4th edn. SAGE, Thousand Oaks, CA (2014) East, E., Nisbet, N.: Analysis of life-cycle information exchange. In: International Conference on Computing in Civil and Building Engineering (2010) Easterby-Smith, M., et al.: Management and Business Research. SAGE, Los Angeles, (2021) Eastman, C., et al.: BIM handbook: A Guide to Building Information Modeling for Owners, Managers, Designers, Engineers and Contractors. John Wiley & Sons Inc (2011) Fusch, P., Ness, L.: Are we there yet? Data saturation in qualitative research. Walden Faculty and Staff Publications, 455 (2015) Gioia, D., Corley, K., Hamilton, A.: Seeking qualitative rigor in inductive research: notes on the gioia methodology. Organ. Res. Methods 16(1), 15–31. (2013). Available at: https://doi.org/10. 1177/1094428112452151 Hackitt, J.: Building a Safer Future Independent Review of Building Regulations and Fire Safety: Final Report. U.K. Government (2018) Hoffmann, E.: Open-ended interviews, power, and emotional labor. J. Contemp. Ethnogr. 36(3), 318–346 (2007) Mayo, G., Issa, R.: Nongeometric building information needs assessment for facilities management. J. Manag. Eng. 32(3), 04015054 (2016) Pirila, S.: The client’s problem space, construction, and Artificial Intelligence. University of Cambridge, Cambridge (2021) Roberts, C., et al.: Digitalising asset management: concomitant benefits and persistent challenges.In: Int. J. Build. Pathol. Adapt. 36(2), 152–173. (2018) Available at: https://doi.org/ 10.1108/IJBPA-09-2017-0036 Sacks, R., et al.: BIM handbook: A guide to Building Information Modeling for owners, designers, engineers, contractors, and facility managers, 3rd edn. John Wiley & Sons Inc., Hoboken, New Jersey (2018) Sadeghi, M., et al.: Developing Building Information Model (BIM) for building handover, operation and maintenance. J. Facil. Manage. 17(4) (2019)

570

J. Chang et al.

Saunders, M., et al.: ‘Chapter 4: Understanding Research Philosophy and Approaches to Theory Development’, in Research Methods for Business Students, 8th edn. Pearson Education Limited, Harlow (2019) Schmidt, F., et al.: Impact of job experience and ability on job knowledge, work sample performance, and supervisory ratings of job performance. J. Appl. Psychol. 73(1), 46–57 (1988) Thabet, W., Lucas, J.: Asset data handover for a large educational institution: case-study approach. J. Constr. Eng. Manage. 143(11), 12 (2017) Thabet, W., Lucas, J.: Asset data handover for a large educational institution: case-study approach. J. Constr. Eng. Manage. 143(11), 05017017. Available at: https://doi.org/10.1061/(ASCE)CO. 1943-7862.0001389 The Institute of Asset Management. Asset Management - an anatomy. The Institute of Asset Management (2015) The International Organization for Standardization. ISO 55000: asset management: overview, principles and terminology. The International Organization for Standardization (2014) Vermunt, D., et al.: Exploring barriers to implementing different circular business models. J. Clean. Prod. 222, 891–902 (2019) Zadeh, P., et al.: Information quality assessment for facility management. Advanced Engineering Information 33, 181–205 (2017)

Perspectives on Smart Maintenance Technologies – A Case Study in Small and Medium-Sized Enterprises (SMEs) Within Manufacturing Industry San Giliyana1(B) , Marcus Bengtsson2,3 , and Antti Salonen2 1 Mälardalen Industrial Technology Center AB, Eskilstuna, Sweden

[email protected]

2 Mälardalen University, Eskilstuna, Sweden

{marcus.bengtsson,antti.salonen}@mdu.se 3 Volvo Construction Equipment Operations, Eskilstuna, Sweden

Abstract. Industry 4.0 consists of nine technological pillars: IIoT, Cloud Computing, Big Data and Analytics, AR, etc. Some of the pillars play an essential role in maintenance development. Previous research presents many technologies for smart maintenance, but one prevailing problem is that there are still challenges to implementing smart maintenance technologies cost-effectively in the manufacturing industry. Therefore, we explore perspectives on smart maintenance technologies from respondents within 15 manufacturing SMEs. We start by investigating whether the companies had implemented smart maintenance technologies, if so, in what context. Then, we explore perspectives from the manufacturing SMEs on added values, challenges, opportunities, advantages, and disadvantages of smart maintenance technologies. However, as none of the case companies had implemented any Smart Maintenance Technologies, only implementation challenges could be investigated.

1 Introduction The fourth industrial revolution, also named industry 4.0, was presented at the Hanover Fair 2011, which focuses on combining production, Information Technology (IT), and the internet (Matt et al. 2020), and according to Monostori et al. (2016) and Thoben et al. (2017), it has resulted in a significant change in the manufacturing industry. According to Liu and Xu (2017), industry 4.0 deals with Information and Communication Technologies (ICT), smart and intelligent factories, and the development of the internet and embedded system technologies, resulting in several new technologies. Vaidya et al. (2018) and Alcácer and Cruz-Machado (2019) have presented nine technological pillars of industry 4.0: 1) Industrial Internet of Things (IIoT), 2) Big Data and Analytics, 3) Augmented Reality (AR), 4) Simulation, 5) Autonomous Robots, 6) Additive Manufacturing (AM), 7) Cyber Security, 8) Cloud Computing, and 9) Horizontal and Vertical System Integration. According to Mittal et al. (2018), industry 4.0 is created by or for larger organizations or Multi-National Enterprises (MNEs). However, considering the © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 571–581, 2023. https://doi.org/10.1007/978-3-031-25448-2_53

572

S. Giliyana et al.

importance of SMEs to European Union (EU), the introduction of industry 4.0 is necessary to compete nationally and internationally (Masood and Sonntag 2020). A company is classified as SME, having less than 250 employees and a turnover of less than or equal to 50 million (EU 2016). Moeuf et al. (2020) have shown that SMEs offer 67.1% of the jobs in the private sector in Europe. In addition, Matt et al. (2020) state that producing faster, cheaper, and with higher quality is no longer enough, manufacturing SMEs need to introduce new innovative industry 4.0 production to meet the competition in the long term (Matt et al. 2020). Therefore, many researchers are writing about industry 4.0 for SMEs, and the number of scientific publications has been growing in recent years (Matt et al. 2020; Masood and Sonntag 2020). Moreover, the European Commission (EC) financially supports SMEs in introducing innovations and even supports research and development projects to create an innovation ecosystem for SMEs (Matt et al. 2020). Ghobakhloo and Ching (2019) have examined which of the nine technological pillars can be implemented in manufacturing SMEs: 1) Cloud Computing for data storage in cloudbased services, 2) IIoT for machine connection, which allows manufacturing SMEs to monitor the status of the machines in real-time, 3) Cyber Security, such as firewall and antivirus, to protect their data from cyber-attacks, and 4) Horizontal and Vertical System Integration using Enterprise Resource Planning (ERP). Masood and Sonntag (2020) have examined which of the nine industry 4.0 technological pillars have high benefits and which have low benefits for SMEs, and how complex these are to implement in SMEs. The industry 4.0 technological pillars which have low complexity and high benefits are AM and Simulation. Big Data and Analytics and Autonomous Robots are in the group of high benefits/high complexity, while in low benefits/low complexity, Cloud Computing and IIoT are located. One function that will be affected by the implementation of industry 4.0 is maintenance (Bokrantz 2017). According to Shin and Jun (2015), several definitions of maintenance exist. In general, maintenance is defined as the following (SS-EN13306, 2017, p. 8): “combination of all technical, administrative and managerial actions during the life cycle of an item intended to retain it in, or restore it to, a state in which it can perform the required function”. Maintenance consists of preventive maintenance and corrective maintenance (CM) (SS-EN13306, 2017). Preventive maintenance consists of conditionbased maintenance (CBM) and predetermined maintenance (PM) (SS-EN13306, 2017). According to SS-EN13306 (2017, p. 35), the definition of CBM is: “preventive maintenance which include assessment of physical conditions, analysis and the possible ensuing maintenance actions”. PM is defined as (SS-EN13306, 2017, p. 35): “preventive maintenance carried out in accordance with established intervals of time or number of units of use but without previous condition investigation”. According to the SS-EN13306 (2017, p. 38), CM is defined as: “maintenance carried out after fault recognition and intended to restore an item into a state in which it can perform a required function”. In addition, autonomous maintenance, which is one of the Total Productive Maintenance (TPM) pillars (Nakajima 1988), focuses on keeping the equipment in basic condition, through maintenance activities performed by an operator. In SS-EN13306 (2017, p. 40), autonomous maintenance is defined as: “maintenance actions carried out by an operator”. Many studies, such as Salonen (2009), Bengtsson (2007), and Algabroun et al. (2020), have shown the importance of maintenance. Another study, performed by Zarreh et al.

Perspectives on Smart Maintenance Technologies

573

(2019), has shown that maintenance is the main function of keeping production systems running. According to Rastogi et al. (2020), low costs, quality control, increased flexibility, and productivity are objectives that manufacturing companies aim for, including SMEs. Maintenance, as a process, plays an essential role in achieving these objectives (Rastogi et al. 2020). Furthermore, in SMEs, maintenance plays a major financial role (Rastogi et al. 2020). The area of maintenance can be divided into four industrial generations. In the first industrial generation, the equipment was run to failure. During the second industrial generation, planned maintenance was implemented, such as scheduled overhauls and systems for planning and control. The third one deals with Condition Monitoring, Design for reliability, and total quality maintenance (Singh et al. 2013). Industry 4.0 places a new demand on maintenance, and some of the nine technological pillars play an essential role in maintenance development (Silvestri et al. 2020). For example, machines can be connected using IIoT and collect maintenance data, such as vibration, pressure, temperature, acoustics, and viscosity (Amruthnath and Gupta 2018). Big Data and Analytics can support advanced data analysis (Witkowski 2017), real-time decision-making (Subramaniyan et al. 2018), and maintenance planning (Silvestri et al. 2020). Simulation can be used to predict a production system’s behavior and thereby support maintenance scheduling and decision-making (Goodall et al. 2019). AR may offer step-by-step guidance for diagnostics and training (Roy et al. 2016). Regarding AM, Chong et al. (2018) have mentioned that 3D-CAD can be used to learn how the equipment is built and thereby plan maintenance activities based on that. However, Lundgren et al. (2021), Campos et al. (2021), and Giliyana et al. (2022) state that further research is needed to support the manufacturing industry in smart maintenance. Masood and Sonntag (2020) have discussed the manufacturing SMEs’ challenges when implementing industry 4.0 technological pillars: 1) training of the workforce, which is the most common one, 2) support from experts, 3) difficulties dedicating time for new technology development, 4) awareness of a wide amount of technologies, and 5) investment needed to implement the technologies, in training and in right people (Masood and Sonntag 2020). Based on that, in this paper, we explore perspectives on smart maintenance technologies from manufacturing SMEs. The exploration is done through 3 steps: 1) we investigate whether there have been any implemented smart maintenance technologies, 2) if so, in what context, 3) we explore perspectives from the SMEs on added values, challenges, opportunities, advantages, and disadvantages with the smart maintenance technologies.

2 Methodology This paper is based on a case study, including 15 SMEs in the Swedish manufacturing industry. 2.1 Data Collection The empirical data was collected through an online open questionnaire as a primary data collection method. Case studies are typically used when a research project is about exploration, theory building, testing, and elaboration/refinement (Karlsson et al. 2016).

574

S. Giliyana et al.

Several data collection techniques can be used in case studies: interviews, questionnaires and observations (Säfsten and Gustavsson 2020). In this research, to explore perspectives on smart maintenance technologies, Google Forms has been used to visualize the questions to the respondents as well as to collect the data. The questionnaire consists of 7 questions (shortened): 1) What types of maintenance are implemented?, 2) What types of smart maintenance technologies have been implemented or tested in your maintenance processes and in what context?, 3) Any added value?, 4) Any challenges?, 5) Any opportunities?, 6) Any advantages, and 7) Any disadvantages? The questionnaire was sent to 26 manufacturing SMEs in Sweden. The questionnaire was sent to the case companies in February 2022. After four weeks, a reminder was sent to the case companies that had not answered the questionnaire, after which the total response rate amounted to 58 percent, see Table 1. Säfsten and Gustavsson (2020) have discussed several weaknesses of questionnaires. One being is that respondents cannot get support. This weakness in this study was offset by performing semi-structured interviews with respondents from six of the case companies that had not implemented any smart maintenance technologies and not mentioned any challenges to implementing smart maintenance technologies. Taking both the questionnaire and the interview into account, respondents from twelve companies answered the question regarding the challenges of implementing smart maintenance technologies. Respondents from the remaining three companies had not answered the question regarding challenges and were unfortunately not available for an interview. The interviews were also performed to offset weaknesses that might arise from only one data collection technique, i.e., triangulation (Yin 2018). Table 1. Shows the case companies and data collection methods Case company

Employees

Type

Data collection methods

A

25

Contract manufacturing

Questionnaire and interview

B

50

Contract manufacturing

Questionnaire

C

20

Manufacturing of latches and quick release battery connectors

Questionnaire and interview

D

53

Manufacturing of installation-ready components to the energy, automotive and process industries

Questionnaire

E

35

Contract manufacturing

Questionnaire

F

37

Contract manufacturing

Questionnaire and interview

G

41

Manufacturing of impact sockets and accessories

Questionnaire

H

10

Contract manufacturing

Questionnaire and interview (continued)

Perspectives on Smart Maintenance Technologies

575

Table 1. (continued) Case company

Employees

Type

Data collection methods

Surface treatment

Questionnaire

I

80

J

43

Contract manufacturing

Questionnaire

K

30

Contract manufacturing

Questionnaire

L

65

Contract manufacturing

Questionnaire

M

100

Manufacturing of car heating and battery charging products

Questionnaire

N

65

Manufacturing of components for gas and steam turbines

Questionnaire and interview

O

14

Contract manufacturing

Questionnaire and interview

2.2 Data Analysis The data was analyzed through three steps. The handwritten notes from the semistructured interviews were typed into a computer in the first step. The semi-structured interviews and questionnaire data were initially analyzed and coded as: implemented maintenance types, implemented smart maintenance technologies, context, added value, challenges, opportunities, advantages, and disadvantages. In the second step, the data was visualized in the form of a matrix using Microsoft Excel and sorted based on nine columns: 1) case company, 2) implemented maintenance types, 3) implemented smart maintenance technologies, 4) context, 5) added value, 6) challenges, 7) opportunities, 8) advantages, and 9) disadvantages. In the third step, the conclusion was made by looking for explanations and comparing the different views and case companies (Miles et al. 2019; Säfsten and Gustavsson 2020). Finally, the thematic analysis (Braun and Clarke 2006) was used to organize the data into four categories: Knowledge, Time and resources, Cost, and Age of the machines.

3 Empirical Findings Interestingly, none of the 15 respondents state that their company has implemented any smart maintenance technologies. Some of the challenges mentioned by the respondent at case company A are “Getting the costs into the budget.”, “Knowing what to measure.”, and “Knowing what kind of data to collect and why.”. Moreover, the respondent at case company A mentioned that there is no time to think about smart maintenance technologies due to the small number of employees. Regarding cost, the respondent at the case company F also mentioned”Start-up costs” as a challenge to implementing smart maintenance technologies. Furthermore, the respondent at case company F mentioned, “Industry 4.0, in general, is created for large companies with necessary implementation resources.”, and “The time between implementing smart maintenance technologies

576

S. Giliyana et al.

and benefits, since SMEs depend on fast results”. Another challenge, mentioned by the respondent at case companies C and N, is that their machines are of older models, and these technologies may not be available for their machines. “The time it takes to implement smart maintenance technologies”, is a challenge mentioned by the respondent at case company K. Additionally, the respondent at case company E mentioned, “We are not familiar with the technologies in the context of our maintenance”. Some other mentioned challenges are,”Learning of the system, diffusion into operations”, case company D,”Competence and experience in senior positions” and”technical and financial resources”, case company I, and”to make these technologies work and make everyone understand and work after them.”, case company J. Table 2 presents all identified challenges for each case company and the type of maintenance implemented. Only two of the companies are on a full corrective maintenance strategy. Twelve companies have implemented some sort of predetermined maintenance, three of these companies have also implemented some autonomous maintenance, and two of these companies even utilize some sort of CBM (the respondent of one company did not answer which maintenance types they were using). In addition, at case companies A, J and N, the respondents mentioned that a Computerized Maintenance Management Systems (CMMS) is implemented as a module in their ERP, and the plan is to have all predetermined maintenance tasks in this module. Furthermore, a tablet is used at case company J to organize autonomous maintenance, spare part management, and work order system. That none of Table 2. Summary of the empirical findings Case Company

Challenges mentioned by the respondent at the case companies

Implemented maintenance types

A

1) the company is small, and due to a small number of employees, there is no time to think about smart maintenance technologies. There must be a full-time person who only works with these technologies, 2) gets the costs into the budget, 3) know what kind of data to collect, and 4) know what to measure

PM

B C

PM 1) the machines are older, and these technologies may not be available for older machines

PM, CM, and Autonomous Maintenance (continued)

Perspectives on Smart Maintenance Technologies

577

Table 2. (continued) Case Company

Challenges mentioned by the respondent at the case companies

Implemented maintenance types

D

1) learning of the system, diffusion into operations

CM

E

1) not familiar with these technologies in CM, and PM the context of maintenance

F

1) start-up costs, 2) the time between implementing these technologies and benefits, since SMEs are dependent on fast results, 3) technical knowledge, and 4) industry 4.0, in general, is created for large companies with necessary implementation resources

PM, CBM, and CM

H

1) time

CM

I

1) competence and experience in senior positions, and 2) technical and financial resources

J

1) to make these technologies work and PM, CM, and make everyone understand and work after Autonomous Maintenance them

K

1) time it takes to implement these technologies

PM, and CM

L

1) knowledge

PM, and CM

G

PM, and CM

M

PM, CM, and Autonomous Maintenance

N

1) knowledge, 2) time, and 3) older machines

PM, and CM

O

1) the technology is not mature. The machine manufacturers do not have the opportunity to offer these types of technologies for maintenance

PM, CBM, and CM

these companies have implemented any smart maintenance technologies can be viewed as rather remarking. As such, the respondents have mentioned nothing on added values, opportunities, advantages, or disadvantages.

578

S. Giliyana et al.

Table 3. Shows the challenges related to Knowledge, Time and resources, Cost, and Age of the machines.

Knowledge

Time and resources

Cost

Know what kind of data to collect.

The company is small, and due to a small number of employees, there is no time to think about smart maintenance technologies. There must be a full-time person who only works with these technologies.

Gets the costs into the budget.

The time between implementing these technologies and benefits, since SMEs are dependent on fast results.

Financial resources.

Know what to measure. Learning of the system, diffusion into operations. Not familiar with these technologies in the context of maintenance. Technical knowledge. Competence and experience in senior positions.

Age of the machines Older machines.

Start-up costs.

Industry 4.0, in general, is created for large companies with necessary implementation resources. Technical resources.

To make these technologies work and make everyone understand and work after them. Knowledge in general.

Time it takes to implement these technologies. Time in general. The technology is not mature. The machine manufacturers do not have the opportunity to offer these types of technologies for maintenance.

4 Discussions and Conclusions The empirical findings show that no smart maintenance technologies have been implemented at the case companies. Therefore, no respondents mentioned nothing on added values, opportunities, advantages, or disadvantages. Respondents from twelve case companies have mentioned challenges to implementing smart maintenance technologies. The remaining three have not answered the question regarding challenges and had no opportunity for an interview. In Table 3, the challenges are organized into four categories. The fact that none of the 15 respondents of the case companies state that their companies have implemented any smart maintenance technologies can be viewed both as remarking and not remarking. It can be viewed as remarking that industry 4.0 and Smart Maintenance have been discussed for more than ten years. It can be viewed as not that remarking as recent research, as previously expressed in the introduction, states that further research is needed to support manufacturing companies in implementing smart

Perspectives on Smart Maintenance Technologies

579

maintenance (Lundgren et al. 2021, Campos et al. 2021, and Giliyana et al. 2022). In the study presented by Giliyana et al. (2022), a similar study, as presented in this paper, with eleven large manufacturing companies showed that seven of the companies had implemented some Smart Maintenance Technologies, one company had only implemented some pilots, and three companies had not implemented any Smart Maintenance Technologies. Thus, to some extent, also in large manufacturing companies implementing Smart Maintenance Technologies have not become commonplace either. Considering the importance of SMEs, the empirical findings show that further research is needed to support manufacturing SMEs in implementing smart maintenance technologies cost-effectively. Based on the empirical findings and previous research, one recommendation is that the manufacturing SME should start with the challenges related to Knowledge, which is the most common when implementing new technologies (Masood and Sonntag 2020).

References Alcácer, V., Cruz-Machado, V.: Scanning the industry 4.0: a literature review on technologies for manufacturing systems. Eng. Sci. Technol. Int. J. 22(3), 899–991 (2019). https://doi.org/10. 1016/j.jestch.2019.01.006 Algabroun, H., Bokrantz, J., Al-Najjar, B., Skoogh, A.: Development of digitalised maintenance – a concept. J. Qual. Maint. Eng. (2020). https://doi.org/10.1108/JQME-04-2019-0039 Amruthnath, N., Gupta, T.:A research study on unsupervised machine learning algorithms for early fault detection in predictive maintenance. In IEEE: 5th International Conference on Industrial Engineering and Applications (ICIEA). Singapore, 26–28 Apr 2018, pp. 355–61 (2018). https:// doi.org/10.13140/RG.2.2.28822.24648 Bengtsson, M.: On Condition Based Maintenance and its Implementation in Industrial Settings. D. Diss., Mälardalen University Press, Ph (2007). https://doi.org/10.13140/RG.2.2.29844.71047 Bokrantz, J.: On the transformation of maintenance organisations in digitalized manufacturing. Licentiate Thesis Diss., Chalmers University of Technology (2017) Braun, V., Clarke, V.: Using thematic analysis in psychology. Qual. Res. Psychol. 3(2), 77–101 (2006). https://doi.org/10.1191/1478088706qp063oa Campos, J., Kans, M., Salonen, A.: A Project Management Methodology for the Digitalisation of the Industrial Maintenance Domain. In: Ball, A., Gelman, L., Rao, B.K.N. (eds.) Advances in Asset Management and Condition Monitoring. SIST, vol. 166, pp. 621–629. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-57745-2_52 Chong, S., Pan, G.-T., Chin, J., Show, P.L., Yang, T.C.K., Huang, C.-M.: Integration of 3D printing and industry 4.0 into engineering teaching. Sustainability 10(11), 1–13 (2018). https://doi.org/ 10.3390/su10113960 EU, E.C.: User Guide to the SME Definition. Luxembourg: Publications Office of the European Union (2016). https://ec.europa.eu/regional_policy/sources/conferences/state-aid/sme/smedef initionguide_en.pdf. Accessed 1 Sep 2022 Ghobakhloo, M., Ching, N.T.: Adoption of digital technologies of smart manufacturing in SMEs. J. Ind. Inf. Integr. 16, 100107 (2019). https://doi.org/10.1016/j.jii.2019.100107 Giliyana, S., Salonen, A., Bengtsson, M.: Perspectives on smart maintenance technologies – a case study in large manufacturing companies. In: IOS Press Ebooks: 10th Swedish Production Symposium 2022 (SPS2022). Skövde, Sweden, 26–29 Apr 2022, pp. 255–66 (2022). https:// doi.org/10.3233/ATDE220145

580

S. Giliyana et al.

Goodall, P., Sharpe, R., West, A.: A data-driven simulation to support remanufacturing operations. Comput. Ind. 105, 48–60 (2019). https://doi.org/10.1016/j.compind.2018.11.001 Karlsson, C., et al.: Research Methods for Operations Management. Taylor and Francis (2016) Liu, Y., Xu, X.: Industry 4.0 and cloud manufacturing: a comparative analysis. J. Manuf. Sci. Eng. 139(3), 1–8 (2017). https://doi.org/10.1115/MSEC2016-8726 Lundgren, C., Bokrantz, J., Skoogh, A.: A strategy development process for smart maintenance implementation. J. Manuf. Technol. Manag. 32(9), 142–166 (2021). https://doi.org/10.1108/ JMTM-06-2020-0222 Masood, T., Sonntag, P.: Industry 4.0: adoption challenges and benefits for SMEs. Comput. Ind. 121, 103261 (2020). https://doi.org/10.1016/j.compind.2020.103261 Matt, D.T., Modrák, V., Zsifkovits, H. (eds.): Industry 4.0 for SMEs: Challenges, Opportunities and Requirements. Springer International Publishing, Cham (2020). https://doi.org/10.1007/ 978-3-030-25425-4 Miles, M.B., Huberman, A.M., Saldana, J.: Qualitative Data Analysis: An Expanded Sourcebook. SAGE Publications, Thousand Oaks (2019) Mittal, S., Khan, M.A., Romero, D., Wuest, T.: A critical review of smart manufacturing and industry 4.0 maturity models: implications for small and medium-sized enterprises (SMEs). J. Manuf. Syst. 49, 194–214 (2018). https://doi.org/10.1016/j.jmsy.2018.10.005 Moeuf, A., Lamouri, S., Pellerin, R., Tamayo-Giraldo, S., Tobon-Valencia, E., Eburdy, R.: Identification of critical success factors, risks and opportunities of industry 4.0 in SMEs. Int. J. Prod. Res. 58, 1384–1400 (2020). https://doi.org/10.1080/00207543.2019.1636323 Monostori, L., et al.: Cyber-physical systems in manufacturing. CIRP Ann. 65(2), 621–641 (2016). https://doi.org/10.1080/00207543.2019.1636323 Nakajima, S.: Introduction to TPM: Total Productive Maintenance. Productivity Press, Cambridge (1988) Rastogi, V., Srivastava, S., Mishra, M., Thukral, R.: Predictive maintenance for sme in industry 4.0. In: IEEE: 2020 Global Smart Industry Conference (Glosic), 17–19 Nov 2020, pp.382–390 (2020). https://doi.org/10.1109/GloSIC50886.2020.9267844 Roy, R., Stark, R., Tracht, K., Takata, S., Mori, M.: Continuous maintenance and the future – foundations and technological challenges. CIRP Ann. 65(2), 667–688 (2016). https://doi.org/ 10.1016/j.cirp.2016.06.006 Salonen, A.: Formulation of Maintenance Strategies. Mälardalen University Press, Licentiate Theses Diss. (2009) Shin, J.-H., Jun, H.-B.: On condition based maintenance policy. J. Comput. Des. Eng. 44, 119–127 (2015). https://doi.org/10.1016/j.jcde.2014.12.006.urn:nbn:se:mdh:diva-4836 Silvestri, L., Forcina, A., Introna, V., Santolamazza, A., Cesarotti, V.: Maintenance transformation through industry 4.0 technologies: a systematic literature review. Comput. Ind. 123, 1–16 (2020). https://doi.org/10.1016/j.compind.2020.103335 Singh, S., Galar, D., Baglee, D., Björling, S.-E.: Self-maintenance techniques: a smart approach towards self-maintenance system. Int. J. Syst. Assur. Eng. Manage. 5(1), 75–83 (2013). https:// doi.org/10.1007/s13198-013-0200-7 Swedish Standards Institute. Maintenance - Maintenance terminology, SS-EN13306 (2017) Subramaniyan, M., Skoogh, A., Salomonsson, H., Bangalore, P., Bokrantz, J.: A data-driven algorithm to predict throughput bottlenecks in a production system based on active periods of the machines. Comput. Ind. Eng. 125, 533–544 (2018). https://doi.org/10.1016/j.cie.2018. 04.024 Säfsten, K., Gustavsson, M.: Research Methodology - for Engineers and Other Problem-Solvers. Lund: studentlitteratur AB, Lund (2020) Thoben, K.-D., Wiesner, S., Wuest, T.: “Industrie 4.0” and smart manufacturing – a review of research issues and application examples. Int. J. Autom. Technol. 11(1), 4–16 (2017). https:// doi.org/10.20965/ijat.2017.p0004

Perspectives on Smart Maintenance Technologies

581

Vaidya, S., Ambad, P., Bhosle, S.: Industry 4.0 – A glimpse. Procedia Manufact. 20, 233–238 (2018). https://doi.org/10.1016/j.promfg.2018.02.034 Witkowski, K.: Internet of Things, big data, industry 4.0 – innovative solutions in logistics and supply chains management. Procedia Eng. 182, 763–769 (2017). https://doi.org/10.1016/j.pro eng.2017.03.197 Yin, R.K.: Case Study Research and Applications: Design and Methods. SAGE Publications, Los Angeles (2018) Zarreh, A., Wan, H., Lee, Y., Saygin, C., Janahi, R.A.: Cybersecurity concerns for total productive maintenance in smart manufacturing systems. Procedia Manufact. 38, 532–539 (2019). https:// doi.org/10.1016/j.promfg.2020.01.067

Improving Maintenance Data Quality: Application of Natural Language Processing to Asset Management Mathieu Payette1(B) , Georges Abdul-Nour1 , Toualith Jean-Marc Meango2 , and Alain Côté2 1 Université du Québec à Trois-Rivières (UQTR), Trois-Rivières, Québec, Canada

{mathieu.payette,georges.abdulnour}@uqtr.ca 2 Hydro-Québec (HQ), Montreal, Québec, Canada {meango.toualithjean-marc,cote.alain7}@hydroquebec.com

Abstract. Artificial intelligence techniques are increasingly used for asset management. The abundance of data available in large electrical utility offers many application opportunities. The use of data-driven models can address some of the biases of physical models traditionally used in reliability engineering. However, in this context, as in many other fields of operation, the quality of data is often questioned by domain experts. Operational data are entered manually by maintenance technicians, and data entry errors are common. One of the errors that is observed is mislabeling of maintenance types, which can lead to poor statistical estimates of failure rate. This paper aims to improve the quality of historical maintenance data, to increase the accuracy of deployed models. To this end, the text fields available in the maintenance history is analyzed to predict the type of maintenance performed. Natural language processing (NLP) techniques are applied to solve this text classification problem. The models are applied to Hydro-Québec TransÉnergie’s power transmission assets. The application of such techniques allows the enrichment of databases and thus reduces uncertainty in decision-making for asset management.

1 Introduction The increasing complexity of industrial systems requires a high level of expertise, especially in the field of reliability engineering and asset management. Recent machine learning techniques now makes it possible to process Big Data and is a good opportunity for large organization to improve decision-making. The goal of Engineering of Asset Management (EAM) is to realize the value of company’s assets. The asset management system (EAMS) aims to implement strategies to maximize the value of the asset throughout its life cycle (International Organisation for Standardization 2015; Amadi-Echendu et al. 2011). Thus, the EAM decision-making process is strongly reliant on information availability. Maintenance data is therefore crucial to the delivery of the asset’s life cycle. Furthermore, some argue that data itself is an asset that can generate value for an organization (Perrons and Jensen 2015). Consequently, historical data should be maintained and improved, to extract the maximum value from them and for the physical asset’s it describes. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 582–589, 2023. https://doi.org/10.1007/978-3-031-25448-2_54

Improving Maintenance Data Quality

583

Hydro-Quebec operates one of the most important transmission systems in America and is one of the world’s largest producers of hydroelectricity. The company was founded in 1944 and its assets are very diverse; some dating back to before the foundation are still in service. Maintenance histories are sometimes incomplete due to the age of the equipment and the various changes in data acquisition systems and coding makes it difficult to trace them. To improve the company’s maintenance policies, it is necessary to model the reliability of equipment. In many cases, reliability is represented by qualitative models using methods such as the Failure Mode, Effects and Criticality Analysis (FMECA) or statistical lifetime distribution. However, this approach has its drawbacks, since it requires a deep knowledge of the equipment, its maintenance history and its operational environment. On the other hand, data-driven models can eliminate some of these disadvantages and provide an estimate of failure rates from the recorded history. Since data are manually entered by technicians, the quality of data is often questioned by experts. It has been reported that mislabelling of data is common, leading to biased estimation of the failure statistics. Indeed, it would be relevant to address the problem at the source, by improving data entry methods for example. This would not change the quality of the data that is already available. As mentioned above, these historical data are assets, and should be treated as such. To increase the accuracy of the models deployed by Hydro-Quebec experts, this research seeks to improve the quality of historical maintenance data. One of the strengths of Hydro-Quebec is that the company has been historicizing maintenance data for more than 50 years. The information contained in the databases is very detailed and free text fields contain a significant amount of information that could be used to correct the labelling of maintenance types.

2 Literature Review 2.1 Types of Maintenance To understand the problem, it is imperative to define the different types of maintenance performed. Corrective or reactive maintenance is where the system is repaired when there is a failure that stops the system functions (Lewis 1996). Preventive maintenance (PM) is when actions are planned and performed to avoid or reduce the occurrence of failures. It can be divided into several types. Systematic or interval maintenance is a predetermined periodic maintenance (ex: change of oil on a car). The interval is often specified by the system’s manufacturer. Condition-based maintenance is performed when the system shows signs of wear that could lead to a failure (Islam 2010). This type of maintenance is triggered by an inspection, or anomaly detected by a set of sensors. Finally, predictive maintenance is where maintenance is performed when mathematical models predict an imminent risk of failure (Jimenez-Cortadi et al. 2020). Figure 1, adapted from (Islam 2010), summarizes each maintenance type presented above. 2.2 Machine Learning Machine learning (ML) is a set of mathematical modelling techniques that allow a computer to learn the structure of a system from data and not by explicitly programming

584

M. Payette et al.

Fig. 1. Maintenance classification

its structure (Hurwitz and Kirsch 2018). There is a multitude of machine learning techniques, which are often grouped according to the way the computer learns from the data. The most commonly known learning methods are supervised learning and unsupervised learning. In supervised learning, the input data are labelled data, meaning that the dataset contains both the predictors and the variable to be predicted. The learning process is divided in three steps: training, validation and tests. In training, a sample of the data is sent, and the algorithm makes predictions on the predictors. Knowing the expected value, the algorithm parameters are modified so that the model is perfected at each iteration. Another sample of the data is used to find the best hyperparameters and the test sample evaluate the ability of the model to make predictions on new data (Russell and Norvig 2002). In general, supervised learning is applied to classification problems, where the variable to be predicted is a class, or to regression where the variable to be predicted is continuous. Unsupervised learning, on the other hand, uses unlabelled data as input. The goal is to find relationships between the variables themselves. In clustering, for example, the objective is to group data points into meaningful clusters, according to similarity metrics (Fahad et al. 2014). Supervised learning requires a large amount of labelled data, and even more for complex modelling tasks. As such, the amount of data required to build the models is very difficult to generate. To overcome this problem, two new branches of artificial intelligence (AI) have emerged: transfer learning and self-supervised learning. In both approaches, a model is pre-trained, with large public datasets for example, to perform a predictive task. Then the model is fine-tuned using the application specific dataset. In transfer learning, the model is pre-trained with labelled data as oppose to self-supervised learning where it is trained on unlabelled data (LeCun and Misra 2021; Yang et al. 2020). 2.3 Text Pre-processing Text classification encompasses two problems in itself. Indeed, to perform classification with common ML methods (decision trees, vector support machines, etc.), the data must be understandable for a machine. A computer recognizes text as a sequence of characters with no particular meaning, so the text must be converted in a numerical form that encapsulates the words and their meaning. Natural language processing (NLP) is

Improving Maintenance Data Quality

585

the field of computer science whose goal is for machines to interpret language in the same way as a human (Beysolow II 2018). Tokenization is the process of separating each element of a sentence into a single fragment. Fragmentation can be done at the sentence level (cutting paragraphs into individual sentences) or at the word level. Lemmatization is the process of transforming words into their canonical forms (Ex: “I am being” converted to “I be”). It also includes the meaning of the word in its transformation (Thanaki 2017). Part-of-Speech tagging consists of evaluating words and their function in a sentence. The first step is to identify the category of the word (noun, verb, adjective, etc.) and then to identify the gender and number of the word. For verbs, it is a matter of identifying the mode and tense of the verb. Then, it is a question of analysing the function of the words and their links between them. 2.4 Word Representation As mentioned above, the goal of NLP is to represent text in a way that can be interpreted by a computer. There are several techniques for digitally representing text. The simplest technique is the bag-of-words. The principle is to establish a vocabulary from a corpus (set of texts) and, for each sentence, create a vector that counts each word. The bag-ofwords can also be a binary representation, similar to one-hot-encoding. A more advanced technique is “word embedding”, which consists in making a vector representation of words in several dimensions. The closer the words are, in the sense of the corpus, the closer they will be in the vector space. Vectors are generated using neural networks (NN), by taking each word and trying to predict its neighbourhood. The best-known approaches are the continuous-bag-of-words and the skip gram model. The drawback of these methods is that they do not take into account the meaning of sentences and words in their contexts. Each word has its own vector, even if the same word can have several meanings (Thanaki 2017). To address this problem, recurrent neural networks (RNR), in particular long shortterm memory models (LSTM) allow generating a vector considering the words and their context in a sentence. The transformers, a recent neural network architecture, can also be used to create such a representation. The advantage of the transformers is that, unlike LSTMs and RNRs which operate sequentially, the architecture of the transformers allows for parallel operations, which greatly reduces the computation time. In addition, the attention mechanisms reduce the effect of vanishing gradient, compared to other RNRs, and thus allow for better understanding of dependencies in text (Vaswani et al. 2017). Finally, the advantage of using LSTM and transformers is that there are pre-trained networks easily available, and it is possible to use them for different machine learning tasks. Popular models based on Google’s Bidirectional Encoder Representations from Transformers (BERT) are generally trained in a self-supervised manner (LeCun and Misra 2021).

586

M. Payette et al.

3 Methodology 3.1 Case Study and Problem Formalization This section provides a formalization of the problem as well as the methodology that is used to solve it. As previously mentioned, experience with enterprise data has shown that historical maintenance data is often misclassified. When examining failure data, many data labelled as failures (corrective maintenance orders) turn out to be preventive maintenance, and vice versa. The goal of this work is to improve the quality of the data by correcting the type of maintenance recorded in the history. The dependent variable is the type of maintenance performed (preventive vs. corrective) and the predictors (independent variables) are extracted from the text fields contained in the maintenance orders and notices. This case study will only focus on those two categories, without distinction for the different types of preventive maintenance. From Sect. 2.2, it’s clear that this problem can be defined as supervised learning (learning from labelled data), particularly a classification problem (categorical predictor) (Russell and Norvig 2002). Document classification is an application of NLP and since there are only two categories, it is a binary classification problem. 3.2 Data Acquisition and Preprocessing At Hydro-Quebec, there are two types of documents that detail maintenance work; the maintenance notice and maintenance order. The first is created following the detection of a problem by a technician and describes the problems of the faulty device. A maintenance order is issued if it is deemed that the device needs to be repaired and contains the details of the work done. Another type of data is called forced equipment downtime. The electrical network is under constant surveillance, since electricity is an essential commodity. When there is a power outage, the equipment in the affected area are all considered unavailable. The employees in charge of monitoring the network will enter this unavailability data into the acquisition system. In theory, by cross-referencing work notices and orders with unavailability data, corrective maintenance should be easily identified. However, identifying correctly which equipment is failing, which is collateral and labelling the cause of failure is highly time consuming, considering the number of events and components involved. Furthermore, the equipment identification code differs from one database to the other. A lot of effort has been put to manually cross reference the two data sources by Hydro-Quebec experts, and a year of pre-processed failure data is now available. A cross-referencing method is developed to automate this process by (Liang et al. 2022), to unify the two data sources. The first step of the project is to extract all the maintenance notices and orders, and to examine the text fields. Then, the data cross-referenced by the experts is used as ground-truth to build the model. 3.3 Modelling, Validation and Testing Once the training data is ready, different classification models will be tested. The results of two pre-trained models are used for the case study as a proof of concepts. The validation of the models is done through cross validation; a validation sample will allow comparing the two models and a test sample will allow to make sure of their quality.

Improving Maintenance Data Quality

587

From the confusion matrix, the accuracy and recall will be calculated as performance metrics for the models. The models are built with the Python library Spacy (Honnibal and Johnson 2015). Figure 2 shows the general pipeline for NLP. The raw text is the input and the Doc object is a representation of the text that is usable for modeling.

Fig. 2. Spacy pipeline for NLP

For this case study, the general models cannot be used, since the text fields analysed are in French. Specific models designed for this purpose are applied. Figure 3 shows the architecture of the French model:

Fig. 3. Text classification pipeline

The pipeline consists of a pre-trained network to which the components can be modified or added. In this case, the pipeline is modified to add text classification.

4 Results and Discussion After extracting data from different databases, work orders are examined for a oneyear period. Corrective and preventive maintenance are labelled according to previous studies made on equipment unavailability. Due to class imbalances, the sample from PM is reduced, to have equal samples from corrective maintenance. Both models are trained on a sample of 120 entries. The first model implemented is the small model from the Spacy French model package and the second is the large model. Respective results are shown below (Figs. 4 and 5). As results show, there is a slight improvement from the small to the large model, without a major impact on the runtime. The precision metrics indicate the proportion of positive identifications of corrective maintenance, which exceed 90% for both models. Recall indicates the percentage of overall correctly classified entry. At 64.54%, the small model is slightly better than a weak classifier. Recall increases at 70.21% for the large model (Table 1).

588

M. Payette et al.

Fig. 4. Confusion Matrix (small model)

Fig. 5. Confusion Matrix (large model)

Table 1. Comparison of the text classification models Model fr_core_news_sm

fr_core_news_lg

Precision

94,79%

97,05%

Recall

64,54%

70,21%

Accuracy

64,74%

71,15%

2,56

5,60

Runtime (sec)

5 Conclusion The goal of this study was to create and compare text classification models to improve data quality in the context of an asset management program. Two models were trained to use a few text samples from two databases. Impressive results are shown from both the small and the large classification models built with Spacy pipelines. Indeed, the two models were trained with Spacy French models and when considering sample size, the

Improving Maintenance Data Quality

589

prediction accuracy is relatively high. This case study proves nonetheless, improvement needs to be made on the dataset as well as the models. The next steps of this work are to build better models by increasing the size of the input data, using data from 5 years of maintenance work orders. In addition, the Spacy pipeline were mostly used as is; an implementation of the models with an optimization of the pipeline’s hyperparameters will yield better results. In the next study, we also plan to test transformer architectures, which are the most advanced models for NLP. Acknowledgement. This research was supported by Hydro-Québec, the Natural Sciences and Engineering Research Council of Canada (NSERC) and the Université du Québec à Trois-Rivières through the Hydro-Québec Asset Management research Chair.

References Amadi-Echendu, J.E., Brown, K., Willett, R., Mathew, J.: Definitions. London, Springer-Verlag, London, Concepts and Scope of Engineering Asset Management (2011) Beysolow II, T.: Applied Natural Language Processing with Python: Implementing Machine Learning and Deep Learning Algorithms for Natural Language Processing, Apress, New York (2018) Fahad, A., et al.: A survey of clustering algorithms for big data: taxonomy and empirical analysis. IEEE Trans. Emerg. Top. Comput. 2, 267–279 (2014) Honnibal, M., Johnson, M.: An improved non-monotonic transition system for dependency parsing. In: Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pp. 1373–1378 (2015) Hurwitz, J., Kirsch, D.: Machine Learning for dummies. Wiley (2018) International organisation for standardization. ISO 55000, 55001 and 55002 Asset Management Standards. BSI Standards (2015) Islam, H.: Reliability-centered maintenance methodology and application: a case study. Engineering 2(11), 863–873 (2010) Jimenez-cortadi, A., Irigoien, I., Boto, F., Sierra, B., Rodriguez, G.: Predictive maintenance on the machining process and machine tool. Appl. Sci. (2076–3417), 10, 224 (2020) Lecun, Y., Misra, I.: Self-supervised learning: the dark matter of intelligence. Meta A I, 23 (2021) Lewis, E.E.: Introduction to Reliability Engineering, Wiley, New York (1996) Liang, Y.P., Blancke, O., Gaha, M., St-jean, G., Aïmeur, E.: Automatic Database Alignment Method to Improve Failure Data Quality. RAMS 2022 (In press) (2022) Perrons, R.K., Jensen, J.W.: Data as an asset: what the oil and gas sector can learn from other industries about “Big Data.” Energy Policy 81, 117–121 (2015) Russell, S. Norvig, P.: Artificial Intelligence: A Modern Approach (2002) Thanaki, J.: Python Natural Language Processing, Packt Publishing Ltd., Birmingham (2017) Vaswani, A., et al.: Attention is all you need. Adv. Neural inf. Proc. Syst. 30 (2017) Yang, X., He, X., Liang, Y., Yang, Y., Zhang, S. Xie, P.: Transfer learning or self-supervised learning? A tale of two pretraining paradigms (2020). arXiv preprint arXiv:2007.04234

RQCM: Risk Qualitative Criticality Matrix. Case Study: Ophthalmic Lens Production Systems in Costa Rica Carlos Parra1(B) , Juan Rodríguez2 , Adolfo Crespo Márquez3 , Vicente González-Prida3 , Pablo Viveros4 , Fredy Kristjanpoller4 , and Jorge Parra5 1 Ingeniería Mecánica, Universidad Técnica Federico Santa María, Valparaíso, Chile

[email protected]

2 Escuela de Ingeniería Electromecánica, Instituto Tecnológico de Costa Rica (ITCR),

Ciudad de Cartago, Costa Rica 3 Escuela Superior de Ingeniería, Universidad de Sevilla, Sevilla, Spain

{adolfo,vgonzalezprida}@us.es

4 Ingeniería Industrial, Universidad Técnica Federico Santa María, Valparaíso, Chile

{pablo.viveros,fredy.kristjanpoller}@usm.cl

5 Escuela de Ingeniería Mecánica, Universidad Tecnológica de Panamá, Ciudad de Panamá,

Panamá [email protected]

Abstract. The use of prioritization analysis techniques allows identifying the level of criticality of physical assets and helps to manage resources: human, economic and technological in a more efficient way. In other words, the process of criticality analysis helps to determine the importance and consequences of the failures of productive equipment in the operational context in which they perform. This article explains the basic theoretical aspects of the equipment prioritization analysis process based on risk matrices (failure frequency and consequences); and the development of the model named Risk Qualitative Criticality Matrix (RQCM). Finally, are presented and analysed the results of a case of application of the RQCM in the sector of ophthalmic lenses (new factory built in Costa Rica - PRATS Laboratory).

1 Introduction Qualitative hierarchizing techniques based on risk analysis are tools that can be used to determine the criticality of industrial business assets. These techniques allow us to evaluate and know the level of importance of industrial assets considered two factors: frequency and consequences of failures and help those involved in decision-making processes effectively guide resources: human, economic and technological in the areas of maintenance, operations, logistics, quality, safety, environment, etc. In other words, the process of criticality risk analysis (CRA), helps determine the importance of assets according to the consequences caused by failure events in the operational context in which they work (Parra et al. 2021a; Neurohr et al. 2021). The following article, takes as a specific reference, the risk-based ranking proposal developed in Phase 2 of the MMM © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 590–601, 2023. https://doi.org/10.1007/978-3-031-25448-2_55

RQCM: Risk Qualitative Criticality Matrix. Case Study

591

(Maintenance Management Model), presented in Fig. 1 (Crespo et al. 2009; Parra and Crespo 2015, Parra and Crespo 2020a, 2020b). The term “critical” and the definition of criticality may have different interpretations depending on the goal that is trying to hierarchize (Parra and Crespo 2015; González-Prida et al. 2012; Parra et al. 2021b and Parra and Crespo 2019). The objective of a critical analysis is to establish a method that serves as a generic instrument in maintenance; and help to determine the hierarchy of plants, systems, equipment, components, etc., from a complex production process, allowing the elements in sections that they can be handled in a controlled and auditable manner.

2 Criticality Model RQCM: Risk Qualitative Criticality Matrix Risk assessment techniques can be used to prioritize equipment/assets and align maintenance actions to key business objectives (Parra et al. 2021a; Li and Wright 2019). When carrying out, it ensures that maintenance actions are effective from the point of view of the main costs associated with maintenance and most importantly be efficient to minimize the consequences on safety, environment, production (Parra et al. 2021b; Junietz et al. 2018). The decision-making process behind the determination of the criticality of assets requires a hierarchical structure and the application of some mathematical models that allow weights and priorities of assets to be evaluated. In this study, the steps to be followed to design the risk-based criticality model would be the following (Parra and Crespo 2018; Parra and Crespo 2019; Parra et al. 2021c): 1. Define a scope and purpose for criticality analysis based on the risk model. This will be defined according to maintenance goals aligned to business goals and management. 2. Define the level of detail of the analysis (Taxonomy Reference - ISO 14224 Standard). 3. Importance criteria of the risk model should be established: ranges of fault frequencies (FF) and the consequence factors (C) to be evaluated (aligned with the business objectives) within the selected risk model. 4. Selecting or developing a risk assessment method that allows the systems within the industry or department. The criticality model taken as a reference for this article is called Risk Qualitative Criticality Matrix (RQCM), originally it was designed for the off-shore production assets of the Magallanes area, ENAP Sipetrol, (ENAP Sipetrol 2015; Enap Sipterol 2016; Crespo 2007; Parra and Crespo 2020a, 2020b; Viveros-Gunckel et al. 2020). For this case, the model of the qualitative criticality risk: RQCM is used in the ophthalmic lens company, hoping to be the door to future risk assessments, directing the maintenance management of PRATS Laboratory towards continuous improvement with the proposed recommendations and the lessons learned. The RQCM (Parra et al. 2020a; Chiu et al. 2017), is a simple analysis process, which is supported in the concept of risk: frequency of a failure by the consequences, the expression used for hierarchy systems from the RQCM model is:

592

C. Parra et al.

Total Criticality Risk (TCR) TCR = FF × C

(1)

where, • TCR: Total Criticality Risk • FF: Failure frequency (failure range in a certain time (failures/year) • C: Consequences of failure events

Effectiveness Phase1: Definition of maintenance objectives , strategiesand responsibilities Phase8: Implementation of the process of continuous improvement and adoption of new tecnologies Phase7: Life cycle analysis and the possible equipmentrenewal

Phase2: Equipmentranking according tothe Importance of its function

Improvement

Phase3: Analysis of weak points in high impact equipment

Phase4: Preventive Maintenanceplans design and resource needs identification

Phase6: Evaluationand maintenance execution control

Evaluation

Information Technologies support SAP PM, MAXIMO, MERIDIUM, MP7i, etc…..

Phase5: Maintenance programing and optimizationin the allocation of resources

Efficiency Fig. 1. Maintenance management process model.

Where the value of the consequences (C) is obtained from the following expression: Estimation of Consequences C = (PI × OF) + MC + HSE where, • • • •

PI: Impact on production factor OF: Operational flexibility factor MC: Maintenance costs factor HSE: Health, Safety and Environment factor The final expression of the TCR prioritization model will be as follows:

(2)

RQCM: Risk Qualitative Criticality Matrix. Case Study

593

Developed TCR TCR = FFx((PI × OF) + MC + HSE)

(3)

The weighted factors of each of the criteria for being evaluated by the expression of risk are presented below: • Fault frequency factor (FF) (scale 1–4) 4: Frequent: greater than 2 events per year 3: Average: 1 and 2 events per year 2: Good: Between 0.5 and an event a year 1: Excellent: Less than 0.5 events a year • Factors of consequences or operational impact (PI) (scale 1–10) 10: Production losses greater than 75% 7: Production losses between 50% and 74% 5: Production losses between 25% and 49% 3: Production losses between 10% and 24% 1: Production losses less than 10% • Impact by operational flexibility (OF) (scale 1–4) 4: No backup units are available to cover production, long repair times and complicated logistics 2: There are backup units that they can be partially cover the impact of production, average repair times and logistics 1: It has standby, there is no affectation in the process • Impact on maintenance costs (MC) (scale 1–2) 2: Repair costs, materials and labour exceeding 20,000 dollars 1: Repair costs, materials and labour less than 20,000 dollars • Impact on Health, Safety and Environment (HSE) (Scale 1–8) 8: High risk of lives losses, serious health damage, higher environmental incident (Catastrophic) that exceeds the allowed limits 6: Average risk of loss of life, important damage to health, environmental incident of difficult restoration 3: Minimum risk of loss of life and health condition (recoverable in the short term) and/or minor environmental incident (controllable), easy-to-contain and repeated leakage 1: There is no risk of loss of life, no health condition, or environmental damage

594

C. Parra et al.

The selection of the weighted factors is carried out in work meetings with the participation of the different persons involved in the operational context of the asset in study (operations, maintenance, processes, safety and environment). To obtain the level of criticality of each equipment/business system, the total values of each of the main factors are taken: frequency and consequences of the failures and are located in the 4 × 5 criticality matrix (Fig. 2). The failure frequency value is located on the vertical axis and the consequence value is located on the horizontal axis (the final result of the consequence expression is taken: ((PI × OC) + CM + HSE calculated). The criticality matrix shown next, allows systems to be ranked in three areas (see Fig. 2): • Area of Non-Critical systems (NC) • Medium Criticality Systems Area (MC) • Critical Systems Area (C) The result of the equation is located in the matrix (Fig. 2) to determine which area the equipment under study is located. The maximum value of risk criticality that can be obtained is 200 points distributed in 3 possible levels of hierarchizing systems (critical, semi critical and not critical). With regard to the development of the risk matrix, it has the following configuration: • Vertical axis (Failure Frequency): 4 rows, maximum value 4 points (scale: 1 to 4) • Horizontal axis (Failure Consequences): 5 columns, maximum value: 50 points (scale: 1 to 50) For subcategories of consequences (C = (PI × OF) + MC + HSE)), the percentage of importance assigned to each factor of the consequences, is aligned with the business objectives and they were approved by company management: • PI × OF = represents 80% of the total weight of the consequences → (40/50 = 80%) • MC = represents 4% of the total weight of the consequences → (2/50 = 4%) • HSE = represents 16% of the total weight of the consequences → (8/50 = 16%)

RQCM: Risk Qualitative Criticality Matrix. Case Study

595

Non Crititcality Area (NC) Medium Crititcality Area (MC) Criticality Area (C)

Fig. 2. Equipment criticality matrix. Source: Parra et al. 2021a

3 Case Study: Application of the RQCM Model in Production Equipment of the PRATS Costa Rica Laboratory The critical analysis was developed at the level of systems (level 5, according to ISO 14224) in the new laboratory of Grupo PRATS. Specifically, were evaluated the 12 more relevant systems that are included as technical locations in the hierarchical structure of the laboratory (Orders Workshop (TE), Treatments (TRA), Control (CT) and Beveling and assembly (BM)) (Rodríguez 2021). In this study, the natural work team was composed of five members, including the facilitator and the people of the following departments: technical management, maintenance management, operations management, operators trained in Brazil and Italy with extensive experience of equipment and maintenance. 3.1 Description of the Productive Process and Operational Context PRATS Laboratory of Costa Rica, specializes in the production of ophthalmic lenses and sunglasses, together with their respective treatments (against scratch, reflection, glare; mirrored, tinted, among others). This laboratory produces 1000 daily lenses between finished and semi-finished (equipment availability required = 88%). This is proposed for a good start of the factory and a competitive insertion in the Costa Rican market. PRATS manufacturing process is divided in the following areas: • • • •

Orders Workshop (TE) → TE Treatments (TRA) → TRA Control (CT) → CT Beveling and assembly (BM) → BM Simplified manufacturing process (see Fig. 3):

596

C. Parra et al.

PRODUCTION ORDER

Provision (manual location change)

Picking

Production (SAT input, TAG recording)

Lenses in STOCK

Manual Blocking (reader, generator, polishing and laser SAT with discar d in generator and polishing

Directly to MEI beveling machine

Before washing machine - Final carving (manual location change)

Control (Buckets with only one TAG) A new Albaran recorded

TAG

is

Treatments (Manual input and output)

Finished lenses Beveling – MEI (TAG)

Control Assembly (Final Control and Labelling)

Fig. 3. PRATS productive process flow diagram.

1. Choice of the necessary block and tools: the most suitable block is chosen, its material (resins CR-39, MR-8 or polycarbonate) and the molds for the tuning and polishing of each surface. 2. Generation of the anterior surface of the lens: consists of four stages: clamping, generation, tuning and polishing. 3. Intermediate Control: the first surface of the sagitta lens and the thickness are controlled. 4. Generation of the posterior surface of the lens: consists of four stages: clamping, generation, tuning and polishing. 5. Treatment: the lenses are washed and healed in ultrasonic washing machine before entering and the baking white room at approximately 120 °C do the treatment of multilayer (against scratches, reflection, glare; mirrored, tinted, etc.). This is done in clean room laboratory with controlled environment Grade C/ISO 7, air quality PM 2.5 and at 22 °C in the environment.

RQCM: Risk Qualitative Criticality Matrix. Case Study

597

6. Beveling of the lens to give it the desired shape by the client, this is carried out with the Italian machine of cutting MEI and is a crucial step, it is the final step before mounting the pair of lenses in the desired frame (semi-finished lenses). 7. Final control: Controls the quality of the surfaces, the mass and the recipe formulated for the client (Robotic AR) 8. Packaging and storage: Later it is delivered to the messengers for shipping to the customer. 3.2 Results and Analysis of the RQCM Application The results of the application of the RQCM (Table 1 and 2) are summarized, the maximum value of risk criticality that can be obtained is 200 points (TCR: Total Criticality Risk) and is distributed in 3 possible levels of hierarchizing systems (critical, semi critical and not critical for the organization). Next, the results of the RQCM tool practical application are summarized: 12 equipment evaluated in PRATS Costa RICA (November 2021). • 1 item in the critical system area (C) (8,33%) • 4 equipment in the semi critical area (SC) (33,33%) • 7 equipment in the non-critical area (NC) (58,33%) The selection of the RQCM model was justified by the effectiveness and ease of implementation of this technique in the process of the priorization. The RQCM, allows to quickly estimate the factor of frequency and consequence of failures, which can help guide the effective selection of critical equipment. The main limitation of RQCM is associated with the minimum existing information and its low quality (because it is a new production line). It is important to mention that the organization understands and recognizes the technical limitations of the RQCM and its impact on the final results of the case study presented in this report (Rodríguez 2021). For the application of the RQCM, the organization formed a working group made up of the following people (4 people): – A leader of the RQCM application (Reliability Engineering). – Two experts in the types of equipment to be evaluated (Process and Quality Engineering) – An expert in industry 4.0 (Automation and Control Engineering) The results of the 12-equipment criticality evaluation are the beginning of an optimization process. After this analysis, in the next phase of the Operational Reliability Optimization Project, the equipment: MEI Bisphera XDD-TBA beveling machine, that remained in the area of high criticality (C), will be taken; and the following reliability and risk engineering methods will be applied (Rodríguez 2021; Yoon et al. 2019): – – – –

RCA (Root Cause Analysis) RCM (Reliability Centered Maintenance) RAM-A (Reliability, Availability and Maintainability-Analysis) CRBA (Cost Risk Benefit Analysis)

598

C. Parra et al.

Table 1. Summary of results obtained in the criticality calculation (TCR). (12 systems evaluated).

Equipment

FF

PI

OF

MC

HSE

C

TCR

Satisloh Layoutblocker-PRA blocking machine

3

1

1

1

1

3

9

Laser Micromac 3D UV Laser Rxe 200

1

3

4

1

1

14

14

Satisloh Auto-Flex polishing machine

3

3

1

1

1

5

15

Satisloh VFT -ORBIT generator

2

5

4

1

3

24

48

LOH Lens lacquer machine

2

3

2

1

3

10

20

FISA CS20 4 Plus + R02 ultrasonic washer machine

1

5

4

1

3

24

24

SAT T E band Orders Workshop

2

1

2

1

1

4

8

Satisloh Multilayer MC-380-X

2

5

4

1

3

24

48

MEI Bisphera XDD-TBA beveling machine

3

10

4

1

3

44

132

SAT Eudepro T ecnic band Beveling and Assembly

2

5

2

1

1

12

24

Enduro Coating SCL CDC 1000PP2

1

5

4

1

3

24

24

Robotic AR T ype MCVP8_V2

2

3

2

1

1

8

16

4 Final Considerations When carrying out a correct application of qualitative methods of criticality analysis can help both managing and technical levels to make more efficient decisions, directly addressing both economic and human resources in the processes related to the operation and maintenance of industrial assets (Parra and Crespo 2020a, 2020b, Villar et al. 2018). It is important that maintenance management understand that criticality models to design or use should be aligned with business objectives and not make the mistake of developing criticality tools where only particular maintenance process factors are included. With respect to the latter point, using criticality models based on the Risk factor analysis, it is very interesting, since the Risk Analysis process allows to evaluate the impact of the factors inherent in the maintenance process and to add the assessment of factors such as: production, quality, production losses costs, safety, and environment, among others. Regarding maintenance management, the results of a semi-quantitative criticality

RQCM: Risk Qualitative Criticality Matrix. Case Study

599

Table 2. Summary of results obtained in the RQCM matrix. (12 systems evaluated). Department

Equipment

CTR

Ranking

BEVELING AND ASSEMBLY

MEI Bisphera XDD-TBA beveling machine

132

CRITICAL

ORDERS WORKSHOP

Satisloh Multilayer MC-380-X

48

SEMI-CRITICAL

ORDERS WORKSHOP

Satisloh VFT-ORBIT generator

48

SEMI-CRITICAL

ORDERS WORKSHOP

FISA CS20 4 Plus + R02 ultrasonic washer machine

24

NON CRITICAL

BEVELING AND ASSEMBLY

SAT Eudepro Tecnic band Beveling and Assembly

24

NON CRITICAL

TREATMENTS

Enduro Coating SCL CDC 1000PP2

24

NON CRITICAL

ORDERS WORKSHOP

LOH Lens lacquer machine

20

NON CRITICAL

CONTROL

Robotic AR Type MCVP8_V2

16

NON CRITICAL

ORDERS WORKSHOP

Satisloh Auto-Flex polishing machine

15

SEMI-CRITICAL

ORDERS WORKSHOP

Laser Micromac 3D UV Laser Rxe 200

14

NON CRITICAL

ORDERS WORKSHOP

Satisloh Layoutblocker-PRA blocking machine

9

SEMI-CRITICAL

ORDERS WORKSHOP

SATTE band Orders Workshop

8

NON CRITICAL

analysis process will enable the development of maintenance strategies and management tools with a risk-based optimization approach and its impact on the business. Finally, the results obtained from the effective application of the methodology RQCM (Risk Qualitative Criticality Matrix), guide the Industrial Assets Managers to make decisions more efficiently and with a lesser degree of uncertainty, helping to maximize the profitability of manufactured products in the PRATS Costa Rica factory throughout their entire life cycle.

References Crespo Márquez, A., Moreu de León, P., Gómez Fernández, J.F., Parra Márquez, C., López Campos, M.: The maintenance management framework. J. Qual. Maint. Eng. 15(2), 167–178 (2009). https://doi.org/10.1108/13552510910961110 Crespo Márquez, A.: The Maintenance Management Framework. Models and Methods for Complex Systems Maintenance. Springer, London (2007). https://doi.org/10.13140/RG.2.2.16765. 38884 Chiu, Y.C., Cheng, F.T., Huang, H.C.: Developing a factory-wide intelligent predictive maintenance system based on Industry 4.0. J. Chin. Inst. Eng. 40(7), 562–571 (2017). https://doi.org/ 10.1080/02533839.2017.1362357 González-Prida, V., Parra, C., Gómez, J.F., Crespo, A.: Audit to a specific study scenario according to a reference framework for the improvement of the guarantee management. In: Berenguer, G., Soares, G. (eds.) Advances in Safety, Reliability and Risk Management (2012). https://doi. org/10.13140/RG.2.2.35353.65123 ENAP SIPETROL: Matriz de Evaluación y Gestión de Riesgos, Yacimientos Pampa del Castillo – La Guitarra. ENAP INF-10-2015-CONF1, Santiago de Chile, Chile (2015) ENAP SIPETROL: Proceso de definición de criticidad desarrollado para los activos de producción Off-Shore, Magallanes. ENAP INF-10-2015-CONF2, Santiago de Chile, Chile (2016)

600

C. Parra et al.

Li, W., Wright, M.: Negative emission energy production technologies: a techno-economic and life cycle analyses review. Energy Technol. (1900871) (2019). https://doi.org/10.1002/ente.201 900871 Neurohr, C., Westhofen, L., Butz, M., Bollmann, H., Eberle, U., Galbas, R.: Criticality analysis for the verification and validation of automated vehicles. IEEE Access 9, 18016–18041 (2021). https://doi.org/10.1109/ACCESS.2021.3053159 Junietz, P., Bonakdar, F., Klamann, B., Winner, H.: Criticality metric for the safety validation of automated driving using model predictive trajectory optimization. In: Proceedings 21st International Conference on Intelligent Transportation Systems (ITSC), pp. 60–65 (2018) Parra, C., Crespo, A.: Ingeniería de Mantenimiento y Fiabilidad Aplicada en la Gestión de Activos. Modelo de Gestión del Mantenimiento (MGM). 2da ed. INGEMAN, Escuela Superior de Ingenieros Industriales, Sevilla, España (2015). https://doi.org/10.13140/RG.2.2.29363.66083 Parra, C., Crespo, M.: Nota técnica 5: Métodos de Análisis de Criticidad y Jerarquización de Activos. INGEMAN, ETSI, Sevilla, España (2018). https://doi.org/10.13140/RG.2.2.21197. 87524 Parra, C., Crespo A.: Nota técnica 4: Técnicas de Auditoría aplicadas en los procesos de Gestión del Mantenimiento. INGEMAN, España (2019). https://doi.org/10.13140/RG.2.2.10169.60003 Parra, C., Crespo, A.: Nota técnica 1: Introducción a un modelo integral de Gestión del Mantenimiento (MGM). INGEMAN, ETSI, España (2020a). https://doi.org/10.13140/RG.2.2.13046. 63049 Parra, C., Viveros, P., Kristjanpoller, F., Crespo, A., González-Prida, V.: Modelos de auditoría para los procesos de gestión de activos, mantenimiento y confiabilidad. Caso de estudio: Sector de Transmisión de Electricidad. INGEMAN, Sevilla, España (2020a). https://doi.org/10.13140/ RG.2.2.32132.14721/1 Parra, C., González-Prida, V., Candón, E., De la Fuente, A., Martínez-Galán, P., Crespo, A.: Integration of asset management standard ISO55000 with a maintenance management model. In: Crespo Márquez, A., Komljenovic, D., Amadi-Echendu, J. (eds.) WCEAM 2019. LNME, pp. 189–200. Springer, Cham (2020b). https://doi.org/10.1007/978-3-030-64228-0_17 Parra, C., Crespo, A.: Introducción al Modelo Integral de Gestión del Mantenimiento y de la Confiabilidad alineado con el enfoque de la norma: UNE 16646 (Mantenimiento en la Gestión de Activos). INGEMAN - Universidad de Sevilla, España. Artículo no. 3, pp. 5–10 (2020b) Parra, C., et al.: Metodología básica de análisis de riesgo para evaluar la criticidad de activos industriales. Caso de estudio: Línea de Manufactura de Envases Biodegradables. Edita INGEMAN, ETSI, Sevilla, España (2021a). https://doi.org/10.13140/RG.2.2.10422.52802/2 Parra, C., Viveros, P., Kristjanpoller, F., Crespo, A., González-Prida, V., Gómez, J.: Técnicas de auditorías para los procesos de: mantenimiento, fiabilidad operacional y gestión de activos (AMORMS & AMS-ISO 55001). INGEMAN, ETSI, España (2021b). https://doi.org/10.13140/ RG.2.2.35842.61124/4 Parra, C., et al.: Técnica de Jerarquización de Activos MCCR: Matriz de Criticidad Cualitativa de Riesgo. Caso de estudio: Unidad de Craqueo Catalítico. INGEMAN, ETSI, Sevilla, España (2021c). https://doi.org/10.13140/RG.2.2.31889.15209/1 Rodríguez, J.: Propuesta de diseño para un modelo de gestión de mantenimiento con acercamiento a técnicas de la Industria 4.0 para la organización: PRATS C/ Industrias de Óptica S.A. Instituto Tecnológico de Costa Rica (ITCR/TEC), Cartago, Costa Rica, pp. 44–84 (2021) Villar Fidalgo, L., Crespo Márquez, A., González Prida, V., De La Fuente, A., Martínez-Galán, P., Guillén, A.: Cyber physical systems implementation for asset management improvement: A framework for the transition. In: Safety and Reliability–Safe Societies in a Changing World, pp. 3063–3069. CRC Press, London (2018). ISBN: 9781351174664. https://doi.org/10.1201/ 9781351174664

RQCM: Risk Qualitative Criticality Matrix. Case Study

601

Viveros-Gunckel, P., Kristjanpoller-Rodríguez, F., Parra-Marquez, C., Crespo-Márquez, A., González-Prida-Díaz, V.: Audit models for asset management, maintenance and reliability processes. Case study: electricity transmission sector. DYNA Manag. 8(1), 14 p. (2020). https:// doi.org/10.6036/MN9826 Yoon, J.T., Yooung, B., Yoo, M., Kim, Y., Kim, S.: Life-cycle maintenance cost analysis framework considering time-dependent false and missed alarms for fault diagnosis. Reliab. Eng. Syst. Saf. 184, 181–192 (2019). https://doi.org/10.1016/j.ress.2018.06.006 Grupo PRATS: Sobre nosotros (2021). https://www.grupoprats.com/lentes/espanol/portada_120_ 1_ap.html MEI SYSTEM: Máquinas para el corte de lentes/MEI BISPHERA XDD-TBA (2020). https://mei system.com/es/ SATISLOH: The art of making lenses. Recuperado de (2020). https://www.satisloh.com/

Economic and Environmental Indicators for Assessing Energy Efficiency Improvements in the Smart Manufacturing Processes Minna Räikkönen1(B) , Teuvo Uusitalo1 , Saara Hänninen1 , Andrea Barni2 , Claudio Capuzzimati2 , Alessandro Fontana2 , and Marco Pirotta2 1 VTT Technical Research Centre of Finland Ltd., P.O. Box 1300, 33101 Tampere, Finland

{minna.raikkonen,teuvo.uusitalo,saara.hanninen}@vtt.fi 2 Scuola Universitaria Professionale della Svizzera Italiana, via la Santa 1, CH-6962 Viganello, Switzerland {andrea.barni,claudio.capuzzimati,alessandro.fontana, marco.pirotta}@supsi.ch

Abstract. Since the manufacturing industry is one of the major global energy consumers and carbon emitters, energy efficiency has emerged as one of the industry’s key drivers. Additionally, digital technologies offer companies significant opportunities to boost productivity and generate cost savings while simultaneously reducing the environmental impact. This study establishes a framework for economic and environmental indicators supporting smart manufacturing and asset management operations. The framework contributes to the sustainability assessment of digital solutions focused on increasing energy efficiency. In this setting, the emphasis is particularly on LCC (Life Cycle Costing) and LCA (Life Cycle Assessment) indicators. The approach is being tested in four different types of pilot companies in the manufacturing sector. The study examines the indicators from several angles at the process and machine levels. In the next phase of our research, software tools for the energy efficiency-oriented online LCC and LCA will be developed to make the indicator framework practical.

1 Introduction Energy efficiency has become one of the key drivers in the manufacturing sector. This is because manufacturing represents a large part of global energy consumption and is responsible for 20% of global carbon emissions (WEF 2020). In addition, the increase in energy prices and the growing importance of sustainability have put new pressure on companies to improve their energy management strategies and systems as well as the energy efficiency of manufacturing processes. In this context, investing in novel technologies and adopting technology to intelligently control energy use can significantly reduce energy consumption (Mawson and Hughes 2019). Increasing energy efficiency driven by novel digital technologies provides companies with opportunities to reduce both the cost and greenhouse gas emissions of manufacturing and asset management processes. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 602–611, 2023. https://doi.org/10.1007/978-3-031-25448-2_56

Economic and Environmental Indicators

603

Indicators are vital when defining targets for energy and asset performance management. These are typically numerical measures that provide quantitative information that goes beyond simple data to show trends or cause-and-effect relationships (Bhanot et al. 2020). In the context of cost and environmental indicators, assessment methods like life cycle costing (LCC) and life cycle analysis (LCA) are of great importance (IEC 2017; ISO 2021, 2018, 2006). Both LCA and LCC belong to the group of sustainability assessment tools (Hoogmartens et al. 2014). Life Cycle Assessment (LCA) quantifies all relevant emissions, resources consumed, environmental and health impacts, and resource depletion issues associated with any good or service (“system”). (European Commission, Joint Research Centre et al. 2016; ISO 2006). Life cycle costing (LCC) can be defined as an iterative process of planning, estimating and monitoring all expenses associated with a product, process, sub-process, or project, including acquisition and all associated costs, operation and maintenance (O&M), refurbishment and retirement (Gibson et al. 2005; ISO 2021). Thus, both the LCA and LCC provide valuable information on possibilities to minimize costs and improve manufacturing processes’ energy efficiency where digital technologies are deployed. However, currently, these methods and related indicators typically focus on the calculation of the environmental and cost impact of manufactured products, not the processes with multiple assets which are in different stages of their lifecycle (ISO 2021, 2006; Saad et al. 2019). Although several indicators-based frameworks have been developed, they differ in their goal and scope and are intended for various uses and stakeholders such as companies, consumers, or authorities (Bhanot et al. 2020). Furthermore, it is necessary to assess both the economic and environmental impacts of novel technologies from the point of view of energy efficiency (Apostolos et al. 2013; Kluczek 2019). Environmental and cost considerations should also be part of any asset management plan and included also in asset and repair-replacement decisions (Abdi and Taghipour 2019; Räikkönen et al. 2016). Furthermore, despite advances in databases and software platforms that support the gathering and calculation of the data for LCC and LCA and related parameters, the procedure, is still remarkably labour-intensive and time-consuming. (Barni, et al. 2018).

2 Research Method The study establishes a framework for cost and environmental impact indicators on the process and machine levels and focuses especially on LCC and LCA indicators in this context (Fig. 1). The research approach is challenge-driven and follows the principles of strategic technology management research by examining how to optimize asset energy consumption and energy efficiency, and simultaneously diminish the cost and environmental emissions. The approach is inspired by the principles of design science (DSR), which is a qualitative research approach that simultaneously generates knowledge about the method used to design an artefact and the design of the artefact itself (Simon, 1996). It emphasizes problem-solving as a source of innovation and thus provides an opportunity to create new knowledge and to make managerial design propositions (Hevner and Gregor 2022). The paper is based on the research carried out in the DENiM project (https:// denim-fof.eu/) co-funded by the European Union under the H2020 Programme. In the

M. Räikkönen et al.

Production site level Production line level Production process level

Life cycle costing - LCC * Cost impact

Production asset level Life cycle assessment - LCA * Environmental impact

ECONOMIC AND ENVIROMENTAL INDICATORS

604

Fig. 1. Research approach: Energy efficiency and indicators at different levels.

project case studies, we had an opportunity for close manufacturing industry-research collaboration within four pilot companies. Pilot 1 is a medical device manufacturer conscious of its environmental impact that will leverage technologies to achieve sustainable production planning and organization the use of renewable energy. Currently, evaluating indicators is a heavily manual and flawed task in terms of data collection and analysis. Pilot 2 consists of two companies that are part of the same value network in the steelmaking industry for the automotive sector. The research project will support the improvement of the current production by addressing the energy-efficient management of the steelmaking and forging processes and related sustainability assessment. Pilot 3 is a manufacturer of tools for sheet metal processing working with very high energy-intensive processes. The project will support the identification of the most promising optimisations in terms of energy usage by extensively integrating digital twins of machining processes for production planning optimisation. Pilot 4 is an SME focusing on the production of mechanical components for the machinery industry. Acting on its highly variable production process the project will support the integration of IoT-based solutions and sustainability assessment for energy management, combining digital twin and energy modelling and optimisation technologies. The aim of the indicator framework and indicators in the case studies is to measure and calculate the sustainability impact of the digital technologies developed in the project and to ensure that the sustainability targets (see Table 1) set for the companies’ manufacturing processes are achieved. In addition, the environmental and economic indicators linked to energy efficiency will be integrated into the LCC and LCA tools to be developed in the project and support companies in reducing the need for manual data collection.

Economic and Environmental Indicators

605

Table 1. Exploring cost savings and positive environmental impact of energy efficiency improvements in industry and academy collaboration – four practical pilot cases. Cases

Production

Pilot 1 Medical device manufacturing

Sustainability target in the project Cost reduction of 15–20% based on overall efficiencies, Energy reduction of 30–35%, Waste & scrap reduction: − 5–10%

Pilot 2 Steel and steel components manufacturing Cost reduction of 15% across the value chain, Energy reduction of 20–25% across the value chain, Waste & scrap reduction through defect detection: −5–10% Pilot 3 Tools production for metals processing

Cost reduction of 12% extending the lifetime, Energy reduction of 20% of the process, Scrap reduction: 10%

Pilot 4 Plastics processing

Cost reduction of 15–20%, Energy reduction of 40% of the entire process

3 Proposed Indicators for Assessing Energy Efficiency Improvements During our research, workshops with pilot companies were held and used to lay the foundation for the development of the indicator framework. In the workshops, an online MURAL board was used to facilitate the discussion. Each workshop was organized with one company at a time to be able to consider its focus areas and company-specific aspects. The proposed framework considers key economic and environmental indicators that ensure achieving the energy efficiency targets of the selected manufacturing process and its assets. By implementing smart energy consumption control systems and maintenance solutions, energy efficiency improvements are assumed and expected across all dimensions of the profit and loss statement and environmental reporting. These include reduced costs (e.g., minimized energy consumption, automated processes, shorter processing times), revenue generation, reduced environmental impacts related to energy consumption, and improved risk management (e.g., reduced risk by using precise and timely data). The originality of the proposed model is mainly related to the integration of asset energy efficiency and economic and environmental sustainability into manufacturing processes and to be able to optimize the asset operations from a sustainability point of view (Fig. 2). 3.1 Economic Indicators The scope of the pilot companies in our research was to evaluate the production costs of the targeted processes and, in some pilots, the sub-processes and selected products, considering also indirect expenses and analysing different time horizons, based on information already available and obtained from the pilot workshops.

M. Räikkönen et al.

Energy efficient manufacturing process Pilot 2 Production asset 1

Pilot 1 Content analysis: literature and earlier projects

Pilot 3

Pilot 4

Production asset 2

Production asset n

Energy efficiency oriented O&M performance measures

606

Economic & Cost Indicators * CAPEX (Ownership cost: production assets) * OPEX (Direct & Indirect Operation and Maintenance (O&M) cost) * profitability of novel digital technologies improving energy efficiency

Enviromental Indicators * Inventory (waste, emissions, material, water, energy) * Impact (Product Environmental Footprint (PEF) & Organisation Environmental Footprint (OEF))

Fig. 2. The high-level categorization of economic and environmental indicators for energy efficient digitalized manufacturing process.

The cost indicators in our study were aligned with the LCC standards and the cost breakdown structure was used to divide the main cost categories into more concrete and easily estimated cost elements, which aided in the definition of cost functions that were used in the calculation of the KPIs under consideration. The most important direct Operation and Maintenance (O&M) cost (OPEX) in the pilot companies is the energy cost, especially the electricity cost of the production assets and the cost of electricity needed to produce the compressed air. As electricity consumption increases with the production volumes, it is considered a direct cost. Energy costs related to the production process can be decreased, for example, by using off-peak-energy hours for the most electricity consuming operations, performing scheduled maintenance periodically, upgrading ageing equipment and by buying the best commercial electricity rates. In general, lowering the demand for kW while using kWhs for efficient manufacturing aids in cost reduction. There are also options for energy cost accumulation and assignment in cost accounting that can be applied in the companies depending also on their current cost accounting systems. The proposed indicator framework explicitly includes indirect production costs as well. This results in three main cost categories: 1. Scrap and waste, 2. Rework and 3. Unavailability cost, all of which are affecting energy efficiency (Fig. 3). In addition, there are also other indirect impacts, such as impacts on inventory cost. From the CAPEX point of view, especially the investment cost for digital technology developments, as well as the cost of major asset enhancement and replacements during an asset’s lifetime, are of interest. When calculating the economic indicators, cost functions must, in most cases, be tailored to meet the needs of the company in question. The LCC results in the pilot companies were summarized and presented both in numerical and graph forms. If several production assets were evaluated at the same time, the comparison of the results of different assets at the process level over the calculation term was also of interest. The main LCC result indicators calculated in the pilot companies were: e/machine operating hour, e/month, e/year, e/machine lifetime, e/calculation period (selected number of years), yearly and cumulative cash flows for each machine and the process including several machines (e), both discounted and nondiscounted values. During the study, it was also extremely important to ensure that the

Economic and Environmental Indicators

607

cost functions developed and applied are consistent with the data that was available and accessible, as well as in a usable form that allows compatibility with the same type of data from different sources. When assessing the LCC for the selected processes, it is evident that a wide range of data - both economic and process-related technical data - is required. Different data collection methods can be used to estimate the life cycle cost, primarily depending on data availability and accessibility, as well as the costs to be analysed: 1. Using constant values previously defined by the pilot company or other stakeholders, such as the cost of maintenance services or the cost of electricity. 2. Calculating cost and other parameters based on technical details of the process and its assets, such as energy consumption, personnel working hours and salary. 3. Parameter value estimation based on available statistics, such as maintenance intervals for various machines. 4. Expert judgments for values with no other data source, such as the price of an unavailability hour.

Fig. 3. Direct and indirect O&M cost indicators of the framework.

608

M. Räikkönen et al.

3.2 Environmental Indicators The environmental inventory indicators in our study were aligned with the GRI (Global Reporting Initiative) standards and thus consider all key dimensions of environmental sustainability: materials, energy, water, emissions, and waste (Barni et al. 2022). However, the focus is on energy at the process and asset level emissions. The impact indicators were selected from the list provided by PEF (Product Environmental Footprint) and OEF (Organisation Environmental Footprint), after which screened considering EPD (Environmental Product Declarations) (Mayer et al. 2020). The inventory indicators within the framework consider only direct data, which means that only resource use and emissions involved in production will be calculated. This approach guarantees that pilot companies do not collect LCI data throughout their entire supply chain. Furthermore, this provides a more direct overview of the measures that companies can use to influence their actions. Direct material: Non-renewable materials, such as minerals, metals, oil, gas, or coal, can be used to manufacture and package a company’s products and services, as can renewable materials, such as wood or water. In both cases, the input materials could be virgin or recycled. The amount of materials used and their sources can reveal the company’s reliance on natural resources and the effects it has on their availability (GRI standards 2016.a). Direct energy: Energy sources used in manufacturing facilities and processes include electricity, fuel, heating, and steam. It can be purchased or self-generated from either renewable (wind, hydro, or solar) or non-renewable sources (coal, natural gas). Implementing energy-saving strategies and increasing the share of renewables is critical to combating climate change and reducing an organization’s overall environmental footprint. Direct water: The amount of water withdrawn and consumed by manufacturing processes, as well as the quality of discharges, may have varying negative effects on the ecosystem. Water consumption, for example, can have an impact on the area’s quality of life, with social and economic consequences for local communities. Furthermore, an organization can assess the impacts it has on water resources through knowledge and measured data. This benefits the ecosystem, other users, and the organization itself. Direct emissions: Emissions into the air are defined as the discharge of substances from a source into the atmosphere. These pollutants damage the climate, ecosystem, air quality, agriculture, and human health. Deterioration of air quality and acidification have resulted in their regulation under international conventions and national laws, including those listed on an organization’s environmental permits (GRI standards 2016.b). Direct waste: Waste is produced by the activities of the company during production processes or service delivery. When waste treatment and disposal are not properly managed, they have a significant negative impact on humans, and these effects frequently extend beyond the locations where waste is generated. To accelerate depletion, waste resources and materials are incinerated or landfilled and lost for future use (GRI standards 2016.c). In terms of impact KPIs, PEF and OEF provide a comprehensive list of 16 indicators that cover all environmental aspects from raw material extraction to use and final disposal. These methodologies provide a standardised measurement of a system’s performance,

Economic and Environmental Indicators

609

allowing improvement actions to be used to improve impact categories. Furthermore, using Product Category Rules, EPD identifies a subset of 7 indicators and methods that are the most significant in the context of the energy efficiency of the manufacturing processes (see Table 2). Table 2. The most significant environmental impact indicators in the context of the energy efficiency of the manufacturing processes. Impact category

Impact category indicator

Unit

Climate change, total

Radiative forcing as global warming potential (GWP100)

kg CO2 eq

Photochemical ozone formation, human health

Tropospheric ozone concentration increase

kg NMVOC eq

Acidification

Accumulated Exceedance (AE)

kg SO2 eq

Eutrophication, freshwater

Fraction of nutrients reaching freshwater end compartment (P)

kg PO43 eq

Water use

User deprivation potential (deprivation-weighted water consumption)

m3 H2 O eq

Resource use, minerals and metals

Abiotic resource depletion (ADP kg Sb eq ultimate reserves)

Resource use, fossils

Abiotic resource depletion – fossil fuels (ADP-fossil)

MJ

The indicators selected and included in the framework can be seen as the ideal combination of indicators suitable for various types of manufacturing industry. During our study, we applied the indicator framework in each pilot company, gathered the data needed and made the calculations to test and to validate the framework. The selection of the indicators was carefully done together with the pilot companies. In the next phase of our project, we will incorporate the indicators into the LCA and LCC tools that will be developed for the pilot companies, as well as enhance the tools to enable online continuous assessment of the environmental and cost impacts of manufacturing processes.

4 Conclusions The developed framework discussed briefly in this paper contributes to an increased understanding of the economic and environmental aspects of improving energy efficiency in smart manufacturing processes. The indicator framework supports energy and asset management and the assessment of cost savings and environmental impact generated by digital technologies that aim to improve energy efficiency both at the process and asset levels. The proposed model’s novelty stems primarily from the incorporation

610

M. Räikkönen et al.

of asset energy efficiency and economic and environmental sustainability into manufacturing processes, as well as the ability to optimise asset operations from a sustainability standpoint. Because LCA and LCC are commonly used to calculate the environmental and economic impact of manufactured products, our framework focuses on processes involving multiple assets at various stages of their lifecycle. Choosing economic and environmental indicators that influence energy efficiency is a complex process. Several indicators can be measured, but not all of them are relevant. Monitoring the inaccurate indicator is a waste of time and resources when there is something more important to measure. Knowledge and understanding of which indicators have the greatest impact enable businesses to direct their expertise toward the activities that influence such indicators. Economic and environmental indicators, from the perspective of energy and asset management, can generate information that is different, if not contradictory, to those related to the overall process and asset effectiveness and performance. Decision makers must balance sustainability and process performance targets while also understanding the cause-and-effect relationships between critical indicators. Furthermore, economic and environmental indicators support the development of sustainable processes, implying that LCC and LCA analyses are critical for improving the energy efficiency of manufacturing processes. For example, the need to calculate and communicate the positive environmental benefits at the process and asset level is clear, yet there has been a clear lack of suitable methods to achieve this. Acknowledgments. This paper is supported by the European Union’s H2020 research and innovation programme under grant agreement Nº 958339, project DENiM (Digital intelligence for collaborative ENergy management in Manufacturing (https://denim-fof.eu/).

References Abdi, A., Taghipour, S.: Sustainable asset management: a repair-replacement decision model considering environmental impacts, maintenance quality, and risk. Comput. Ind. Eng. 136, 117–134 (2019). https://doi.org/10.1016/j.cie.2019.07.021 Apostolos, F., Alexios, P., Georgios, P., Panagiotis, S., George, C.: Energy efficiency of manufacturing processes: a critical review. Procedia CIRP 7, 628–633 (2013). https://doi.org/10.1016/ j.procir.2013.06.044 Barni, A., et al.: Design of a lifecycle-oriented environmental and economic indicators framework for the mechanical manufacturing industry. Sustainability 14, 2602 (2022). https://doi.org/10. 3390/su14052602 Barni, A., Fontana, A., Menato, S., Sorlini, M., Canetta, L.: Exploiting the digital twin in the assessment and optimization of sustainability per-formances, In: Intrenational Conference on Intelligent Systems (IS). Presented at the International Conference Intell. Syst.: Theory, Res. Innov. Appl., Institute of Electrical and Electronics Engineers Inc., pp. 706–713 (2018) https:// doi.org/10.1109/IS.2018.8710554 Bhanot, N., Qaiser, F.H., Alkahtani, M., Rehman, A.U.: An integrated decision-making approach for cause-and-effect analysis of sustainable manufacturing indicators. Sustainability 12, 1517 (2020). https://doi.org/10.3390/su12041517

Economic and Environmental Indicators

611

European Commission, Joint Research Centre, Cristobal-Garcia, J., Pant, R., Reale, F., Sala, S.: Life cycle assessment for the impact assessment of policies. Publications Office of the European Union, LU. (2016). https://doi.org/10.2788/318544 Gibson, R.B., Holtz, S., Tansey, J., Whitelaw, G., Hassan, S.: Sustainability Assessment - Criteria and Processes. Routledge (2005) ISBN 978–1–84407–051–0 GRI standards. GRI 301: Materials 2016 (2016a) ISBN: 978–90–8866–104–4 GRI standards. GRI 305: Emissions 2016 (2016b) ISBN: 978–90–8866–108–2 GRI standards. GRI 306: Effluents and waste 2016 (2016c) ISBN: 978–90–8866–109–9 Hevner, A., Gregor, S.: Envisioning entrepreneurship and digital innovation through a design science research lens: a matrix approach. Inf. Manage. 59, 103350 (2022). https://doi.org/10. 1016/j.im.2020.103350 Hoogmartens, R., Van Passel, S., Van Acker, K., Dubois, M.: Bridging the gap between LCA, LCC and CBA as sustainability assessment tools. Environ. Impact Assess. Rev. 48, 27–33 (2014). https://doi.org/10.1016/j.eiar.2014.05.001 IEC. IEC 60300. Dependability management – Part 3–3: Application guide – Life cycle costing (2017) ISBN 978–2–8322–3886–8 ISO. ISO 15663 - Petroleum, petrochemical and natural gas industries — Life cycle costing (2021) ISO. ISO 14067: Greenhouse gases — Carbon footprint of products — Requirements and guidelines for quantification (2018) ISO. ISO 14040: Environmental management — Life cycle assessment — Principles and framework (2006) Kluczek, A.: An energy-led sustainability assessment of production systems – an approach for improving energy efficiency performance. Int. J. Prod. Econ. 216, 190–203 (2019). https://doi. org/10.1016/j.ijpe.2019.04.016 Mawson, V.J., Hughes, B.R.: The development of modelling tools to improve energy efficiency in manufacturing processes and systems. J. Manuf. Syst. 51, 95–105 (2019). https://doi.org/10. 1016/j.jmsy.2019.04.008 Mayer, F.D., Brondani, M., Vasquez Carrillo, M.C., Hoffmann, R., Silva Lora, E.E.: Revisiting energy efficiency, renewability, and sustainability indicators in biofuels life cycle: analysis and standardization proposal. J. Clean. Prod. 252, 119850 (2020). https://doi.org/10.1016/j.jclepro. 2019.119850 Räikkönen, M., Välisalo, T., Shylina, D., Tilabi, S.: Supporting Asset Management DecisionMaking—New Value Creation Perspective. In: Koskinen, K.T., Kortelainen, H., Aaltonen, J., Uusitalo, T., Komonen, K., Mathew, J., Laitinen, J. (eds.) Proceedings of the 10th World Congress on Engineering Asset Management (WCEAM 2015). LNME, pp. 479–486. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-27064-7_46 Saad, M.H., Nazzal, M.A., Darras, B.M.: A general framework for sustainability assessment of manufacturing processes. Ecol. Ind. 97, 211–224 (2019). https://doi.org/10.1016/j.ecolind. 2018.09.062 Simon, H.: The Sciences of the Artificial. The MIT Press, MA, Cambridge (1996) WEF. How manufacturing can thrive in a digital world and lead a sustainable revolution. This article is related to the World Economic Forum’s Annual Meeting in Davos-Klosters, Switzerland, 21–24 January 2020 (2020)

Reliability and Resilience Engineering

Resilience Exposure Assessment Using Multi-layer Mapping of Portuguese 308 Cities and Communities Seyed M. H. S. Rezvani1(B) , Nuno Almeida1 , Maria João Falcão Silva2 , and Damjan Maletiˇc3 1 CERIS, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais 1,

1049-001 Lisboa, Portugal {seyedi.rezvani,nunomarquesalmeida}@tecnico.ulisboa.pt 2 Laboratório Nacional de Engenharia Civil, Av. do Brasil 101, 1700-075 Lisboa, Portugal [email protected] 3 Faculty of Organizational Sciences, University of Maribor, Kidriˇceva 55a, 4000 Kranj, Slovenia [email protected]

Abstract. Resilience evaluation systems have been used to aggregate and calibrate multi-dimensional resilience-related inputs. But these evaluation systems do not offer comprehensive outputs nor sufficient analytical capacity to support the design and implementation of city and community recovery and resilience plans. Urban resilience research covers multiple fields of study like earthquakes, floods, and tsunamis, amongst other types of disaster risks. This study focuses on the strategic needs of resilience and recovery plans for Portuguese cities and communities and proposes a sophisticated mapping system to address this gap. The main output of the proposed approach is a multi-layer heatmaps with scores based on various disasters for all Portuguese cities. The resilience score is obtained through real statistical geo-data analysis.

1 Introduction From 2016 till 2020, more than $600 billion, corresponding to about $1,800 per person in the US, has been spent to cover the total cost of disaster worldwide, and, unfortunately, it is not possible to avoid the continued growth of these expenses for the next years. In this circumstances, rapid and effective recovery of the lost performance is an important issue for the Urban Resilience of sustainable cities and societies. On the other hand, based on the United Nations (UN) sustainable development goals to enhance innovation in infrastructure (Goal 9), establishing sustainable cities and communities (Goal 11), it will be need preparedness for climate change as climate actions (Goal 13), the need for a decision tool is identified (Duarte et al. 2021; Rezvani et al. 2021). Urban resilience evaluation is becoming a progressively significant subject, with senior management of public and private organizations focusing on the requirements to protect and maximize the value derived from the urban built environment and its constructed assets (Almeida et al. 2021; Falcão Silva et al. 2020). The authors have © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 615–623, 2023. https://doi.org/10.1007/978-3-031-25448-2_57

616

S. M. H. S. Rezvani et al.

previously proposed an Urban Resilience Evaluation System (URES) combined with an Automated Rational and Consistence Decision Making (ARCDM) tool attempting, in this manner, to address the subjective bias of different influencers and decision-makers involved in improving urban resilience (Duarte et al. 2022; Rezvani et al. 2022). The pilot-test applications of this combined solution showed limitations for achieving a comprehensive result in the context of urban resilience on a large scale (a country). The purpose of the research study presented is to propose a solution to an existing resilience measurement challenge and to facilitate the move from firm and static to multi-disciplinary and dynamic event resilience measurement by considering GIS based information and mapping through whole country. The research developed comprises, besides the present introduction (1), a brief literature review (2) incorporating the conceptual background of the main topics addressed, namely asset management and risk management, as well as urban resilience. The following section corresponds to the presentation of the methodology (3) developed and used. Afterwards are presented and discussed the main preliminary results obtained by the development of multi-level mapping of urban resilience for the Portuguese territory (4), ensuring efficient decisionmaking for all of those engaged in life - cycle management and resilience improvement of constructed assets. By the end of the paper the main conclusions and future work developments (5) are highlighted.

2 Multi-level Mapping of Urban Resilience for the Portuguese Territory Natural and man-made catastrophes have become more common every year across the world, putting a strain, and adversely affecting economies and societies. To lessen the susceptibility and danger of these disruptions, it is necessary to foresee and prepare the buildings and infrastructures as asset systems for Natural and man induced catastrophes, becoming more resilient. The interconnectedness of aging building assets and infrastructures, especially with significant population boom, worsens the issue for a resilient society confronting diverse sources crises (ECCE 2020). Resilience may be described relying at the term‘s context. Regarding the resilience of built property, a probable definition resulting from a mixture of different proposals, is the intrinsic capacity of built property to take in and adapt to the disruption and restoration of its purposeful performance (Hosseini et al. 2016) (Fig. 1). Resilience is a collaborative effort between individuals and organizations that are conscious and competent to engage, and the authorities, which should contribute to maximize urban resilience (Komljenovic 2020). For this purpose, policy-driven, regulatory, or programmatic measures are all possible. A well-formulated urban resilience strategy’s implementation process should establish techniques for successful delivery of agreed-upon goals via policy gateway, accountable parties, and assessment of financial, social, and political viability (Maleti´c et al. 2021). Even though the resilience domain has recently received a lot of attention, (Raškovi´c et al. 2020), across many cases there is still a lack of broad agreement including how to assess resilience at the built assets level (i.e., buildings and infrastructures). Therefore, in this area, it is noteworthy several approaches to incorporate resilience-related aspects

Resilience Exposure Assessment Using Multi-layer Mapping

617

Fig. 1. Level of performance: more vs less resilient, adapted from (Maleti´c et al. 2021).

in built-environment assessment methods, like: i) REDi (Almufti and Willford 2013); ii) ARMS (Burroughs 2017); iii) RELi (USGBC 2018); iv) LiderA (Pinheiro 2011, 2020). Developing resilience evaluation systems is nowadays a trend on a global scale, with several nations, such as Australia (ARMS, BRRT) as well as the United States of America (RELi, FORTIFIED, ANCR, BRLA), having achieved noteworthy progress, particularly concerning natural disasters caused by climate and seismic threats. The growing body of research that relate the principle of urban resilience is increasing at an exponential rate, a trend that is notably prominent in the disciplines of climate change, global warming, and natural disasters (Meerow and Newell 2019). A systematic framework to urban resilience can provide a unified system and shared knowledge for responding to the changing environment, enabling local authorities, corporations, service providers, individuals, and communities in pursuit of a common objective. This needs a holistic approach to concerns ranging from education and health care to social bonding, security, and safety. The construction of a unique tool and a comparison of urban resilience scenarios developed for a local case study with the offered instrument are presented in dynamic evaluation of urban resilience to natural hazards. In particular, the use of stochastic modelling in a system dynamics model with output in the form of an index score enables for the assessment of urban resilience possibilities. The calibration, validation and simulation suggest the tool’s suitability for evaluating various urban resilience scenarios, and therefore for developing urban resilience strategies within urban policy planning (Feofilovs and Romagnoli 2021), focusing on social dimensions within a framework to improve resilience. However, the tools could be more effective if more environmental dimensions and indicators were included in the proposed scenario analysis to connect more globally, as well as having location-based solutions based on GIS data to provide decision-makers with a more robust tool. The mapping provides a visual representation of how to transfer rigid numbers into a visualized tool.

618

S. M. H. S. Rezvani et al.

Developing resilience evaluation is now a trend on a global scale. The incorporation of GIS information modelling, a multi-criteria decision method, and a composite indicator approach into a single instrument allows for the elimination of the limitations of present resilience measuring methods. The GIS-based method enables the definition of dynamic urban resilience behaviour in dimensions of urban settings, including interaction and feedback between dimensions, indicators, and parameters, including social, economic, and environmental factors, alongside illustrating short-term and long-term urban resilience views. The GIS-based method with ARCDM allows for the scenario simulation of natural hazards inside the system dynamics model, allowing the uncertainty of disaster risk management to be represented within the evaluation of urban resilience (Rezvani 2021).

3 Methodology The Automated Rational and Consistent Decision Making (ARCDM), based on Multi criteria Decision analysis (MCDA) especially Analytical hierarchy process (AHP), is a process for simulating scenarios through stochastic analysis (Rezvani and Almeida 2021). The methodology proposed comprises i) transforming national statistical data to indicators; ii) establishing intrinsic information of cities as indicators; iii) developing, for country-level indicators, an Automated Rational and Consistent Decision Making (ARCDM) process by simulating different scenarios, which is based on Multicriteria Decision Analysis (MCDA) complemented by an Analytical Hierarchy Process (AHP); iv) developing multi-level mapping of cities using QGIS. The scenarios stochastic and statistical analysis is performed on the indicator level to ensure that all the possible scenarios have been counted as exposure risk of the disruptions. The proposed method is applied to the most recent national statistical official data from INE (https://www.ine.pt/) for all 308 Portuguese municipalities. The outputs are analysed and presented in Sect. 4. Results and discussions. The datasets for these municipalities are processed using both URES and ARCDM methodologies and are finally presented as GIS maps for a more comprehensible perception and better support of decision making. These indicators are scalable and may include extensive aspects of society. Nevertheless, to obtain a full model, it needs to be expanded to multivariate and multi-scenario analysis employing spatial data to provide geoinformation results with greater resolution. To achieve stability and robustness, additional decision-making tools (Rezvani and Almeida 2021) and statistical analysis will be used. These maps were generated based on data availability and the most important indicators associated to fire and flood, with more relevant indicators being included in future research.

Resilience Exposure Assessment Using Multi-layer Mapping

619

4 Result and Discussion The preliminary effects of the study under development show that the disaster risks, like fire and flood, have direct and extremely important impacts on the performance of a city and/or a community in the context of urban resilience. However, by considering statistical national data from INE and performing an ARCDM, it is possible to contribute and enlarge the Environmental Dimension (D1) considered in the resilience classification scoring system developed and proposed in previous research (Duarte et al. 2022; Rezvani et al. 2022). The criteria used in the ARCDM are calibrated for the Portuguese context, considering a wide awareness of environmental concerns, with a focus on the area’s vulnerability to moderate and elevated levels of natural catastrophe risks, offering an overview of potential dangers as well as finding the intrinsic aspects of the research territory, such as elevation, proximity to the ocean and streams, slopes, among others. When a fire breaks out, cities and communities that lack strategy and planning perform worse in terms of resilience than those that do, especially when taking into account green areas in or near the cities and communities under study that may also play a role in fire emergencies and improve resilience. This aims to contribute to the discussion on the role of green areas in fire emergencies. The best indicators for assessing each city and/or community exposure to fire disasters may be considered the combination of available green areas and forests (Fig. 2a), as well as the fact of the fire’s trigger, based on the highest temperature ever recorded in the hottest months of the previous year (Fig. 2b) which can happen again as a heat wave, a very common problem in the recent effects of global warming. In the context of Urban Resilience other indicators with major importance corresponds to the demographic structure of the Portuguese population. Furthermore, age group combinations have a considerable impact on exposure, vulnerability, and resilience. According to INE statistic records, the age group combinations in Portuguese territory comprise four different categories (0–14, 15 to 24, 25 to 64, >65) which present major importance in the presence of a disaster and considering the cities and/or communities’ ability of recovery. The first (0–14) and last groups (>65) are inactive or non-working populations, which correspond that is, those aged 0 to 14 and 65 and older are unable to cope with disruptions on their own and, as a result, require assistance to recover from the effects of natural disasters (Fig. 2c). These groups naturally increase the vulnerability during natural disasters and slow down the recovery phase, which takes longer to allow society to return to its functional performance. The results obtained from the ARCDM with AHP performed allow us to better understand the population distribution and its implications in the cities and/or communities’ resilience, in the present situation which also concerns not only fire risk exposure but also flood hazard index. By taking normalized mean of these three geo-indicators (vegetation cover, the highest recorded temperature in the warmest months of the last year as well as the demographic structure of the population) the fire risk exposure, with direct impacts on the cities and communities’ resilience, may be considered and divided into five sequential categories (very low, low, medium, high, very high) which play a very relevant role in the one-to-nine odd number AHP decision analysis calculation for fire risk exposure for 308 municipalities (Fig. 3a).

620

S. M. H. S. Rezvani et al.

Fig. 2. Indicators for assessing city and/or community exposure to fire disasters: a) Forest coverage by percentage; b) maximum temperature in the hottest month; c) Age ratio distribution

Flood is one of the other hazards under preliminary evaluation in the scope of the present research implementation, being its assessment based on the corresponding indicator achieved from the results available on the INE website, added with some other intrinsic and location-based assessing parameters. Considering the flood hazard index determined for each one of the 308 Portuguese municipalities (Fig. 3b) were used, as

Resilience Exposure Assessment Using Multi-layer Mapping

621

complementary parameters, the: i) water % area; ii) precipitation (mm) and precipitation anomaly, and iii) maximum elevation. Additionally, and to improve the obtained results, were also considered, in the flood index reverse parameters, namely: i) minimum elevation, and iii) distance sea by km. The flood hazard map is generated by using the normalized mean of the scores acquired from the 2021 statistical geo-data.

Fig. 3. Hazards under evaluation for assessing city and/or community: a) Fire risk exposure; b) flood hazard index

5 Conclusions In this study each intrinsic feature is related with natural disaster risks at the national level, and the exposure and susceptibility of the cities under study have been estimated to receive a greater insight of each municipality’s intrinsic characteristics. Cities along the shore are more vulnerable to tsunamis and floods, but they have more resilient economic and demographic qualities that allow them to return quickly to their previous performance as the resilience factor. Municipal fire exposure and vulnerability are major concerns for regions with denser inhabitants and forests together with the highest temperatures during the warmest months. This serves as a reminder to be more exposed and vulnerable-prepared while dealing with fire. A city’s high population density and high density of wooded regions both raise the risk of fire exposure. The likelihood of being exposed to high temperatures during the warmest months of the year rises when a forest geographical area is present in a city, which is a risk factor for fire exposure. On the basis of comparable rainfall

622

S. M. H. S. Rezvani et al.

anomalies from the previous year, concerns have been expressed about probable floods brought on by low elevation and heavy precipitation, as well as the density of waterbodies, including such rivers and streams. As part of this study series, more characteristics that can significantly affect these indices will be investigated in the future. The closeness of areas to rivers and streams puts them at a significant danger of flooding. Due to the existence of waterbodies combined with an anomalous amount of rain, is a risk factor for floods. Actually, the concept of urban resilience is founded on the assumption that cities are highly vulnerable to natural disasters and man-made hazards. Having the concept of adaptive indicators based on GIS-based modelling can aid in the identification of vulnerable regions in urban areas. For example, analysing flood risk in metropolitan areas may be done using GIS-based flood and fire modelling, as demonstrated in this study based on specific flood and fire indicators. Another strategy is to encourage bottom-up approaches rather than top-down approaches. Because even if all of the data were aggregated in one superior decision maker, they cannot take into account the city’s particular context and prospective capacities. Domestic capacity of cities in Flood and Fire Resilience should be considered when developing a model based on the availability and amenity of data and resources in the city. These data should be gathered from various city sources as needed to improve the decision-making process.

References Almeida, N.M., Silva, M.J.F., Salvado, F., Rodrigues, H., Maletiˇc, D.: Risk-informed performancebased metrics for evaluating the structural safety and serviceability of constructed assets against natural disasters. Sustainability 13(11), 5925 (2021). https://doi.org/10.3390/su13115925 Almufti, I., Willford, M.: REDiTM rating system resilience-based earthquake design initiative for the next generation of buildings (2013) Burroughs, S.: Development of a tool for assessing commercial building resilience. Procedia Eng. 180, 1034–1043 (2017). https://doi.org/10.1016/j.proeng.2017.04.263 Duarte, M., de Almeida, N.M., Falcão, M.J., Rezvani, S.: Resilience rating system for buildings and civil engineering works. In: 15th WCEAM 2021 (2021a). https://d322cbc3-d0a9-4a0d87e9-aa9414df6e27.filesusr.com/ugd/4d4145_82522a050aa9486eb137754639bad2db.pdf Duarte, M., Almeida, N., Falcão, M.J., Rezvani, S.M.H.S.: Resilience rating system for buildings against natural hazards. In: Pinto, J.O.P., Kimpara, M.L.M., Reis, R.R., Seecharan, T., Upadhyaya, B.R., Amadi-Echendu, J. (eds.) WCEAM 2021. LNME, pp. 57–68. Springer, Cham (2022). https://doi.org/10.1007/978-3-030-96794-9_6 ECCE: ECCE Position paper 2020 - The need for integrating structural/seismic upgrade of existing buildings, with energy efficiency improvements (2020) Falcão Silva, M.J., de Almeida, N.M., Salvado, F., Rodrigues, H.: Modelling structural performance and risk for enhanced building resilience and reliability. Innov. Infrastruct. Solut. 5(1), 1–20 (2020). https://doi.org/10.1007/s41062-020-0277-1 Feofilovs, M., Romagnoli, F.: Dynamic assessment of urban resilience to natural hazards. Int. J. Disaster Risk Reduct. 62, 102328 (2021). https://doi.org/10.1016/J.IJDRR.2021.102328 Hosseini, S., Barker, K., Ramirez-Marquez, J.E.: A review of definitions and measures of system resilience. Reliab. Eng. Syst. Saf. 145, 47–61 (2016). https://doi.org/10.1016/j.ress.2015.08.006

Resilience Exposure Assessment Using Multi-layer Mapping

623

Komljenovic, D.: Engineering asset management at times of major, large-scale instabilities and disruptions. In: Crespo Márquez, A., Komljenovic, D., Amadi-Echendu, J. (eds.) WCEAM 2019. LNME, pp. 255–270. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-642280_22 Maleti´c, D., Marques de Almeida, N., Komljenovic, D., Lovrenˇci´c, V., Maletiˇc, M.: Digitalizing predictive maintenance to improve asset management: are we ready? September, pp. 413–437 (2021). https://doi.org/10.18690/978-961-286-388-3.34 Meerow, S., Newell, J.P.: Urban resilience for whom, what, when, where, and why? Urban Geogr. 40(3), 309–329 (2019). https://doi.org/10.1080/02723638.2016.1206395 Pinheiro, M.D.: LiderA Sistema voluntário para a sustentabilidade dos ambientes construídos (2011) Pinheiro, M.D.: Avaliação pelo LiderA do grau de resiliência de Lisbon Green Valley (Sintra) (2020). https://doi.org/10.13140/RG.2.2.34017.25446 Raškovi´c, M., et al.: Performance based planning of complex urban social-ecological systems: the quest for sustainability through the promotion of resilience. Sustain. Cities Soc. 48, 91–97 (2020). https://doi.org/10.1016/j.scs.2015.07.004 Rezvani, S.: Multi-criteria decision analysis of subcontractors selection for infrastructure projects: a case study of an electrified railway project. In: Online Event: 11th IMA International Conference on Modelling in Industrial Maintenance and Reliability (MIMAR) (2021). https://doi. org/10.13140/RG.2.2.22610.07369 Rezvani, S.M., de Almeida, N.M., Falcão, M.J., Duarte, M.: Urban resilience dimensions, indicators, and parameters for constructed assets evaluations by automate rational and consistent decision making using AHP. Sustain. Cities Soc., 1–12 (2021). Artigo em desenvolvimento Rezvani, S.M., de Almeida, N.M., Falcão, M.J., Duarte, M.: Enhancing urban resilience evaluation systems through automated rational and consistent decision-making simulations. Sustain. Cities Soc. 78 (2022). https://doi.org/10.1016/j.scs.2021.103612 USGBC: RELi 2.0 rating guidelines for resilient design + construction (2018). https://www.usgbc. org/resources/reli-20-rating-guidelines-resilient-design-and-construction

Use of Survival Analysis and Simulation to Improve Maintenance Planning of High Voltage Instrument Transformers in the Dutch Transmission System Swasti R. Khuntia1(B) , Fatma Zghal1 , Ranjan Bhuyan1 , Erik Schenkel1 , Paul Duvivier2 , Olivier Blancke2 , and Witold Krasny2 1 Asset Management Onshore, TenneT TSO B.V, Arnhem, The Netherlands

{swasti.khuntia,fatma.zghal,ranjan.bhuyan, erik.schenkel}@tennet.eu 2 Cosmo Tech, Lyon, France {paul.duvivier,olivier.blancke,witold.krasny}@cosmotech.com

Abstract. This paper describes the use of survival analysis and simulation to model the lifetime of high voltage instrument transformers in the Dutch transmission system. To represent asset aging, the non-parametric Kaplan-Meier method is used to enable the fitting of Weibull distribution. Such an approach is implemented on three different voltage levels, namely 110 kV, 150 kV, and 220/380 kV. Real failure and inspection data is used to achieve a realistic failure model of the instrument transformers. Failure and maintenance data occurring between 1989 and 2021 have been used for this study. In spite of missing and low-quality data, a rich failure database could still be prepared. This study also offers insights into factors (i.e., voltage level, in-service age) influencing the remaining life from both graphical survival function and parametric Weibull distribution analysis. Based on the derived statistics, future possible maintenance planning scenarios are simulated under a complex system modelling framework in a digital twin enabled platform. Eventually, the scenarios are evaluated in terms of replacement costs (CAPEX), inspection hours, and unavailability hours.

1 Introduction TenneT, as European transmission system operator, is facing power supply reliability challenges that originate in a globally aging infrastructure and increasing complexity of business operations in the context of energy transition. While power transformers, due to the criticality of their function on the grid have been the focus of many studies, concerns have been raised recently on the lack of focus on long-term asset management of Instrument Transformers (ITs). ITs play an important role in the metering of electrical quantities and protection of other system components. Due to their importance, any unplanned unavailability due to failures can cause considerable outage costs to utilities. Consequently, it is crucial to properly characterize the aging of ITs using statistical approaches that will enable to predict the evolution of the IT population failure over the © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 624–635, 2023. https://doi.org/10.1007/978-3-031-25448-2_58

Use of Survival Analysis and Simulation to Improve Maintenance

625

next years. In addition, it will yield valuable perspectives in terms of optimizing maintenance and replacement policies accordingly. The reliability analysis of ITs is very much dependent on the defined maintenance strategies which will provide a reliable and safe power supply. By definition, asset management involves strategies to explore, identify, plan, invest, utilize, maintain, replace, and dispose of assets while maximizing their value and performance under some prescribed financial constraint (Khuntia et al. 2016). Since ITs play such an important role, it is expected that statistical failure analysis will give a better insight on actual maintenance planning performance to the asset management team at TenneT. Technically, in the reliability analysis of IT, it is interesting to identify the independence or dependence of the specific covariates that indicate the operation of the IT. For any kind of data-driven methodology and, in particular, asset reliability characterization, a robust database is needed, both in terms of volumetry and quality (Balzer and Neumann 2011). However, it can be argued that there should be a preference for robust data and that there are techniques that could be used to cope with data discrepancies. In our case, the historical failure data play an important role in understanding the behavior of ITs. Literature study reveals that explosion is one of the highest reported failure modes. Impact of explosion not only relates to direct cost of IT replacement but also chances of replacement of neighboring equipment damaged in the explosion. CIGRE reports are one of the primary sources for publicly available failure databases of ITs. Three series of CIGRE reports are available online. The first report was published in 1990 which covered failures of ITs (voltage >72.5 kV) in about 15 countries. The survey covered 136033 transformers in the period from 1970 to 1986 (CIGRÉ WG 23.07: The paper-oil insulated measurement transformer 1990). The second report published results for 131207 ITs (voltage > 60 kV) in the period from 1985 to 1995 in the year 2009 (CIGRÉ SC A3: State of the art of instrument transformers 2009). The third results of a wider international survey was published in 2012. It collected population and failure data for ITs of voltage > 60kV and excluded AIS ring current transformers that were in service during the years 2004 to 2007 inclusive (CIGRE, 2012). Some other failure investigations were reported (Poljak and Bojani´c 2010; Raetzke et al. 2012; Tee et al. 2021), where authors focus on reduction of IT explosion and better condition monitoring of ITs. Nonetheless, the truth is that failure is probabilistic in nature, and it needs investigations on the relationship with asset data and failure cause. The use of semi-parametric Cox model was reported in (Tee et al. 2021). The authors elaborated the factors influencing the probability of failures through analysis on the lifetime data from both graphical survival function plots and semi-parametric Cox model. With the use of Simulation Digital Twin technology from Cosmo Tech, TenneT analyzed various maintenance strategies. The Digital Twin has been calibrated based on the historical failure data that it recorded with statistical technique relying on survival analysis. Literature study shows that survival analysis was used for power transformer reliability studies of around 2000 nos. in the Canadian and around 6000 nos. in the Australian utility (Picher et al. 2014; Martin et al. 2018). Ref. (Picher et al. 2014) described the data of Canadian utility Hydro-Quebec where they adopted a good match using the Kaplan-Meier and Weibull distribution. Finally, the method concluded that Weibull distribution is a better fit and the results looked promising. Similarly, ref. (Martin

626

S. R. Khuntia et al.

et al. 2018) followed a similar strategy for Australian data. The authors deduced the choice of Kaplan-Meier or Weibull distribution based on the different voltage classes. In practice, Weibull distribution fitted to empirical failure data are commonly used to calculate life expectancy. However, the challenge in applying such a distribution to electrical assets is that often the root cause of failure is not related to the normal aging of the asset, but rather external factors. The aim of this paper is three-fold: (1) use of real failure data to model a time-varying failure rate based on Weibull parameters obtained from Kaplan-Meier survival analysis, (2) investigate extrapolation methods to maximize value of existing inspection results across IT population, and (3) use digital twin enabled simulation to tune the required resources necessary to realize TenneT’s strategy for considered substation equipment maintenance and renewals.

2 Data and Methodology 2.1 Description of Data As of the date of writing this paper, TenneT owns and maintains a large fleet of ITs in the Dutch high voltage AC network (i.e., 110, 150, 220 and 380 kV) as shown in Fig. 1(a). It is of interest to see the age profile of the existing population, in terms of years since manufacture because reliability is often related to age. However, lifetime data can be complicated as some ITs often extend over several decades. At TenneT, the expected design life of an IT is 45 years. This age is affected and reduced, sometimes substantially, depending on the design or utilization of the IT, i.e. its loading or the environment to which it is exposed. In some cases, a good maintenance scheme can even increase the replacement age. Although there is no prescribed replacement age, it is the responsibility of the asset management department to formulate the maintenance policies based on failure history. For this study, failure data was obtained from various sources, starting from failure records, reports to talking to experts. Fortunately, TenneT did not record a high number of major failures since the 1989. A major failure is defined as a sudden explosive event that has caused an immediate emergency system outage or trip. Figure 1(b) lists the failure events with respect to manufacturer (coded for confidentiality) and IT age. The failure list was not adequate to come up with a statistical model. In addition, maintenance reports (or work orders) and expert knowledge was used to populate the list and gain utmost information. A work order is a document that provides all the information about a maintenance task and outlines a process for completing that task. In case of IT, corrective work orders are used (the others being periodic maintenance and inspection work orders). Discussion with experts led us to use the work orders when an IT was out of service for any kind of maintenance. Figure 1(c) shows the total recorded failures for the IT population. In the recent years, one observation worth noticing is that the number of failures has increased significantly.

Use of Survival Analysis and Simulation to Improve Maintenance

627

(a)

(b)

(c)

Fig. 1. (a) Voltage-based IT population, and (b) Actual failure list until July 2021, (c) Populated failure from work order and expert opinion until July 2021

628

S. R. Khuntia et al.

2.2 Survival Analysis and Failure Rate Modelling Survival analysis is a statistical technique used to estimate the lifespan of a particular population under study. It is an analysis of time-to-event data (Wang et al. 2019). One of the widely used survival analysis technique is the Kaplan-Meier (KM) estimate (CIGRE, Germany Bland, J.M. and Altman, D.G. 1998). The KM estimator uses lifetime data to perform survival analysis. Although it is widely used in medical research to gauge the part of patients living for a specific measure of time after treatment, it has been used in the power systems sector to model the survival of electric assets (Martin et al. 2018). The use of KM estimate is supported by two reasons: one is that it does not assume that the data fits a statistical distribution, and second is that it allows the inclusion of censored data (when an IT had not failed by mid-2021). ˆ is defined as: For a population, the survival function S(t)   di ˆ = 1− S(t) ni i:ti t) = 1 − P(X > t + 3|X > t) =1−

P(X > t + 3) P(X > t + 3 ∩ X > t) =1− P(X > t) P(X > t)

=1−

1 − F(t + 3) 1 − P(X < t + 3) =1− 1 − P(X < t) 1 − F(t)

where, • F is the cumulative distribution function of the reliability law • X is a random variable representing the occurrence of a failure.

632

S. R. Khuntia et al.

Table 2. (a) TenneT Asset Health Index (AHI) definition, (b) Classification of Resources (FTE: Full Time Employment).

Table 3. Different Scenarios under Study.

Replacement Inspections on bay every 3,6,12 months

Condition-based

Time-based

220/380 kV

45 years

45 years

110/150 kV

AHI score red or purple 45 years

220/380 kV

No inspections

No inspections

110/150 kV

Time-based starting at 25 years

Time-based starting at 25 years

3.3 Simulation The reliability law was used to evaluate the different scenarios for an efficient maintenance planning. A simulation period of 100 years is chosen for this study since it is assumed that the most recent IT replacements will be in operation until the end of this century. Time-based scenario is the current maintenance planning at TenneT. It is compared against a condition-based scenario. Both the scenarios are explained in detail in Table 2. The resources are listed in Table 1(b). In principle, both scenarios are very similar in the sense that the same simulation model dataset is used. The difference lies in the trigger for the replacement activities of the 110/150 kV assets. In fact, in time-based scenario, which represents the current way of working, the trigger is based on the real age of the asset. As soon as the asset reaches 45 years of age, replacement is triggered, and action is performed as resources are unlimited. On the other hand, in the condition-based scenario, the trigger is based on the apparent age of the asset. The apparent age is an attribute of every asset that reflects its degradation rate and it can be different from the real age of the asset. If the apparent age is higher than the real age, the asset degrades faster than normal. If the apparent age is lower than the real age, the asset degrades slower than normal. When the apparent age of the asset reaches 50 or 54, it means that the asset is reaching AHI score of respectively 6 or 3 that is red or purple (see Table 1(a)), and the replacement action is triggered.

Use of Survival Analysis and Simulation to Improve Maintenance

633

Fig. 3. Unconstrained scenarios simulation.

Fig. 4. 40 FTE constrained scenarios simulation.

Fig. 5. 60 FTE constrained scenarios simulation.

From the figures, two conclusions can be made: (1) replacement activities are the major cost driver in the TOTEX (Total Expenses), and (2) Human resources (HR) costs are the major cost driver in the replacement costs. Simulation results show that in case HR availability is restricted, there is no significant difference between the time-based and condition-based replacement strategies. In fact, switching to a condition-based strategy

634

S. R. Khuntia et al.

might not be beneficial in that case since it comes with change and investments for little to no reward. If HR availability is guaranteed for the foreseeable future, then it is highly beneficial to switch from a time-based replacement strategy to a condition-based strategy as this would contribute to flattening the curve. Also, this would represent a lot of work at the beginning to prepare the necessary processes and investments for the new strategy but would lead to significant gains on the long term.

4 Conclusion Maintenance planning of high voltage ITs using real data from the Dutch transmission system operator was illustrated in this study. The study aimed at understanding how digital twin enabled technology along with failure data can help TenneT to make better future maintenance strategies. The strategies aimed at easing financial decisions related to replacements (in terms of flattening the replacement curve) and unavailability of ITs in the network. Working on real data uncovered several challenges including missing data (both quantity and quality) and outliers. The non-parametric Kaplan-Meier survival analysis helped in parameter estimation of Weibull distribution. TenneT data could be translated to the data format to be used in the digital twin CTA tool, meaning that our data could be easily adapted to other software platforms. It is worth to mention that in this study, both data ownership as well as data confidence did not hinder the progress. Data confidence was built upon although multiple data sources had to be aligned together. TenneT partnered with Cosmo Tech to build the data ownership philosophy for successful digital twin implementation for maintenance planning.

References Balzer, G., Neumann, C.: Asset simulation and life cycle assessment for gas insulated substation (2011) CIGRE, Germany Bland, J.M. and Altman, D.G., 1998. Survival probabilities (the Kaplan-Meier method). Bmj, 317(7172), 1572–1580 (1998) CIGRÉ WG 23.07: The paper-oil insulated measurement transformer, CIGRÉ Technical Brochure 57, (1990) CIGRÉ SC A3: State of the art of instrument transformers, CIGRÉ Technical Brochure 394 (2009) CIGRE Final Report of the 2004 – 2007 International Enquiry on Reliability of High Voltage Equipment Part 4 - Instrument Transformers. Working Group A3.06 (2012) CIGRE: Guidelines for the Use of Statistics and Statistical Tools on Life Data, Working Group D1.39 (2017) Davidson-Pilon, C.: Lifelines: survival analysis in Python. J. Open Source Softw. 4(40), 1317 (2019) Khuntia, S.R., Rueda, J.L., Bouwman, S., van der Meijden, M.A.: A literature survey on asset management in electrical power [transmission and distribution] system. Int. Trans. Electr. Energy Syst. 26(10), 2123–2133 (2016) Martin, D., Marks, J., Saha, T.K., Krause, O., Mahmoudi, N.: Investigation into modeling Australian power transformer failure and retirement statistics. IEEE Trans. Power Delivery 33(4), 2011–2019 (2018) Picher, P., et al.: Use of health index and reliability data for transformer condition assessment and fleet ranking. A2–101, CIGRE (2014)

Use of Survival Analysis and Simulation to Improve Maintenance

635

Poljak, M., Bojani´c, B.: Method for the reduction of in-service instrument transformer explosions. Eur. trans. Electr. power 20(7), 927–937 (2010) Raetzke, S., Koch, M. Anglhuber, M.: Modern insulation condition assessment for instrument transformers. In: 2012 IEEE International Conference on Condition Monitoring and Diagnosis, pp. 52–55. IEEE (2012) Rinne, H.: The Weibull Distribution: A Handbook. Chapman and Hall/CRC (2008) Tee, S., Liu, Q., Wang, Z., Hafid, F., Tournet, P.: Failure investigation and asset management of combined measuring instrument transformers. High Voltage 6(1), 61–70 (2021) Wang, P., Li, Y., Reddy, C.K.: Machine learning for survival analysis: a survey. ACM Comput. Surv. (CSUR) 51(6), 1–36 (2019)

Resilience Assessment of Public Treasury Elementary School Buildings in Lisbon Municipality João Garcia, Seyedi Rezvani, Maria João Falcão Silva(B) , Nuno Almeida, Cláudia Pinto, Rui Gomes, Mónica Amaral Ferreira, Filipe Ribeiro, Filipa Salvado, and Carlos Sousa Oliveira LNEC, Buildings, Av. Brasil 101, 1700-066 Lisboa, Portugal

Abstract. The resilience of buildings and civil engineering infrastructures has increasingly attracted the attention of different stakeholders, including engineering professionals from different areas, scientists, standardization bodies, investors and financial institutions, regulatory agencies, different user groups, as well as national and regional administrative services. This interest has motivated the development of methods for classifying the resilience of built assets, to identify aspects that can be improved and establish investment priorities, to increase their resilience when facing extreme events or other types of risks. The paper presents contributions to improve a system for classifying the resilience of built assets organized in dimensions, indicators and parameters that cover not only the intrinsic characteristics of buildings, but also their exposure to natural and man-made hazards, with the community and with the users. The paper specifically focuses on the resilience of public treasury buildings elementary schools in the municipality of Lisbon, some of which have been object, throughout their life cycle, of different interventions carried out within the scope of different public investment programs.

1 Introduction The safety of people and the high value of buildings and their contents, as well as their criticality in meeting the basic needs and well-being of communities, have always raised concerns regarding their resilience and reliability (Cutter et al. 2003). Between 1970 and 2013, natural disasters around the world generated economic losses of nearly 3 trillion dollars (Cerè et al. 2019). The World Bank estimates that the cost of city and community vulnerabilities due to natural disasters, extreme events and/or pandemics could reach more than US$300 billion per year by 2030. (Falcão Silva et al. 2020; Almeida et al. 2021). The classification of the resilience of built assets is increasingly becoming a topic of the greatest importance and relevance for both asset managers and their users, with the need for its operationalization through a system increasingly emerging. That is consensual and easy to implement and use (VRS 2017; Falcão Silva et al. 2021). The urban resilience of built assets can be seen as the ability of these physical assets to withstand severe damage within acceptable degradation parameters and to recover in reasonable © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 636–644, 2023. https://doi.org/10.1007/978-3-031-25448-2_59

Resilience Assessment of Public Treasury Elementary School Buildings

637

time intervals. No definition for this has been unified yet, but strength, absorption, and recovery characteristics are generally recognized as the basis for resilience assessment systems. The striking advantages of increased resilience have increasingly attracted the attention of managers and engineers to its use in various aspects related to risk reduction and prioritizing the budget allocated to assets especially built national assets that need to be preserved for future generations (Vugrin et al. 2010; Rahi 2019; Burroughs 2017). There were previous efforts from the world’s most important stakeholders, as well as different collaborative projects in this area such as: a) The World Bank Group; b) Global Facility for Disaster Reduction and Recovery; c) 100 Resilient Cities program to promote urban resilience from Rockefeller Foundation; d) UNISDR; e) City Resilience Index (CRI), created by Arup and financed by The Rockefeller Foundation; f) Inter-American Development Bank; g) ICLEI, and Cities; among others (URH 2022). There are initiatives that are already being developed and successfully implementing specific resilience rating systems to support quality management and governance methodologies for different types of buildings, based on different approaches, being referred as examples the Hyogo Framework for Action (HFA), which is a work developed by the United Nations (UN) and the World Risk Report (WRR) which is a work performed by the United Nations University for Environment and Human Security (UNU-EHS) (Kammouh et al. 2017). However, some of them are not yet compatible or standardized, making it difficult for them to be widely implemented and used in the future. In Portugal, there is still a lack of a specific system of classification of resilience of generalized implementation and a methodology of governance to support the management of the quality of the buildings, accepted by all the intervening ones, therefore it seems to be of the utmost importance to develop a specific system of classification of resilience to support decisions and the use of innovative solutions for a sustainable and resilient AECO sector in line with European Standards and Community Policies already published and to be complied with in the 21st century. Considering the identified gaps, resilience classification models should cover not only environmental, economic, technical, physical, social aspects, but also political, regulatory, and organizational aspects, seeking to contribute to the calibration of performance and resilience metrics for new buildings and infrastructure or existing (VRS 2017) (Duarte et al. 2021a, 2021b, 2021c). The classification system now presented and resulting from an expansion of a previous proposal (Duarte 2021) is expected to provide major contributions to the pressing need to achieve the UN sustainable development goals, namely by giving due consideration to the concept increasingly most important of the resilience of built assets when impacted by a wide range of natural and man-made disasters.

2 Research and Method This paper proposes an approach based on the expansion of a previously proposed classification system (Duarte 2021). The improved approach comprises: i) An innovative method of assessing resilience covering multiple dimensions (economic-financial, social, environmental, organizational, and technical); ii) around 100 indicators that can be incorporated at any stage of the life cycle of various types of assets; iii) real-case pilot test in the municipality of Lisbon covering some critical built assets for the system validation and implementation.

638

J. Garcia et al.

The method considered for the expansion of the proposal allows: i) Establish dimensions and resilience indicators for a scoring and weighting system to classify and assess the vulnerability of built assets; ii) develop correlation and scalability of the basic dimensions of economic, environmental, social, organizational and technical resilience for built assets and built asset systems, as well as networks; iii) develop a scoring system and ranking of resilience dimensions and indicators to provide the basis for comparison and assessment platform; iv) carrying out a multi-criteria decision analysis for weighting criteria and alternatives to compose a resilience weighting model; v) conducting risk analysis through the stochastic approach to uncertainty assessment to reduce model risk and increase reliability; vi) implement sensitivity, robustness and application case analysis of the built-in asset resilience model for existing critical power, transport, water and sanitation and interrelated communication network infrastructure and residential, commercial and other types of buildings to classify them and extract resilience matrix; vii) establish management guidelines to improve the resilience of built assets, covering policies, institutions and processes, specialists, operations and maintenance, planning and dimensioning, financial support and incentives. The expanded resilience rating system allows for the differentiation and recognition of different levels of resilience of existing or new buildings and will provide a high standard benchmark for the empowerment of different actors in the AECO sector for a resilient and sustainable 21st century. By promoting higher quality products and buildings, the proposed research will increase the reduction of constructive defects and, consequently, will motivate improvements in the existing warranty systems for construction works. In addition, the resilience rating system should be developed so that it can also be used for building resilience certification and technical control for different types of buildings. The definition of dimensions, indicators and parameters aims to assess resilience and facilitate communication and consultation procedures. The parameters subdivide the indicators, and, in turn, each set of indicators expresses in more detail each of the dimensions mentioned above. Its selection was confirmed through the result of the literature review, considering that: i) the selected parameters are measurable; ii) there is information available for its quantification and iii) it is desirable to avoid overlapping or repetition of metrics (Duarte et al. 2021c). The evaluation criteria defined for each parameter were initially established based on the limits of different metrics. The process of reviewing and calibrating indicators, parameters and evaluation criteria is expected to be iterative. The process must be monitored for the influence of judgments or opinions, lack of data and difficulty of quantification.

3 Resilience Rating System The previously proposed and validated resilience rating system (Duarte 2021) was developed and expanded to better suit the intended objective (Table 1). The work presented contributes to an expansion of ways to measure the resilience of built assets, based on a previously proposed resilience classification system. The proposed expansion of the system comprises 5 dimensions, 18 indicators and 95 parameters. The indicators and parameters added with the expansion of the resilience rating system now presented are marked in green in Table 1. In addition to the expansion proposals, easily identifiable in green, we find identified by (*) and (**) the cases in which 1

Resilience Assessment of Public Treasury Elementary School Buildings

639

Table 1. Resilience classification system – dimensions expansion D1 - ENVIRONMENTAL I1 - Earthquake P1 - Seismic zoning - type 1 EC8 P2 - Seismic zoning - type 2 EC8 P3 - Seismic vulnerability of PDM soils P4 - Terrain slope P5 - EC8 soil type (1) P6 - Distance to cliffs P7 - Distance to geological faults P8 - Population density I2 - Tsunami and tidal effect P9 - Land altitude P10 - Distance to the coast P11 - Distance to the river P12 - Natural barriers in the surroundings P13 - Man-made barriers in the surroundings P14 - Moving objects P15 - Rows built between the coast and the building P16 - Susceptibility to the direct tidal effect PDM P17 - Relative location I3 - Flood P18 - Distance to the river P19 - Natural barriers in the surroundings P20 - Man-made barriers in the surroundings P21 - Vulnerability to PDM Floods P22 - Distance to vegetation I4 - Fire P23 - Density of vegetation P24 - Vegetation maintenance status P25 - Type of vegetation P26 - Adjacent buildings P27 - Proximity to the industrial zone I5 - Landslides P28 - Terrain slope P29 - Precipitation P30 - Groundwater level position D2 - ECONOMIC-FINANCIAL I6 - Insurance P31 - Insurance against natural disasters I7 - Financial and strategic implications P32 - Financial plan P33 - Economic assessment of downtime P34 - Existence of disaster funds P35 - Access to External/Internal credit P36 - Access to titles D3 ORGANIZATIONAL I8 - Internal organization P37 - Business continuity plan P38 - Risk analysis and management P39 - Post disaster recovery plan P40 - Routine P41 - Simulacra P42 - Learning and updating P43 - Destructive event data P44 - Responsible I9 - External organization P45 - Compliance with the existing regulatory scenario P46 - External standards for resilient construction P47 - Responsible entity P48 - Relationship between the community and stakeholders P49 - Monitoring

D4 - SOCIAL I10 - Emergency infrastructure Parameter P50 - Access to police stations P51 - Access to fire stations P52 - Access to shelters P53 - Access to hospitals and health centers I11 - Social responsibility P54 - Occupants P55 - Disclosure P56 - Social vulnerability P57 - Existence of mutual help programs with neighbors P58 - No. of social defense organizations D5 - TECHNICAL I12 - Conservation P59 - Year of construction P60 - Structural system P61 - State of conservation P62 - Maintenance, faults, and updates history I13 Accessibility P63 - Building density (1) P64 - Alternate routes (*) P65 - Street characteristics I14 - Seismic safety of the building P66 - Plant irregularity P67 - Irregularity in height P68 - Interaction with adjacent buildings P69 - Slabs unevenness P70 - Expansion joint I15 - Building fire safety P71 - State of conservation of electrical installations (2) P72 - Gas installations P73 - Distance between overlapping spans P74 - Existence of fire compartmentalization (*2) P75 - Fire detection and alarm (*) P76 - Existence of emergency signs and lighting (2) P77 - Existence of security team (2) P78 - Escape paths P79 - Existence of smoke control and evacuation systems (*2) P80 - Existence of intrinsic means of combat (*2) P81 - Existence of fire extinguishers (**2) P82 - Existence of external hydrants (2) I16 - Building Flood Safety P83 - Existence of barriers (2) P84 - Existence of pumping systems against flooding (*2) P85 - Vulnerability and exposure of facades (2) P86 - Number of floors P87 - Street characteristics P88 - Vulnerability of underground floors P89 - Waterproofing solutions (basements) P90 - Wastewater drainage systems I17 - Building safety against tsunamis P91 - Number of floors P92 - Guidance P93 - Ground floor hydrodynamics (*) I18 - Building safety against landslides P94 - Degree of Waterproofing (soils) P95 - Slope stability

640

J. Garcia et al.

evaluation criterion was added and those in which 2 were added, respectively. With the identification (1) and (2), the cases in which the evaluation criteria were reformulated and those in which the parameter name was reformulated to make it clearer, respectively, are highlighted. It is also worth noting the change in the name of D2 to Economic-Financial, to better express the contents, it encompasses.

4 Case Study Implementation To validate the proposed resilience classification system, 4 buildings for collective use (CUB) belonging to the public purse, intended to house research facilities, and located in the municipality of Lisbon, were considered. The E1 building was built and inaugurated in the 1950s, respecting the regulations in force at the time, it has a reinforced concrete structure and exterior and interior walls simply filled with brick masonry structural mesh and facades mostly covered with marble. And integrate limestone ashlar elements in the basements, sills, and windowsills. The building underwent expansion works in the 90s of the 20th centuries (in 2 phases), having been the subject of improvement works, which consisted of the restoration of the facade, preventive interventions and maintenance and remodeling of several sanitary facilities. The E2 building comprises three interconnected bodies as a whole and was built and inaugurated in the early 1960s, respecting the regulations in force at the time. In the construction of the building, the exposed concrete structures, the masonry walls, and the glazed brick walls in a natural and cream tone are highlighted. The E3 building, built, and inaugurated in the early 1970s, respecting the regulations in force at the time, has a reinforced concrete structure and exterior and interior walls simply filled with the brick masonry structural mesh. In relation to the E2 building, ceramic elements were chosen for the exterior cladding of the building’s facades. The building underwent expansion works and was improved, including interventions in the interior, especially in the sanitary facilities. The E4 building was built and opened in the mid-1990s in compliance with the regulations in force at the time. It has a reinforced concrete structure and walls (interior and exterior) in brick masonry, not having undergone any interventions, given that it is a more recent building, being in good condition and perfectly up to date.

5 Results Analysis The results of the resilience classification carried out for the 4 CUB located in the municipality of Lisbon for validation of the proposed system are presented in the following figures. The results of the classifications of the 4 CUB target of the study show great similarities, particularly about the dimensions economic-financial (D2), organizational (D3) and social (D4). The distinction, although marginal, can be identified by the remaining dimensions, environmental (D1) and technical (D5) (Fig. 1a). The similarities identified are since all these buildings have the same geographic location, belong to the same

Resilience Assessment of Public Treasury Elementary School Buildings

641

Fig. 1. Resilience classification for the studied CUB: a) dimensions; b) indicators

organization, and have similar types of use. In Fig. 1b we can see a graphic representation of the classification obtained by the buildings analyzed in the different indicators, where: I1 to I5 comprises D1, I7 and I8 correspond to D2, I8 and I9 represent D3, D4 comprises indicators I10 and I11, the remaining indicators I12 to I18 therefore belong to D5. The analysis of the indicators proves the similarities verified in the analysis carried out by dimensions, with the greatest differences being observed in I12 (Conservation), particularly given the different construction periods and corresponding structural systems identified in the different CUB analyzed, as well as in I17 (Building safety against tsunamis) due to the orientation of buildings in relation to the coastal direction. There are also marginal differences in indicator I16 (Safety of the building against flooding) arising from the difference observed in terms of the number of floors and the vulnerability of underground floors, and in I4 (Fire) due to the construction date of E2 and E4 being after 1967, in which a very relevant change was made at the level of the current regulations and, for the same reason, they present a very significant classification at the level of P26 (Adjacent Buildings). To better understand the relationships between the classifications obtained by the buildings, we can observe, in Figs. 2, 3 and 4, the classifications of each parameter in each of the dimensions that make up the system. An analysis of the environmental dimension (D1) allows the validation of the previous statement and a confirmation of the similarity of the level of resilience obtained by each of the EUCs. We can see in Fig. 2a that the only parameter in which it is possible to identify a distinction is P26 (Adjacent Buildings), which evaluates the year of construction of the buildings based on the fire regulations that came into force in 1967. Thus, it appears that E2 and E4 get maximum rating and E1 and E3 the minimum. The remaining parameters do not show differences because they express the characterization of the study site in relation to environmental risks, which is the same for all EUCs. The representation of the classifications obtained by the EUC in D2 present in Fig. 2b, expresses the equality of the objects of study in all the constituent parameters of the dimension in case, arising from the fact that all of them belong to and are injured by the same organization. Parallel to the dimension described above, in Fig. 3a we observe equality in all parameters comprised by the organizational dimension (D3), in this case. At an organizational

642

J. Garcia et al.

Fig. 2. Resilience classification for the studied CUB, for different parameters: a) environmental dimension (D1); b) economic-financial dimension (D2)

Fig. 3. Resilience classification for the studied CUB, for different parameters: a) organizational dimension (D3); b) social dimension (D4)

level, the entity that manages the buildings ends up determining that this equality exists in terms of classification, with the aggravating factor of the location of the buildings and their type of use being the same. Figure 3b expresses the equality of all parameters of the fourth dimension, which addresses the social issue (D4). The fact that it is not possible to observe any kind of difference in parameters P50 (Access to police stations), P51 (Access to fire stations), P52 (Access to shelters) and P53 (Access to hospitals and health centers) belonging to indicator I10 (Emergency infrastructure) and P57 (Existence of mutual aid programs with neighbors) and P58 (Nº social defense organizations) belonging to I11 (Social responsibility) is corroborated by the proximity of the location of the EUC object of study. The parameters P54 (Occupants), P55 (Disclosure) and P56 (Social vulnerability) that are part of I11 (Social responsibility) are influenced by the management and type of use of the buildings, which is similar, considering that the entity that performs this function is the same. In the representation of the parameters related to the technical dimension (D5) (Fig. 4), we observed the greatest asymmetries in P59 (Year of construction), P88 (Vulnerability of underground floors) and P92 (Orientation). As described above when describing the EUC object of study, they are buildings built in different years, which

Resilience Assessment of Public Treasury Elementary School Buildings

643

Fig. 4. Resilience classification for the studied CUB, for different parameters: technical dimension (D5)

explains the difference in P59, belonging to I12 (Conservation). The vulnerability of underground floors (P88), assessed in I16 (Safety of the building against flooding), is conditioned by the number of underground floors and their susceptibility to floods, so only E1 has a moderate vulnerability as it has 2 underground floors. The orientation (P92) of the buildings in relation to the coast evaluated in I17 (Safety of the building against tsunamis) is the same for E1 and E4, which are parallel to the coastline, thus obtaining the worst classification in this parameter. As for the others (E2 and E3) they are both perpendicular, which gives them the maximum classification.

6 Conclusions As mentioned, the work presented contributes to an expansion of ways to measure the resilience of built assets, based on a previously proposed resilience classification system. The expansion comprises a system with 5 dimensions, 18 indicators and 95 parameters. The results of the system calibration for 4 public buildings located in the municipality of Lisbon are presented. However, it is still necessary to develop complementary work to implement the proposed assessment in representative number and diversity of the typologies of constructed assets, as well as to extend the scope of the proposed multivariable classification system about other types of risks. (e.g. human-induced hazards) and the identification of countermeasures and their classification. The level of maintenance of buildings is a very important criterion that could be subdivided, for example, into corrective and preventive dimensions. Different buildings with different functions and uses can, and should, be used as empirical case studies to show how technical performance and risk engineering can be programmed defensively to improve resilience and reliability in a more sustainable environment for future generations. The expansion of this approach may, in the future, include i) an online platform with the objective of rapid and wide dissemination of research results, developments and applications; ii) a roadmap to increase the reach and extended impact of project results for public and private organizations that manage construction assets, such as government agencies, banks, insurance companies, design and construction companies and various professionals in the AECO.

644

J. Garcia et al.

References Almeida, N., Falcão Silva, M.J., Salvado, F., Rodrigues, H., Maletic, D.: Risk-informed performance-based metrics for evaluating the structural safety and serviceability of constructed assets against natural disasters. Sustainability 13, 5925 (2021). https://doi.org/10.3390/su1311 5925,Q1/Q2Scopus Burroughs, S.: Development of a tool for assessing commercial building resilience. University of Canberra, ACT 2601, Australia (2017). https://doi.org/10.1016/j.proeng.2017.04.263 Cerè, G., Rezgui,Y., Zhao, W.: Urban-scale framework for assessing the resilience of buildings informed by a Delphi expert consultation. Int. J. Disaster Risk Red. 36, 101079 (2019). https:// doi.org/10.1016/j.ijdrr.2019.101079 Cutter, S.; Boruff, B.; Shirley, W.: Social vulnerability to environmental hazards. Soc. Sci. Q. 84(2) (2003) Duarte, M.: Sistema de classificação de resiliência para edifícios perante riscos naturais. Dissertação de mestrado em engenharia civil, Instituto Superior Técnico, Universidade de Lisboa (2021) Duarte, M., Almeida, N., Falcão Silva, M.J., Rezvani, S.: Resilience rating system for buildings against natural hazards. In:15WCEAM, Brasil, paper ID 42 (2021a) Duarte, M., Almeida, N., Falcão Silva, M.J., Rezvani, S.: Resilience rating system for buildings against natural hazards. In: 15 WCEAM, Brasil, paper ID 94 (2021b) Duarte, M., Almeida, N., Falcão Silva, M.J., Salvado, F.: Resilience of constructed assets against natural extreme events from the engineering standpoint. In: CEES 2021, Coimbra, Portugal (2021c) Falcão Silva, M.J., de Almeida, N.M., Salvado, F., Rodrigues, H.: Modelling structural performance and risk for enhanced building resilience and reliability. Innov. Infrastruct. Solut. 5(1), 1–20 (2020). https://doi.org/10.1007/s41062-020-0277-1 Falcão Silva, M.J., Almeida, N., Salvado, F., Rodrigues, H.: Structural performance evaluation system for improved building resilience and reliability. In: Farsangi, E., Noori, M., Gardoni, P., Takewaki, I., Varum, H. (eds.) Reliability-Based Analysis and Design of Structures and Infrastructure. CRC Press, Taylor & Francis (2021). https://doi.org/10.1201/9781003194613 Kammouh, O., Dervishaj, G., Cimellaro, G.P.: A new resilience rating system for Countries and States. In: Urban Transitions Conference, Shanghai, September 2016 (2017). Procedia Engineering, 98 (2017), 985–998, Elsevier Rahi, K.: Indicators to assess organizational resilience – a review of empirical literature. Int. J. Disaster Resil. Built Environ. 10(2/3), 85–98 (2019). https://doi.org/10.1108/IJDRBE-112018-0046 Urban Resilience Hub. http://urbanresiliencehub.org/medellin-colaboration/. Accessed 17 July 2022 VRS: Voluntary resilience standards. Meister Consultants Group, Inc. (2017) Vugrin, E.D., Warren, D.E., Ehlen, M.A., Camphouse, R.C.: A framework for assessing the resilience of infrastructure and economic systems. In: Gopalakrishnan, K., Peeta, S. (eds.) Sustainable and Resilient Critical Infrastructure Systems, pp. 77–116. Springer, Heidelberg (2010). https://doi.org/10.1007/978-3-642-11405-2_3

Disaster Risk Mitigation Through Capital Investment in Enhanced Building Resilience Maria João Falcão Silva(B) , Filipa Salvado, and Nuno Almeida LNEC, Buildings, Av. Brasil 101, 1700-066 Lisboa, Portugal [email protected]

Abstract. The recognition of the high costs of the vulnerabilities of cities and communities due to various types of disaster risks have helped build the business case for investments improving the quality and resilience of buildings and civil engineering works. This important issue has attracted the attention of several stakeholders, including engineering professionals from different fields, scientists, standardization bodies, investors and financial institutions, regulatory agencies, user groups of several, as well as administrative services at national and regional level. This research work discusses the application of a resilience rating system to show the impacts of capital investment in building refurbishment and renewal. It highlights aspects in building renovation programs that efficiently and quickly increase their resilience in face of extreme events. The work presented builds on previous discussions on ways to measure the resilience of built assets, namely based on a rating system composed by different dimensions, several indicators, and parameters. It covers not only the building’s intrinsic qualities, but also its interdependence with the community, surroundings, and users in the post-disaster context.

1 Introduction The resilience of buildings and engineering works has awakened institutions from various stakeholders, including engineering professionals from different areas, scientists and financial standardization bodies, regulators, as well as administrative services at national and regional level and management entities. of assets (AM). This interest in correcting the broader vision of resilience is a key issue for achieving the ONU Sustainable Development Goals (SDGs), regarding economic issues and the need to provide the public, including groups, an environment that can better adapt to the risks of future disasters (Sarhosis et al. 2019). The World Bank estimates that the cost of the vulnerabilities of cities and communities due to these types of disaster risks could reach more than USD 300 billion per year by 2030. On the other hand, these estimated costs can be reduced through capital investments for improving the quality and resilience of engineered physical assets that are the backbone of modern societies (e.g., infrastructure, industrial facilities, and buildings). This research work addresses the resilience of Portuguese public-school buildings against different risks from an engineering standpoint, namely regarding to building structural safety and serviceability. These school buildings, constructed through the © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 645–656, 2023. https://doi.org/10.1007/978-3-031-25448-2_60

646

M. J. F. Silva et al.

XXth century, had extended interventions in the scope of a national public investment program, between 2009 and 2011. The buildings were analysed before and after different types of rehabilitation interventions. The materials and constructive solutions to be adopted considered the current needs (regulatory and legislative requirements) as well as the maintenance system to be implemented. The interventions contemplate equipment, facilities and technical designs currently required in legislative environment, structural safety, seismic reinforcement, and fire safety aspects. Related to the economic indicators and the potential investment, it is noted that the rehabilitation interventions have a positive impact in the resilience score, according to the evaluation scale presented. The conclusions comprise the main results obtained, considering the resilience rating model, the different dimensions analysed, and the main representative parameters and indicators, considering the type of structure and the type of intervention performed.

2 Conceptual Framework Considering resilience represents the ability of a building to resist, absorb, accommodate, adapt, transform, and recover from the effects of hazard, it is necessary to understand the importance of disaster risks which are prerequisites for the development of standardization for building and civil engineering works resilience (Duarte et al. 2021a). Natural disasters are not isolated. They are repeated although they may be exacerbated through modern human activities, for example, the propagation of contagious diseases through massive, global travel. Natural disasters should not be considered as unpredictable, transitory events demanding emergency responses, but rather as ongoing risks with lifecycles extending over the years whose mitigation and adaptation should be embedded in urban planning and policy. This framing points to the balance required of policymakers: the need to make large-scale investments or to exclude potential economic developments today for the sake of reducing the impacts of future events or, where possible, to enable the two policies to coincide Harrison and Williams 2016 Jun). The World Economic forum (WEF) publishes annual reports on global risks including natural disaster risk assessment. Recent reports highlight several types of disasters, with influence in the urban resilience, among the 10 major risks: i) Interstate conflict; ii) extreme weather events; iii) failure of national governance; iv) state collapse or crisis; v) unemployment or underemployment; vi) natural catastrophes; vii) failure of climate change adaptation; viii) water crises; ix) data fraud or theft; x) cyber-attacks (World Economic Forum 2015). EMDAT is an online database of international disasters both natural and technological. During the 20th and in early 21st centuries, until mid-century there were fewer than 10 reported events per year, but following World War II, this has grown to several hundred per year. EM-DAT also provides data for fatalities and financial costs (EM-DAT, 2022). The classification of the resilience of built assets is increasingly becoming a topic of the greatest importance and relevance for both asset managers and their users, with the need for its operationalization through a system increasingly emerging. That is consensual and easy to implement and use (Falcão Silva et al. 2020). There are some countries that are already developing and successfully implementing specific resilience rating systems to support quality management and governance

Disaster Risk Mitigation Through Capital Investment

647

methodologies for different types of buildings, based on different approaches. However, some of them are not yet compatible or standardized, making it difficult for them to be widely implemented and used in the future. There were previous efforts from the world’s most important stakeholders, as well as different collaborative projects: a) The World Bank Group; b) Global Facility for Disaster Reduction and Recovery; c) 100 Resilient Cities program to promote urban resilience from Rockefeller Foundation; d) UNISDR; e) City Resilience Index (CRI), created by Arup and financed by The Rockefeller Foundation; f) Inter-American Development Bank; g) ICLEI, and Cities; among others (URH, 2022). Communities evaluating investments aimed at improving their resilience face a tradeoff between short-term costs and benefits that may only be realized if a disturbance occurs during the planning horizon. Thus, traditional estimates of return-on-investment generally assume a hazard event occurs within the analysis time frame. Yet, as mentioned, even in the absence of a disruptive incident, resilience investments may produce returns that are valuable to the community in other ways. Resilience investment options that achieve the same primary goal may differ with respect to co-benefits. A critical part of improving community-level resilience is acknowledging and prioritizing actions or projects for the buildings and infrastructure systems that support important social functions. A given community may assess the hazards it most readily faces and in turn prepare, mitigate risk, and plan recovery narrowly tailored to this assessment. However, it is also important to assess community goals in a broader, perhaps hazard agnostic, setting, as well, and ensure that these goals are addressed while planning for increased resilience (Fung and Helgeson 2017). The most recent efforts in terms of standardization raise awareness of the need for a structured overview of information on the resilience of buildings and civil engineering works, namely regarding the concept itself and the risks and measures of disasters. Regarding fundamental concepts, ISO 22845 classifies resilience in different contexts, as well as the definitions of resilience that are currently under development. For disaster risks, this international standard defines three types: i) weather-induced, ii) earthquakeinduced; and iii) induced by human hand. In Portugal, in alignment with the ISO 22845 and other existing international documents (Burroughs 2017; VRS 2017), there is already a proposal specific system of classification of resilience of generalized implementation and a methodology of governance to support the management of the quality of the buildings, and to promote a sustainable and resilient AECO sector in line with European Standards and Community Policies already published and to be complied with in the 21st century.

3 Proposed Resilience Rating System The proposed tool simplifies the identification of the resilience and weaknesses of the buildings, allowing for easy communication and comparison, whether over time for the same building or for others. It is aimed at all those involved in the processes of construction, maintenance, and management of built assets, such as designers, contractors, project managers, owners and even insurance companies, more specifically those dedicated to municipal matters, whose need to determine the building and community resilience is

648

M. J. F. Silva et al.

high (Duarte et al. 2021.b). The proposed resilience classification model for built assets seeks to be based on the ISO 22845 standard with a focus on natural disasters whose national exposure is high or medium, adapted from: earthquakes, floods (urban, rivers, seas), fires and tsunamis. The proposed model has a hierarchical structure with three layers: dimensions, indicators and parameters and follows the following principles: i) Minimize performance reduction; ii) Minimize recovery time after an event and iii) Maximize recovery capacity. The classification model, which is semi-quantitative, is based on existing resilience classification systems and sustainability classification systems that are reasonably mature. The scale adopted meets the recommendations of ISO 11863, as it considers 5 different levels expressed in single-digit integers on a scale of 1,3,5,7 and 9, where 1 corresponds to the worst performance and 9 for the best (Duarte, et. al., 2021a; Almeida et al. 2021). For a clearer interpretation of the final score, the numerical score can be transposed into resilience classes from F to A++ allowing the differentiation of resilience levels to be understood and intuitive (Table 1). Table 1. Resilience classification system proposal to Portugal. Generic calibration

Average score

Class

Exceptionally demanding

9

A++

Clearly louder than normal, but not exceptionally demanding Typical, medium or normal

[8,9]

A++

[7,8]

A+

Clearly lower than normal, but acceptable in some duly justified situations

[6,7]

A

[5,6]

B

Generic calibration exceptionally demanding

[4,5]

C

[3,4]

D

Clearly louder than normal, but not exceptionally demanding

[2,3]

E

[1,2]

F

The definition of indicators and parameters aims to assess resilience and facilitate communication and consultation procedures. The parameters subdivide the indicators, and, in turn, each set of indicators expresses in more detail each of the dimensions mentioned above. Its selection was confirmed through the result of the literature review, considering that: i) the selected parameters are measurable; ii) there is information available for its quantification and iii) it is desirable to avoid overlapping or repetition of metrics (Duarte et al. 2021.b. 2021b). The resilience classification system proposed assesses resilience according to 5 dimensions: D1-Environmental (Table 2); D2 – Economic (Table 3); D3 – Organizational (Table 4); D4 – Social (Table 5); and D5 – Technical (Table 6), which are subdivided into 16 indicators and 75 parameters. The evaluation criteria defined for each parameter were initially established based on the limits of different metrics. The process of reviewing and calibrating indicators,

Disaster Risk Mitigation Through Capital Investment Table 2. Indicators and parameters for the environmental dimension (D1). Indicator

Parameter

I1 - Earthquake

P1 - Seismic zoning - type 1 EC8 P2 - Seismic zoning - type 2 EC8 P3 - Seismic vulnerability of PDM soils P4 - Terrain slope P5 - EC8 soil type P6 - Distance to cliffs

I2 - Tsunami and tidal effect

P7 - Land altitude P8 - Distance to the coast P9 - Distance to the river P10 - Natural barriers in the surroundings P11 - Man-made barriers in the surroundings P12 - Moving objects P13 - Rows built between the coast and the building P14 - Susceptibility to the direct tidal effect PDM P15 - Relative location

I3 - Flood

P16 - Distance to the river P17 - Natural barriers in the surroundings P18 - Man-made barriers in the surroundings P19 - Vulnerability to PDM Floods P20 - Distance to vegetation

I4 - Fire

P21 - Density of vegetation P22 - Vegetation maintenance status P23 - Type of vegetation P24 - Adjacent buildings P25 - Proximity to the industrial zone

Table 3. Indicators and parameters for the economic dimension (D2). Indicator

Parameter

I5 - Insurance

P26 - Insurance against natural disasters

I6 - Financial and strategic implications

P27 - Financial plan P28 - Economic assessment of downtime

649

650

M. J. F. Silva et al. Table 4. Indicators and parameters for the organizational dimension (D3).

Indicator

Parameter

I7 - Internal organization

P29 - Business continuity plan P30 - Risk analysis and management P31 - Post disaster recovery plan P32 - Routine P33 - Simulacra P34 - Learning and updating P35 - Destructive event data P36 - Responsible

I8 - External organization

P37 - Compliance with the existing regulatory scenario P38 - External standards for resilient construction

Table 5. Indicators and parameters for the social dimension (D4). Indicator

Parameter

I9 - Emergency infrastructure

P39 - Access to police stations P40 - Access to fire stations P41 - Access to shelters P42 - Access to hospitals and health centres

I10 - Social responsibility

P43 - Occupants P44 - Disclosure P45 - Social vulnerability

parameters, and evaluation criteria, for improvement, is expected to be iterative. The process must be monitored for the influence of judgments or opinions, lack of data and difficulty of quantification. The proposed resilience classification system allows classifying and comparing the performance of buildings, identifying their vulnerabilities, essential information to establish investment priorities. Multiple stakeholders are involved in the life cycle of buildings that may benefit from the developed proposal.

Disaster Risk Mitigation Through Capital Investment

651

Table 6. Indicators and parameters for the technical dimension (D5). Indicator

Parameter

I11 - Conservation

P46 - Year of construction P47 - Structural system P48 - State of conservation

I12 Accessibility

P49 - Building density P50 – Alternative routes P51 - Street characteristics

I13 - Building seismic safety

P52 - Plant irregularity P53 - Irregularity in height P54 - Interaction with adjacent buildings P55 – Slope difference P56 - Expansion joint

I14 - Building safety against fire

P57 – Distance between overlapping spans P58 - Gas installations P59 - Smoke control and evacuation systems P60 - Intrinsic means of combat P61 – Electrical installations P62 - Fire compartmentalization P63 - Security team P64 - External hydrants P65 - Emergency signs and lighting P66 - Fire extinguishers P67 - Fire detection and alarm P68 - Escape routes

I15 - Building security against flooding

P69 - Barriers P70 - Pumping systems against flooding P71 - Vulnerability and exposure of facades P72 - Number of floors

I16 - Building safety against tsunami

P73 - Number of floors P74 - Orientation P75 - Ground floor hydrodynamics

4 Case study implementation This case study comprises 3 schools Buildings (E1, E2 and E3) spread over 3 geographical areas of mainland Portugal:

652

M. J. F. Silva et al.

1. Lisbon city 2. Interior north 3. Littoral south These three school buildings underwent extensive interventions between 2009 and 2011, within the scope of a public investment program. The interventions include equipment, installations and technical projects currently required under the legislation, structural safety, seismic reinforcement, and fire safety aspects (Parque Escolar, 2010). Figures 1, 2 and 3 comprises the results for E1, E2 and E3 considering the five dimensions analysed (D1, D2, D3, D4, D5).

Fig. 1. Resilience classification – Dimensions (school building E1)

Analyzing Figs. 1, 2 and 3 it appears that all the 3 school buildings under study, despite their geographical localization, presented improvements in their resilience class in all dimensions. It should be noted that the buildings were all with a high degree of degradation, which justifies the high investment in rehabilitation interventions and consequently in the improvement of their resilience. Figures 4, 5 and 6 corresponds to the results obtained for the main indicators (I1, I2, I3, I4, I5, I6, I7, I8, I9, I10, I11, I12, I13, I14, I15, I16) considering the type of structure, the proposed resilience classification model, the type of interventions and the corresponding costs. The description of the type of interventions is not the object of the work presented in this research work.

Disaster Risk Mitigation Through Capital Investment

653

Fig. 2. Resilience classification – Dimensions (school building E2)

Fig. 3. Resilience classification – Dimensions (school building E3)

Regarding the resilience classes per indicators (Figs. 4, 5 and 6), there are also several improvements. However, some indicators, which are not only dependent from the building’s intrinsic characteristics, but also from more extrinsic characteristics, remain roughly at the same level of resilience class, for the three schools’ buildings under study.

654

M. J. F. Silva et al.

Fig. 4. Resilience classification – Indicators (school building E1)

Fig. 5. Resilience classification – Indicators (school building E2)

Disaster Risk Mitigation Through Capital Investment

655

Fig. 6. Resilience classification – Indicators (school building E3)

5 Conclusions The work presented contributes to a discussion on ways to measure the resilience of built assets, namely based on a resilience rating system composed of 5 dimensions, 16 indicators and 75 parameters. The proposed classification system considered in the scope of the present study contemplates not only the intrinsic qualities of the building, but also its interdependence with the community, with the surroundings and with the users in a post-disaster context. The resilience rating system used allows different stakeholders to identify which aspects of the built assets, and efficiently and quickly, should be improved so that it is possible to establish investment priorities to increase their resilience in the face of the occurrence of events. Extremes. This information can be useful for all stakeholders, that is, the owner, asset managers, insurers, and municipal entities, allowing a better understanding of the important contribution of built assets to the construction of resilient communities. However, there is still under development further work, to improve the resilience classification system and to implement it in a representative number and diversity of building assets, as well as to expand the scope of application of the proposed multivariate classification system with respect to other types of risks (man-induced risks) and the identification of compensatory measures and their classification.

656

M. J. F. Silva et al.

References NM Almeida MJF Silva F Salvado H Rodrigues D Maletiˇc 2021 Risk-informed performancebased metrics for evaluating the structural safety and serviceability of constructed assets against natural disasters Sustainability 13 11 5925 https://doi.org/10.3390/su13115925 Burroughs, S.: Development of a Tool for Assessing Commercial Building Resilience. University of Canberra, ACT 2601 Australia (2017). https://doi.org/10.1016/j.proeng.2017.04.263 Duarte, M.: Resilience classification system for buildings facing natural risks. MSc in Civil Engineering, Instituto Superior Técnico, Universidade de Lisboa (in portuguese) (2021) Duarte, M., de Almeida, N.M., Falcão, M.J., Rezvani, S.: Resilience rating system for buildings and civil engineering works. 15th WCEAM (2021a). https://d322cbc3-d0a9-4a0d-87e9-aa9414 df6e27.filesusr.com/ugd/4d4145_82522a050aa9486eb137754639bad2db.pdf Duarte, M., Almeida, N., Falcão Silva, M.J., Salvado, F.: Resilience of constructed assets against natural extreme events from the engineering standpoint. CEES (2021). Coimbra, Portugal (2021) EMDAT, (2022) The International Disaster Database : http://www.emdat.be. Accessed 13 Aug 2022 MJ Falcão Silva NM Almeida De F Salvado H Rodrigues 2020 Modelling structural performance and risk for enhanced building resilience and reliability Innov. Infrastruct. Sol. 5 1 1 20 https:// doi.org/10.1007/s41062-020-0277-1 Fung, J., Helgeson, J.: Defining the resilience dividend: accounting for co-benefits of resilience planning. NIST Technical Note, 1959 (2017) https://doi.org/10.6028/NIST.TN.1959. C Harrison P Williams 2016 A systems approach to natural disaster resilience Simul. Model Pract. Theo. 65 11 31 https://doi.org/10.1016/j.simpat.2016.02.008 ISO 11863: – Buildings and building-related facilities - Functional and user requirements and performance - Tools for assessment and comparison. International Organization for standardization (2011) ISO/TR 22845:2020 - Resilience of buildings and civil engineering works. International Organization for standardization V Sarhosis D Dais E Smyrou ˙IE Bal 2019 Evaluation of modelling strategies for estimating cumulative damage on Groningen masonry buildings due to recursive induced earthquakes Bull. Earthq. Eng. 17 8 4689 4710 https://doi.org/10.1007/s10518-018-00549-1 VRS: Voluntary resilience standards. Meister Consultants Group, Inc. (2017) World Economic Forum, (2015) Global Risks 2015 Report, World Economic Forum (2015) http:// www3.weforum.org/docs/WEF_Global_Risks_2015_Report15.pdf. Accessed 18 July 2022

Optimized Petri Net Model for Condition-Based Maintenance of a Turbine Blade Ali Saleh(B) , Manuel Chiachio, and Juan Chiachio Department of Structural Mechanics and Hydraulic Engineering, Dr. Severo Ochoa Street, 18071 Granada, Spain {alisaleh,mchiachio,jchiachio}@ugr.es

Abstract. Monitoring a structure’s health is important to avoid catastrophic failures and reduce operating costs by applying the Condition Based Maintenance (CBM) strategy. CBM can reduce the inspections, but cannot replace them because of the probability of failure or error of CBM. Reinforcement learning (RL) is an artificial intelligence technique for optimizing decisions based on unrelated factors as it connects the decision to a final goal without understanding the problem details. Also, it allows for automatic policy updates without any user intervention. On the other hand, Petri nets (PN) are mathematical tools suitable for maintenance modelling since they can model heterogeneous information, parallel operations, and synchronization, and provide a graphical interpretation. In this study, the Petri net model (PN) is combined with the Monte Carlo Reinforcement Learning (MCRL) method to find the optimal maintenance strategy and the optimal inspection intervals for wind turbine blades as a function of the quality of the condition monitoring system (CMS), the health of the blade, and the remaining useful life of the wind turbine.

1 Introduction The offshore wind industry has been growing by around 30% each year from 2010 to 2018 (IEA 2019). At the same time, offshore wind turbines (OWT) are subjected to difficult weather conditions and are prone to risks through mechanical force, erosion, and biofouling (Röckmann et al. 2017). This affects the operation and maintenance (O&M) costs, which represent 25–30% of the offshore wind farms’ lifecycle costs (Röckmann et al. 2017). In recent years, several studies have been made to optimize the energy costs of OWT by investigating the impact of reliability (Dao et al. 2019) on OWT, the causes, and the effects of downtime (Scheu et al. 2012) on turbine productivity. Other works have also considered the effect of environmental conditions, crew transfer vessels, and crane ship availability on OWT O&M (Besnard et al. 2012), and also the effect of weather, the price for vessels, workers’ working hours, type of maintenance, failure rates, and electricity prices (Hofmann and Sperstad 2013). Irrespectively, holistic O&M models that can represent and optimize O&M policies under heterogeneous conditions, are missing in the literature. This study presents a PN model for OWT reliability accounting for degradation, condition monitoring, inspection, and maintenance. The model has been trained by using the © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 657–664, 2023. https://doi.org/10.1007/978-3-031-25448-2_61

658

A. Saleh et al.

Monte Carlo Reinforcement learning (RLMC) method to optimize the condition-based maintenance and inspection intervals while taking into account weather conditions, reliability of CMS, running costs, electricity prices, working hours, components reliability, site accessibility, probability of failures, and lifespan of the wind turbine.

2 Methodology 2.1 Monte Carlo Reinforcement Learning Reinforcement learning (RL) is a machine learning training method based on rewarding or punishing desired or undesired behaviors respectively. It teaches a learning element called the agent from interacting with the environment through trial and error. The agent receives rewards after each action and changes the state of the environment. The rewards are used to evaluate being in a state or taking an action from that state, and these are known as value functions. In the absence of a complete model of the environment, it is required to calculate the value functions of state-action pairs or what is known as the Q-Values (Sutton and Barto 2018). This study uses Monte Carlo Reinforcement learning (MCRL), which is one of the model-free RL methods. The method works by generating episodes following an initial random policy referred to as π : S0 , A0 , R1 , S1 , A1 , R2 , . . . , ST −1 , AT −1 , RT , where St , At , and Rt are the state, action, and reward at time t respectively, and T is the terminating state of the episode. Then, the discounted expected return, Gt , can be obtained as the sum of future rewards starting from the time step t, discounted by the discount rate γ , as follows: Gt = Rt+1 + γ Rt+2 + γ 2 Rt+3 + · · · =

T 

γ k−t−1 Rk

(1)

k=t+1

The expected return at the time step, t, is then used to update the value function Q(St , At ), of the state-action pair encountered at that step, as follows:   Q(St , At ) = Q(St , At ) + α Gt − Q(St , At ) (2) where α ∈ [0, 1] is a learning rate parameter, with α =1 meaning that the effect of the latest update will be dominant. The value functions are evaluated to find the optimal policy that increases the long-term rewards. This can be done by updating the initial policy by favouring the actions with higher Q-Values, which is known as the greedy policy. Indeed, when the policy is updated, the Q-Values no longer satisfy it because they are based on old actions. For that reason, the policy evaluation (updating Q-Values), and the policy improvement steps (updating the policy) should be repeated until the policy becomes stable. A problem that occurs when following a greedy policy is that it prevents the agent from trying actions with lower Q-Values initially, knowing that initial Q-Values are not reliable because they are based on a non-optimal policy. This is known as the exploration-exploitation dilemma that exists in almost all RL methods. One of the methods for solving this is the ε-greedy strategy as follows: (3)

Optimized Petri Net Model for Condition-Based Maintenance

659

where ε is the exploration rate parameter and is the set of actions available at state s. This method keeps a probability equal to ε for exploration, which ensures that all actions will be visited and updated even if they have low Q-Values. The parameter ε should be specified between 0 and 1, with 1 meaning that the policy is fully exploratory. It is recommended to start the problem with a high exploration rate and decay it gradually as the solution converges. 2.2 Reinforcement Learning with Petri Net Model A PN is a directed bipartite composed of connected transitions and places. The places contain tokens, where the number of tokens in a place is the marking of that place, and the markings of all places define the state of the PN. Mathematically, a PN is defined as a tuple N = P, T, F, W, Mo , where P, T, F, W, and Mo are the sets of places, transitions, arcs, arcs’ weights, and initial marking respectively (Murata 1989). The dynamics of a PN are controlled by the firing rule, which says that if the number of tokens in the pre-set places of a transition is greater than or equal to the weights of its pre-set arcs, the transition can fire. When a transition fires it consumes tokens from preset places equal to the pre-set arcs’ weights and produces tokens in the post-set places equal to the post-set arcs’ weights. This causes a change in the markings, which can be described using the state equation defined as: M k+1 = M k + AT uk (4) where uk is the firing vector, which is a binary vector describing the firing states of the transitions, and AT is the incidence matrix, which represents the difference between weights of input and output arcs connecting places and transitions. Additional definitions are added to the PN to model the complexity of practical applications. Time delays can be assigned to transitions, given by τ , and these PNs are then called Timed Petri Nets (TPN). The value of τ can be deterministic or given by a probability density function, which in this case will be called a stochastic Petri Net (SPN). Once the firing rule is satisfied in a TPN or a SPN, the transition is said to be enabled, however, it will not fire until the delay time passes while maintaining the enabled state of the transition. Maintenance models usually require the use of PN variants known as high-level Petri nets (HLPN) (Chiachío et al. 2022). Indeed, this study uses inhibitor arcs and reset arcs. If the transition is connected to an inhibitor arc, which is an arc with an empty circular ending, the transition is prevented to be fired if the marking of the inhibiting place is more than or equal to the weight of the inhibiting arc. If the transition is connected to a place by a reset arc, which is an arc with a filled circular ending, the place marking is changed to the weight of that arc. To combine RL with PN, additional definitions are added in this work. The first one is the action groups, g, which is used to group transitions that represent different actions. This definition is used to allow RL agent to choose one of the transitions to be enabled once all of them are enabled based on the RL policy. The other one is the delay group, d , which controls the delay time of any transition in a delay group by RL each time that transition is enabled. This can be used to optimize any time period, like the periods between consecutive inspections, based on the condition of the system. A

660

A. Saleh et al.

delay group, d, takes three arguments, namely a, b, and m, which represent the minimum delay, maximum delay, and the number of discretized values respectively, and uses these arguments to discretize the delay to a vector of values. Then, RL evaluates the value function of choosing each delay as a separate action.

3 Optimized Petri Net Model for the Blade of an Offshore Wind Turbine The proposed methodology is used here to create a PN model that acts as a decision support system for the condition-based maintenance of the blade and CMS of OWT. Figure 1 shows the PN responsible for simulating the inspection, condition monitoring, degradation, and decision making of the blade of an offshore wind turbine. Transitions marked with blue triangles are the ones with delays (timed or stochastic transitions) while transitions marked with red triangles are the ones that produce reward when they fire. Places p1 and p7 represent the true condition and the known condition of the turbine blade and place p4 represents the reliability of the condition monitoring system (CMS). These places can be marked by three, two, one, or zero tokens, where these markings represent the normal, degraded, critical, and failed states, respectively. Two conditions are defined for the blade because the CMS is assumed to produce errors if its condition deteriorates, hence the known condition, which is the condition available for the user, can be different from the actual one.

Fig. 1. Petri net model for the inspection, condition monitoring, degradation, and decision making of the blade of an offshore wind turbine

Optimized Petri Net Model for Condition-Based Maintenance

661

Transition t1 models the deterioration of the blade as a function of its state according to the distributions of its delay time provided in Table 1. When t1 fires, it consumes one token from p1 and marks p2. Then, if p1 is unmarked (failed state), t35 fires to mark p14 (number of failures) and p13, or if p1 is marked, t36 fires to mark only p13. After that, t2, t3, or t4 fire if the CMS is in a normal, deteriorated, or failed condition respectively, and these cases represent successful, erroneous, or failed updates of the known condition. An erroneous update of the known condition is described by marking p3 which fires one of the transitions t8 to t11 randomly, whereas a correct update is described by marking p5 which fires one of the transitions t12 to t15. The detection of any deterioration of the blade leads to marking, p8, and taking a decision to repair it (t18), or not to repair it (t19), through action group g2. The deterioration of the CMS is modeled by transition t20 according to the distributions provided in Table 1. Also, note that a periodic inspection is modeled in parallel to the condition monitoring through transition t7. The delay time of t7 controls the period between two consecutive inspections, and it is controlled by RL through delay group d 1(1, 20, 10), which works as described in Sect. 2.2. Every time this transition is fired, the known condition of the blade is updated accordingly, and the CMS is checked. Then, decisions for repairing the CMS and the blade are taken through action groups g1 and g2 if any of them is in a deteriorated state.

Fig. 2. Petri net model for checking if the repair can continue once the maintenance team arrives at the site

The decisions are taken by the RL agent based on the state of the RL environment. The RL environment is defined to include the known condition (p7), the CMS condition (p4), and the remaining useful life of the turbine. The remaining useful life is considered by discretizing the lifespan of the turbine into intervals that affect the RL state. The bounds of the intervals are chosen as: [0, 10, 20, 30, 40, 50, 60, 70, 75], where 75 years is the maximum lifespan of the OWT. All the rewards have been defined in monetary terms either as costs or as losses of the revenues. The losses can be due to downtime or decreased efficiency of the turbine, and they are calculated by subtracting the actual revenues, Rv , from the normal revenues, Rv. For a time period, p[yrs.], the normal

662

A. Saleh et al.

revenues have been calculated by multiplying the annual revenues, period, p:

, by the

(5)

where AR is the annual revenues, AE is the annual energy production, ER is the electricity rate, CF is the capacity factor of the OWT, and P is the capacity of the OWT. It has been assumed that each failure decreases the efficiency of the OWT by a factor of 0.8, so the actual revenues can be calculated by multiplying Rv by 0.8M (p14) , where M (p14) is the marking of p14 which represent the number of failures. Then, to account for the downtime, the result must be multiplied by M (p11), which is a binary place representing the working state of the OWT. Transitions t5 and t18 represent the CMS and blade repair decisions, respectively. It is assumed that the repair of CMS can happen on-site without the need for additional equipment, so t5 directly reset p4 to normal. On the other hand, the repair of the blade requires logistic preparation, waiting for good weather, traveling to the site, and doing the actual repair. The first three steps are combined in transition t39 while the actual repair is modeled by the Petri net shown in Fig. 2. Each of these steps requires different time and costs based on the state of the blade. The transitions responsible for updating the known condition (t8 to t15) reset p16 every time they fire to cancel the old preparation if it exists. It is assumed that even if the old preparation was canceled, part of the preparation costs has to be paid once it was started. For this, the preparation costs are divided between transitions t18 which represents that preparation has started and t39 which represents that preparation has ended and the site was reached. All the costs and delay times are listed in Table 1 and are based on the paper published by (Le and Andrews 2016). Once the site is reached, the repair must occur based on the turbine’s true condition. However, the repair may be prepared based on a wrong known condition because of CMS’s erroneous, or the condition of the blade may change before the site is reached. For this, the actual repair can occur only if the known and true conditions are the same, which is described by the “Check Condition PN” shown in Fig. 2. If any of the failed repair scenarios happened, the CMS is checked, and a decision regarding its repair is taken if it was found to have deteriorated. Also, the known condition of the blade is updated correctly regardless of the CMS condition because it can be checked directly on-site, and a decision is taken regarding its repair if it was found to have deteriorated. On the other hand, if the repair was successful, the known and true states of the blade will be reset to the normal condition and the costs of repairing will be counted as rewards.

Optimized Petri Net Model for Condition-Based Maintenance

663

Table 1. Time delays and rewards of the timed and reward transitions. All other transitions are is Weibull distribution with shape β and scale η. instantaneous and without rewards. Parameters Acronyms for years, weeks, and hours are yrs. wks., and hrs. Respectively. Name

Delay

Reward -

Controlled by RL -

-

4 Results and Conclusions The learning process is divided into intervals of 200 episodes to plot the average and uncertainty bounds of different variables. Figure 3 shows that the average rewards have increased, average downtime have decreased, and the average number of failures has decreased as a function of the number of episodes. The rewards are calculated in terms of losses and costs, and the figure shows that they became almost zero when the learning process ended. This reflects that the method was successful in finding an optimal policy. Also, it can be noted that the number of failures and downtime is directly related to the losses, and this is why they were reduced to almost zero by the RL agent.

Fig. 3. Variables of interest as a function of episode number

664

A. Saleh et al.

Besides, the final policy shows that for most cases before 70 years of wind turbine service, it is better to repair the blade and the CMS if any deterioration is detected. However, it is better to follow run to fail strategy after 70 years of its service. This time dependency can be explained by saying that as approaching the end of life of the turbine, it is not important anymore to keep it in a good condition because it’s going to be out of service, so the goal will change to making use of it as much as possible before getting rid of it. On the other hand, the choice of repairing as soon as deterioration is detected is because the minor maintenance costs are less than the major one, so it’s better to repair before a critical condition is reached. Also, the deterioration time from normal to degraded condition is much larger than that from degraded to critical or to failure. Thus, it is like investing a small amount of money to make a component serve for a longer time which is more attractive. These results are affected by the assumption that maintenance can return the component to its pristine state. This is not very realistic and can be considered in future studies. Regarding the inspection intervals, it was found that they were getting shorter when the remaining life or the condition of CMS decreased. This is because it will take some time before the blade starts deteriorating, so inspection will not be very important at the beginning of the service life. Also, there is no need to make an inspection as long as the CMS works properly.

References Besnard, F., Fischer, K., Tjernberg, L.B.: A model for the optimization of the maintenance support organization for offshore wind farms. IEEE Trans. Sustain. Energy 4, 443–450 (2012) Chiachío, M., Saleh, A., Naybour, S., Chiachío, J., Andrews, J.: Reduction of Petri net maintenance modeling complexity via approximate Bayesian computation. Reliab. Eng. Syst. Saf. 222, 108365 (2022) Dao, C., Kazemtabrizi, B., Crabtree, C.: Wind turbine reliability data review and impacts on levelised cost of energy. Wind Energy 22, 1848–1871 (2019) Hofmann, M., Sperstad, I.B.: NOWIcob–a tool for reducing the maintenance costs of offshore wind farms. Energy Procedia 35, 177–186 (2013) IEA 2019: Offshore wind outlook (2019) Le, B., Andrews, J.: Modelling wind turbine degradation and maintenance. Wind Energy 19, 571–591 (2016) Murata, T.: Petri nets: properties, analysis and applications. Proc. IEEE 77, 541–580 (1989) Röckmann, C., Lagerveld, S., Stavenuiter, J.: Operation and maintenance costs of offshore wind farms and potential multi-use platforms in the dutch north sea. In: Buck, B., Langan, R. (eds.) Aquaculture Perspective of Multi-Use Sites in the Open Ocean, pp. 97–113. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-51159-7_4 Scheu, M., Matha, D., Hofmann, M., Muskulus, M.: Maintenance strategies for large offshore wind farms. Energy Procedia 24, 281–288 (2012) Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction, MIT Press, Cambridge (2018)

Multi-disciplinary and Dynamic Urban Resilience Assessment Through Stochastic Analysis of a Virtual City Seyed M. H. S. Rezvani1(B) , Nuno Almeida1 , and Maria João Falcão Silva2 1 CERIS Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais 1,

1049-001 Lisbon, Portugal {seyedi.rezvani,nunomarquesalmeida}@tecnico.ulisboa.pt 2 Laboratório Nacional de Engenharia Civil, Av. do Brasil 101, 1700-075 Lisbon, Portugal [email protected]

Abstract. Natural disasters cause an average of 60,000 deaths per year. There was a total of 416 natural disasters worldwide in 2020. Resilience to disruption and quick recovery of the operational performance of buildings and infrastructure is critical to the continuous functioning of cities and communities. This requires recognition of critical weaknesses of urbanized society. This recognition can be mapped in terms of different dimensions that must be rated and weighted based on their relative importance. There have been various efforts to recognize urban resilience indicators and parameters. This paper aims at contributing to the current body of knowledge by differentiating the resilience scoring of buildings and infrastructure based on their use-type. The study includes a stochastic analysis of a virtual city consisting of 10,000 buildings with various use-types. The virtual city is modelled with earlier case studies undertaken by the authors in the city of Lisbon. This study proves how intrinsic and obtained values can be used to generate a building resilience score in conjunction with use-type.

1 Introduction Senior management of public and private organizations is focusing on the requirements to boost the value given by the urban built environment, making urban resilience a more significant subject (Duarte et al. 2022). Even though the resilience domain has recently received a lot of attention, (Hernantes et al. 2019; Marana et al., 2019; Rasoulkhani et al. 2019; Yao and Wang 2020), there is still a widespread disagreement over how to quantify resilience at the level of constructed assets and asset systems (i.e., buildings and infrastructure). Figure 1 shows the results of a Scopus search for “Urban resilience.” More than 10,000 papers are expected to be published in 2022, according to estimates. A risk-managed performance-based approach to urban constructed assets (built environment) assists the various stakeholders involved throughout their life cycle to enhance their resilience in relation with inherent risks and also risk aggravation factors (Almeida et al. 2015) that need to be properly managed and controlled. For example, assets that have been underutilized or ignored may frequently be managed more effectively by © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 665–673, 2023. https://doi.org/10.1007/978-3-031-25448-2_62

666

S. M. H. S. Rezvani et al.

Fig. 1. Urban resilience keyword appearance in Scopus publications

methodically analyzing and accounting for their inherent risks. This method is especially effective for assets towards the end of their useful life when standard asset management techniques are needed to decide whether the best option is a full replacement or a refurbishment or upgrade. Resilience considerations need to be factored in in these decisions. Different researchers have been proposing urban resilience evaluation methods calibrated in several situations and based on various nation codes (Almufti and Willford 2013; Atrachali et al. 2019; Burroughs 2017; Büyüközkan et al. 2022; RELi 2018). One of the issues with these systems is how to account for and overcome the subjective biases of the many stakeholders and decision-makers engaged in improving urban resilience. There are numerous resilience evaluation systems calibrated in different circumstances, including resilience-related characteristics in built-environment assessment methodologies, and based on different country codes in the context of urban resilience, including REDi, ARMS, RELi, and LiderA, to name a few (Almufti and Willford 2013; Atrachali et al. 2019; Burroughs 2017; Büyüközkan et al. 2022; RELi 2018). REDi™ is a resilience-based scoring scheme for seismic and beyond-code design approaches used in the planning and evaluation phases to reach higher performance design (Almufti and Willford 2013). The Australian Resilience Measurement Scheme for Buildings (ARMS) is a comprehensive approach that applies resilience that includes physical infrastructure, environmental, economic, social, political, regulatory, and organizational resilience. This system assesses the building’s resilience by evaluating many dimensions and sub-dimensions and rating its performance elements (threats and opportunities) (Burroughs 2017). RELi™ 2.0 is a comprehensive, resilience-based scoring system for socially and environmentally resilient design and construction in integrative design processes, based on current sustainable and regenerative criteria for emergency management, adaptability, and social strength (USGBC 2018). LiderA is a sustainability certification framework that enables the evaluation and is directed to the design, construction, and operation stages of all sorts of developed facilities in the building envelope (LiderA 2005; Pinheiro 2011, 2020). Developing resilience evaluation is now a norm on a global scale. Certain nations, including such Australia (ARMS, BRRT) as well as the United States of America (RELi, FORTIFIED, ANCR, BRLA), have achieved noteworthy progress, particularly in the area of natural disasters caused by climate and seismic threats. Learning opportunities in communities provide more opportunities for individuals to communicate and engage. The use of disaster-reduction techniques reduces the amount

Multi-disciplinary and Dynamic Urban

667

of local community knowledge. As officials take over upkeep, development, and safety of public areas, it appears that top-down management encourages people’s separation from the community (Hosseinioon 2019). The numerical results of using a resilience index based on the number of households lacking service to compare different scenario occurrences have evaluated with the virtual city concept (Bianco et al. 2017). In a metropolitan region that includes Bangkok and Tokyo, different research assessed the structural analysis of virtual social capital for urban resilience (Kawamoto et al. 2021). The demands of the urban environment are taken into account when developing decision-making tools, which can be either subjective (based on the opinions of experts) or objective (data-driven and stochastic), working alone or in tandem, with or without weighting of Multi-Criteria Decision Method (MCDM) techniques. It is difficult enough to choose the best evaluation approach without considering the circumstances. In order to manage asset systems that are subject to various risks, such as tangible or intangible, natural, or man-made disasters, decision-makers require a variety of tools and methods to enhance the decision-making process. A mix of data-driven and stochastic analysis will be required to arrive at a unified multidisciplinary approach for sustainable UR. The URES (Urban Resilience Evaluation System) and ARCDM (Automated Rational and Consistent Decision-Making Simulations) seek to address the subjective bias of a varied collection of influencers and decision-makers involved in strengthening urban resilience (S. M. Rezvani et al. 2022). The ARCDM is a stochastic algorithmic approach based on MCDM and the Analytic Hierarchy Process (AHP) that can help alleviate this restriction of typical decision-making tools (S. Rezvani & Almeida, 2021). ARCDM can assist by generating several objective situations that fall within the acceptable subjective consistency range. The goal of the study under development is to address the subjective biases of a wide variety of stakeholders and decision-makers involved in enhancing urban resilience. Additionally, by utilizing stochastic virtual city modelling throughout the entire city, the study aims to enable the transition from firm and static to multi-disciplinary and dynamic event resilience measurement as well as to suggest a solution to an ongoing barrier in resilience measurement. The limitations of existing resilience assessment approaches can be overcome by merging prior case studies and expanding the results to a larger sample population to identify the weakest point of each dimension in terms of structure usage type. The Multicriteria approach makes it possible to define dynamic urban resilience behavior in a variety of urban settings, including interaction and feedback between dimensions, indicators, and parameters, including social, economic, and environmental factors, along with illustrating short-term and long-term urban resilience perspectives. The stochastic location-based method with ARCDM enables scenario modeling of natural hazards inside the system dynamics model, allowing for a clear depiction of the uncertainty of disaster risk management in the evaluation of urban resilience. The ARCDM model generates several decision matrices of the specified size and then runs them through a consistency screening algorithm (Python-based developed code). These extracted AHP decision matrices would be of multiple levels, with the dimensions (Environment, Economic, Organizational, Social, and Technical) being the highest level, followed by indicators and parameters (Duarte et al. 2022; S. M. Rezvani et al. 2022). By regenerating them through uncertainty analysis in the context of the stochastic virtual

668

S. M. H. S. Rezvani et al.

city and isolating each use-type result with their consistency in outputs, these ARCDM level-based analyses enable us to identify the weakest parameter in various settings, which in this study as the novelty the result was made more consistent. The research is based on previous case studies developed for the city of Lisbon. The findings of calibration, validation, and simulation show that this technique is suitable for analyzing various urban resilience scenarios. This could help develop urban resilience strategies within urban policy planning (Chirisa and Nel 2021). The construction of a novel alternative and comparison of urban resilience scenarios across a variety of likely conditions for indicators and parameters in URES is presented in the dynamic and stochastic evaluation of urban resilience to natural disasters (Feofilovs and Romagnoli 2021).

2 Methodology Decision science has applications in many fields of research and is acknowledged as a helpful tool for diverse decision-makers to make a succinct and objective choice (Rezvani & Almeida, 2021). Previous research looked at 11 buildings in Portugal that fit into six types of use (Duarte et al. 2022). Based on the density of the available constructions in Lisbon, the new research incorporates a stochastic distribution that allows the production of a virtual city of 10,000 buildings. This study allows the generation of a stochastic distribution using earlier research data and the investigation of five various types of statistical distributions (Rezvani and Gomes 2021), as stated in Fig. 2.

Fig. 2. – Proposed stochastic approach of five groups for generating virtual assets

The model works as shown in the algorithm below, however deciding how to choose which indicator’s regulator is a broad topic that is outside the scope of this research. Then it will be recalculated till the final result is more than 95% accurate when compared to the previously computed stochastic models. The model’s stability is critical for making the results reproducible for other researchers.

Multi-disciplinary and Dynamic Urban

669

IF Regulator=Global_Index THEN Result=INDEX (Global_Range) ELSEIF Regulator=Local_Index THEN Result=INDEX (Local_Range) ELSEIF Regulator= Global_Randomized THEN Result=RAND (Global_Range) ELSEIF Regulator= Local_Rand THEN Result=RAND (Local_Range) ELSEIF Regulator= Free_Randomized THEN Result=RAND (RANGE (1,9))

The first group is the global-indexed distribution, which is based on the previously surveyed indices to obtain the same resilience score as the entire population analyzed in our case study. The second is the inside use-type local-indexed for items that are connected to inner group information and must be disseminated only based on their use-type. The third is the global-randomized distribution of case studies based on the maximum and minimum of previous cases studied in the study’s entire population, which is typically based on case study building locations and some more intrinsic characteristics so that the parameters can freely vary throughout the city. For example, the slope and distance to the sea. The fourth group is the use-type local-randomized distribution, which might deviate much from the population’s maximum and minimum, and they must get a randomized value based on the maximum and minimum of their use-type group. The use type of a building must be approximated as a variable that must be constrained in order to obtain the value of other groups. The last and fifth groupings are simply items with the lowest and greatest resilience values between 0 and 9. The suggested technique enables for a small number of buildings to support choices, the simulation of various scenarios, and the development of a risk assessment for a whole city. This five-category use-type distribution gives a comprehensive perspective that allows the city’s developed assets to attain a wider range of resilience ratings by including other use-type characteristics when permitted. The following step is to combine the stochastic analysis variability with the risk assessment and weight the resulting parameters to provide a comprehensive result across the centrality theory with the randomized and indexed-randomized integrated by the ARCDM weights. For those parameters with the flexibility to move across the resilience scores, it is useful to study the effect of those parameters on one another under various scenarios. They circumvent the limited access to the scores, which are dependent on their use-type within each group and total scores to be included as an appendix to the complete edition.

670

S. M. H. S. Rezvani et al.

The stochastic virtual city model is normally can be used in various contexts, using URES indicators and parameter or even more advance maintenance and lifecycle cost related indicators like the technical dimension’s indicators in URES model; as some instances are electrical installations, fire compartmentation, security team, outdoor fire hydrants, emergency signage and lighting, fire extinguishers, fire detection and alarm, escape paths, barriers, flood pumping systems.

3 Result and Discussion The distribution of buildings initially was based on population, and the requirement for those structures is 10,000 buildings for 40,000 residents (assuming 4–5 per house or apartment), which is approximately 8% of the actual population of the city of Lisbon. It is an appropriate scale for the virtual city under evaluation, and the distribution for the demands of 40,000 residents is 9,000 dwellings, with 1,000 for the remaining five usetypes (Shopping centers and commercial centers, hospitals, hotels, and leisure centers host). In this study, the URES five major dimensions, 16 indicators, and 75 parameters are employed identically as in the previous study (Rezvani et al. 2022), with a greater emphasis on the influence of use-type in the context of virtual city as clearly outlined previously. However, after studying the above-mentioned data, was discovered that the results did not approach stability (the central theorem of statistics), therefore it was necessary to raise the number of buildings to the point where the result could not be modified based on any re-calculation. The combination of the weight based on the use-type is then to be calculated within each group based on the ARCDM scenario analysis of the previous studies. That is used as a weighted risk assessment analysis of the resilience assessment system. Subsequently, based on the combination of these weights in different scenarios, it can be determined where the exact weaknesses of each asset system are located, and the specialist in each domain can propose various solutions to reduce the risk of natural hazards and increase urban resilience to maintain the city’s performance level at an operational level (Fig. 3). EDUCATION 5.2 5.2 5 1

2

3

4

5

HOSPITAL

6.3

FACTORY

6.1 5.7

SHOPPING CENTER HOTEL

6

7

8

9

HOUSE

Fig. 3. - Use-type Resilience Score distribution through stochastic analysis

The scenario analysis implies that there is always a step toward the system’s reliability analysis. For example, in one use-type, there is a deficit in the social dimension in certain

Multi-disciplinary and Dynamic Urban

671

indicators such as age combinations, whereas in another use-type, the problems are in the economic dimension. The power of scenario analysis is that policymakers or individuals may pinpoint the exact target at which failure may occur. This, in conjunction with stochastic approaches, will increase the ability to uncover undetected instabilities in reliability analysis that the survey or real data did not capture. The generated result could also be used in other cities by just changing the parameters that distinguish those two cities. Figure 4 uses a box and whisker graphic to represent the data distribution into quartiles while also emphasizing the mean and outliers. There is a chance for whiskers, or vertical lines that protrude from the boxes. An outlier is defined as a point outside the whiskers or lines that indicates variability outside the upper and lower quartiles. When the box is larger, such as at a hospital, it signifies that different indicators are rated differently. Some indicators have scores of roughly 5, and such indicators may be improved. Furthermore, we should pay close attention to the lower outliers since they have the potential to lead to failure in the early phases of a disruption and to cascade failure in other weak indicators. Therefore, it is crucial to improve the lower outliers in order to prevent domino failure. To achieve this, work efficiently on just the indicators that need to be improved in order to raise the overall stability of the asset system.

Fig. 4. Use-type Resilience Score distribution in box and whisker

4 Conclusion As a result of their behavior in organizational and economic dimensions, homes and apartments were found to be the least resilient asset systems in a multidisciplinary

672

S. M. H. S. Rezvani et al.

dynamic analysis of the virtual city. Cost-cutting methods and a lack of over-design, such as in the case of public constructions, are responsible for this. Although factories, educational institutions, and schools often perform below averagely, they can be made better by enhancing their level of economic, social, and technical resilience. Due to their well-built structures and careful attention to organizational indicators, shopping centers, commercial centers, hospitals, hotels, and leisure centers typically have resilience scores that are better than average and more reliable. This is especially true considering that these buildings can be located anywhere in a virtual city and can undergo multidisciplinary dynamic analysis while considering different probabilities of their variables. The overall score of the asset systems is calculated at this stage of the study, and the next stage will work stochastically on each indicator and their associated parameters. This will enable a “weak point finder” tool, which will allow policymakers to decide on which aspects of which building types to improve. For example, providing flood barriers and discharge points for neighborhoods near rivers and bodies of water, or pressing residential regions for more rehabilitation work by allocating budget to them, or building higher capacity infrastructure for those specific use-types to withstand natural disasters. In order to address the question to generalize the model, it should be stated that the model can be generalized for any cities all over the world if a sufficient number of case studies are available, and there would be a need to amplify the result to the entire region where we cannot conduct additional case studies.

References Almufti, I., Willford, M.: REDi TM Rating System Resilience-based Earthquake Design Initiative for the Next Generation of Buildings (2013) Atrachali, M., Ghafory-Ashtiany, A.-H., K., Arian-Moghaddam, S.: Toward quantification of seismic resilience in Iran : developing an integrated indicator system. Int. J. Disaster Risk Red. 39 (2019)https://doi.org/10.1016/j.ijdrr.2019.101231 Bianco, M., Cimellaro, G.P., Wilkinson, S.: Virtual city for water distribution research in crisis management. In: COMPDYN 2017 - Proceedings of the 6th International Conference on Computational Methods in Structural Dynamics and Earthquake Engineering, vol. 1, pp. 2075–2088 (2017). https://doi.org/10.7712/120117.5550.18269 Burroughs, S.: Development of a tool for assessing commercial building resilience. Procedia Eng. 180, 1034–1043 (2017). https://doi.org/10.1016/j.proeng.2017.04.263 Büyüközkan, G., Ilıcak, Ö., & Feyzio˘glu, O.: A review of urban resilience literature. Sustain. Cities Soci. 77 (2022).https://doi.org/10.1016/j.scs.2021.103579 Chirisa, I., Nel, V.: Resilience and climate change in rural areas: a review of infrastructure policies across global regions. In: Sustainable and Resilient Infrastructure pp. 1–11. Informa UK Limited (2021). https://doi.org/10.1080/23789689.2020.1871538 de Almeida, N.M., Sousa, V., Alves Dias, L., Branco, F.A.: Managing the technical risk of performance-based building structures. Vilnius Gediminas Tech. Univ. 21(3), 384–394 (2015). https://doi.org/10.3846/13923730.2014.893921 Duarte, M., Almeida, N., Falcão, M.J., Rezvani, S.M.H.S.: Resilience rating system for buildings against natural hazards. In: Pinto, J.O.P., Kimpara, M.L.M., Reis, R.R., Seecharan, T., Upadhyaya, B.R., Amadi-Echendu, J. (eds.) WCEAM 2021. LNME, pp. 57–68. Springer, Cham (2022). https://doi.org/10.1007/978-3-030-96794-9_6

Multi-disciplinary and Dynamic Urban

673

Feofilovs, M., Romagnoli, F.: Dynamic assessment of urban resilience to natural hazards. Int. J. Disaster Risk Red. 62, 102328 (2021). https://doi.org/10.1016/J.IJDRR.2021.102328 Hernantes, J., Maraña, P., Gimenez, R., Sarriegi, J.M., Labaka, L.: Towards resilient cities: a maturity model for operationalizing resilience. Cities 84, 96–103 (2019). https://doi.org/10. 1016/j.cities.2018.07.010 Hosseinioon, S.: Urban resilience and informality: effects of formalisation in Golestan, Iran. In: Brunetta, G., Caldarice, O., Tollin, N., Rosas-Casals, M., Morató, J. (eds.) Urban Resilience for Risk and Adaptation Governance. RC, pp. 111–127. Springer, Cham (2019). https://doi. org/10.1007/978-3-319-76944-8_8 Kawamoto, K., Tontisirin, N., Yamashita, E.Y.: The structural analysis of virtual social capital for Urban resilience in a Metropolitan Area: The case of Tokyo and Bangkok. Nakhara: J. Environ. Des. Plann. 20. (2021). https://doi.org/10.54028/NJ202120101 LiderA – Sistema de avaliação da sustentabilidade (2005). http://www.lidera.info/ Marana, P., et al.: Towards a resilience management guideline — Cities as a starting point for societal resilience. Sustainable Cities Soc. 48, 101531 (2019). https://doi.org/10.1016/j.scs. 2019.101531 Pinheiro, M.D.: LiderA Sistema voluntário para a sustentabilidade dos ambientes construídos (2011) Pinheiro, M.D.: Avaliação pelo LiderA do grau de resiliência de Lisbon Green Valley (Sintra) (2020). https://doi.org/10.13140/RG.2.2.34017.25446 Rasoulkhani, K., Mostafavi, A., Cole, J., Sharvelle, S.: Resilience-based infrastructure planning and asset management: study of dual and singular water distribution infrastructure performance using a simulation approach. Sustain. Cities Soc. 48, 101577 (2019). https://doi.org/10.1016/j. scs.2019.101577 RELi: RELi 2.0. December (2018) Rezvani, S., de Almeida, N.M.: Multi-criteria decision analysis of subcontractors selection for infrastructure projects: a case study of an electrified railway project. In: Online Event: 11th IMA International Conference on Modelling in Industrial Maintenance and Reliability (MIMAR) (2021). https://doi.org/10.19124/ima.2021.01.5 Rezvani, S., Gomes, M. C.: Assessment of pavement degradation through statistical analysis model: A case study of the Department of Transportation (DOT) of Iowa, USA. Online Event: 11th IMA International Conference on Modelling in Industrial Maintenance and Reliability (MIMAR). (2021). https://doi.org/10.19124/ima.2021.01.11 Rezvani, S.M., de Almeida, N.M., Falcão, M.J., Duarte, M.: Enhancing urban resilience evaluation systems through automated rational and consistent decision-making simulations. Sustain. Cities Soc. 78 (2022)https://doi.org/10.1016/j.scs.2021.103612 USGBC. (2018). RELi 2.0 Rating Guidelines for Resilient Design + Construction (2018). https:// www.usgbc.org/resources/reli-20-rating-guidelines-resilient-design-and-construction Yao, F., Wang, Y.: Towards resilient and smart cities: a real-time urban analytical and geo-visual system for social media streaming data. Sustain. Cities Soc. 63, 102448 (2020). https://doi.org/ 10.1016/j.scs.2020.102448

Applications of International and Local Guidelines and Standards

Selecting KPIs in Asset Maintenance of Onshore Wind Farms Using Standard EN 15341:2019 Daniel Gaspar(B) , Odete Lopes, Carlos Rodrigues, and Serafim Oliveira Research Centre in Digital Services – CISeD, Polytechnic Institute of Viseu, Viseu, Portugal {danigaspar,odete,soliveira}@estgv.ipv.pt

Abstract. The need for energy and a sustainable environment means that the renewable energy sector is becoming more critical and in constant technological development. The applicability of maintenance indicators from a European Standard about KPIs nowadays has become an indispensable tool in the decisionmaking process. This paper aims to describe and explain the method that was used to select asset management KPIs in the maintenance area. Indicators were selected based on standard EN 15341:2019, that define the various types of maintenance and the essential maintenance performance indicators (KPIs). For selecting the most critical and relevant maintenance indicators, several factors were considered in operation and maintenance actions, from the definition of objectives in asset management strategic, the SCADA control system and is interconnected with a management and maintenance software (CMMS) and the difficulties to collect data. The maximization of assets is essential for business growth, and a wind farm, despite all its specificities, is still an industrial facility for electricity production.

1 Introduction The claim for energy, and consequently for green energy, means that the renewable energy sector is in constant technological development, where the wind and solar energy will be fundamental in the EU energy system (WindEurope 2022). However, it can be considered that the renaissance of wind energy took place after 1980 due to the huge oil crisis at the time, which forced a rethinking of the sector. Since then, and until today, the sector has always been growing, not always as strong as in the last decade, when wind turbines with a nominal power of 5,000 KW were already being manufactured and with future forecasts of greater powers, as shown in the Fig. 1 (WindBox 2022):

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 677–687, 2023. https://doi.org/10.1007/978-3-031-25448-2_63

678

D. Gaspar et al.

Fig. 1. Technological evolution of wind turbines (WindBox 2022)

The standardization of assets in accordance with international standards and the use of indicators to monitor the performance of assets is fundamental, where ISO and IEC international standards gain particular relevance (ISO 2018). The in-depth study of maintenance activities led to the creation of new aspects and analyses, and data analysis is currently fundamental to support future decisions. In this sense, over the last few years, attempts have been made to standardize the various aspects of maintenance that allow global applicability through the publication of international (ISO, IEC) and European (EN) standards. The concepts defined in the standard EN 13306:2017 - Maintenance Terminology show that maintenance is divided into several fields of action, where maintenance indicators become fundamental in decision making (CEN 2019). A wind farm, despite its singularities, must be seen from the same industrial perspective where all the previous concepts are highly relevant, intending to increase its operating results, despite its final product being intangible since it is energy (kWh).

2 The Selection of KPIs in Asset Maintenance Maintenance performance indicators (Key performance indicators - KPIs) are now a reality and have become an extremely powerful tool for decision making. Gonzalez et al. (2017) review of the major existing indicators used in the O&M of wind farms (WFs), and Carnero and González-Prida (2017) analyze different systems of indicators by which maintenance management is evaluated. These indicators provide accurate information on certain highly relevant aspects of asset management. This is the one of the best way to achieve maintenance excellence, using all resources competitively. The EN 15341:2019 standard illustrates the entire KPI

Selecting KPIs in Asset Maintenance of Onshore Wind Farms

679

improvement process. Most of the indicators have applicability in all industrial facilities, being used to quantify and qualify a given asset. With the indicators, comparisons can be established, internal and external benchmarking (Åhrén and Aditya 2009); measure the status of the asset, make a diagnosis of weaknesses and strengths, identify objectives and define goals to be achieved so that continuous improvement actions are planned and control the development of these actions over time (CEN 2019). 2.1 Method for Classifying Indicators for Maintenance There is a huge amount of data available in the context of the operation and maintenance of a wind farm, so it is essential to select the most critical information to improve the asset’s performance. In this context, maintenance indicators are a valuable help because they allow to quantify what matters (Kutucuoglu et al. 2001). For selecting the most critical and relevant maintenance indicators, several factors were considered in operation and maintenance actions, from the definition of objectives in asset management strategic, to the definition of parameters and the difficulties to collect data. These are the indicators that measure, quantify and qualify the actual and expected results as a complex result of maintenance performance and which can be consulted in the European standard EN 15341:2019. From the current European standard of indicators, the following indicators, in Table 1 can be given as an example of selection of indicators. Table 1. Maintenance indicator. Adapt from (CEN 2019) KPI

Factors

Definitions and notes

M11 Availability based on operating time (%)

Operating time

Time interval throughout which an item is performing as required

Required operating time

Required operating time is the interval throughout which the item is required to be in an up-state and operating. It excludes standby time and idle time. It excludes lost operating time due to external reasons such as market demand, laws, business reasons, lack of resources, working time

Total time to restoration

Time interval, from the instant of failure, until restoration (it includes delays and repair time)

O&S16 MTTR (mean time to restoration) (Hours)

(continued)

680

D. Gaspar et al. Table 1. (continued)

KPI

O&S17 MRT (mean repair time) (%)

PHA8 Operational availability due to maintenance (%)

Factors

Definitions and notes

Number of failures

Total failure restored using internal or external resources

Total repair time

Total repair time spent to adapt or repair components or groups internal or external

Number of failures

Total failure restored using internal or external resources

Total Operating time

Time when the physical asset is performing as required

Total Operating time + Downtime

Time when the physical asset is performing as required plus the time lost due to failures and preventive maintenance activities

The definition of indicators for analysis was obtained through the EN 15341:2019 standard and internal procedures, which allowed the classification of KPIs. For the selection of indicators, the possibility of collecting data by accessing the internal database is essential. Without access to the data, it is impossible to analyze the maintenance indicators; this step is an exclusion factor and, in case of impossibility of data collection, it requires a new definition of the internal objectives. After selecting indicators, their calculation and analysis of the results followed, making it possible to verify the degree of fulfilment of the objectives and the possibility of developing improvement actions. To classify the most critical maintenance indicators to the decision making in the wind farm’s operation, the method of the risk analysis and referred in the IEC 31010:2019 standard as a method of the consequence/Probability Matrix. The consequence/probability matrix combines the qualitative or semi-quantitative classification of the consequence to define a risk level. The format of the matrix and the definitions that apply to it depend on the context in which it is used, and an appropriate design for the circumstances must be used. In this case a scale with 25 points was created, divided into three levels, very similar and adjusted from IEC 31010:2019 standard (IEC 2019) (Tables 2 and 3). • Indicator high relevant – included between the 17 and 25 points; • Indicator medium relevant - included between the 9 and 16 points, • Indicator low relevant - included between the 1 and 8 points.

Selecting KPIs in Asset Maintenance of Onshore Wind Farms

681

Table 2. Scale relevance

Scale Relevance: 1a8

Low

9 a 16

Medium

17 a 25 High

Impact relevant factor

Table 3. Weighting matrix

5

5

10

15

20

25

4

4

8

12

16

20

3

3

6

9

12

15

2

2

4

6

12

10

1

1

2

3

4

5

1 2 3 4 5 Effective/accessibility factor

The scale relevance levels of the table were applied where, according to the weighting matrix, a column was inserted as the next factor of the maintenance indicators with the associated representative. The impact factor was quantified from one to five according to the direct interference in the results and impact on the operational wind farm according to the Table 4. The effective data factor is the process of gathering data for use in KPIs indicators. It’s a crucial part of data analysis and provides the information about the accessibility and reliability of the data to collect in order to calculate the KPIs (Table 5). Table 4. Impact factor Impact relevant KPI factor Very high

5

High

4

Medium

3

Low

2

Very low

1

682

D. Gaspar et al. Table 5. Effective data factor Effective/accessible data factor Very effective

5

Good effective

4

Normal

3

Low effective

2

Very low effective

1

2.2 The Selection of Indicators in Maintenance of Wind Farms Within the operation and maintenance of a wind farm, there is an enormous amount of data available, so it is essential to select the most critical information to improve the asset’s performance. In this context, maintenance indicators are a precious help because they allow to quantify what really matters. The operating and maintenance conditions of a wind farm play an essential role in the overall context of the installation. Low availability and lower production jeopardize investment costs and profitability. A wind farm must have an availability higher than 97% and even higher than 99% for state-of-the-art wind turbines and an adequate operating strategy (Gasch and Twele 2011). The maintenance of a wind farm is quite diversified and involves management and control activities with specific software and the maintenance of several subsystems that build the installation, from the access to the park to the analysis of data collected by sensors placed in specific components. For all this operation and maintenance, the owner companies have their own human resources, and subcontract various services and activities, which makes external resources very important to the maintenance performance. The internal human resources are normally delegated specific operation and maintenance functions, such as predictive maintenance tasks. The subcontracted companies are entrusted with the remaining tasks, in which the maintenance of the wind turbines is highlighted, which, as a rule, is delivered to the manufacturer due to the guarantee of the equipment and its specific knowledge.

3 Case Study: Data Collection and Organization The wind farm under study is located on the southern slope of Serra do Caramulo, in the district of Viseu, Portugal. It consists of 16 wind turbines of 2 MW of unit power, a meteorological tower, a control building and a 20/60 kV elevator substation. The location is very favorable for the production of wind energy, although, in terms of altitude, it is below the theoretical limit of 800 m for the production of wind energy in inland lands (mountains) (Fig. 2).

Selecting KPIs in Asset Maintenance of Onshore Wind Farms

683

Fig. 2. Study wind farm

The wind farm is installed at an altitude between 689 m and 792 m, but benefits from a geographical location favorable to sea currents, creating a small orographic “funnel” with a lot of wind potential from the Atlantic coast, as can be seen in the following figure (Fig. 3).

Fig. 3. Wind rose of the wind farm under study

All data from the wind turbines arrives constantly and updated (online) to the existing SCADA where it is possible to operate and monitor, individually or collectively, the different wind turbines (Fig. 4).

684

D. Gaspar et al.

Fig. 4. Wind turbine visualization example (SCADA SGIPE OR WINDNET for Gamesa Wind Turbines, 2017)

The SCADA control system is interconnected with a management and maintenance software (CMMS), owned by the company that exploits the asset, where all the activities of the wind turbines are recorded and characterized according to the internal manual for characterization of stops. The internal CMMS interconnects with other tools and software (e.g. SAP) and presents several management and operation functionalities through the various panels (dashboards) available. All data collected were structured and coded. All activities carried out and developed in the wind farm are recorded in specific software. The activities carried out are accounted for separately concerning the operation and maintenance of the wind farm. In the same way, all wind turbine and substation stops are registered in a specific software for this purpose, according to different criteria and duly typified with the company’s terminology and coding. According to the European standard EN 15341:2019, the research team with the experts from the owner of the wind farm, select 22 indicators for classification, taking into account the impact and importance of the management of the wind farm under study (Table 6).

Selecting KPIs in Asset Maintenance of Onshore Wind Farms

685

Table 6. Relevant matrix for select indicator Codification (EN 15341)

Impact (1 - 5)

Effective (1 a 5)

Result (factor selection)

Operational availability due to maintenance

PHA8

5

4

20

Frequency of maintenance injury (Number/ Man hours worked))

HSE4

5

5

25

Severity of injuries of maintenance

HSE6

5

5

25

Injury frequency of external maintenance (Event/Hour)

HSE8

4

5

20

Safety control rate (N°/man-hours)

HSE17

3

2

6

Outsourcing degree (%)

M9

2

4

8

Frequency of maintenance shutdown Nº/year

M15

3

4

12

Rate of results (%)

M19

4

3

12

MTBF meantime between failures (hours)

E5

4

5

20

MRT mean repair time (%)

E6

4

5

20

Rate of failures N°/Year

E8

3

4

12

Down time due to corrective maintenance

E9

4

5

20

Down time due to preventive maintenance

E12

4

4

16

Maintenance personnel vs site personnel (%)

O&S1

2

2

4

Rate of indirect vs direct personnel (%)

O&S2

2

2

4

Intensity of works by external maintenance

O&S8

2

3

6

Proportion of corrective maintenance (%)

O&S9

4

4

16

Proportion of preventive maintenance (%)

O&S11

3

3

9

MTTR - mean time to restoration (hours)

O&S16

5

5

25

MRT mean repair time (%)

O&S17

4

5

20

Proportion of total delay time (%)

O&S18

3

5

15

Proportion of scheduled available man hours

O&S22

2

2

4

KPI

686

D. Gaspar et al.

4 Results and Conclusions After all the analysis and execution of the selection process, following all the steps of the method presented, the maintenance indicators used in future reports were determined after approval by the corporate hierarchy and presented to the departments involved. The selected indicators are shown in the following table, which gives the indicator code according to the EN 15341:2019 standard and the calculated value for the selection factor, obtained through the weighting matrix (Table 2). Table 7. Selected maintenance indicators Indicator

Code

Factor Sel.

Operational availability due to maintenance (%)

PHA8

20

Frequency of maintenance injury

HSE4

25

Severity of injuries of maintenance (%)

HSE6

25

Injury frequency of external maintenance

HSE8

20

E5

20

MTBF: meantime between failures (hours) MRT (mean repair time) (%)

E6 / O&S17

20

Down time due to corrective maintenance (%)

E9

20

MTTR (mean time to restoration) (Hours)

O&S16

25

As an industrial facility for the production of electricity, a wind farm has several specificities that distinguish it from typical industrial facilities. External phenomena often condition the application of different techniques and methodologies. It cannot be considered due to their unpredictability, which forces the maintenance process to be quite flexible and dynamic, which is one of the first conclusions to be drawn in this work’s development. A method was developed for the identification and selection of maintenance indicators. For this selection, the European standard on maintenance KPIs and the company’s internal procedures was used. Due to a lot of data and indicators, the method developed made it possible to select and highlight the most important indicators for developing maintenance plans and improvements to be implemented in the operation and maintenance of the wind farm. This work, intend to contribute to increase profitability of the asset in question, where theoretical knowledge will be reconciled with the practical implementation and analysis of all its indicators, intending to make known and define the priorities of the main indicators to maintenance management.

References Åhrén, T., Aditya, P.: Maintenance performance indicators (MPIs) for benchmarking the railway infrastructure: a case study. Benchmarking Int. J. 16, 247–258 (2009). https://doi.org/10.1108/ 14635770910948240

Selecting KPIs in Asset Maintenance of Onshore Wind Farms

687

Carnero, M.C., González-Prida, V. (eds.): Optimum Decision Making in Asset Management. IGI Global, Hershey (2017). https://doi.org/10.4018/978-1-5225-0651-5 CEN, European Committee for Standardization: EN 15341: 2019 Maintenance—Maintenance Key Performance Indicators. European Committee for Standardization Brussels, Belgium (2019) Gasch, R., Twele, J.: Wind Power Plants: Fundamentals, Design, Construction and Operation. Springer, Heidelberg (2011). https://doi.org/10.1007/978-3-642-22938-1. https://books.google. pt/books?id=c-QB9PiC_GMC Gonzalez, E., et al.: Key performance indicators for wind farm operation and maintenance. Energy Procedia 137, 559–570 (2017). https://doi.org/10.1016/j.egypro.2017.10.385 IEC: IEC/ISO 31010: 2019: Risk management—Risk assessment techniques. IEC Geneva, Switzerland(2019) ISO: 55002:2018 Asset Management—Management systems-Guidelines on the application of ISO 55001. ISO Geneva, Switzerland (2018) Kutucuoglu, K.Y., Hamali, J., Irani, Z., Sharp, J.M.: A framework for managing maintenance using performance measurement systems. Int. J. Oper. Prod. Manag. 21(1/2), 173–195 (2001). https://doi.org/10.1108/01443570110358521 WindBox: Componentes-dos-aerogeradores (2022). http://windbox.com.br/blog/componentesdos-aerogeradores/ WindEurope: Making wind farms and the power system more interoperable (2022). https://win deurope.org/intelligence-platform/product/making-wind-farms-and-the-power-system-moreinteroperable/

Mapping Maintenance Related Information Using the MIMOSA CRIS Standard: A Case Study Within Gravel Road Maintenance Mirka Kans1,2(B) and Jaime Campos2 1 Chalmers University of Technology, Göteborg, Sweden

[email protected]

2 Linnaeus University, Växjö, Sweden

[email protected]

Abstract. The efficient maintenance of gravel roads depends on digital tools and data as well as information systems adapted for specific needs while conforming with current standards for better interconnectivity. Within gravel road maintenance, cloud-based solutions are a key alternative. This because necessary information, such as road condition data, could be shared throughout the network and provide the opportunity to coordinate maintenance activities between stakeholders. Effective standards are essential since they permit various services with the ability to work together while at the same time supporting differences that ease competition and innovation. Among several maintenance standards, the OSACBM and MIMOSA CRIS database schema are intended for handling information exchange and communication within and between systems supporting conditionbased maintenance. This paper tests the usability of these standards for translating context-dependent information into a standardized data set. The domain studied is gravel road maintenance. Critical data was gathered and elicited from different stakeholders and represented in the form of a conceptual information model. This paper exemplifies how the conceptual model could be mapped into a logical model with respect to the CRIS information model. The main results are in the form of a conceptual and logical information model for purposes of gravel road maintenance, as well as a procedure suggested for mapping the information model following specific steps.

1 Introduction Gravel road maintenance is planned and carried out periodically according to a preventive maintenance strategy based on the maintenance history. Due to climate and socio-economical changes, current maintenance plans may not be a reliable basis in the future (Kans et al. 2022). Instead, real needs and conditions must be given greater consideration using a predictive and data-driven strategy (Mbiyana et al. 2021). Gravel road maintenance is today characterized by a low level of digitization and a high degree of subjective decision-making. Even though digital solutions have been developed for the management of gravel roads, such applications are used mainly by larger infrastructure service providers, while operational and tactical decision-making is a matter for small © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 688–696, 2023. https://doi.org/10.1007/978-3-031-25448-2_64

Mapping Maintenance Related Information

689

companies. These companies would benefit from standardized and user-friendly applications, either cloud-based or standalone. Moreover, information sharing between different stakeholders is limited. This affects the planning and execution of maintenance activities for the individual contractors as well as for different stakeholders within the value chain (Enkell and Svensson 1999). For this purpose, cloud-based solutions are the best option, as necessary information, such as road conditions, could be shared throughout the network. Cloud-based solutions also provide the opportunity to coordinate maintenance activities between stakeholders. The area of gravel road management and maintenance has great potential to become more efficient with the aid of modern digital technology (Enkell and Svensson 1999; Saarenketo 2005; Alferor and McNiel 2017; Radeshi et al. 2018). This is enabled by information systems adapted for specific needs while conforming with current standards for better interconnectivity (Thurston 2001; www.mimosa.org). The standardization in ICTs (Information and Communication Technologies) is a key factor since systems comprises various modules and components by different firms, therefore, well-matched or interoperability concerning modules and components is crucial for the organizations (Shin et al. 2015; Kharlamov et al. 2019). Thus, effective standards are important since they permit diverse services with the ability to work together while at the same time supporting differences that ease competition and innovation. Several models and methods are applicable within the area of asset and maintenance management, such as the Computer Integrated Manufacturing Open System Architecture (CIMOSA), the General Reference Architecture Model (GERAM), the Common Information Model (CIM), the Open System Architecture for Enterprise Application Integration (OSA-EAI), and the International Standard IEC 62264 (Kans and Ingwald 2008). Other important standards when considering Prognostics and health management (PHM) technologies are specific standards for different modules of the system, for instance, for purposes of data requirements and management, measurement techniques, diagnostics, and prognostics (Vogl et al. 2014; Guillén et al. 2016). These standards are developed first-hand for industrial applications and engineering assets but could be utilized for other types of assets as well, such as road infrastructure. As the application is not fully straightforward, a structured step-by-step approach would be needed. The main purpose of this paper is to test the usability of current standards in the context of gravel road maintenance by translating context dependent information into a standardized data set based on the MIMOSA Common Relational Information Schema (CRIS) database. The paper aim is to propose a procedure for translating context dependent information into standardized data sets. The process is illustrated by translating information requirements for gravel road maintenance into a standardized data set based on the MIMOSA CRIS database.

2 Applying MIMOSA for Gravel Road Maintenance The mapping follows four essential steps: 1) selection of relevant MIMOSA sub models, 2) identification and representation of the context-specific information model on the conceptual level, 3) mapping of the context-specific model with respect to MIMOSA, and 4) transferring the conceptual model into a logical model.

690

M. Kans and J. Campos

2.1 The MIMOSA CRIS Model and Selected Sub Models OSA-CBM (Open System Architecture for Condition Based Maintenance) and MIMOSA (Machinery Information Management Open Systems Alliance) have been working on standards for information exchange and communication among different modules for CBM (Thurston 2001; www.mimosa.org). MIMOSA developed a Common Relational Information Schema (CRIS). It is a relational database model for different data types that need to be processed in a CBM application. The system interfaces are defined according to the database schema based on CRIS. The definitions of interfaces developed by MIMOSA are an open data exchange convention to use for data sharing in today’s CBM systems. The MIMOSA CRIS is an extensive model, whereas this paper focuses on two sub models: the Measurement event database classes, and Work order database classes. These sub models, in turn, are quite extensive, and therefore further simplifications for reducing the number of classes to be included in the solution were made in accordance with e.g., Mathew et al. (2006). Figure 1 illustrates measurement event classes as an extract of the MIMOSA CRIS CMCore model, and Fig. 2 is an extract from the MIMOSA CRIS WorkCore basic model showing classes needed for computerized maintenance management. The boxes represent the classes in the respective system. The boxes are connected through lines, which represent their relationship. The numbers/symbols shown between the classes specify the multiplicity (cardinality) in the specific relation: 0 ... 1 or 1 denotes a relationship with up to one, while an asterisk, *, denotes a multiple relationship. 1 1

Database *

* 1 1

DataSourceType

TransducerType

1

*

1

1

CalculaƟonType

*

*

Transducer

*

DataSource

0..1

0..1

0..1 * 0..1

Transducer_axis_direcƟon_type

*

Asset

MeasurementEvent *

*

0..1

*

MeasurementLocaƟon

MeasurementLocaƟonType *

*

*

0..1 0..1

1 *

Fig. 1. Extract from the CMCore: measurement event database classes.

Several classes relate to the class Database through association, i.e., they form parts of the database. In Fig. 1, the measurement event database classes are, for instance, the Asset class connected with the Measurement event and location, e.g., the type of measurement performed and the location of the sensors. Next, all the data and information from the different classes are stored in the database. In Fig. 2, the relationship between WorkOrder and WorkStep is made through composition, which illustrates that WorkStep is a part of the WorkOrder class and cannot exist separately.

Mapping Maintenance Related Information Database

*

*

WorkTaskType

0..1

1

1

0..1

0..1 *

*

*

0..1

*

WorkOrder

1

691

*

WorkRequest

OrderedListRequestForWork

0..1 * *

1

PriorityLevelType

* 1

*

1

Asset

0..1

WorkStep *

Fig. 2. Extract from the WorkCore: work order database classes.

2.2 Establishing the Context-Depended Information Model The conceptual gravel road information model is based on the results of a collaborative project aiming at developing digital tools for gravel road maintenance, see Kans et al. (2022). In total 17 classes were identified: 8 representing maintenance management (Road, road condition, Sieve curve, Work order, Weather condition, Resource, Maintenance plan, Failure report), 2 representing users and organizations (User, Organization), and 7 representing the management of customers (Customer, Invoice, Contract, Request, Cost specification, Quotation, Document). For detailed information of the information model, see Kans and Campos (2022).

RoadCondiƟon

Road 1

*

1

1 *

FailureReport

*

*

WorkOrder

*

Fig. 3. Extract from the conceptual gravel road information model.

In this paper, four of the classes representing maintenance management are in focus for further mapping. These are selected as they correspond with the CMCore and WorkCore sub models in the MIMOSA CRIS. Their interrelationships are illustrated in Fig. 3 while descriptions of the selected classes are found in Table 1.

692

M. Kans and J. Campos Table 1. Class descriptions.

Class

Description

Road

Road identification information including coordinates and technical specifications of the road, such as gravel length and wear course. Owner information and whether the road receives subsidies or not is captured for administrative purposes

Road condition Condition information is captured from sensors and saved in aggregated form in the system. IRI, slope, crossfall, road edge reflects two out of four condition classes used in Sweden. The class could be extended to cover all four condition scores or modified to local conditions Work order

Information regarding maintenance type, priority, and status are examples of direct work order attributes. Information for planning and reporting are captured by the attributes start and end date, planned and real material and time consumption, and condition score before and after maintenance. In addition, the work order class is related to several other classes, such as road and road condition, weather conditions, recourses, and the maintenance plan

Failure report

Failure information consists of type, priority, and failure description. The reporter is in this context the maintainer, i.e., a system user, but could be extended to be a road owner or road user, requiring additional attributes for capturing the identity of the reporting person

2.3 Mapping of Gravel Road Information According to MIMOSA Figure 4 illustrates the mapping of the context-depended model with respect to the CMCore and WorkCore database classes. The four classes to be mapped have been colour coded to enable better reading of the figure.

Fig. 4. Mapping of the information model.

Mapping Maintenance Related Information

693

The Road class is associated with the Asset class in the CMCore and WorkCore databases (marked with green in the figure). RoadCondition is associated with main part of the CMCore classes (marked with orange in the figure). The WorkOrder class is associated with four of the classes in the WorkCore database (marked with blue in the figure): WorkOrder, WorkOrderSteps, WorkTaskType, and PriorityLevelType. FailureReport, finally, is associated with three of the classes in the WorkCore database (marked with red in the figure): WorRequest, OrderedListRequestForWork, and PriorityLevelType. Figure 5 presents the mapped information model. WorkOrder from Fig. 3 is represented by the classes WorkOrder, WorkOrderSteps, WorkTaskType, WorkRequest, OrderedListRequestForWork, and PriorityLevelType. FailureReport is a type of work request and is represented by the classes WorkRequest, and PriorityLevelType. Road is connected with the Asset and Road classes. The Asset class represents both the assets to be maintained and the assets that are used for measurement purposes, and therefore we choose to illustrate this with aggregations (i.e., the unfilled diamond) of the Road and DataSource classes. RoadCondition relates to DataSource, DataType, MeasurementEvent, MeasurementLocation, MeasurementLocationType, and CalculatonType. A simplification was made with respect to the measurement data, as the classes Transducer and TransducerType were not included in the final information model. Instead, the classes DataSource and DataSourceType will represent transducer input. * 0..1

WorkTaskType

*

1

WorkOrder

1

Database 1

DataSourceType 0..1

0..1

*

0..1

1

1

1

*

WorkStep

*

1 *

CalculaƟonType *

1

DataSource

*

WorkRequest

* *

PriorityLevelType

*

*

*

1

0..1 *

*

TransducerAxisDirecƟonType

0..1

0..1

Asset

MeasurementEvent *

OrderedListRequestForWork

*

*

*

MeasurementLocaƟonType * *

*

0..1 1

MeasurementLocaƟon

0..1

1 *

0..1

Road

* 1

Fig. 5. Mapped information model.

From the results of the mapping, it is obvious that road condition information in the conceptual gravel road information model was on a highly aggregated abstraction level, and the mapping will enable a more realistic understanding of the required data to be captured. This will as well facilitate the implementation phase, e.g., when developing the database, as the conceptual model quite easily could be translated into a logical model. 2.4 Transferring the Conceptual Model into a Logical Model In the fourth step, attributes were added to the information model. Attributes are characteristics of the class that makes it distinguishable and unique from other classes. In a

694

M. Kans and J. Campos

data model, primary and foreign keys are specific types of attributes that create a unique identifier for the class and connect classes with each other. While adding foreign keys, the cardinality of the model is also checked. WorkOrder

WorkTaskType 0..1

wtTypeID: long userID: string 0..1

*

*

WorkStep *

workStepID: long user ID: string workOrderID: long wtTypeID: long plTypeID: long

1

*

workOrderID: long userID: string date: dateƟme startDate: dateƟme endDate: dateƟme status: string remark: string wtTypeID: long

1

Database 1

DataSourceType 1

1

*

*

*

PriorityLevelType

*

1

plTypeID: long userID: string

OrderedListRequestForWork

*

workReqID: long userID: string date: dateƟme failureTypeID: long descripƟon: string

*

*

DataSource dsID: long userID: string dsTypeID: long

*

*

*

TransducerAxisDirecƟonType trAxisDirTypeID: long userID: string measLocID: long

* 0..1

*

mlocTypeID: long userID: string measLocID: long

1

Asset *

utcEvent: string Ɵmestamp: dateƟme dataQualityType: short remarks: string dsID: long

0..1

MeasurementLocaƟon measLocID: long userID: string

0..1

MeasurementEvent

*

0..1

MeasurementLocaƟonType *

*

calcTypeID: long userID: string utcEventID: long

*

*

workOrderID: long workReqID: long

1

CalculaƟonType 0..1

0..1

WorkRequest 1

dsTypeID: long userID: string

1

0..1 1

*

0..1

* 1

Road roadID: long userID: string naƟonalRoadID: long length: long gravelLength: long wearCourse: string coordinates: string

Fig. 6. Logical information model.

The first attribute listed in the logical information model presented in Fig. 6 is used as the primary key, while the foreign keys are duplicated primary keys of other classes added at the end of the attribute list. The class OrderedListRequestForWork uses the foreign keys from WorkOrder and Workrequest as combined primary keys. The information model in Fig. 6 is not the final one. Additional modifications will be necessary, e.g., the attribute position for the Road class might be handled with classes for location tracking, such as GeoPosition. This could be associated with the MeasurementEvent class as well. In addition, as userID is frequently used as an attribute, this should be associated with a User class.

3 Conclusions Digital transformation is often viewed as a technology innovation project, which in the worst case, results in high-tech equipment and systems that do not meet the digital maturity or business needs of the users (Campos et al. 2021). This highlights the importance of considering the already existing standards and best practices when developing ICT systems, including databases, as is the case of the current work. In this paper, the context-dependent information of gravel road maintenance was successfully mapped with respect to the CMCore Measurement event and WorkCore database classes following a four-step procedure. The mapping showed that the suggested procedure is suitable for this kind of work. The main results achieved in the paper are confirming the applicability and usability of MIMOSA CRIS for the modelling of infrastructure assets and proposing a procedure for the mapping process. The novel step-by-step procedure is

Mapping Maintenance Related Information

695

perceived as a useful tool for similar mapping activities, when developing industrial as well as infrastructure asset management systems. During the mapping process it was noted that points of iteration might appear between the steps, for example between steps 1 and 2, or between steps 3 and 4, for best result, see Fig. 7. The authors also recognize that the process itself might be of iterative nature, as it could be wise to develop the information model in several sub steps for easy handling. In this paper, for example, the mapping focused on two of the sub models of MIMOSA CRIS, while other sub models, not explicitly found in the original context-depended information model, should be included as well.

Fig. 7. The proposed mapping process.

Further work on the database model includes extensions with other relevant sub models of the MIMOSA CRIS. The GeoLocation, for instance, is a crucial part of the solution to be able to identify the geographical coordinates of the gravel road in question, which are not highlighted in the current mapping. Therefore, in later stages of the work the GeoLocation will be added into the solution. Acknowledgements. The research has been conducted partly within the project Sustainable maintenance of gravel roads funded by The Kamprad Family Foundation. The project develops new methods and technologies for gravel road maintenance.

References Alferor, R.M., McNiel, S.: Method for determining optimal blading frequency of unpaved roads. Transp. Res. Rec. 1252, 21–32 (2017) Campos, J., Kans, M., Salonen, A.: A project management methodology to achieve successful digitalization in maintenance organizations. Int. J. COMADEM 24, 3–9 (2021) Enkell, K., Svensson, J.: Grusvägsstyrsystem Förstudie. VTI notat 44-2000, VTI, Linköping (1999) Guillén, A.J., González-Prida, V., Gómez, J.F., Crespo, A.: Standards as reference to build a PHMbased solution. In: Koskinen, K.T., et al. (eds.) WCEAM 2015. LNME, pp. 207–214. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-27064-7_20 Kans, M., Ingwald, A.: Common database for cost-effective improvement of maintenance performance. Int. J. Prod. Econ. 113(2), 734–747 (2008) Kans, M., Campos, J.: A novel database model for gravel road maintenance. In: 2021 International Conference on Maintenance and Intelligent Asset Management (ICMIAM), pp. 1–5 (2022) Kans, M., Campos, J., Håkansson, L.: The development of a cloud-based information system for gravel road maintenance. Int. J. COMADEM 25(2), 31–38 (2022)

696

M. Kans and J. Campos

Kharlamov, E., Mehdi, G., Savkovi´c, O., Xiao, G., Kalaycı, E.G., Roshchin, M.: Semanticallyenhanced rule-based diagnostics for industrial Internet of Things: the SDRL language and case study for Siemens trains and turbines. J. Web Semant. 56, 11–29 (2019) Mbiyana, K., Kans, M., Campos, J.: A data-driven approach for gravel road maintenance. In: 2021 International Conference on Maintenance and Intelligent Asset Management (ICMIAM), pp. 1–6 (2021) Mathew, A., Zhang, L., Zhang, S., Ma, L.: A review of the MIMOSA OSA-EAI database for condition monitoring systems. In: Mathew, J., Kennedy, J., Ma, L., Tan, A., Anderson, D. (eds.) Engineering Asset Management. Springer, London (2006). https://doi.org/10.1007/9781-84628-814-2_88 Radeshi, R., Maher, M., Barakzai, K.: Defining needs for optimized management of gravel road networks. In: Transportation Association of Canada Conference - Innovation and Technology: Evolving Transportation, TAC 2018 (2018) Saarenketo, T.: Monitoring, communication and information systems & tools for focusing actions. Roadex II Report (2005) Shin, D.H., Kim, H., Hwang, J.: Standardization revisited: a critical literature review on standards and innovation. Comput. Stand. Interfaces 38, 152–157 (2015) Thurston, M.G.: An open standard for web-based condition-based maintenance systems. In: 2001 IEEE Autotestcon Proceedings. IEEE Systems Readiness Technology Conference, pp. 401– 415. IEEE (2001) Vogl, G.W., Weiss, B.A., Donmez, M.A.: Standards for prognostics and health management (PHM) techniques within manufacturing operations. In: Annual Conference of the Prognostics and Health Management Society 2014 (2014)

Perceived Relevance of Asset Management Topics in Industry and Academia Nuno Almeida1(B) , Joe Amadi-Echendu2 , Daniel Gaspar3 , Edmea Adell4 , Joana Torcato5 , João Vieira6 , and Eduardo Leite7 1 CERIS, IST, CT 204, Lisbon, Portugal [email protected] 2 University of Pretoria, Pretoria, South Africa [email protected] 3 IPV, IST, CT 204, Lisbon, Portugal [email protected] 4 Assetsman, IFRAMI, Paris, France [email protected] 5 IST, Lisbon, Portugal 6 CERIS, IST, IP, Lisbon, Portugal [email protected] 7 OSEAN, Uma, Funchal, Portugal [email protected]

Abstract. A better understanding of the relative importance of the various topics of the asset management discipline can help guide the efforts of the asset management community in various fronts: education, research, and industrial practice. This article presents a study on the perceived relevance of the various topics and themes of the asset management body of knowledge. The article provides reflections based on (i) the results of surveys conducted in Portugal covering the communities of participants in the national congress of engineering asset management (CongrEGA) and the members of the national “mirror” committee to ISO/TC 251 Asset Management and; (ii) the results of snap surveys conducted with participants in the WCEAM, plus those of a pilot-test study of a scheme proposed by ISEAM for the Recognition of higher education academic programs in Engineering Asset Management (REAM). The results are discussed in the light of the planned revision to the second version of Asset Management Landscape published by the GFMAM.

1 Introduction The asset management approach is used in organizations of different types and can take several interpretations in different geographies or sectors. The expression was initially used in the financial sector and related to items which provided value to their owners (such as investment portfolios), not without having its own roots and preponderant developments in different areas of engineering. The first scientific studies on the management of engineered physical assets emerged in the 1970s, as an extension and broadening of © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 697–707, 2023. https://doi.org/10.1007/978-3-031-25448-2_65

698

N. Almeida et al.

the scope of engineering studies, started in the previous decade, focused on the maintenance of industrial facilities (Wijnia 2016). The experiences of applying this new approach in the industry gained momentum in the following decade and, in the 1990s, several associative movements, regular conferences, commercial initiatives and technical guides helped consolidating the discipline of physical asset management (Almeida et al. 2021). The two versions of PAS 55 (2004 and 2008) were important milestones in the systematization of the approach to physical asset management. Terotechnology combined economic and engineering management and was the forerunner of the business thinking implicit in PAS 55. The 2008 release of PAS 55 brought out the importance of risk management in asset management. In 2014, the ISO 55000 series of standards helped diffuse asset management worldwide in the broadest sense of the term, i.e. covering both physical and non-physical assets. It brought this important business function into the realm of ISO management system standards, linking it with other well-known management system standards dealing with quality, safety, environment etc. After the publication of the ISO 55000 series of standards in 2014, several major publications in the field aligned with the principles and terminology established in those standards (AMC 2014; GFMAM 2014; NAMS 2015; IAM 2015). The Global Forum on Maintenance & Asset Management (GFMAM) published the second version of Asset Management Landscape in 2014 (GFMAM 2014). This is a foundation document for many organizations and businesses, including those external to GFMAM. It provides guidance on the breadth of asset management covering topics in six major areas, namely: (1) strategy & planning, (2) decision making, (3) lifecycle delivery, (4) knowledge enablers, (5) organization & people, and (6) risk & review. This paper discusses the relative importance of the six major areas and the 39 topics of asset management described in the GFMAM Asset Management Landscape. The discussion is based on the perceptions of a representative sample of the asset management community in Portugal as well as the international asset management community of participants in the World Congress on Engineering Asset Management (WCEAM), plus a pilot-test study conducted with higher education institutions from South Africa, Canada and Australia, utilising a scheme proposed by the International Society of Engineering Asset Management (ISEAM) for the Recognition of academic programs in Engineering Asset Management (REAM).

2 Surveys This study is implemented through surveys that were carried out involving asset management professionals, academics, and researchers. Questionnaires were distributed to all registered members of the Portuguese “mirror” committee to ISO/TC 251 Asset Management (CT 204), in 2020 and 2022, and to all registered authors of submissions to the 1st Portuguese Congress on Engineering Asset Management in 2022 (CongrEGA). The questionnaire comprises two parts: A and B. The first (A) part contains questions about the general aspects of the organization and the functions, roles, and experiences of the participant. The second part (B) begins with questions about the asset management terminology and definitions as established in ISO

Perceived Relevance of Asset Management Topics

699

55000 and then focuses on the relevance of AM topics. Table 1 shows the main characteristics of the respondents to this questionnaire. Column “PT (2020)” characterizes the distribution of the Portuguese economic activity as a whole, suggesting that the Portuguese “mirror” committee to ISO/TC 251 Asset Management (CT 204) is considerably well represented by the different sectors of the national economy. Table 1. Characterization of participants in the questionnaire. Characteristics of the organization of participants

CongrEGA (2022)

CT 204 (2022)

CT 204 (2020)

PT (2020)

Sample size

98

60

42

–

Returned responses

18%

38%

60%

–

Sector

Public

61%

65%

44%

–

Private

39%

35%

56%

–

Services

50%

35%

33%

39%

Non-profit

33%

35%

24%

31%

Economic activity

Mission or main activity

Asset portfolio

Industry

11%

22%

29%

19%

Construction

6%

8%

10%

6%

Agriculture, fishing, etc.

–

–

4%

5%

Asset management

33%

44%

32%

–

Supplier or service provider

28%

22%

40%

–

Influencer

17%

26%

28%

–

Other

22%

8%

–

–

Physical

68%

76%

73%

–

Non-physical

32%

24%

27%

–

The study reported here follows from previous snap survey results presented during 11th and 12th WCEAM in Jiuzhaigou (China) and Brisbane (Australia), respectively. This snap survey aimed at encouraging self-evaluation and peer review, and thus, the Recognition of academic programs in Engineering Asset Management. Recognition, rather statutory accreditation, has two fundamental aims: i) assure that academic programs so recognised address the Engineering Asset Management (EAM) body of knowledge, and ii) encourage institutions so recognized to influence further advance the EAM body of knowledge. There is a distinction between a professional society such as ISEAM’s recognition plus endorsement of an EAM program at a higher education institution and the accreditation of such a program based on, e.g., the Washington Accord Accreditation Protocol. Recognition implies that the curriculum content (i.e., body of knowledge),

700

N. Almeida et al.

methods of delivery and outcomes of a particular program at a higher educational institution are endorsed by a learned professional society like ISEAM. Such an endorsement provides credibility towards accreditation of a particular program if the provider wishes to seek accreditation according to, e.g., the Washington Accord.

3 Results and Discussion Studies reported on the perceived relevance of asset management topics are grounded on the agreement of respondents with the definition of asset as established in ISO 55000 (2.3): “item, thing or entity that has potential or actual value to an organization.” The next sections summarize the results and offer a discussion about the perceptions of a representative sample of academics, researchers, and practitioners of the asset management community in Portugal, as well that of the international asset management community. 3.1 Perceptions of the Portuguese Asset Management Community Figure 1 shows that the majority in all three groups of respondents in the three surveys conducted in Portugal (CongrEGA in 2022 and CT 204 in 2022 and 2020) strongly agree with the definition of asset as established in ISO 55000 and that there is no significant disagreement with this definition.

Fig. 1. Level of agreement with the definition of asset in ISO 55000

The participants in the questionnaire were asked to rate the relevance of each of the 39 topics of asset management described in the GFMAM Asset Management Landscape (see Fig. 2). The rating scale ranges from 1 to 5, where 1 is “not relevant”, 2 is “slightly relevant”, 3 is “moderately relevant”, 4 is “relevant” and 5 is “very relevant”. Figure 3 presents the results in their order of perceived relevance. The top 10 topics belong mostly to the group areas of Strategy & Planning (4 occurrences), Risk & Review (3 occurrences), Decision making (2 occurrences) and Organization & People (1 occurrence). There is no topic from the group areas of Lifecycle Delivery and Asset Information in this short list. On the other hand, the last 10 topics in the rank belong mostly to the Lifecycle Delivery group area (7 occurrences). There is no topic perceived with a relatively low importance in the group areas of Asset Information and Risk & Review. Figure 4 shows the results obtained for each of the three groups of participants in the questionnaire.

Perceived Relevance of Asset Management Topics

701

Fig. 2. List of Asset Management Landscape topics (GFMAM 2014).

It is worth mentioning that, as can be seen in Fig. 4, the topic Operations & Maintenance Decision-Making presents significantly different results depending on the sample of participants. CongrEGA (22) and CT 204 (22) participants rank this topic in the 7th and 22nd (out of 39) positions, respectively.

702

N. Almeida et al.

The fact that Demand Analysis got low should be further explored. If assets are there to create value, then this relies on them meeting a level of demand. The relatively low score of Configuration Management is somehow expected since many organisations are immature in this area to their abiding loss as assets fail to perform their intended function. Figure 5 illustrates the overall perception of importance for each group of topics and the gap between this perception and the asset management functions performed by the participants providing the feedback (asset managers, finance, project engineers, maintenance personnel, etc.). It is worth noting that respondents with roles in the areas of Lifecycle Delivery and Asset Information somehow undervalued the relative importance of these same areas. On the contrary, areas such as Organisation & People and Risk & Review are, on average, perceived as having higher relevance even by respondents that do not perform functions in those areas. 3.2 Perceptions of the International Engineering Asset Management Community It is also worth contrasting the results presented above with the ongoing efforts of higher education institutions to establish asset management academic programs. Following its commitment to continuously improve education and research in Engineering Asset Management, the International Society of Engineering Asset Management (ISEAM) started formally encouraging the recognition of academic programs focusing on the discipline in 2016. The ISEAM proposes the use of well-known guidelines as a background for developing its Recognizing scheme for academic programs in Engineering Asset Management (REAM), namely the GFMAM Landscape (GFMAM 2014) or the IAM Asset Management Anatomy (IAM 2015), the AMBoK (AMC 2014) and relevant publications included in the books of proceedings of the World Congress on Engineering Asset Management (1st edition: WCEAM 2006). The ISEAM conducted two snap surveys involving 20 and 22 participants registered in the WCEAM 2016 and 2017, respectively. The intent of these surveys was to capture the perceived relevance of topics and themes of Engineering Asset Management as these are covered in academic programs in a small sample of international universities. Results showed that, on average, the most covered topics at that time were: – Strategy & planning: policy; strategic planning; strategy and objectives; asset management planning. – Decision making: capital investment decision making, operations & maintenance decision making; life cycle value. – Lifecycle activities: asset creation and acquisition; maintenance delivery; fault and incidence response; reliability engineering; asset operations; asset decommissioning and disposal. – Knowledge enablers: data and information management; asset information and strategy. – Organization & people: asset management leadership; organizational culture.

Perceived Relevance of Asset Management Topics

703

Fig. 3. Overall relevance of the 39 topics of asset management in the GFMAM Asset Management Landscape.

704

N. Almeida et al.

Fig. 4. Relevance of the 39 topics of asset management in the GFMAM Asset Management for each group of participants in the questionnaire

– Risk & review: risk assessment and management; asset performance and health monitoring; asset management systems monitoring; asset costing and valuation; stakeholder engagement. These results prompted a trial recognition scheme to be pilot-tested in a few higher education institutions. In this pilot-test, the REAM covered and scored content, learning outcomes, delivery methods, publications record, industry contact and qualification & awards. See Fig. 6 for some of these results.

Perceived Relevance of Asset Management Topics

705

Fig. 5. Relationship between the relevance of the areas and the scope and functions performed by the respondents

It is also worth mentioning that, in 2019, the international standardization committee ISO TC 251 Asset Management established an internal Product Improvement Proposal (PIP) aiming at providing guidelines to asset management teaching. The point of departure of this PIP initiative by the international standardization committee ISO TC 251 Asset Management was the recognition that asset management is taught at many levels and in many different forms, covering different topics, with varying levels of depth and with different support materials. Although the PIP was not fully implemented and incorporated in the outputs of ISO/TC 251, the discussions were useful to develop an overview of the state of the art of instruction in asset management, foster a better comprehension of the teaching materials that are currently being used and what new or improved materials are needed, and to identify conceptual difficulties in professors and students understanding of asset management. The outcomes of the discussion also suggested that the underlying intellectual capital in this area, and the international academic and professional development programs, will probably increase considerably in the next few years.

706

N. Almeida et al. Scale:

0 - No Value ; 1-3 - Weak ; 4-6 - Adequate ; 7-9 - Good ; 10 - Very Good

Topics: 0

1

2

3

4

5

6

7

8

Asset Management Policy (#1) Asset Management Strategy & Objectives (#2) Demand Analysis (#3) Strategic Planning (#4) Asset Management Planning (#5) Capital Investment Decision Making (#6) Operations & Maintenance Decision Making (#7) Life Cycle Value Realisation (#8) Resourcing Strategy (#9) Shutdown & Outage Strategy (#10) Technical Standards & Legislation (#11) Asset Creation & Acquisition (#12) Systems Engineering (#13) Configuration Management (#14) Maintenance Delivery (#15) Reliability Engineering (#16) Asset Operations (#17) Resource Management (#18) Shutdown & Outage Management (#19) Fault & Incidence Response (#20) Asset Decommissioning & Disposal (#21) Asset Information Strategy (#22) Asset Information Standards (#23) Asset Information Systems (#24) Data & Information Management (#25) Procurement & Supply Chain Management (#26) Asset Management Leadership (#27) Organisational Structure (#28) Organisational Culture (#29) Competence Management (#30) Risk Assessment & Management (#31) Contingency Planning and Resilience Analysis (#32) Sustainable Development (#33) Management of Change (#34) Assets Performance & Health Monitoring (#35) Asset Management System Monitoring (#36) Management Review, Audit & Assurance (#37) Asset Costing & Valuation (#38) Stakeholder Engagement (#39)

Fig. 6. Sample of results of the REAM pilot-test.

9

10

Perceived Relevance of Asset Management Topics

707

4 Conclusions As a result of parallel developments in different areas of application, the perceived importance of different asset management topics can vary amongst academics, researchers, and practitioners of the asset management communities. The extent of these variations is analysed in this paper based on results of questionnaires distributed to a representative sample of asset management practitioners and academics and researchers in Portugal in 2020 and 2022, and also during snap surveys and a pilot-test study sponsored by the ISEAM. The results thus obtained are to be further explored with a continuous analysis involving experienced asset management professionals, namely to identify systemic gaps in how people think about asset management and identify risks to the delivery in good asset management if some of the topics are not universally valued. These insights can be considered as inputs for the ongoing revision that GFMAM Landscape. The purpose of the 3rd edition of the GFMAM Landscape is to describe a common understanding of asset management, particularly highlighting its breadth. It aims at providing a framework against which knowledge bases and practices can be aligned, compared, and contrasted. The Landscape complements the ISO55000 suite of standards and related material, which is primarily intended to address management systems. Future studies will also address how subjects are uniformly considered to be the most valuable or less valuable in different environments (e.g. academia vs industry) and contexts (e.g. country level, regions, etc.) and explore ways of bridging discrepancies between the different sets of results.

References de Almeida, N.M., Vieira, J., Silva, J.G., e Castro, C.: The impact of asset management development programs in infrastructure organizations. In: Rodrigues, H., Gaspar, F., Fernandes, P., Mateus, A. (eds.) Sustainability and Automation in Smart Constructions. ASTI, pp. 247–258. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-35533-3_29 AMC (ed.): Framework for Asset Management, 2nd edn. A. M. Council (2014) GFMAM: The Asset Management Landscape, 2nd edn. Global Forum of Maintenance and Asset Management (2014) NAMS: International Infrastructure Management Manual (IIMM): Institute of Public Works Engineering Australia; Association of Local Government Engineers of New Zealand, 5th edn. National Asset Management Steering Group (2015) The IAM: Asset Management – An Anatomy. The Institute of Asset Management (2015). https:// theiam.org/media/1486/iam_anatomy_ver3_web-3.pdf Wijnia, Y.: Processing risk in asset management: exploring the boundaries of risk based optimization under uncertainty for an energy infrastructure asset manager. Ph.D. thesis, TU Delft Repositories (2016)

The Concession Contract as an Instrument to Safeguard the Long-Term Condition of Logistics Infrastructure Assets Monica Lopez-Campos1(B) , Raúl Stegmaier1 , and Eduardo Candón2 1 Universidad Técnica Federico Santa María, Valparaíso, Chile

{monica.lopezc,raul.stegmaier}@usm.cl

2 Department of Industrial Management, University of Seville, Seville, Spain

[email protected]

Abstract. Given the great importance of logistics infrastructure for the development of a country, it is in the interest of both the public and private sectors to have a framework to regulate and facilitate the maintenance of this infrastructure throughout its life cycle. The objective is to harmonize the long-term interests of the State as the mandating entity with the shorter-term interests of the concessionaires, ensuring that the conditions of the assets are safeguarded to the benefit of the level of service for the end-users. Based on the Chilean case, a methodology is proposed that aims to create a reference framework. Although it is a project under development, there are already concrete results on how maintenance concession contracts should be managed, as the key instrument that allows the proper exploitation of the assets by the concessionaires and that in turn helps the State not to evade its responsibility in the proper use of the logistics assets throughout its life cycle.

1 Introduction A country’s logistics infrastructure is a key element in guaranteeing the well-being of its population, and the development of its commercial, scientific, technological, and social, exchange. Those countries that have reliable airports, roads, ports, and railway systems, for example, are much more attractive for international trade and investment. Thus, the interest in the proper construction, operation, and maintenance of these strategic logistics assets comes not only from the private sector but also from the public sector (Arvis et al. 2018). The State, as the entity that must provide conditions for the quality of life and development of its citizens, is particularly interested in the proper management of national logistics assets. A common and successful strategy, followed by several developing countries, has been the concession or public-private cooperation for the construction, operation, and maintenance of this infrastructure (Bitran et al. 2013). This model has provided some nations with modern and functional infrastructure while reducing the associated risks and costs (Guasch et al. 2003). Among the countries that have adopted this model, the case of Chile is an outstanding example. Thanks to this modality, Chile has had the opportunity to materialize transcendental works and to project a strategy of © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 708–715, 2023. https://doi.org/10.1007/978-3-031-25448-2_66

The Concession Contract as an Instrument

709

territorial integration and international connection, transforming its urban centers into friendlier cities, boosting its productive development, and with a better quality of life for each of its inhabitants (Dirección General de Concesiones n.d.). Since the 1990s, there have been more than 80 concession contracts in different phases throughout the Chilean territory, where the construction of more than 3,000 km of highways stands out. See Fig. 1 (Torres et al. 2016). The renovation of the country’s road infrastructure, through urban highways and interurban routes, is the best example of the achievements of the concessions industry. Likewise, there are 11 airports in Chile whose infrastructure has been modernized thanks to the concession system (Dirección General de Concesiones n.d.). The design and construction, which are carried out at the beginning of each infrastructure development project, together with the maintenance management, which must be carried out continuously throughout the life cycle of the asset, are key elements that make the difference between an extended, safe and efficient operation over time and the deterioration and risky conditions of this type of infrastructure, which is generally used by very massive amounts of users per year and whose cost for lost profits is generally considerable. In this way, we have that the good performance of the maintenance of this infrastructure is a matter of concern for the State, which places the operation and maintenance of the assets in the hands of concessionary companies but cannot evade the responsibility of ensuring that the conditions in which the national assets are operated are the best to safeguard the entire useful life in adequate conditions with respect to the image, functionality, safety, productivity, comfort, and conservation.

2 Problem Statement The construction, operation, and maintenance of countries’ key logistics infrastructure are frequently carried out by concession contracts or operating subsidiaries. However, the State or public entity, should not abdicate its responsibility for the use and maintenance of the infrastructure during its entire life cycle. The challenge is therefore to be able to harmonize the long-term objectives of the holding (State) with the more shortterm objectives of the concessionaires, achieving a win-win agreement where the main beneficiary should be the end-user. It is here where the basic tool to solve these conditions arises: the concession contract. Although in general, governments that operate with concession contracts have defined legal and regulatory schemes in this regard, operated by Public Works Concession Laws, there is still interest in the public sector to know and formalize the “best practices” in the contracts that ensure that the final objectives of preservation and fulfillment of a basic level of service to the end-users, as well as the satisfaction of all stakeholders, are achieved. There is even another level of interest in this issue, on the part of the concessionaires themselves, who in turn contract third-party companies to perform more specific tasks. These concessionaires are also interested in knowing how contracts can be an element that guides the management, supervision, and control of the tasks, being an element of negotiation and agreement between the parties involved.

710

M. Lopez-Campos et al.

In summary, what is sought as a topic of interest for all parties involved in logistics asset management is to know the best practices regarding contract management, if possible structured in a framework that offers a diversity of practical tools and allows to ensure their application in a variety of contexts.

Fig. 1. Example of the PPP Chilean highways (Torres et al. 2016)

3 Methodology Thus, this paper presents a project that seeks to create a framework, which should contain a guide for the process and tools for the negotiation, design, evaluation, and renewal of maintenance contracts with subsidiary companies, from a technical and management perspective, ensuring the conservation of the infrastructure asset throughout its life cycle, and making the interests of the agency or holding compatible with those of the contractor. To achieve the development of the mentioned framework, the following project stages are proposed. The main field of application and analysis of this research corresponds to the Chilean environment, without leaving out that the final scope of the framework is to a variety of cases in different logistics industries around the world. The stages are: 1. 2. 3. 4.

Study of the features of concession models. Analysis of international best practices. Analysis of the Chilean economic and legal context related to concessions. Case study in Chilean companies

The Concession Contract as an Instrument

711

5. Design of the framework 6. Validation of the framework and inference of general conclusions These stages have been worked on during the last two years, having achieved progress, and expecting more results and conclusions to come. Not all the findings, but the main ones until now are reported in the following section.

4 Preliminary Results 4.1 Features of Concession Models and Analysis of International Best Practices In order to understand the implications of concession contracts, it is necessary to start from the general context, analyzing their most relevant terms and vocabulary, together with the main developments in the world on this topic. This understanding and analysis have been based on specialized literature, a systematic literature review (Alvesson and Sandberg 2011), and particularly from documents generated by government entities such as the Ministry of Public Works and relevant institutions such as the World Bank. From the systematic literature review, it was found that the first scientific articles analyzing how concession contracts for infrastructure development should be formulated date back to the 1990s. Their authors make general recommendations on contract design, award process procedures, and optimal contract duration (Crampes and Estache 1998; Kerf et al. 1998). Focusing more precisely on the area of asset management and infrastructure maintenance, the early work by Al-Hammad (1995) reviews the main problems between the client and the infrastructure maintenance performers, mainly the lack of direct supervision by the contractor, the lack of precision in specifications and standards, and low budget allocations by the owners. Since these early studies, and with the aim of contributing to mitigating the aforementioned challenges, several researchers have published proposals mainly over the last twenty years. In general, we noticed that a good categorization for the study of concession contracts refers to the stage of their life cycle, with studies focusing specifically on their design, measurement, and control or on their renewal. After an exhaustive literature review presented in the thesis of Castro (2020) and in the conference paper of López-Campos et al. (2021), we can confirm the evidence that PPPs (Public-Private Partnership) offer great advantages to contractors to provide society with high-quality service infrastructure. And additionally, some points of interest to carry out a concession contract in good shape were identified: 1) Taking care of financial viability, 2) Procuring collaborative work, 3) Shielding against political instability in the region or even at the worldwide level, and 4) Being careful in the estimation of the contract concession period. (López-Campos et al. 2021). Abounding on the mentioned interest points, we can say that taking care of the financial viability, is an issue widely treated by various authors. As an example, Ghayal and Salgude (2019) make an interesting study taking as a reference the management of the road network in India. They conclude that to avoid excessive time and costs in the projects, the PPP contract must include options to reduce the financial burden of the contractor. In their proposal, 60% of the financing is provided by the authority during the construction phase and the remaining 40% is granted annually. The possibility of

712

M. Lopez-Campos et al.

loans and ways to conduct a financial analysis is studied by several authors (Liu and Lee 2016; Lara-Galera et al. 2016). Regarding the importance of collaborative work, we found that Sabri et al. (2019) examine the causes of the conflicts present in the construction of Norwegian roads and tunnels. They found that the main cause of conflict between the parties corresponds to the specification of the bidding and the understanding of the contract. So, the collaborative work for the understanding of the client and the anxiety of the contractors is the basis for reducing and preventing this type of conflict. Notwithstanding the foregoing, conflicts during PPP projects can also come from abroad. Liu and Lee (2016) recount the Taiwanese experience of what happens when the environmental conditions change, for example when the estimate in the use of the concessioned resource changes suddenly due to social or economic factors, providing guidelines based on their experience in the face of financial loss by the State. Case that since the covid-19 pandemic has been of emerging interest. Regarding the estimation of the concession period, Yang et al. (2016) study the duration of the concessions using game theory. They argue that the duration of the concession contract is relevant since it directly affects PPP interests but also influences the cost and quality of construction. For example, they mention that if the concession period is short, companies tend to choose the cheapest construction costs to the detriment of quality, which in the future increases maintenance costs. Finally, as Yan et al. (2019) mention, a fair distribution of benefits and risks is not only one of the key factors in deciding the concession period but also an important prerequisite for good cooperation between the government and the private sector in PPP projects. On the other hand, in addition to the literature analysis, during this stage, it was possible to identify some internationally-recognized success cases implemented and still in force, such as the Queensland Government Contract and Procurement Management Framework (Queensland Government 2022), the applications carried out by Prof. Sara Cullen of the University of Melbourne (Cullen 2015), and international standards of interest, such as ISO 10845: 2010, which sets the tone for the development of a procurement system in the construction sector. 4.2 Analysis of the Chilean Context An essential part of the knowledge base required herein is the analysis of the economic and legal context in which the management of concession contracts is inserted. The main sources of information consist of data from the national government and indicators and studies carried out by renowned institutions. A valuable source of information is the World Bank Group site, the Public-Private Partnership in Infrastructure Resource Center for Contracts, Laws and Regulations (World Bank 2020). In the case of Chile, there is a General Directorate of Concessions under the Ministry of Public Works, which has a very clear mission and functions as a provider and manager of public infrastructure works and services through public-private partnerships. On its website, complete information can be found with the aim of presenting the current state of the different concessions, the functioning of the Concessions Council, as well as making reports and publications transparent (Dirección General de Concesiones n.d). The Ministry of Public Works website has the rules and regulations for foreign and national investment in infrastructure, written in Chinese, English, and Spanish (Ministerio de

The Concession Contract as an Instrument

713

Obras Públicas n.d.), highlighting Chile as a country of investment opportunities in infrastructure. Chile’s interest in this subject stems from the good experience it has had, the public-private partnership system has been a State-driven policy that has allowed the development of infrastructure that otherwise would not have been possible, contributing to economic and social development, improving service standards, productivity and quality of life in the country (Torres et al. 2016). According to the Chilean Ministry, the achievements have been possible because the system has been built on the basis of a stable legal framework, which has provided the industry with indispensable standards of transparency and credibility, as well as a modern financial system that has provided investors with tools for project execution (Torres 2016). Despite these achievements, there is still huge interest on the part of the various actors to ensure that they operate according to best practices and to have a reference model. To complete a diagnosis and find out the interests of the related entities, based on the findings made in the previous step of the methodology, a survey consisting of 50 questions has been designed with the aim of finding out the level of adoption that the companies currently have with respect to the best practices found. This survey has been applied to date to three companies in the Chilean port sector and is in the process of being applied to a few more in the railway sector. The idea is to generate case studies that can characterize the functioning of contracts in Chile and evaluate the gaps, to generate a framework to assist in the development of the best contracts management practices.

5 Expected Results and Conclusion A framework of reference refers to a structured methodology, with well-defined objectives, in this case, the achievement of win-win agreements for the operation and maintenance of infrastructure concessions. The framework should be composed of sequential stages that will correspond to the natural life cycle of the contracts, including a prior analysis of the context of the case to provide robustness to the entire process and adapt it to the country’s environment. Each stage of the model will support the achievement of a necessary precedent for fulfilling the general objectives, and each stage will have one or more associated methodological tools. These tools may be qualitative or quantitative and will be detailed in the same framework. Therefore, the desired structure is dynamic, sequential, and focused on continuous improvement. The case studies in the previous stage had the objective of characterizing the general management of some concession and maintenance contracts for Chilean logistics infrastructure. Based on these cases, a diagnosis is made, specifying the needs and interests of the parties involved and the context of the country. From this diagnosis begins the methodology for the creation of the framework of reference. Finally, the validation mechanism for the proposed framework of reference will be its use in the design and implementation of an operation and maintenance concession contract, initially for the case of a Chilean entity. A comparison of the results obtained with respect to the expected results will be made. In conclusion, it is expected that the resulting framework will be of interest to various stakeholders: to the State as a mandating entity, to the State itself acting also as an operating company, to the concessionaires, and to third party private sector companies subject to maintenance contracts.

714

M. Lopez-Campos et al.

Acknowledgment. This research has been funded by the Fondecyt project number 11180964 of the Chilean National Council for Science and Technology.

References Al-Hammad, A.: Interface problems between owners and maintenance contractors in Saudi Arabia. J. Perform. Const. Facil. 9(3), 194–205 (1995) Alvesson, M., Sandberg, J.: Generating research questions through problematization. Acad. Manag. Rev. 36(2), 247–271 (2011). https://doi.org/10.5465/amr.2009.0188 Arvis, J.-F., et al.: Connecting to Compete 2018: Trade Logistics in the Global Economy. World Bank, Washington, DC (2018). http://hdl.handle.net/10986/29971 Bitran, E., Nieto-Parra, S., Robledo, J.S.: Opening the Black Box of Contract Renegotiations: An Analysis of Road Concessions in Chile, Colombia and Peru. OECD Development Centre Working Papers, 317. OECD Publishing, Paris (2013). https://doi.org/10.1787/5k46n3wwx xq3-en Castro, C.: Panorama de los contratos de infraestructura y modelos de concesión: revisión sistemática de la literatura y análisis de hallazgos. Memoria de grado Ingeniero Civil Industrial. Universidad Técnica Federico Santa María, Chile (2020) Crampes, C., Estache, A.: Regulatory trade-offs in the design of concession contracts. Util. Policy 7, 1–13 (1998) Cullen, S.: The 12 Best Practices in Contract Management. Open Windows (2015) Dirección General de Concesiones: Dirección General de Concesiones. Ministerio de Obras Públicas (n.d.). http://www.concesiones.cl/quienes_somos/Paginas/default.aspx. Accessed June 2022 Ghayal, M.S., Salgude, R.R.: Effect of hybrid annuity model on road project. Int. J. Eng. Adv. Technol. 8(6), 1525–1530 (2019). https://doi.org/10.35940/ijeat.F8149.088619 Guasch, J.L., Laffont, J.-J., Straub, S.: Renegotiation of concession contracts in Latin America. Policy Research Working Paper, 3011. World Bank, Washington, DC (2003). http://hdl.handle. net/10986/18224 ISO 10845-1:2010. Construction procurement – Part 1: Processes, methods and procedures Kerf, M., Gray, R., Irwin, T., Levesque, C., Taylor, R.: Concessions for infrastructure: a guide to their design and award. World Bank Technical Paper Núm, 399, Finance, Private Sector and Infrastructure Network (1998) Lara-Galera, A.L., Sánchez-Soliño, A., Galindo-Aires, R.: First generation highways. Participation loans valuation in the framework of real options. Revista de la Construcción 15(2), 115–124 (2016) Liu, L.-R., Lee, Y.-M.: Remedial measures for erroneous environmental policies: assessing infrastructure projects of waste-to-energy incineration in Taiwan with a case study of the Taitung incinerator. Sustainability 8(12), 1284 (2016). https://doi.org/10.3390/su8121284 López-Campos, M., Tapia, L., Castro, C., Stegmaier, R.: Safeguarding the long-term condition of logistics infrastructure assets: an analysis of concession contracts. In: Castanier, B., Cepin, M., Bigaud, D., Berenguer, C. (eds.) Proceedings of the 31st European Safety and Reliability Conference, ESREL 2021 Organizers. Research Publishing, Singapore (2021). https://doi.org/ 10.3850/981-973-0000. ISBN: 981-973-0000-00-0. esrel2021-paper Ministerio de Obras Públicas: Normas y reglamentos para la inversión extranjera y nacional en infraestructura (n.d.). https://www.mop.cl/Prensa/Paginas/InversionOOPP.aspx. Consulted June 2022

The Concession Contract as an Instrument

715

Queensland Government: Contract Management Framework (2022). https://www.forgov.qld. gov.au/search-procurement-resources/contract-management-framework. Consulted June 2022. Accessed June 2022 Sabri, O.K., Lædre, O., Bruland, A.: Why conflicts occur in roads and tunnels projects in Norway. J. Civ. Eng. Manag. 25(3), 252–264 (2019). https://doi.org/10.3846/jcem.2019.8566 Tapia, L.: Diagnóstico de la operación actual de contratos de concesión de mantenimiento de infraestructura: casos representativos de Chile. Memoria de grado Ingeniero Civil Industrial. Universidad Técnica Federico Santa María, Chile (2021) Torres, J., Chackiel, J., Riveros, N., Hernánez, P.: Concesiones de obras públicas en Chile. 20 años. Ministerio de Obras Públicas de Chile (2016). https://concesiones.mop.gob.cl/Docume nts/libro-Concesiones_obras-publicas-chile-20.pdf World Bank: Public-Private-Partnership in Infrastructure Resource Center (2020). https://ppp.wor ldbank.org/public-private-partnership/. Consulted June 2022 Yan, X., Chong, H.-Y., Zhou, J., Li, Q.: Concession model for fair distribution of benefits and risks in build-operate-transfer road projects. J. Civ. Eng. Manag. 25(3), 265–275 (2019). https://doi. org/10.3846/jcem.2019.8649 Yang, P., Wang, S., Zhou, T.: Strategies of the game on the use of BOT model for the ecotourism enterprises in the infrastructure development of ecological tourism. MATEC Web Conf. 61 (2016). https://doi.org/10.1051/matecconf/20166102001

Agile Methods in Industrial Maintenance Lasse Metso1(B) and Nils E. Thenent2 1 Industrial Engineering and Management, LUT University, Yliopistonkatu 34,

53850 Lappeenranta, Finland [email protected] 2 Lufthansa Technik, Hamburg, Germany [email protected]

Abstract. Agile has become a popular term in organisations and businesses, beyond the IT sector. Characteristics of agile ways of working such as the assumption that boundary conditions and requirements might (and will) change make them attractive to applications in dynamic environments in general. Methods such as Scrum or Xtreme Programming turn agile values into actionable descriptions, originally intended for software development. This work offers a literature-based assessment of the suitability of agile methods for the maintenance of industrial production systems. In addition a process draft for the introduction of agile methods into maintenance practices is outlined. The findings show that so far there has been a relatively limited attention to agile methods in an industrial maintenance context but at the same time suggest the potential for their application. Keywords: Agile · Lean · Industry 4.0 · Maintenance · Industrial maintenance

1 Introduction The term “agile” has become a popular buzzword in organisations and businesses, not only in the IT sector. In contrast, to the best of the authors’ knowledge the extent to which industrial maintenance might benefit from agile ways of working has so far not been addressed extensively. For example, Srivastava et al. (2018) make the point that agile thinking in maintenance is a prerequisite of lean and agile manufacturing systems, without however outlining agile thinking specifically or clarifying maintenance conditions. This work seeks to identify where agile principles and techniques can be considered suitable to the environment in which industrial maintenance is performed, and where not. While there might not be a single definition of “agile”, the Agile Manifesto from 2001 sets key values: • • • •

“Individuals and interactions over processes and tools Working software over comprehensive documentation Customer collaboration over contract negotiation Responding to change over following a plan” (Beck et al. 2001)

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 716–725, 2023. https://doi.org/10.1007/978-3-031-25448-2_67

Agile Methods in Industrial Maintenance

717

Initially developed for software development, agile methodology is nowadays adopted by organisations seeking flexibility in various other fields (Annosi et al. 2020). Examples include the application of agile principles in human resource management (Denning 2018), business intelligence (Hughes 2013; Krawatzeck and Dinter 2015) and digital transformation (Ghezzi and Cavallo 2020). Since agile methodologies have their origins in the development of products in environments where requirements can change, they can help teams to meet business goals under uncertainty (Livermore 2008). Highsmith (2002) describes agile as methodology that helps development teams to hit a moving target. Complementing the agile manifesto, agile principles provide guidance for the implementation in practices (Cobb 2015): • • • • • • • • • • • •

Early and continuous delivery of valuable software Welcome changing requirements Deliver working software frequently Business people and developers must work together Build projects around highly motivated individuals Face-to-face conversation Working software is the primary measure of progress Promote sustainable development Continuous attention to technical excellence and good design Simplicity is essential Self-organizing teams Continuous improvement

Maintenance strategies in an industrial context such as production lines, milling or power plants include preventive, corrective and condition-based approaches. Preventive maintenance may be time or usage oriented. Depending on the specific case and the consequences of failures different weights may be given to these approaches. Since reliably planned production is important in most industries, detailed maintenance plans are the norm (Shyjith et al. 2008). Accordingly, purely corrective strategies are less common. Nevertheless, whenever unplanned events occur ad-hoc interventions are required. So, maintenance processes are also subject to dynamic situations that require prioritized operations. In the following we present an overview of agile methods and methodologies that originate from software development, followed by a brief introduction to agile manufacturing. Then, industrial maintenance is discussed to pinpoint some key characteristics. The succeeding discussion addresses how agile methods fit industrial maintenance and outlines a draft process for the introduction of agile methods to industrial maintenance practices. The paper ends with the conclusion that includes future research areas.

2 Agile Methodology in Software Development The Agile Manifesto was proposed as a response to experiences with “waterfall” development processes. These are based on detailed requirements and execution plans at the

718

L. Metso and N. E. Thenent

beginning of a project and then executed sequentially without going back once a step is completed (Cobb 2015). In contrast, agile principles embrace ongoing evolution during the development through an incremental approach. Development frameworks that follow values and principles of the Agile Manifesto are known as agile techniques (Rigby et al. 2016). Agile development teams concentrate on the features and functions that provide value to the user (or customer) and deliver them fast to receive feedback (Abrahamsson et al. 2002). Changes are accepted as a fact and are implemented as evolving requirements. An agile development process can be seen as a cycle where all phases of the process are visited in each iteration (Fig. 1). The goal of each cycle is to deliver a working product that is incrementally better than the product before the start of the cycle. In this way changing requirements can be considered in each cycle (Kuhrmann et al. 2016).

Fig. 1. Phases of agile process (Adapted from Kuhrmann et al. 2016).

Agile values and principles are turned into compact and actionable guidelines in the Scrum framework or Xtreme Programming (XP), originally for software development. Typical features of XP such as pair programming prevent errors and support knowledge sharing. Scrum focuses on the management of activities, for example daily status meetings. Agile development in software adds interaction and enables the incremental delivery of software as well as collaboration with customers and fast response to changes (Kajko-Mattsson and Nyfjord 2009).

3 Agile Manufacturing Already discussed in the 1990s, agile manufacturing (AM) intends to improve the competitiveness of companies through organizational flexibility and responsiveness in manufacturing (Gunasekaran 1999). AM can be considered a development of lean manufacturing (LM) and as such has relationships to Total Quality Management (TQM) and Just in Time (JIT) production. TQM, JIT and total productive/preventive maintenance (TPM) practices are seen as positively influencing AM practices and operational performance (Iqbal et al. 2018; Khalfallah and Lakhal 2021), e.g. through the TQM focus

Agile Methods in Industrial Maintenance

719

on continuous process improvement. AM is also described as the next step of LM and flexible manufacturing (Potdar et al. 2017). Recent developments such as Industry 4.0 emphasize connected production systems supplemented by virtual twin enabled simulation to further reduce the time to market of new products. This so called smart factory can be modular and agile because it is flexible and allows for production lines to be replaced or expanded (Gilchrist 2016). In summary, AM aims at competitive advantages on a business level without comparable process descriptions as for example Scrum.

4 Industrial Maintenance Whether maintenance could also benefit from agile ways of working needs a closer look at the conditions and requirements that apply to an industrial context. Metso et al. (2016) highlight that industrial maintenance exhibits multiple problem fields such as organisational, procedural, environmental as well as hardware and human-related aspects. Interactions of humans with IT-systems and among each other suggest that the agile values of direct communication and collaboration with the customer (or user) are highly relevant in such context. In the maintenance management framework eight phases are outlined that are connected to specific maintenance management techniques. Phase one begins with maintenance objectives and KPI’s, following definition of assets priority and maintenance strategy. The next phase includes weak point intervention and the design of preventive maintenance. Then follows the preventive maintenance planning and optimization of schedules and resources. Subsequently the maintenance execution is controlled and assessed as well as the asset life cycle analysed, which concludes the seventh phase. This leads into phase eight, continuous improvement and the utilization of new techniques. Since the maintenance management describes a circle, phase eight again continuous into phase one (Márquez et al. 2009). In the light of the technological evolution of production systems and increased integration of hard- and software including analytical capabilities, often associated with the term “Industry 4.0”, the approaches to maintenance evolve, too. In particular conditionmonitoring and predictive capabilities enable a shift away from general predefined plans to more targeted condition-specific maintenance actions (Metso and Thenent 2020). Raji et al.(2021) point out that Industry 4.0 technologies enable lean and agile supply chain strategies. These comprise: – – – – – –

centralized and collaborative planning, frequent introduction of new product developments, response speed to customers’ needs, IT integration, flexibility and ability to changes, the capability to change the production mix and a flexible use of equipment and an increased level of product customization.

Industry 4.0 supports the implementation of an agile supply chain strategy in management. The customers’ need for rapid changes increase the organizations willingness to develop agile capabilities (Raji et al. 2021).

720

L. Metso and N. E. Thenent

Maintenance functions and processes are defined in maintenance standards. Here is a brief description on maintenance processes, management and strategies. Standard EN 13306 defines maintenance terminology (Suomen Standardisoimisliitto SFS 2017a, b). It defines maintenance as a combination of technical, administrative and managerial actions and maintenance management contain activities that determine the maintenance requirements, objectives, strategies and responsibilities. When implementing those it involves maintenance planning, control and improve maintenance activities and economics. Maintenance strategy is defined as management methods used to achieve maintenance objectives. The most important maintenance types when discussing agile maintenance are: preventive maintenance, condition-based maintenance, predictive maintenance, and potentially also active maintenance, improvements, modernization and opportunistic maintenance. The Standard EN 16646: maintenance within physical asset management specifies more details about asset management and maintenance process management (Suomen Standardisoimisliitto SFS 2014). EN 17007 defines maintenance policy as a general approach to the provision of maintenance and maintenance support based on the objectives and policies of owners, users and customers. The maintenance policy includes instructions about methods, programme, budget, etc. The goals can be to maximize availability and useful life of the items, to improve/guarantee safety of the items and individuals, product quality, environmental protection, and to optimize maintenance costs. The maintenance policy gives guidance on maintenance strategy (corrective and/or preventive, predetermined or condition-based maintenance) and in-house or outsourced maintenance. Maintenance procedures describe preventive and corrective maintenance tasks. They also include resources required, the company’s health and safety rules as well as the regulatory aspects (e.g. environment, safety). The above, as well as other maintenance-related standards, have to be taken into account in maintenance related tasks, planning and management. Consequently, maintenance has to be considered a highly standardized and depending on the industry also highly regulated domain.

5 Discussion: To What Extent Do Agile Methods Fit Industrial Maintenance? The question in the headline of this section points towards the suitability of agile methods depending on the specific context. Several authors have highlighted that agile methods can only be effective in particular conditions. For example, Preußig (2020) links the costs of changes to the suitability of agile methods. The higher the costs for changes during the (agile) product development become, the less effective agile approaches are. Transferred to a maintenance context this relation means that: 1. whenever clear target conditions can be described upfront, i.e. there will certainly be no changes during the maintenance process and 2. whenever changes will lead to high costs, for example through the extended downtime of a production line, agile methods are less suitable.

Agile Methods in Industrial Maintenance

721

Further criteria for assessing the suitability of agile methods within a situation at hand are based on the Cynefin framework (Fuchs et al. 2019). The following list describes the four situational types, adapted to maintenance tasks or environments: – In a situation where relationships are clear and existing knowledge can be used, well established approaches can successfully be employed for example by following best practices. When routine tasks and procedures are clearly described and the environment is static (for example during the night stop of an aircraft). – Wherever domain specific expert knowledge and analysis is required a good practices approach is appropriate. Such situation reflects for example step-by-step trouble shooting where the appropriate tools and processes are known but the specific case might differ from prescribed examples. – When cause and effect relationships are unclear, ad-hoc learning is required: an appropriate strategy comprises probing and adaption. In such context agile methods are particularly effective. For example, in practice, troubleshooting is not always successful only by following the known and prescribed processes. Sometimes innovation and even good luck play a role (Kinnison 2004). When it comes to production system improvements, incremental approaches allow careful adjustment of the technical system under real-world conditions. – In chaotic situations, rapid interventions are required to move into a more manageable situation that can only then be approached in a more structured way. During an emergency in a production system where no procedures exist it is essential to quickly stabilize the situation, in particular when large (mechanical) forces are involved. Only once the immediate danger is under control maintenance and operations experts may derive strategies for recovery. Similar to the Cynefin-based approach, but particularly aimed at agile methods in maintenance, Heitz et al. (2020) propose an assessment of different kinds of work to identify the most suitable maintenance team structure. Their focus lies on specific tasks, comparable to the assessment of the situation at hand: – Activities for optimization and improvement for example the design of maintenance strategies, the planning of major shutdowns, and continuous activities for reliability improvements, require intensive coordination. Here, cross-functional teams are considered most effective. – Operational support activities as for example services for problem-solving or the implementation of new processes, involve some regularity and can follow an a combined lean and agile thinking approach where personnel with different expertise join teams as required. – Frontline execution activities such as day-to-day preventative-maintenance, troubleshooting, and emergency maintenance as well as planned activities during shutdowns can effectively be carried out by self-organized that work according to lean principles. While the Cynefin-based approach provides a strong indication the suitability of agile methods in general, the assessment by Heitz et al. provides more specific guidance

722

L. Metso and N. E. Thenent

for a suitable structure and staffing of the maintenance team. Since the need for on the spot decision making and the degree of standardization can vary from case to case, it is important to take into account the individual situation with respect to technical, operational and human resources-related boundary conditions before an agile approach can be considered. Pharaon (2022) demonstrates how Total Productive Maintenance (TPM) can benefit from agile principles such as iteration, workflow management and cross-functional collaboration with specific adaptions of Scrum and a scaled agile framework. The results of the case study show a reduction of unplanned downtime by 20% in 12 months. Despite this example of a successful adoption of agile methods in a maintenance organization, there are multiple barriers. Hierarchical structures of maintenance organizations can create a tendency for specialized functional teams and trades that work isolated from each other, teams such as planning, execution, and reliability teams operating in silos (Heitz et al. 2020). Further examples can be outlined through similarities between the implementation of agile methods and Industry 4.0, as identified by Raji et al. (2021). Brozzi et al. (2021) noticed that lager companies were better prepared to support the transition toward the Industry 4.0 than SMEs. The size of company affects the ability to utilize new methods. SMEs do not have the high turnover to invest into new technology and methods. Furthermore, SMEs lack a clear vision how to implement new technology, exacerbated by limited knowledge and skills about digital transformation, and a lack of decision-making to support high-level technology implementation (Chonsawat and Sopadang 2020; Brozzi et al. 2021). Also, different industrial sectors are on different levels of digitalization. An example from the ceramics industry shows a of lack automation, monitoring systems and IT systems, and consequently difficulties in the implementation of condition-based maintenance (Kellner et al. 2020). Aside from technology, it has to be expected that turning agile principles into maintenance practices requires new employee skills as well as significant changes affecting culture and processes, similar to the implementation of digitalization (Kohnová et al. 2019). To come back to the question raised in the section headline “how suitable are agile approaches for industrial maintenance?” there is no black and white answer. While there are examples in the literature that demonstrate a successful implementation, experiences from Industry 4.0 implementation show multiple barriers to such fundamental management and way of working changes. Therefore, we propose a draft process that can guide the introduction of agile practices into industrial maintenance: 1. Task assessment: An overview of all maintenance tasks that shall be considered for a more agile way of working including an assessment e.g. based on the Cynefin framework can reveal a general suitability. 2. Skills assessment: Is the effected personnel, and that includes managers, planners and technicians willing and able to adapt new ways of working that may include less managerial control and more local decision making? 3. Pilot introduction: The selection of a pilot cases or example maintenance package can increase the chances for a quick implementation and suitability demonstration.

Agile Methods in Industrial Maintenance

723

4. Learn sufficiently long time: Starting with a selected set of methods, for example Scrum experience is gathered “sufficiently” long to reach a certain stability in the process that includes overcoming the first difficulties. 5. Assess and refine: When reaching a state that is stable enough, the selected approach is assessed. For example with regards to effectiveness, efficiency, suitability for other cases and need for improvement. 6. Expand: Look for more cases, other organisational units or tasks where the above learned approach seems suitable. For the new cases again, start with the assessments 1 and 2. 7. Continuously adapt and improve: In line with lean and agile principles continuous learning and improvements shall be an intrinsic characteristic of the established individual approach.

6 Conclusion From the values of the Agile Manifesto, literature on agile methodologies such as Scrum, and agile manufacturing we found only a few examples on the implementation of agile methods in an industrial maintenance context. These show that, transferring agile principles such as the continuous delivery of incremental product elements of value from software development to the maintenance of production systems are at least partly feasible and promising. One study clearly demonstrates the improvements that have been achieved through the Agilized TMP Model (Pharaon 2022). Ideally, the introduction of agile methods is founded on a strong organizational support in an environment where cross-functional teams have the need to learn while they fulfil value-adding tasks to their customer. Such situation can for example be found in a company-wide transformation from production to service-provision (Pawar et al. 2009). In an Industry 4.0 context where lean practices together with interconnected machines and condition-monitoring in-time and ad-hoc maintenance interventions allow (and require) flexibility. It is therefore reasonable to assume that the importance of methods that support decision-making in dynamic environments – as agile methods promise – will further increase. Rather than planning for long downtime periods where all maintenance activities are scheduled, prioritized work package can be scheduled for shorter maintenance “sprints”. On the other hand, the lack of both, agile skills and the ability to invest into new technologies and methods can hamper the proliferation of more agile maintenance practices. This work offers an impression of the possibilities agile principles and techniques can hold for the maintenance of industrial production systems. It is limited by the reviewed literature and therefore further research including a more systematic literature review is needed to investigate the application of agile methods in practice through empirical studies. In addition to the literature overview we propose a process draft that can guide the introduction of agile methods into maintenance practices. Since this draft is primarily based on the literature further research including empirical evidence is required for validation and refinement. The authors gratefully acknowledge the reviewers’ comments that helped to improve the article.

724

L. Metso and N. E. Thenent

References Abrahamsson, P., Salo, O., Ronkainen, J.: Agile Software Development Methods: Review and Analysis. VTT, Espoo (2002) Annosi, M.C., Foss, N., Martini, A.: When agile harms learning and innovation (and what can be done about it). Calif. Manag. Rev. 63(1), 61–80 (2020). https://doi.org/10.1177/000812562094 8265 Beck, K., et al.: Manifesto for Agile Software Development (2001). http://agilemanifesto.org/. Accessed 6 January 2014 Brozzi, R., Riedl, M., Matta, D.: Key readiness indicators to assess the digital level of manufacturing SMEs. Procedia CIRP 96, 201–206 (2021). https://doi.org/10.1016/j.procir.2021. 01.075 Chonsawat, N., Sopadang, A.: Defining SMEs’ 4.0 readiness indicators. Appl. Sci. 10(24), 8998 (2020). https://doi.org/10.3390/app10248998 Cobb, C.G.: The Project Manager’s Guide to Mastering Agile: Principles and Practices for an Adaptive Approach. Wiley, Hoboken (2015) Denning, S.: The emergence of Agile people management. Strategy Leadersh. 46(4), 3 (2018). https://doi.org/10.1108/SL-04-2018-0042 Fuchs, C., Barthel, P., Winter, K., Hess, T.: Agile Methoden in der digitalen Transformation – mehr als ein Konzept für die Softwareentwicklung. Wirtschaftsinformatik Manag. 11(4), 196–207 (2019). https://doi.org/10.1365/s35764-019-00192-8 Ghezzi, A., Cavallo, A.: Agile business model innovation in digital entrepreneurship: lean startup approaches. J. Bus. Res. 110, 519–537 (2020). https://doi.org/10.1016/j.jbusres.2018.06.013 Gilchrist, A.: Industry 4.0: The Industrial Internet of Things. Apress, Berkeley (2016) Gunasekaran, A.: Agile manufacturing: a framework for research and development. Int. J. Prod. Econ. 62, 87–105 (1999) Heitz, S., Maestu, M., Marcote, M., Thibert, J.: The right tools for every job: lean and agile in maintenance. Operations Practice (2020) Highsmith, J.A.: Agile Software Development Ecosystems. Addison-Wesley, Boston (2002) Hughes, R.: Agile Data Warehousing Project Management: Business Intelligence Systems Using Scrum and XP. Morgan Kaufmann, Waltham (2013) Iqbal, T., Huq, F., Bhutta, M.K.S.: Agile manufacturing relationship building with TQM, JIT, and firm performance: an exploratory study in apparel export industry of Pakistan. Int. J. Prod. Econ. 203, 24–37 (2018) Kajko-Mattsson, M., Nyfjord, J.: A model of agile evolution and maintenance process. In: Sprague, R.H., Jr. (ed.) Proceedings of the 42nd Annual Hawaii International Conference on System Sciences: Abstracts and CD-ROM of Full Papers, Waikoloa, Big Island, Hawaii, 5–8 January 2009, pp. 1–10. IEEE, Los Alamitos (2009) Kellner, T., Necas, M., Kanak, M., Kyncl, M., Kyncl, J.: Assessment of readiness for Industry 4.0 implementation in ceramic industry. Manuf. Technol. 20(6), 763–770 (2020). https://doi.org/ 10.21062/mft.2020.110 Khalfallah, M., Lakhal, L.: The impact of lean manufacturing practices on operational and financial performance: the mediating role of agile manufacturing. Int. J. Qual. Reliabil. Manag. 38(1), 147–168 (2021). https://doi.org/10.1108/IJQRM-07-2019-0244 Kinnison, H.A.: Aviation Maintenance Management. McGraw-Hill Companies, Inc., New York (2004) Kohnová, L., Papula, J., Salajová, N.: Internal factors supporting business and technological transformation in the context of Industry 4.0. Bus. Theory Pract. 20, 137–145 (2019). https://doi. org/10.3846/btp.2019.13

Agile Methods in Industrial Maintenance

725

Krawatzeck, R., Dinter, B.: Agile business intelligence: collection and classification of agile business intelligence actions by means of a catalog and a selection guide. Inf. Syst. Manag. 32(3), 177–191 (2015). https://doi.org/10.1080/10580530.2015.1044336 Kuhrmann, M., Münch, J., Richardson, I., Rausch, A., Zhang, H. (eds.): Managing Software Process Evolution. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-31545-4 Livermore, J.A.: Factors that significantly impact the implementation of an agile software development methodology. J. Softw. 3(4), 31–36 (2008). https://doi.org/10.2307/j.ctt46nrzt.12 Márquez, A.C., de León, P.M., Fernández, J.G., Márquez, C.P., Campos, M.L.: The maintenance management framework: a practical view to maintenance management. J. Qual. Maint. Eng. 15, 167–178 (2009) Metso, L., Marttonen, S., Thenent, N.E., Newnes, L.B.: Adapting the SHEL model in investigating industrial maintenance. J. Qual. Maint. Eng. 22(1), 62–80 (2016) Metso, L., Thenent, N.E.: Characteristics of Maintenance 4.0 and their reflection in aircraft engine MRO. In: Ball, A., Gelman, L., Rao, B.K.N. (eds.) COMADEM 2019, pp. 499–509. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-57745-2_42 Pawar, K.S., Beltagui, A., Riedel, J.C.K.H.: The PSO triangle: designing product, service and organisation to create value. Int. J. Oper. Prod. Manag. 29(5), 468–493 (2009). https://doi.org/ 10.1108/01443570910953595 Pharaon, J.W.: An agilized total productive maintenance framework to reduce unplanned downtime. Praxis in partial fulfillment of the requirements for the Doctor of Engineering, The George Washington University (2022) Potdar, P.K., Routroy, S., Behera, A.: Agile manufacturing: a systematic review of literature and implications for future research. Benchmarking Int. J. 24(7), 2022–2048 (2017). https://doi. org/10.1108/BIJ-06-2016-0100 Preußig, J.: Agiles Projektmanagement, 2nd edn. Haufe-Lexware; Haufe, Freiburg im Breisgau (2020) Raji, I.O., Shevtshenko, E., Rossi, T., Strozzi, F.: Industry 4.0 technologies as enablers of lean and agile supply chain strategies: an exploratory investigation. Int. J. Logist. Manag. 32(4), 1150–1189 (2021). https://doi.org/10.1108/IJLM-04-2020-0157 Rigby, D., Sutherland, J., Takeuchi, H.: Embracing agile. Harvard Business Review, pp. 40–48, May 2016. https://hbr.org/2016/05/embracing-agile. Accessed 29 May 2022 Shyjith, K., Ilangkumaran, M., Kumanan, S.: Multi-criteria decision-making approach to evaluate optimum maintenance strategy in textile industry. J. Qual. Maint. Eng. 14(4), 375–386 (2008) Srivastava, P., Khanduja, D., Agrawal, V.P.: Integrating agile thinking into maintenance strategy performance analysis. Int. J. Process Manag. Benchmarking 8(2), 228–245 (2018) Suomen Standardisoimisliitto SFS: SFS-EN 16646: Maintenance - Maintenance within physical asset management, Brussels. European Committee for Standardization (2014) Suomen Standardisoimisliitto SFS: SFS-EN 13306: Maintenance. Maintenance terminology, Brussels. European Committee for Standardization (2017a) Suomen Standardisoimisliitto SFS: SFS-EN 17007 Maintenance process and associated indicators, Brussels. European Committee for Standardization (2017b)

Standards-Based Interoperable Digital Twin in Industry 4.0 – A Pilot Demonstration Karamjit Kaur1(B) , Matt Selway1 , Markus Stumptner1 , Alan Johnston2 , and Joseph Mathew3 1 Industrial AI Research Centre, University of South Australia, Adelaide, Australia {karamjit.kaur,matt.selway,markus.stumptner}@unisa.edu.au 2 MIMOSA, 2200 Jack Warner Pkwy, Suite 300, Tuscaloosa, AL, USA [email protected] 3 The Asset Institute, Brisbane, Australia [email protected]

Abstract. In the era of Industry 4.0, digital twins bridge the gap between the physical and digital worlds, enabling early-detection of issues, increased production and efficiency, among other benefits. The development of digital twins begins early in the life cycle of a system/plant by using the data from various information systems as it passes through its life cycle, from as-designed to asbuilt to as-maintained state. These information systems which need to provide data for the construction of digital twins are mostly proprietary systems and thus act as stumbling blocks in the way of industry gaining benefits associated with industrial digital transformation. The industry should be able to develop and use standards-based adaptors to alleviate this issue. In this paper, we elaborate on this standards-based interoperable approach and its challenges and provide implementation details of an industry-led pilot where digital twin was established during the capital projects, then maintained, synchronized, and leveraged throughout the rest of the asset life cycle, including the brownfield information remediation. In the pilot, we have validated Open Industrial Interoperability Ecosystem (OIIE) specifications, published as part of ISO 18101, which brings multiple standards and specifications together in an interoperable fashion to enable a supplier neutral industrial digital ecosystem.

1 Introduction Industry 4.0 has introduced the concept of smart industry where cyber-physical systems observe the physical processes of the industrial plant and make decentralized decisions. The Industrial Internet of Things (IIoT) connects machines and sensors, and these linked resources send rich data which can be used to improve collaboration between performance, operations, and asset maintenance (Mumtaz et al. 2017). In an Industry 4.0 environment, sensors, devices, machines, and components are becoming equipped with digital twins that contextualize these components and their generated information to provide increased business value. Further, digital twins of individual components must be integrated into those of larger systems and facilities. As such, Digital twins empower © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 726–735, 2023. https://doi.org/10.1007/978-3-031-25448-2_68

Standards-Based Interoperable Digital Twin

727

factory/facility operators to continuously capture machine states and condition, combine it with information captured in other systems, and further analyse the combined data to predict the optimal point in time at which to initiate maintenance, reduce costs, and maximize production. Undoubtedly, the concept of digital twins offers great opportunities and benefits to the industry; however, the industry is facing challenges in their realization and implementation during both their creation and in keeping the digital twin up to date with respect to the changes made in the physical world. These challenges primarily emerge because of lack of interoperability between systems (Greives and Vickers 2017). No single system of an enterprise typically tracks all the necessary information to keep digital twins up to date, instead it is scattered between different systems of record. To utilize this data effectively it needs to be shared between systems, systems of systems, and enterprises, which can only be done by achieving interoperability across systems at different levels (Sjarov et al. 2021). For industries to utilise the Industry 4.0 concepts such as IIoT, there needs to be strong convergence between IT (Information Technology) and OT (Operational Technology) domains in the industry to enable digital transformation (Kuusk et al. 2019). The ideal way to achieve this kind of interoperability is through the use of (open) standards (Semeraro et al. 2021). This is made difficult, however, due to the number of standards applicable to industrial interoperability (each with a different focus), the adoption of different standards in different areas of industry, the evolution of standards to fulfil the requirements of their user-bases, the inconsistencies of overlaps and gaps left when combining standards, and a lack of agreement between users of different standards (Platenius-Mohr et al. 2020). The Open Industrial Interoperability Ecosystem (OIIE) architecture (published as part of ISO/TS 18101-1:2019) provides a solution for achieving standards-based interoperability for Digital Twins. The OIIE specifications enable devices and systems to communicate effectively in both inter- and intra- enterprise contexts using a variety of standards, data models, and exchange protocols (Kaur et al. 2018). In this paper, we provide details of OIIE OGI (Oil and Gas Industry) pilot demonstration executed along with industry partners (Yokogawa, Worley, PdMA) and utilising industry standards and specifications (MIMOSA, CII, IOGP, ISA) which spans over multiple phases of an asset’s life cycle and illustrate how digital twins are kept up to date by leveraging standards-based interoperability. While originating in the Oil & Gas Industry, the OIIE and associated pilots are applicable to the broader asset intensive industries, such as transport and utilities associated with critical infrastructure.

2 Digital Twin for Asset Life Cycle – Pilot To create and maintain digital twins across the asset life cycle, data needs to be exchanged, validated and processed, which is an unresolved challenge in most of the industries (Doubell et al. 2021). Digital twins can enact as a medium to retain information between various life cycle phases of an asset. The information is provided by different organisations along the way, for example, manufacturers provide model details and specification of the asset while owner/operators provide operational and maintenance information. Thus, this information requires transforming to a common standardized format for

728

K. Kaur et al.

exchange, federation, and processing. The MIMOSA CCOM information model is one of the standard information models used for exchange of asset life cycle information (Hietala et al. 2022).

Design & Engineering • As-designed P&ID informaƟon in Proteus XML • TranformaƟo n to CCOM XML

Greenfield Make/Model Match-up • RFI/RFI Response based on funƟonal requirements • Standardised digital datasheets

Procurement

InstallaƟon

• RFQ/RFQ Response

• Capital project asset installaƟon • CII AWP Data Requirements - InstallaƟon Work Packages

• Purchase order

InformaƟon Hand-over to O&M • As-built informaƟon handover to O&M

OperaƟons & Maintenance • CBM Data AcquisƟon • CBM Trigerring • Remove & Replace

Brownfield InformaƟon RemediaƟon • RFI/RFI Response based on Asset data

Fig. 1. OIIE OGI pilot demonstration

The OIIE OGI pilot is an ongoing public interoperability testbed which demonstrates the use of multiple standards and specifications across the asset life cycle for achieving a digital ecosystem. The pilot is run in a series of phases, with a major milestone demonstrated in December 2019 and a recent extension demonstrated in April 2022. In this paper we summarise the details of this ongoing pilot and what has been demonstrated to date. The OIIE OGI Pilot is based on sets of OIIE Use Cases which capture the requirements of system-of-system interactions and their standards-based implementation. Figure 1 shows the various OIIE Use Cases demonstrated in the pilot, where information in the digital twin starts building up from the design and engineering phase and continues throughout the operations and maintenance phase, with information handovers between organisations occurring during the process. The pilot involves all supply-chain actors EPCs (Engineering Procurement & Construction), OEMs (Original Equipment Manufacturers), O/Os (Owner/Operators), and software vendors. We provide details below of each step shown in Fig. 1: • Design and Engineering Phase – The example used in the pilot is of a debutanizer fractionator whose process and instrumentation diagram (P&ID) were provided by Advisian Digital (Worley) in Proteus XML format. This format is generally used by EPC’s engineering systems for P&ID exchange1 . The P&ID information was transformed to MIMOSA CCOM XML using transformation engine hosted at Industrial AI Research Centre, University of South Australia. • Make/Model match-up – After completion of material take-off, the EPC determines the best make and model meeting or exceeding the functional requirements 1 Maintenance of the Proteus XML format has relatively recently been adopted by DEXPI.

Standards-Based Interoperable Digital Twin

•

•

•

•

•

729

of functional locations/tags in the P&ID by requesting information from OEMs. In the demonstration, a datasheet published by ISA for temperature transmitters was used for specifying functional requirements, which was transformed to MIMOSA CCOM for interoperable exchange. The OEM (Yokogawa) responded with possible model numbers and model datasheet information in MIMOSA CCOM format. Procurement – There may be multiple RFI/RFI response exchanges between EPC and OEM before EPC send out RFQs to OEMs asking for pricing and other technical details, including details on which documents will be supplied and when. In the recent demonstration, we showcased sending of RFQs for a low-voltage motor using IOGP JIP 33 procurement specifications. Installation - The procured assets were installed as part of the capital project, and the as-built information (which asset is installed at which functional location at what time) thus created, was handed over to the O/O in MIMOSA CCOM XML format. The status of the work was monitored against CII AWP IWP (Installation Work Packages), keeping scheduling and constructions systems up to date, with the work orders defined conforming to AWP IWP data requirements. Information Hand-over to O&M (Operations & Maintenance) – A digital information handover from EPC to O/O was demonstrated using MIMOSA CCOM XML format. To overcome the issue of inconsistent terminologies used by organizations, reference data libraries (RDLs) such as IOGP JIP36 CFIHOS and ISO 15926-4 RDL are used. Operations and Maintenance – Two aspects of condition-based maintenance were demonstrated. Initially, condition and operation data were collected for monitoring and analysis. The second aspect was to produce health assessments and advisories so that maintenance activity is performed before failures occur. The PdMA Corporation provided advisories based on the data analytics performed on data collected on one of the motors in the debutanizer. The advisory was to replace the motor. SAP PM (Plant Maintenance) module was used in the pilot to demonstrate the replace activity showing near real-time updates in SAP PM based on work order updates made by the (simulated) technician in the field. Brown-field information remediation – Over the period of operations, some information about the asset may be worn-off or lost over time or may not been provided during hand-over. We demonstrated how O/Os can re-gain the information about an asset by sending a request for information on the serialized asset, including as much information they have on the asset. The OEM (Yokogawa) returned appropriate model asset-specific data back to the O/O in the pilot. This is an essential aspect of upgrading Brownfield sites to be Digital Twin capable and was performed leveraging the same elements as for Make/Model match-up.

The dataset and messages used in the pilot, along with the pilot’s application architecture and recorded demonstration, are available on the MIMOSA website. The use cases above incorporate many interoperable interactions between systems supported by the following ecosystem components:

730

K. Kaur et al.

• OpenO&M ISBM specification provides supplier-neutral middleware-based message exchange and data transport in conformance with the ISA-95/IEC 62264 Messaging Service Model. It supports publish/subscribe and request/response interactions, for which agreed channel/topic configurations according to OIIE guidelines were used. Security is defined on a per-channel basis, allowing only authorized applications to participate in necessary system interactions. • Information and message models for the representation of messages and service inputs/outputs such as MIMOSA CCOM XML, Proteus XML, BatchML. • Reference data for common interpretation of information such as CFIHOS RDL, ISO 15926-4 RDL, IEC 61360 (Common Data Dictionary), and MIMOSA RDL (where not covered by others). • MIMOSA SDAIR specification for an (asset) interoperability register containing identifiers for all physical and logical assets and an unlimited number of relationships between them. This component provides federating functionality across the systems participating in the ecosystem, including relationships that cannot be captured by any individual system. Future, and in progress, iterations of the OIIE OGI Pilot will include additional OIIE primary component specifications: the OpenO&M Service Directory, for more dynamic configuration of system interactions, and OpenO&M CIR (Common Interoperability Registry), for broader reconciliation of object identifiers.

3 Standards Used by the Pilot The OIIE approach and the OIIE OGI Pilot aim to incorporate a number of fit-forpurpose standards into the ecosystem to achieve interoperability. These standards come from a multitude of sources and address different aspects of Digital Twins. Figure 2

Standards Bodies

Incorporated Standards

Reference Data Libraries

MIMOSA

ISO 18435 (OSA-EAI)

ISO 15926-4

OpenO&M IOGP JIP36

ISO 8000 (Data Quality)

IOGP

ISO 16812 (API 660)

REST/JSON PIP UN/CEFACT CCT NORSOK

IEC 61360 (CDD) ISO 20922 (MQTT)

OPC FoundaƟon ISO 9834-8 (UUID)

ISA-TR20 ISO 13709 (API 610)

IEC

ISO 22745 (Terminological DicƟonaries)

ISA

Product Datasheets

SOAP/XML

IEC 62264 Part 6 (MSM) ECCMA OTD (eOTD)

IEC

ISO 20022 (XML)

ISO 21778 (JSON)

ISO 13374 (OSA-CBM) OAGi ISO

Data Protocol & Formats

EnergisƟcs UoM DicƟonary

ISO 19464 (AMQP)

ASME IOGP JIP33

Fig. 2. Summary of standards identified by the OIIE and OIIE OGI pilot

Standards-Based Interoperable Digital Twin

731

identifies many of the standards publishing bodies and related standards that are currently incorporated in the OIIE or are planned to be incorporated in future Pilots. These are listed across several dimensions. Each of these typically include one or more data models.

4 Requirements for Digital Twins Platenius-Mohr et al. 2020 identified a list of requirements to fulfil to enable interoperable digital twins, with respect to information models not mathematical models. Here we discuss how the OIIE approach helps fulfil these requirements and thus facilitate the realisation of standards-based interoperable digital twins. (R1) Supporting different formats: This requirement states the importance of ability to transform between different formats. In the OIIE architecture and specifications, transformations are an essential part of the ecosystem. It acknowledges that there are many different (standard) information models that focus on different aspects of the Asset Life Cycle and aims to identify appropriate models that fill the necessary capabilities and identify what is necessary to have them work together, such as appropriate mappings. Moreover, the concept of adaptors at either end of the ISBM embody the local transformations, such as those Yokogawa implemented between MIMOSA CCOM and their proprietary format in the pilot. The OIIE approach also supports the concept of delivering transformations as a service by plugging systems, such as the Transform Engine (Berger et al. 2010) used to transform Proteus XML into MIMOSA CCOM, into the ecosystem. (R2) Variety of information deployment: This requirement elicits that the transformation between information models should be independent of information deployment mechanism, i.e., they could exist in the cloud, on devices themselves, or anywhere in between. An OIIE adaptor may provide a local transformation, i.e., between standard model and specific proprietary model. However, that does not require that it exist at the location of the target system: depending on requirements and capabilities, an OIIE adaptor may be deployed in the cloud, on device, or anywhere in between. Similarly, when delivering transformations as a service, the service is by-definition independent of the deployment of target systems. The pilot illustrated this with adaptors co-located with the target system, such as in the CBM Use Case; some adaptors were deployed ‘at the edge’ within the same infrastructure but communicating via API; and the Transform Engine was deployed in the cloud. (R3) Evolution of source and target formats: This requirement asserts the need to consider evolution in information models over time by the transformation rules/engine. The evolution can be either in the source or target information model. OIIE Specifications, including OIIE Scenarios document basic data requirements as well as basic mappings between standards that have been identified for use with that Scenario. Moreover, the specifications document versions against which conformance is to be measured. While this may constrain evolution in the short term, it ensures interoperability by preventing implementations from claiming conformance against untested

732

K. Kaur et al.

or not yet agreed formats. The OIIE Piloting process can then be used to review and test new updated models. OIIE Scenarios can be updated in a version-controlled manner, and implementations can declare their conformance to the new (and old) versions of the specifications and/or models. In addition, transformation as-a-service can assist with evolved models as different services could provide the transformation for new models. The OIIE approach does not aim to eliminate evolution, as it is an integral part of industrial growth; however, it is important to constrain it so that the ecosystem can maintain interoperability rather than fragment into pockets of non-interoperable systems. (R4) Semi-automatic definition: This requirement specifies the need to have tool supported means for defining mappings and transformations between models. The OIIE specification does not specify any specific means defining transformations. There are many approaches to data and model transformation, such as that used in the Transform Engine (Berger et al. 2010) or the more recent work of Selway et al. 2018, which has built on the literature of model-driven transformations. There is also diverse literature on schema/ontology matching and mapping that can be leveraged by transformation tools (Ochieng and Kyanda 2018; Sabou et al. 2020). It is left to researchers and software vendors to provide the best tools to support the definition and implementation of transformations. However, by providing solutions that are OIIE conformant, chiefly the ISBM specification, such tools can be plugged into the ecosystem and leveraged by users of many organisations as meet their needs. (R5) Bidirectional mapping: This requirement states that transformations need to be able to be performed in both directions as information exchanges may occur in both directions. Transformations can be complex and even more so when made to be bidirectional (Selway et al. 2018). The OIIE approach helps manage the complexity in several ways: 1) OIIE Scenarios identify whether a transformation is unidirectional or requires bidirectionality, for example, in the pilot the edge device mapping measurements to MIMOSA CCOM in the CBM Use Case requires only unidirectional mapping, while Work Management aspects require bidirectionality to support requests to create or update work orders and broadcast the work order updates; 2) supporting specifications such as the OpenO&M CIR support the tracking of objects identifiers and their mappings, which is often necessary to achieve bi-directionality; 3) a federated model, such as that managed by the MMOSA SDAIR, tracks meta-data that assists with traceability required for bidirectional transformations. (R6) Composite structures: This requirement asserts that not only flat structures, but complex nested structures and their relations need to be addressed by mappings. The OIIE approach recognises the need for complex transformations. One of the earliest Use Cases that was addressed by previous pilots was Information Handover to O&M, which involves the transformation of design information including P&IDs and breakdown structures from different viewpoints. Such a transformation captures complex hierarchies and relations between elements of the models and is a well-known requirement for Digital Twins of assets all sizes, from equipment to plants, platforms, and facilities.

Standards-Based Interoperable Digital Twin

733

(R7) Maintain semantics: This requirement indicates that the reference data libraries also be transformable to maintain references to the common definitions. The OIIE approach has maintained the need to reference many different sources of reference data as, just like the models themselves, no one reference data library (RDL) captures all that is necessary. Moreover, the combination of Enterprise RDLs, which extend the core reference data, as well as different RDLs being adopted by different domains or world regions, means that the need to relate and transform between different RDLs is an accepted requirement for an OIIE ecosystem. The Pilot, for example, has investigated ISO 15926 and CFIHOS RDLs, in addition to RDLs necessary for standardised datasheet definitions. Moreover, the OpenO&M CIR provides object identifier mapping and querying services, which can be used to related and find mappings between referenced data items. The OIIE also references ISO 29002 (and its derivatives) for electronic dictionaries, which can also be used to link reference data item definitions to a certain extent. (R8) Make target models accessible: This requirement indicates that the agreed target models of transformations be readily accessible, e.g., via APIs, both online or offline. As the OIIE specifications identify standard information models that are fit-forpurpose, this generally means that they are backed by machine readable schemas. Such schemas are usually distributed with the standard (for offline use) and can typically be retrieved from online locations. In cases where the schemas are not accessible, the OIIE specifications can potentially be used to define a location or mechanism on top of the original standard that allows retrieval of such schemas. Furthermore, the OIIE infrastructure services can be used to set up a means of sharing the schemas, as well as providing their own for their services. For example, the ISBM services advertise their schemas for both the SOAP/XML interface and REST/JSON interface (through an OpenAPI specification). In addition to supporting the data transformation requirements identified by Platenius-Mohr et al. 2020, the OIIE approach supports additional requirements that are necessary for interoperable Digital Twins such as model federation across systems of record, management of change, and service registration and discovery.

5 Future Pilots Work is underway for future OIIE OGI pilot demonstrations where the emphasis is on industrial digital ecosystem seamlessly spanning intra- and inter-enterprises and the associated cyber-security and ecosystem administration requirements. To provide secure inter-enterprise connectivity, mappings of OpenO&M ISBM specification to Advanced Message Queuing protocol (AMQP) v1 are being documented and implemented. In addition, as the first in a potential series of specifications defining how other messaging protocols and standards can be integrated into OIIE ecosystems, this work will enable integration of AMQP-enabled systems and edge devices. This work is performed in the OpenO&M ISBM Joint Working Group (JWG) which operates under the OpenO&M initiative. It is an initiative where multiple industry standards organizations work collaboratively to produce information standards and specifications for exchange of O&M data and include members like ISA, MESA, OAGi and OPC foundation.

734

K. Kaur et al.

Further supporting the inter-enterprise efforts, ongoing pilots work is incorporating the discovery of services across organisational boundaries. To that end, the OpenO&M Service Directory specification is being updated with an extended service and capability metamodel. This will improve the registration and discovery of systems participating in a digital ecosystem via models of the capabilities and services provided.

6 Conclusion A continuous and iterative information handover during a capital project to an O/O with the intent of building the digital asset alongside the physical asset, was demonstrated in the OIIE OGI Pilot. Different standards, specifications, RDLs were incorporated and applied in the pilot where and when they were appropriate based on requirements and adoption by industry. This demonstration captures some of the essential processes for building and maintaining a Digital Twin that is synchronized with changes occurring in the real-world and supporting analytics. These essential elements include ensuring continuity of Engineering Data, the association of accurate Product Model Data to the designed and constructed Plant, event-driven updates to the model state of the Plant during both Capital Project and Operations, and the capture and analysis of Condition information leading to maintenance activities.

References Grieves, M., Vickers, J.: Digital twin: mitigating unpredictable, undesirable emergent behavior in complex systems. In: Kahlen, J., Flumerfelt, S., Alves, A. (eds.) Transdisciplinary Perspectives on Complex Systems, pp. 85–113. Springer, Cham (2017). https://doi.org/10.1007/978-3-31938756-7_4 Sjarov, M., Kißkalt, D., Lechler, T., Selmaier, A., Franke, J.: Towards “design for interoperability” in the context of systems engineering. Procedia CIRP 96, 145–150 (2021). 8th CIRP Global Web Conference – Flexible Mass Customisation (CIRPe 2020). https://www.sciencedirect.com/sci ence/article/pii/S2212827121000913 Mumtaz, S., Alsohaily, A., Pang, Z., Rayes, A., Tsang, K.F., Rodriguez, J.: Massive internet of things for industrial applications: addressing wireless IIoT connectivity challenges and ecosystem fragmentation. IEEE Ind. Electron. Mag. 11(1), 28–33 (2017) Semeraro, C., Lezoche, M., Panetto, H., Dassisti, M.: Digital twin paradigm: a systematic literature review. Comput. Ind. 130, 103469 (2021). https://www.sciencedirect.com/science/article/pii/ S0166361521000762 Kaur, K., Selway, M., Grossmann, G., Stumptner, M., Johnston, A.: Towards an open-standards based framework for achieving condition-based predictive maintenance. In: Proceedings of the 8th International Conference on the Internet of Things, pp. 1–8 (2018) Platenius-Mohr, M., Malakuti, S., Grüner, S., Schmitt, J., Goldschmidt, T.: File-and API-based interoperability of digital twins by model transformation: an IIoT case study using Asset Administration Shell. Future Gener. Comput. Syst. 113, 94–105 (2020) Doubell, G.C., Kruger, K., Basson, A.H., Conradie, P.: The potential for digital twin applications in railway infrastructure management. In: Pinto, J.O.P., Kimpara, M.L.M., Reis, R.R., Seecharan, T., Upadhyaya, B.R., Amadi-Echendu, J. (eds.) WCEAM 2021. LNME, pp. 241–249. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-96794-9_22

Standards-Based Interoperable Digital Twin

735

The, Y.-L., Kuusk, A.G.: Aligning IIoT and ISA-95 to improve asset management in process industries. In: Crespo Márquez, A., Komljenovic, D., Amadi-Echendu, J. (eds.) WCEAM 2019. LNME, pp. 153–163. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-64228-0_14 Hietala, J., Jokinen, T. and Kaarmila, P.: Applying MIMOSA CCOM to modern web interfaces, VTT Technical Research Centre of Finland (2022). https://cris.vtt.fi/en/publications/applyingmimosa-ccom-to-modern-web-interfaces Berger, S., Grossmann, G., Stumptner, M., Schrefl, M.: Metamodel-based information integration at industrial scale. In: Petriu, D.C., Rouquette, N., Haugen, Ø. (eds.) MODELS 2010. LNCS, vol. 6395, pp. 153–167. Springer, Heidelberg (2010). https://doi.org/10.1007/978-3-642-161292_12 Selway, M., Grossman, G., Stumptner, M., Owen, K.R., Dexter, R.M.: Integration of visual contracts and model transformation for enhanced MDE development. In: Proceedings of MODELS 2018, pp. 415–425. ACM, New York (2018) Ochieng, P., Kyanda, S.: Large-scale ontology matching: state-of-the-art analysis. ACM Comput. Surv. 51(4), 75:1–75:35 (2018). https://doi.org/10.1145/3211871 Thiéblin, E., Haemmerlé, O., Hernandez, N., Trojahn, C.: Survey on complex ontology matching. Semant. Web 11(4), 689–727 (2020). https://doi.org/10.3233/SW-190366

Author Index

A Abdul-Nour, Georges, 508, 582 Adell, Edmea, 697 Ahmadi, Hamid, 521 Akyazi, Tugce, 343 Alberdi, Elisabete, 343 Almeida, Nuno, 34, 439, 615, 636, 645, 665, 697 Alonso, Pilar Jiménez, 87 Amadi-Echendu, Anthea P, 292 Amadi-Echendu, Joe, 292, 697 Amadi-Echendu, Joe E., 487 Antomarioni, Sara, 14 Aránguiz, Andrés, 323, 375 Arroyo, Alberto, 195 B Baharshahi, Mohammad, 219 Barandier, Pedro, 154 Barni, Andrea, 602 Bengtsson, Marcus, 571 Bevilacqua, Maurizio, 14 Bhuyan, Ranjan, 624 Biard, Gabrielle, 56 Biedermann, Hubert, 301 Bikaun, Tyler, 45 Blain, Michel, 261 Blancke, Olivier, 624 Brasen, Lucas Peter Høj, 312 Brooks, Richard, 162 Bustamante, Sergio, 195 C Campos, Jaime, 688 Candón, Eduardo, 76, 87, 205, 708

Capuzzimati, Claudio, 602 Cárcel-Carrasco, Javier, 3 Cardoso, Antonio João Marques, 154 Carnero, María Carmen, 3 Carriazo, José Antonio Marcos-Alberca, 145 Castellano, Eduardo, 273 Cavaiuolo, Ivan, 405 Chang, Janet, 561 Chiachio, Juan, 657 Chiachio, Manuel, 657 Ciarapica, Filippo Emanuele, 14 Costa, João, 463 Côté, Alain, 582 Crespo del Castillo, Adolfo, 162 Crespo Márquez, Adolfo, 67, 76, 145, 205, 416, 533, 590 D da Silva, Renan Favarão, 108 Davies, Rhys, 448 de Almeida, Nuno Marques, 544 de Croon, John, 448 de Souza, Gilberto Francisco Martha, 108 Di Pasquale, Lorenzo, 405 Diop, Issa, 508 Duque, Pablo, 323, 375 Duvivier, Paul, 624 E Emmanouilidis, Christos, 252 F Fernandez, Eduardo Candon, 416 Ferreira, Mónica Amaral, 636

© ISEAM 2023 A. Crespo Márquez et al. (Eds.): WCEAM 2022, LNME, pp. 737–739, 2023. https://doi.org/10.1007/978-3-031-25448-2

738 Fochesato, Donatella, 405 Fontana, Alessandro, 602 G Garcia, João, 636 Garcia, Jorge Merino, 561 Gaspar, Daniel, 463, 677, 697 Giliyana, San, 571 Gomes Morgado, João, 439 Gomes, Rui, 636 Gómez, Juan Fco., 87 González, Ana Gómez, 283 González, Antonio, 195 González, Manuel, 474 González-Prida, Vicente, 354, 590 Goti, Aitor, 343 Grilo, Elson, 463 Guillén, Antonio, 205, 354 Guillén, Antonio J., 76, 87 Gutsche, Katja, 385 H Hadjidemetriou, Georgios, 397 Hänninen, Saara, 602 Heaton, James, 397 Heikkilä, Eetu, 499 Hernández, Mauricio Rodríguez, 416 Herrera, Manuel, 162 Hodkiewicz, Melinda, 45 Hutabarat, Windo, 98 I Izquierdo, Juan, 273 J Johnston, Alan, 726 K Kallio, Topias, 172 Kans, Mirka, 688 Kaur, Karamjit, 726 Khuntia, Swasti R., 624 Komljenovic, Dragan, 508 Koskinen, Kari, 172 Koskinen, Kari T., 182 Krasny, Witold, 624 Kristjanpoller, Fredy, 354, 590 Kuganesan, Srijeyanthan, 118 Kuganesan, Thanansan, 118 L Laitinen, Jouko, 172, 182 Lapshe, Refiloe R., 487 Leite, Eduardo, 697

Author Index Leturiondo, Urko, 23, 76, 205, 283 Liu, Loretta, 162 Lopes, Odete, 463, 677 López, Antonio Guillen, 416 Lopez, Urko, 273 Lopez-Campos, Monica, 708 Lucantoni, Laura, 14 Luciano, Margherita, 405 M Mäkiaho, Teemu, 172 Maletiˇc, Damjan, 615 Manana, Mario, 195 Männistö, Vesa, 428 Marcos, José A., 76 Martínez-Corral, Aurora, 3 Mathew, Joseph, 726 Maurice, Richard, 195 Meango, Toualith Jean-Marc, 582 Merino, Jorge, 162 Metso, Lasse, 130, 716 Miranda, Alexandre, 154 Mofrad, Meysam Esmaeilzadeh, 521 Montiel, Virginia, 229 Moretti, Nicola, 397, 561 Mvele, Jedial O., 487 N Nieto, Estela, 283 Nour, Georges Abdul, 56 Nourelfath, Mustapha, 261 Ntoyanto-Tyatyantsi, Nonceba, 292 O Okeyia, Charles, 544 Oliveira, Carlos Sousa, 636 Oliveira, Serafim, 677 Oyarbide, Aitor, 343 Oyekan, John, 98 Öz, Emrehan, 499 P Parlikad, Ajith, 397, 561 Parlikad, Ajith Kumar, 162 Parra, Carlos, 323, 354, 375, 590 Parra, Jorge, 590 Patrício, Hugo, 439 Payette, Mathieu, 582 Peiravi, Abdossaber, 261 Picado Arguello, B., 240 Pinto, Cláudia, 636 Pirotta, Marco, 602 Pizarro, Félix, 323, 375 Poças Martins, João, 439

Author Index Polenghi, Adalberto, 405, 428 Poulter, Karen, 162 Prabhu, Vinayak, 98 R Räikkönen, Minna, 602 Ramezani, Saeed, 219 Rezvani, Seyed M. H. S., 615, 665 Rezvani, Seyedi, 636 Ribeiro, Filipe, 636 Rica, Elena, 252 Roda, Irene, 405, 428 Rodrigo-Muñoz, Francisco, 533 Rodrigues, Carlos, 677 Rodríguez, Ángel, 23 Rodríguez, Juan, 590 S Saleh, Ali, 657 Salonen, Antti, 366, 571 Salvado, Filipa, 636, 645 Sánchez-Herguedas, Antonio, 533 Sabapathipillai, Gowrishankar, 118 Sasidharan, Manu, 162, 397 Schenkel, Erik, 624 Schlögel, Santina, 385 Schmiedbauer, Oliver, 301 Schooling, Jennifer, 397 Schwabe, Oliver, 34 Sedghi, Abolfazl, 521 Seecharan, Turuna S., 333 Selway, Matt, 726 Silva, Maria João Falcão, 615, 636, 645, 665 Sonntag, Sören Dominik, 98 Stegmaier, Raúl, 708 Stewart, Michael, 45 Stumptner, Markus, 726

739 T Tambo, Torben, 312 Thenent, Nils E., 130, 716 Thibault, Denis, 261 Tila, Tahrim Zaman, 333 Tiusanen, Risto, 499 Tiwari, Ashutosh, 98 Toikka, Tauno, 182 Torcato, Joana, 697 Tordi, Isabella, 405 Turner, Chris, 98 U Uribetxebarria, Jone, 23 Uusitalo, Teuvo, 602 V Vainio, Henri, 172 Välisalo, Tero, 499 Vega, Emanuel, 323, 375 Velásquez, Matías, 354 Vieira, João, 439, 697 Viveros, Pablo, 354, 590 W Wijnia, Ype, 448 X Xie, Xiang, 561 Y Yousofi Tezerjan, Mostafa, 219 Z Zanjani, Masoumeh Kazemi, 261 Zghal, Fatma, 624 Zubizarreta, Ainhoa, 23