Information and Communication Technologies of Ecuador (TIC.EC) [1st ed. 2020] 978-3-030-35739-9, 978-3-030-35740-5

Table of contents :
Front Matter ....Pages i-xiii
Front Matter ....Pages 1-1
Hearing Loss and Its Association with Clinical Practice at Dental University Students Through Mobile APP: A Longitudinal Study (Juan-Carlos Cobos-Torres, Ronald Ramos, Juan Carlos Ortega Castro, Miriam Fernanda Ortega Lopez)....Pages 3-17
Front Matter ....Pages 19-19
A Bayesian Network Approach to Identity Climate Teleconnections Within Homogeneous Precipitation Regions in Ecuador (Renato Ávila, Daniela Ballari)....Pages 21-35
Front Matter ....Pages 37-37
Design and Implementation of an Automatic System for the Monitoring and Monitoring of a Prototype Refrigeration Plant with Parallel Compressors (Elsy del Rocio Villamar Garcés, Jorge Luis Gonzalez Murillo, Jacinto Gabriel Lino Sánchez, Monica Karina Jaramillo Infante, Oswaldo Villamar Chele)....Pages 39-52
Design of Emergency Call Record Support System Applying Natural Language Processing Techniques (Andrea Trujillo, Marcos Orellana, María Inés Acosta)....Pages 53-65
Integrating ISA-95 and IEC-61499 for Distributed Control System Monitoring (Jairo D. Llamuca, Carlos A. Garcia, Jose E. Naranjo, Cesar Rosero, Edison Alvarez-M, Marcelo V. Garcia)....Pages 66-80
Front Matter ....Pages 81-81
Genomic Databases Exploration Using Conceptual Models (C. Vanessa Solis, P. Ana León, Oscar Pastor Lopez)....Pages 83-96
Management of Humanitarian Logistics in the Stages Prior to Natural Disasters in Canton Ambato, Ecuador (Santiago Velastegui, Rosa Galleguillos-Pozo, Cesar Rosero, Marcelo V. Garcia)....Pages 97-108
Variability Features in Building Approaches for Context-Aware Mobile Applications (Estevan Gómez-Torres, Cecilia Challiol, Silvia E. Gordillo)....Pages 109-123
Wheelchair Controlled by Eye Movement Using Raspberry Pi for ALS Patients (Jorge Buele, José Varela-Aldás, Franklin W. Salazar, Angel Soria, Víctor H. Andaluz)....Pages 124-136
A Data Quality Model for AAL Systems (Lenin Erazo-Garzon, Jean Erraez, Lourdes Illescas-Peña, Priscila Cedillo)....Pages 137-152
A Novel Approach to Detect Missing Values Patterns in Time Series Data (Juan-Fernando Lima, Patricia Ortega-Chasi, Marcos Orellana Cordero)....Pages 153-166
Detection and Classification of Urban Actors Through TensorFlow with an Android Device (Andres Campoverde, Gabriel Barros)....Pages 167-181
Support Vector Regression to Downscaling Climate Big Data: An Application for Precipitation and Temperature Future Projection Assessment (Stalin Jimenez, Alex Aviles, Luciano Galán, Andrés Flores, Carlos Matovelle, Cristian Vintimilla)....Pages 182-193
Prediction of Imports of Household Appliances in Ecuador Using LSTM Networks (Andrés Tello, Ismael Izquierdo, Gustavo Pacheco, Paúl Vanegas)....Pages 194-207
Front Matter ....Pages 209-209
Integrating Corporate Social Responsibility to a Process-Based Cost Analysis System (Erik Sigcha, Lorena Siguenza-Guzman, Villie Morocho)....Pages 211-224
Models, Guidelines and Trends for Process Quality Management: A Literature Review (Anthony Guacho, Ana Jara, Erik Sigcha, Rodrigo Arcentales-Carrion, Lorena Siguenza-Guzman)....Pages 225-238
Modeling and Ranking External Stakeholders in Open Source Software Adoption (Lucía Méndez Tapia, Juan Pablo Carvallo)....Pages 239-254
Front Matter ....Pages 255-255
Landslide Monitoring System Through Wireless Sensor Network Using RTK Technique: Case of Study Basin of Loja City (Darío Valarezo, Gabriela Mendieta, Manuel Quiñones-Cuenca, Belizario Zarate, Verónica Quiñonez, John Soto)....Pages 257-269
Front Matter ....Pages 271-271
A Design for a Secure Malware Laboratory (Xavier Riofrío, Fernando Salinas-Herrera, David Galindo)....Pages 273-286
Back Matter ....Pages 287-288

Citation preview

Advances in Intelligent Systems and Computing 1099

Efraín Fonseca C Germania Rodríguez Morales Marcos Orellana Cordero Miguel Botto-Tobar Esteban Crespo Martínez Andrés Patiño León   Editors

Information and Communication Technologies of Ecuador (TIC.EC)

Advances in Intelligent Systems and Computing Volume 1099

Series Editor Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Advisory Editors Nikhil R. Pal, Indian Statistical Institute, Kolkata, India Rafael Bello Perez, Faculty of Mathematics, Physics and Computing, Universidad Central de Las Villas, Santa Clara, Cuba Emilio S. Corchado, University of Salamanca, Salamanca, Spain Hani Hagras, School of Computer Science and Electronic Engineering, University of Essex, Colchester, UK László T. Kóczy, Department of Automation, Széchenyi István University, Gyor, Hungary Vladik Kreinovich, Department of Computer Science, University of Texas at El Paso, El Paso, TX, USA Chin-Teng Lin, Department of Electrical Engineering, National Chiao Tung University, Hsinchu, Taiwan Jie Lu, Faculty of Engineering and Information Technology, University of Technology Sydney, Sydney, NSW, Australia Patricia Melin, Graduate Program of Computer Science, Tijuana Institute of Technology, Tijuana, Mexico Nadia Nedjah, Department of Electronics Engineering, University of Rio de Janeiro, Rio de Janeiro, Brazil Ngoc Thanh Nguyen , Faculty of Computer Science and Management, Wrocław University of Technology, Wrocław, Poland Jun Wang, Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Shatin, Hong Kong

The series “Advances in Intelligent Systems and Computing” contains publications on theory, applications, and design methods of Intelligent Systems and Intelligent Computing. Virtually all disciplines such as engineering, natural sciences, computer and information science, ICT, economics, business, e-commerce, environment, healthcare, life science are covered. The list of topics spans all the areas of modern intelligent systems and computing such as: computational intelligence, soft computing including neural networks, fuzzy systems, evolutionary computing and the fusion of these paradigms, social intelligence, ambient intelligence, computational neuroscience, artificial life, virtual worlds and society, cognitive science and systems, Perception and Vision, DNA and immune based systems, self-organizing and adaptive systems, e-Learning and teaching, human-centered and human-centric computing, recommender systems, intelligent control, robotics and mechatronics including human-machine teaming, knowledge-based paradigms, learning paradigms, machine ethics, intelligent data analysis, knowledge management, intelligent agents, intelligent decision making and support, intelligent network security, trust management, interactive entertainment, Web intelligence and multimedia. The publications within “Advances in Intelligent Systems and Computing” are primarily proceedings of important conferences, symposia and congresses. They cover significant recent developments in the field, both of a foundational and applicable character. An important characteristic feature of the series is the short publication time and world-wide distribution. This permits a rapid and broad dissemination of research results. ** Indexing: The books of this series are submitted to ISI Proceedings, EI-Compendex, DBLP, SCOPUS, Google Scholar and Springerlink **

More information about this series at http://www.springer.com/series/11156

Efraín Fonseca C Germania Rodríguez Morales Marcos Orellana Cordero Miguel Botto-Tobar Esteban Crespo Martínez Andrés Patiño León •

•

•

•

•

Editors

Information and Communication Technologies of Ecuador (TIC.EC)

123

Editors Efraín Fonseca C Universidad de las Fuerzas Armadas ESPE Sangolqui, Ecuador Marcos Orellana Cordero Escuela de Ciencias de la Computación Universidad del Azuay Cuenca, Ecuador Esteban Crespo Martínez Universidad del Azuay Cuenca, Ecuador

Germania Rodríguez Morales Departamento de Ciencias de la Computación y Electrónica San Cayetano Alto Universidad Técnica Particular de Loja Loja, Ecuador Miguel Botto-Tobar Department of Mathematics and Computer Eindhoven University of Technology Eindhoven, Noord-Brabant The Netherlands Andrés Patiño León Universidad del Azuay Cuenca, Ecuador

ISSN 2194-5357 ISSN 2194-5365 (electronic) Advances in Intelligent Systems and Computing ISBN 978-3-030-35739-9 ISBN 978-3-030-35740-5 (eBook) https://doi.org/10.1007/978-3-030-35740-5 © Springer Nature Switzerland AG 2020 This work is subject to copyright. All rights are reserved 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

Preface

The Sixth Conference on Information and Communication Technologies “TIC.EC” will be held in Cuenca-Ecuador from November 27 to 29, 2019. This academic event is considered as one of the most important conferences about ICT in Ecuador, as it brings scholars and practitioners from the country and abroad to discuss the development, issues, and projections of the use of information and communication technologies in multiples fields of application. In 2019, the “TIC.EC” conference was organized by Universidad del Azuay (UDA) and its Engineering School and the Ecuadorian Corporation for the Development of Research and Academia (CEDIA). The content of this volume is related to the following subjects: • • • • • • • • • •

Software engineering Security Data Networks Architecture Applied ICTs Technological entrepreneurship Links between research and industry High-impact innovation Knowledge management and intellectual property

In its 2019 edition, the TIC.EC conference received 75 submissions in English from 203 authors coming from nine different countries. All these papers were peer-reviewed by the TIC.EC 2019 Program Committee consisting of 52 high-quality researchers coming from 23 different countries. To assure a high-quality and thoughtful review process, we assigned each paper at least three reviewers. Based on the results of the peer reviews, 19 full papers were accepted (they were written in English), resulting in a 25% acceptance rate, which was within our goal.

v

vi

Preface

We would like to express our sincere gratitude to the invited speakers for their inspirational talks, to the authors for submitting their work to this conference, and the reviewers for sharing their experience during the selection process. November 2019

Efrain Fonseca C Germania Rodríguez Morales Marcos Orellana Cordero Miguel Botto-Tobar Esteban Crespo Martínez Andrés Patiño León Gabriela Parra Robles

Organization

Honorary Committee Nicolay Samaniego Erazo (Presidente) Francisco Salgado Arteaga Juan Pablo Carvallo Vega (Director)

CEDIA, Ecuador Rector Universidad del Azuay, Ecuador Ejecutivo CEDIA, Ecuador

Director Committee Efraín Fonseca Germania Rodríguez Morales Marcos Orellana Cordero

Universidad de las Fuerzas Armadas, Ecuador Universidad Técnica Particular de Loja, Ecuador Universidad del Azuay, Ecuador

Organizing Committee Esteban Crespo Andrés Patiño Alejandra Zarama Cristian Alvarracín Priscila Calderón Belén Valdez Galia Rivas Toral Francisco Toral Fernanda Chica Erick Brito Ximena Lazo Gabriela Parra Robles

Universidad del Azuay, Universidad del Azuay, Universidad del Azuay, Universidad del Azuay, Universidad del Azuay, Universidad del Azuay, CEDIA, Ecuador CEDIA, Ecuador CEDIA, Ecuador CEDIA, Ecuador CEDIA, Ecuador CEDIA, Ecuador

Ecuador Ecuador Ecuador Ecuador Ecuador Ecuador

vii

viii

Organization

Program Committee Ángel Alberto Magreñán Belen Bermejo Angela Díaz Cadena Bilal Alatas Camelia Delcea Carlos Juiz Carme Quer Antonio Franco Chi-Hua Chen Álvaro Jiménez Sánchez Claudia P. Ayala Clemente Izurieta Coral Calero Diego Brandao Diego Riofrío-Luzcando Diego Alfonso Almeida Galárraga Avireni Srinivasulu Noor Zaman Eduardo Almentero Elena Avram Emil Pricop Fabio Castro-Llanos Firas A. Raheem Francisco J. Lopez-Pellicer Gerardo Matturro Gleison Santos Ibraheem Kasim Ibraheem Samanta Cueva Ioan Viorel Banu Isabel Sofia Sousa Brito Janneth Chicaiza John W. Castro Sandra Sanchez-Gordon Jose Tenreiro Machado Juan Herrera Lidia Lopez Marcelo Zambrano Maria Hallo

Universidad de La Rioja, Spain University of the Balearic Islands, Spain Universitat de Valencia, Spain Firar University, Spain Bucharest University of Economic Studies, Romania University of the Balearic Islands Universitat Politècnica de Catalunya, Spain Escuela Politecnica Nacional, Ecuador Fuzhou University Universidad Técnica de Ambato, Ecuador Universitat Politècnica de Catalunya, Spain Montana State University, USA Universidad de Castilla-La Mancha, Spain Centro Federal de Ensino Tecnologico Celso Suckow da Fonseca, Brazil Universidad Internacional SEK del Ecuador SENESCYT, Ecuador JECRC University, Jaipur-303905, Rajasthan (State), India Taylor’s University, Selangor, Malaysia, Malaysia UFRRJ, Brazil Gheorghe Asachi Technical University of Iasi Petroleum-Gas University of Ploiesti, Romania CIAT University of Technology-Iraq, Iraq University of Zaragoza Universidad ORT Uruguay, Uruguay UNIRIO, Brazil Baghdad University/College of Engineering, Iraq Universidad Técnica Particular de Loja, Ecuador Gheorghe Asachi Technical University of Iasi, Romania Instituto Politécnico de Beja, Portugal UTPL, Ecuador Universidad de Atacama, Chile Escuela Politecnica Nacional, Ecuador ISEP, Portugal Escuela Politécnica Nacional, Ecuador Universitat Politècnica de Catalunya, Spain Universidad Técnica del Norte, Ecuador Escuela Politécnica Nacional, Ecuador

Organization

Miguel Botto-Tobar Tania Calle-Jimenez Mohamed Kamel Nathaly Orozco Pablo Palacios Pablo Torres-Carrion Patricio Galdames Roberto Murphy Rommel Torres Jerwin Prabu A. Tatiana Gualotuña Xavier Franch María Perez Yves Rybarczyk

Sponsoring Institutions

Universidad del Azuay https://www.uazuay.edu.ec/

CEDIA

https://www.cedia.edu.ec/es/

ix

Eindhoven University of Technology, the Netherlands Escuela Politecnica Nacional, Ecuador Concordia University, Canada Universidad de las Américas, UDLA, Ecuador Universidad de Chile Universidad Tecnica Particular de Loja, Ecuador Universidad del Bio-Bio, Chile Instituto Nacional de Astrofísica, Óptica y Electrónica (INAOE), Mexico UTPL, Ecuador Bharati Robotic Systems India Pvt Ltd., India Universidad de las Fuerzas Armadas ESPE, Ecuador Universitat Politècnica de Catalunya, Spain Escuela Politécnica Nacional, Ecuador Universidade NOVA de Lisboa, Portugal

Contents

Applied TIC’s Hearing Loss and Its Association with Clinical Practice at Dental University Students Through Mobile APP: A Longitudinal Study . . . . . Juan-Carlos Cobos-Torres, Ronald Ramos, Juan Carlos Ortega Castro, and Miriam Fernanda Ortega Lopez

3

Data A Bayesian Network Approach to Identity Climate Teleconnections Within Homogeneous Precipitation Regions in Ecuador . . . . . . . . . . . . Renato Ávila and Daniela Ballari

21

High Impact Innovation Design and Implementation of an Automatic System for the Monitoring and Monitoring of a Prototype Refrigeration Plant with Parallel Compressors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Elsy del Rocio Villamar Garcés, Jorge Luis Gonzalez Murillo, Jacinto Gabriel Lino Sánchez, Monica Karina Jaramillo Infante, and Oswaldo Villamar Chele Design of Emergency Call Record Support System Applying Natural Language Processing Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrea Trujillo, Marcos Orellana, and María Inés Acosta Integrating ISA-95 and IEC-61499 for Distributed Control System Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jairo D. Llamuca, Carlos A. Garcia, Jose E. Naranjo, Cesar Rosero, Edison Alvarez-M, and Marcelo V. Garcia

39

53

66

xi

xii

Contents

Knowledge Management and Intellectual Property Genomic Databases Exploration Using Conceptual Models . . . . . . . . . . C. Vanessa Solis, P. Ana León, and Oscar Pastor Lopez Management of Humanitarian Logistics in the Stages Prior to Natural Disasters in Canton Ambato, Ecuador . . . . . . . . . . . . . . . . . Santiago Velastegui, Rosa Galleguillos-Pozo, Cesar Rosero, and Marcelo V. Garcia

83

97

Variability Features in Building Approaches for Context-Aware Mobile Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 Estevan Gómez-Torres, Cecilia Challiol, and Silvia E. Gordillo Wheelchair Controlled by Eye Movement Using Raspberry Pi for ALS Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 Jorge Buele, José Varela-Aldás, Franklin W. Salazar, Angel Soria, and Víctor H. Andaluz A Data Quality Model for AAL Systems . . . . . . . . . . . . . . . . . . . . . . . . 137 Lenin Erazo-Garzon, Jean Erraez, Lourdes Illescas-Peña, and Priscila Cedillo A Novel Approach to Detect Missing Values Patterns in Time Series Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Juan-Fernando Lima, Patricia Ortega-Chasi, and Marcos Orellana Cordero Detection and Classification of Urban Actors Through TensorFlow with an Android Device . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Andres Campoverde and Gabriel Barros Support Vector Regression to Downscaling Climate Big Data: An Application for Precipitation and Temperature Future Projection Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 Stalin Jimenez, Alex Aviles, Luciano Galán, Andrés Flores, Carlos Matovelle, and Cristian Vintimilla Prediction of Imports of Household Appliances in Ecuador Using LSTM Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 Andrés Tello, Ismael Izquierdo, Gustavo Pacheco, and Paúl Vanegas Linkage of Research with Industry Integrating Corporate Social Responsibility to a Process-Based Cost Analysis System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 Erik Sigcha, Lorena Siguenza-Guzman, and Villie Morocho

Contents

xiii

Models, Guidelines and Trends for Process Quality Management: A Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 Anthony Guacho, Ana Jara, Erik Sigcha, Rodrigo Arcentales-Carrion, and Lorena Siguenza-Guzman Modeling and Ranking External Stakeholders in Open Source Software Adoption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239 Lucía Méndez Tapia and Juan Pablo Carvallo Networks Landslide Monitoring System Through Wireless Sensor Network Using RTK Technique: Case of Study Basin of Loja City . . . . . . . . . . . 257 Darío Valarezo, Gabriela Mendieta, Manuel Quiñones-Cuenca, Belizario Zarate, Verónica Quiñonez, and John Soto Security A Design for a Secure Malware Laboratory . . . . . . . . . . . . . . . . . . . . . 273 Xavier Riofrío, Fernando Salinas-Herrera, and David Galindo Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

Applied TIC’s

Hearing Loss and Its Association with Clinical Practice at Dental University Students Through Mobile APP: A Longitudinal Study Juan-Carlos Cobos-Torres(B) , Ronald Ramos , Juan Carlos Ortega Castro , and Miriam Fernanda Ortega Lopez Catholic University of Cuenca, Cuenca 010107, Ecuador [email protected]

Abstract. In Ecuador, hearing loss ranks third amongst the most common chronic physical condition. Dentists are health professionals who are exposed to harmful noises from equipment in their clinics and may suffer hearing loss. This research seeks to prove if the persistent high-frequency sounds produced by dental equipment could cause hearing decrement among dental students during the fifth semester of their studies. The present study was conducted on 70 students from the Faculty of Odontology belonging to the Academic Unit of Health and Welfare at the University of Cuenca in their fifth cycle during the second semester 2016–2017. In a soundproofed room, audiometry was done on ten students per day through the application “hearing test” by Marcin Malaski, 1.1.3 version. Based on studies, they revealed that 40% of the students who did not have any hearing loss started to experience some form of hearing impairment. The 10% of students that had a mild hearing loss reached a moderate hearing loss. In the case of gender, 5% of male students changed their hearing level from a mild hearing loss to a moderate after their clinic activities. Among women 1.18% it had expressed a change from a mild hearing loss to a moderate. According to the project results, the noise exposure for five or six months in university dental clinic may cause hearing impairments. Therefore, it is highly recommended the use of hearing protectors for students as well as teachers. Keywords: Noise · Turbine · Clinic · Hearing loss · Mobile health

1 Introduction 1.1 A Subsection Sample The ear is a sensory organ that not only helps in hearing but also to maintain the balance of the human body. Under favorable and normal conditions, it helps the interaction of the individual with the environment and society. The brain processes in parallel in the cortex auditory signals: one is the processing of the sign content “what”, and the other is the spatial processing “where” [1]. Hence the vital importance of hearing in exchanging information as well as spatial processing. © Springer Nature Switzerland AG 2020 E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 3–17, 2020. https://doi.org/10.1007/978-3-030-35740-5_1

4

J.-C. Cobos-Torres et al.

The World Health Organization (WHO) estimates that 466 million people around the world have deafness or disabling hearing loss, which represents 6% of the population. Half of these audition loss cases could be prevented by applying public health measures [2]. In Ecuador the National Council for the Equality of Persons with Disabilities shows the next figures: 289366 disabled people aged 18 to 65 years old (11.34% representing people with hearing impairments); in the province of Azuay there are 16823 disabled people (10.71% has hearing impairment) and in the parish of Cuenca 12245 there are persons with disability (11.26% has hearing disabilities). The statistical data for hearing impairment in Ecuador are high since it ranks as third in place with great prevalence among the population with 13.31 [3]. Ecuador applies the Safety and Health Regulation for Workers and Working Environment Improvement. Ordinance Number 2393 states that all companies should guarantee to all the workers a suitable and conducive working environment to develop their physical and mental abilities. The same regulation regarding with “Occupational Noise” establishes that the permissible exposure limit for continuous or intermittent noises in a working day is as follows: 85 dB – 8 h, 90 dB – 4 h, 100 dB – 1 h, 110 dB – 0.25 h, and 115 dB – 0.125 h. Odontologists in everyday clinical practice are constantly exposed to potentially harmful sounds that are generated by the air turbine; such exposure does not start at the beginning of their professional career, but from the earliest pre-professional training, which has the effect of increasing the exposure time [4]. Hearing range varies between individuals, and it depends on factors such as volume and tone. Human beings have auditory perception from 20 Hz to 20 kHz in frequency and more than 0 dB up to 120 dB in sound pressure level. The most sensitive area differs from 100 Hz to 8 kHz for human speech [5]. A sound at about 125 dB depending on its frequency reaches to threshold pain. In addition, if the frequency is measured, 90 dB may already result inner ear damage at 3 to 6 kHz range. All this information is shown in Fig. 1.

Fig. 1. Hearing range (6) vs. high-speed air turbinehes.

Hearing Loss and Its Association with Clinical Practice

5

According to [7] the United States Department of Health, a study carried out in 1959 confirms that the noise coming from the turbines is above 84 dB. This is in a frequency range from 4.8 kHz to 9.6 kHz without being in contact with the tooth surface. Five decades later dental turbines have improved substantially; however, it is important to realize that old turbines are still used in many disadvantaged countries. A current study [8] shows in its results that the maximum sound level was 85.8 dB in dental offices, and 92.0 dB in laboratories. In the dental clinics, the highest noise was produced by ultrasonic scaler (85.8 dB), and the lowest noise (49.7 dB) by a high-volume aspirator. While in the laboratory, the highest noise was caused during grinding by the stonecutter (92.0 dB), and the lowest by the denture-polishing unit (41.0 dB). As for the sound of turbines, without being in contact with tooth surface is about 62 dB and 82 dB. The daily dental practice requires the use of air turbines in each of its areas as in oral rehabilitation for tooth decay treating. Also, in the selective grinding, endodontics allows for the opening of the cavity and its subsequent instrumentation among other areas. With all this, the working times or high noise exposure is 45 min. Therefore, it can now be affirmed that in the operative dentistry area, the level of sound to operator´s ear height is 83.13 dB; in the dental prosthesis placement procedure, the level of sound to the operator is 81 dB; and in endodontics therapy, to a lesser degree, the level of sound to the operator is 65.57 dB [9].

2 Related Work The dentistry specialists in their everyday practice are exposed to sounds potentially harmful. Equipment such as turbines, air compressors, among others cause high noise levels in clinics and dental offices. On this point, any profession can have a risk factor to a greater or lesser degree. In the particular case of deafness, it can end up being a cause of an inability to work. Therefore, [10] provides a set of strategies and methods of diagnoses to detect deafness at an early stage. Only thus, it is possible to have suitable conduct to slow the disease progression. Additionally, authors state that high-frequency noises are extremely harmful than the sound of low frequencies. Also, it should be noted that prolonged exposure to the noise causes temporary or permanent changes in the auditory perception [11]. For these reasons, future dental professionals should get familiar with topics related to Occupational Health. Thus, [12] demonstrates the absence of coherent contents about Occupational Health into the curriculum of a Venezuelan University. There are different kinds of occupational hearing loss. Concerning dentistry, it would be caused by exposure to noise. For this reason, there are countless studies carried out in university dental clinics. Many reflect disagreements with their conclusions, although they always point out that the lack of protection causes hearing loss prevalence while the staff is working. Some transversal approaches have been made through surveys and audiometry where the subjectivity of the survey respondent is involved. Then, [13] in a transversal study carried out through surveys and a sample of 135 students found out that 95.6 have not suffered deafness. Another study conducted with a teaching staff [14] concludes that educational professionals could suffer hearing loss over time. Conversely, [15] presents a transversal research performed through a survey and audiometry based on 82 individuals which included dentists, interns, and dental assistants. Results showed that

6

J.-C. Cobos-Torres et al.

40% of the population had sensorineural hearing loss and acoustic trauma. Additionally, the measurement of the levels of noise was performed using sonography. In another transversal study, [16] demonstrates that 38 Saudi dentists belonging to different areas and who were randomly selected about 15.8% of dentists and 2.6% of the control group had some auditory loss. Moreover, [17] compares pure-tone audiometric threshold of dental assistants and dentistry students. The results show that clinical assistants who regularly used high-speed dental pieces had a worse audition than other study group members. Additionally, the data shows that safeguard implementation helps to reduce the occupational auditory loss prevalence. Nowadays, the close connection between tinnitus and audition loss has been demonstrated [18] through a brief questionnaire to all the members of the Dental Association of Oklahoma. It is demonstrated the risk, tinnitus prevalence, and hearing loss caused by noise to most dental professionals. Furthermore, [19] presents a study with sound spectrum generated by turbines ranging from audible up to ultrasonic frequencies (0–70 kHz). This study relies on the fact that ultrasonic sound due to turbines may be harmful. These measurements were performed using a microphone and a sonometer. Four main picks were 5.6 kHz (±0.73) in the audible range, and 20.1 kHz (±2.16), 35.7 kHz (±2.56), and 46.5 kHz (±0.71) in the ultrasound range. Authors tell that in the audible range, the pick corresponds to the fundamental frequency emitted by the turbine. Its amplitude is lower or equal to the risk limit of hearing loss of 85 dB (Fig. 1), but during the cutting, the generated noise level reaches 94 dB. Finally, [20] from 2003 to 2012 audiograms were applied to 1.4 million workers (8702 within medical care and social assistance). The 19%, general hearing loss prevalence was found in this sector. In the subsectors case as Medical laboratories and other doctors´ offices, the prevalence was 31% and 24%, respectively. Besides, [21] presents a longitudinal study in a ten years range which collects data about the effects on health among dentists through a questionnaire and audiometric test in both ears. This showed a hearing loss of 4 kHz on the left ear which shows a possible occupational acoustic trauma. Therefore, this research work arises from the conclusions mentioned above. Several authors suggest the need for an in-depth study to provide accurate details about the risks of occupational hearing loss. It is imperative to establish if the facts as described before are related to the odontological practice. It is clear that a longitudinal study on odontologists during ten years will ensure more compelling information, but it may also be affected by different factors since dentists could have experienced different external exposures during their professional practice. These exposures in a long-term time may also have caused hearing loss. Consequently, a short-time longitudinal study carried out on the Dentistry Students in their fifth semester aims to confirm these researches. The less likely it is that in a shorter period of time may exist external influences different from dental practice to cause deafness. A mobile application is used for the measurements.

3 Subjects and Methods The goal of this study is to verify if there is hearing impairment among Dentistry Students exposed to loud noises during a semester of studies. When arriving, the student reads

Hearing Loss and Its Association with Clinical Practice

7

and signs the agreement. Then, the initial diagnostic record is filled, after responding a brief test, the audiometry is carried out. This is a basic test which is used to determine the degree of hearing loss in connection with the frequency and sound amplitude. To begin, the application will be adapted to the student’s auditive perception and the external environment. After this, the student will listen a vast amount sounds in different amplitude (decibel) and frequencies (hertz). The student points out his/her auditory perception during the test (Fig. 2), and when the test has been finished, the information is stored. For comparisons, all procedure repeats after five months once the course of studies has finished. The application used is called “Hearing Test” version 1.1.3 by Marcin Masalski www.e-audiologia.pl/HearingTest.com.

Fig. 2. Experiment configuration in soundproof room.

4 Population A longitudinal study was carried out during the second semester in the school year 2016 to 2017. Students from the Faculty of Odontology belonging to the Academic Unit of Health and Welfare of the Catholic University of Cuenca were invited to participate in this research. The hearing test was performed in a soundproof room (soundproofed and sound dampening) with the students in their fifth semester who have started their clinical activity. Ten students per day were chosen a sample of seventy which included forty-seven women and twenty-three men corresponding to 67% and 32% respectively. To avoid problems due to possible changes in thresholds, the hearing tests were only performed on students who explained in advance that they had not been exposed to noise in the previous 24 h (recreational noise, earphones, concerts among others), or noises caused by dental laboratory devices.

5 Results Several audiometric assessments at different frequencies in both ears were performed. All the data were tabulated and analyzed. Firstly, a review of possible anomalous or

8

J.-C. Cobos-Torres et al.

incorrect values was done as a result of the students and their untrusted responses to the audiometric test. Even in the event that they had been exposed to noise, this fact would not have been reported. Of a total of four hearing tests series (measurement frequency 125 Hz, 250 Hz, 500 Hz, 1 kHz, 2 kHz, 4 kHz, 8 kHz, 12 kHz, 16 kHz, 18 kHz), a series in duplicate for each ear, since two series cover the beginning, and the other two the end of the semester. A total of 4800 samples were analyzed. Since it becomes impossible that students have improved their hearing at the end of the course of study, a weighing was carried out to find measures that are not consistent. Twelve samples were excluded, thus the sample included 38 women and 20 men which corresponds to 66% and 34% of the total sample. The students of the fifth semester who start their clinical activities were chosen. According to the degree of hearing loss, there may be a greater or lesser ability to hear the sounds. OMS classify it in five scales (Table 1). Table 1. Deafness levels according to the OMS. Level of hearing difficulty Decibels Mild

26–40

Moderate

41–55

Moderately severe

56–70

Severe

71–91

Profound

>91

Audition loss can be due to genetic causes, delivery complications, some infectious illnesses, chronic ear infections, use of certain drugs, the exposure to excessive noise, and aging [22]. On the whole, under this classification it is seen that at the beginning of the study of a total of 1160 measurements (ten series in all the frequency measured ranges) 5.17% did not have any auditory deficiency, 93.36% has a mild hearing loss and 1.47% moderate hearing loss. At the end of the study (the same number of measurements and frequencies), only 1.12% did not have any hearing impairment, 96.36% had a mild hearing loss, and 2.41% had a moderate hearing impairment, which is shown in Fig. 3. At the beginning of the study, it is further observed that among the male gender – a total of 400 measurements (ten series in all frequency-measured ranges) – 6.75% does not have any hearing impairment, 93.25% has a mild hearing loss, and no male student had moderate hearing loss. At the end of the study (the same number of measurements and frequencies), only 1.25% does not have hearing impairment; 98.25% had a mild hearing loss, and 0.50% a moderate hearing loss. In contrast to this, with the trends in the female gender at the beginning of the study of a total of 760 measurements (the same number of measurements and frequencies), 4.34% did not have hearing loss, 93.42% had a mild hearing impairment, and 2.24% had a moderate hearing loss. At the end of the study (the same

Hearing Loss and Its Association with Clinical Practice

9

Fig. 3. Series tabulated.

number of measurements and frequencies), only 1.5% displayed not experiencing hearing impairment, 95.53% had a mild hearing loss, and 3.42% had a moderate hearing loss. All this information is shown in Fig. 4.

Fig. 4. Tabulate series by gender.

Measurement calculation, confidence interval, variance, standard deviation, and minimum and maximum values have been calculated for each series. This is shown in Table 2. All the cases in which the variation is above 10% are set forth below. At the beginning of the study, results show that at a frequency of 125 Hz in the case of the right and left ear, the average of the threshold of sensitivity was 98.97 dB and 29.91 dB respectively. At the end of the study, the average of the threshold sensitivity was 33.97 dB and 34.83 dB which demonstrates its increase of about 17% in both ears. At 1000 Hz the

10

J.-C. Cobos-Torres et al.

average of the threshold sensitivity on the right and left ear was 30.86 dB and 30.25 dB, respectively. When finishing the study, the average of threshold sensibility was 35.69 dB and 34.57 dB with an increase of 15% approximately. At 4000 Hz the average of the threshold sensitivity on the right and left ear was 31.72% and 33.79% with an increase of 10% and 4% for each ear. At 8000 Hz on the right and left ear, the average of threshold sensibility was 29.91 dB and 30.95 dB respectively. When finishing the study, the average was 35.00 dB and 35.17 dB with an increase in the threshold of 17% and 14% for each ear. Additionally, it is important to note that the variance of the sample decreases in almost all the frequencies – though this is not only in the high frequencies from 12 kHz. Likewise, the minimum and maximum thresholds of the samples vary in the frequencies of 125 Hz, 1 kHz, 2 kHz, 4 kHz, and 8 kHz. Table 2. Statistical analysis of the audiometry tests to the different frequencies at the beginning and end of the semester.

N Mean 95% CI Variance SD Min Max

N Mean 95% CI Variance SD Min Max

N Mean 95% CI Variance SD Min Max

BEFORE 125 HzLeft 58 28.97 27.668 30.263 24.35 4.93 15 40

BEFORE 125 HzRight 58 29.91 28.447 31.381 31.13 5.58 20 40

AFTER 1 25 HzLeft 58 33.97 32.871 35.060 17.33 4.16 25 40

AFTER 1 25 HzRight 58 34.83 33.784 35.871 15.76 3.97 25 40

BEFORE 250 HzLeft 58 31.64 30.326 32.950 24.90 4.99 20 40

BEFORE 250 HzRight 58 33.19 31.803 34.576 27.81 5.27 20 40

AFTER 2 50 HzLeft 58 36.12 35.266 36.975 10.56 3.25 30 45

AFTER 2 50 HzRight 58 36.03 35.057 37.012 13.82 3.72 30 45

BEFORE 500 HzLeft 58 35.26 33.856 36.661 28.44 5.33 20 45

BEFORE 500 HzRight 58 35.69 34.632 36.747 16.18 4.02 25 45

AFTER 500 HzLeft 58 36.38 35.153 37.606 21.75 4.66 20 45

AFTER 500 HzRight 58 36.72 35.753 37.695 13.64 3.69 25 45

BEFORE 1000 HzLeft 58 30.862 29.627 32.097 22.0508 4.6958 20 40

BEFORE 1000 HzRight 58 30.259 28.617 31.900 38.967 6.2424 10 40

AFTER 1000 HzLeft 58 35.69 34.632 36.747 16.1827 4.0228 20 45

AFTER 1000 HzRight 58 34.569 33.516 35.622 16.039 4.0049 25 40

BEFORE BEFORE AFTER AFTER BEFORE BEFORE AFTER AFTER 2000 Hz2000 H2000 Hz- 2000 Hz- 4000 Hz- 4000 Hz- 4000 Hz- 4000 HzLeft Right Left Right Left Right Left Right 58 58 58 58 58 58 58 58 31.29 31.81 34.40 35.86 31.724 33.793 35.172 35.259 30.073 - 3 30.491 - 3 33.321 - 3 34.730 - 3 30.408 - 3 32.579 - 3 34.158 - 3 33.969 - 3 2.514 3.130 5.472 6.994 3.040 5.007 6.187 6.548 21.54 25.17 16.73 18.54 25.0454 21.3249 14.882 24.0547 4.64 5.02 4.09 4.31 5.0045 4.6179 3.8577 4.9046 20 20 25 25 20 25 25 20 45 45 45 45 40 45 45 45

(continued)

Hearing Loss and Its Association with Clinical Practice

11

Table 2. (continued)

N Mean 95% CI Variance SD Min Max

N Mean 95% CI Variance SD Min Max

BEFORE 8000 HzLeft 58 29.914 28.181 31.646 43.4135 6.5889 15 40

BEFORE 8000 HzRight 58 30.948 29.288 32.609 39.8745 6.3146 15 40

AFTER 8000 HzLeft 58 35 33.674 36.326 25.4386 5.0437 15 45

AFTER 8000 HzRight 58 35.172 33.894 36.451 23.654 4.8635 20 45

BEFORE 12000 Hz -Left 58 33.448 31.906 34.990 34.392 5.8645 20 45

BEFORE 12000 Hz -Right 58 34.052 32.447 35.656 37.2429 6.1027 20 45

AFTER 12000 Hz -Left 58 33.534 31.996 35.073 34.2181 5.8496 20 45

AFTER 12000 Hz -Right 58 34.052 32.447 35.656 37.2429 6.1027 20 45

BEFORE 16000 Hz -Left 58 38.017 36.763 39.271 22.7541 4.7701 15 45

BEFORE 16000 Hz -Right 58 38.879 38.139 39.620 7.9325 2.8165 30 45

AFTER 16000 Hz -Left 58 38.017 36.763 39.271 22.7541 4.7701 15 45

AFTER 16000 Hz -Right 58 38.879 38.139 39.620 7.9325 2.8165 30 45

BEFORE 18000 Hz -Left 58 39.31 38.638 39.982 6.5336 2.5561 30 45

BEFORE 18000 Hz -Right 58 39.569 38.904 40.234 6.3899 2.5278 30 45

AFTER 18000 Hz -Left 58 39.31 38.638 39.982 6.5336 2.5561 30 45

AFTER 18000 Hz -Right 58 39.569 38.904 40.234 6.3899 2.5278 30 45

Similarly, for each series the same previous statistical analysis was developed in terms of gender. This information is shown in Table 3. Regarding gender, the increase in the threshold changes according to the side of the ear. Thus, men have an increase in the audition thresholds above 10% in the frequencies of 125 Hz, 1000 Hz, 2000 Hz, and 4000 Hz in the right ear; and frequencies of 125 Hz, 250 Hz, and 8000 Hz in the case of the left ear. When speaking about women, the increase in hearing threshold is more than 10% for frequencies at 250 Hz, 1000 Hz, 4000 Hz, and 8000 Hz on the left ear. On the right ear, the increase is above 10% in a frequency of 2000 Hz. Moreover, the minimum hearing thresholds of the samples vary in both ears, and gender at frequencies of 125 Hz and 250 Hz. At 500 Hz, the minimum hearing threshold changes in women in the left ear, while in men, it changes in the right ear. At 1000 Hz, the minimum hearing threshold varies in the left ear among women, and in both ears in men. At 2000 Hz, the minimum hearing threshold changes in both ears among women, and in the right ear in the case of men. At 4000 Hz, the minimum hearing threshold varies in both ears in women, and in the right ear in men. And at 8000 Hz, the minimum hearing threshold changes in the left ear, and right ear in women. Finally, the maximum hearing threshold varies randomly in both genders. It varies at 250 Hz in both ears in women, at 1000 Hz in men and women in their left ears, at 2000 Hz in men in the right ear, and 4000 Hz in both ears in women, at 8000 Hz in both ears among men and in the left ear in women.

12

J.-C. Cobos-Torres et al.

Table 3. Statistical analysis of hearing test carried out by gender at different frequencies at the beginning and end of the semester.

SEX N Mean 95% CI Variance SD Min Max

SEX N Mean 95% CI Variance SD Min Max

SEX N Mean 95% CI Variance SD Min Max

SEX N Mean 95% CI Variance SD Min Max

BEFORE 125 HzLeft Man Woman 20 38 29 28.947 26.774 27.275 31.226 30.620 22.6316 25.889 4.7573 5.0881 20 15 40 40

BEFORE 125 HzRight Man Woman 20 38 29 30.395 26.099 28.670 31.901 32.120 38.4211 27.5427 6.1985 5.2481 20 20 40 40

AFTER 125 Hz-Left

AFTER 125 Hz-Right

Man 20 34.25 32.506 35.994 13.8816 3.7258 25 40

Man 20 35.25 33.474 37.026 14.4079 3.7958 25 40

BEFORE 250 HzLeft Man Woman 20 38 31.5 31.711 29.087 30.077 33.913 33.344 26.5789 24.6977 5.1555 4.9697 20 20 40 40

BEFORE 250 HzRight Man Woman 20 38 32.5 33.553 29.511 32.029 35.489 35.077 40.7895 21.4972 6.3867 4.6365 20 25 40 40

AFTER 250 Hz-Left

AFTER 250 Hz-Right

Man 20 36 34.559 37.441 9.4737 3.0779 30 40

Man 20 35 33.302 36.698 13.1579 3.6274 30 40

BEFORE 500 HzLeft Man Woman 20 38 35.25 35.263 32.913 33.433 37.587 37.094 24.9342 31.01 4.9934 5.5687 20 20 40 45

BEFORE 500 HzRight Man Woman 20 38 35 36.053 32.991 34.776 37.009 37.329 18.4211 15.0782 4.292 3.8831 25 25 40 45

AFTER 500 Hz-Left

AFTER 500 Hz-Right

Man 20 35.75 33.702 37.798 19.1447 4.3755 25 40

Man 20 36.25 34.574 37.926 12.8289 3.5818 25 40

BEFORE 1000 HzLeft Man Woman 20 38 30 31.316 27.991 29.709 32.009 32.923 18.4211 23.8976 4.292 4.8885 20 20 35 40

BEFORE 1000 HzRight Man Woman 20 38 28.75 31.053 25.281 29.254 32.219 32.851 54.9342 29.9431 7.4118 5.472 10 20 40 40

AFTER 1000 Hz-Left Man 20 34 31.774 36.226 22.6316 4.7573 20 40

Woman 38 33.816 32.364 35.267 19.5057 4.4165 25 40

Woman 38 36.184 35.075 37.294 11.3976 3.376 30 45

Woman 38 36.711 35.122 38.299 23.3464 4.8318 20 45

Woman 38 36.579 35.491 37.667 10.9531 3.3095 30 45

Woman 38 34.605 33.261 35.950 16.7319 4.0905 25 40

Woman 38 36.579 35.364 37.794 13.6558 3.6954 30 45

Woman 38 36.974 35.733 38.214 14.2425 3.7739 30 45

AFTER 1000 HzRight Man Woman 20 38 33.75 35 31.910 33.676 35.590 36.324 15.4605 16.2162 3.932 4.0269 25 25 40 40

(continued)

Hearing Loss and Its Association with Clinical Practice

13

Table 3. (continued)

SEX N Mean 95% CI Variance SD Min Max

SEX N Mean 95% CI Variance SD Min Max

SEX N Mean 95% CI Variance SD Min Max

SEX N Mean 95% CI Variance SD Min Max

BEFORE 2000 HzLeft Man Woman 20 38 31.5 31.974 28.971 30.373 34.029 33.574 29.2105 23.702 5.4047 4.8685 20 20 40 45

BEFORE 2000 HRight Man Woman 20 38 29.5 32.237 27.656 30.676 31.344 33.798 15.5263 22.564 3.9403 4.7502 20 25 35 45

AFTER 2000 Hz-Left

BEFORE 4000 HzLeft Man Woman 20 38 32.75 31.184 30.818 29.415 34.682 32.953 17.0395 28.9651 4.1279 5.3819 25 20 40 40

BEFORE 4000 HzRight Man Woman 20 38 32.75 34.342 30.818 32.759 34.682 35.925 17.0395 23.2041 4.1279 4.8171 25 25 40 45

AFTER 4000 Hz-Left

BEFORE 8000 HzLeft Man Woman 20 38 28.5 30.658 25.096 28.630 31.904 32.686 52.8947 38.069 7.2729 6.17 15 20 40 40

BEFORE 8000 HzRight Man Woman 20 38 31.75 30.526 28.687 28.476 34.813 32.576 42.8289 38.9047 6.5444 6.2374 20 15 40 40

AFTER 8000 Hz-Left Woman 38 35.658 34.170 37.146 20.5014 4.5278 25 45

AFTER 8000 HzRight Man Woman 20 38 35.75 34.868 33.438 33.271 38.062 36.466 24.4079 23.6309 4.9404 4.8612 20 20 45 45

BEFORE 12000 HzLeft Man Woman 20 38 32.75 33.816 29.861 31.927 35.639 35.705 38.0921 33.0192 6.1719 5.7462 20 20 40 45

BEFORE 12000 HzRight Man Woman 20 38 32.5 34.868 29.925 32.794 35.075 36.943 30.2632 39.8471 5.5012 6.3125 25 20 40 45

AFTER 12000 HzLeft Man Woman 20 38 32.75 33.947 29.861 32.069 35.639 35.825 38.0921 32.6458 6.1719 5.7137 20 20 40 45

AFTER 12000 HzRight Man Woman 20 38 32.5 34.868 29.925 32.794 35.075 36.943 30.2632 39.8471 5.5012 6.3125 25 20 40 45

Man 20 34 32.559 35.441 9.4737 3.0779 30 40

Man 20 35 32.991 37.009 18.4211 4.292 30 45

Man 20 33.75 31.026 36.474 33.8816 5.8208 15 40

Woman 38 34.605 33.107 36.104 20.7859 4.5592 25 45

Woman 38 35.263 34.058 36.468 13.4424 3.6664 25 45

AFTER 2000 HzRight Man Woman 20 38 35 36.316 33.302 34.803 36.698 37.829 13.1579 21.1949 3.6274 4.6038 25 25 40 45 AFTER 4000 HzRight Man Woman 20 38 35.25 35.263 33.318 33.514 37.182 37.012 17.0395 28.3073 4.1279 5.3205 25 20 40 45

(continued)

14

J.-C. Cobos-Torres et al. Table 3. (continued)

SEX N Mean 95% CI Variance SD Min Max

SEX N Mean 95% CI Variance SD Min Max

BEFORE 16000 HzLeft Man Woman 20 38 38.5 37.763 36.785 36.025 40.215 39.501 13.4211 27.9694 3.6635 5.2886 30 15 40 45

BEFORE 16000 HzRight Man Woman 20 38 39 38.816 37.559 37.925 40.441 39.707 9.4737 7.3435 3.0779 2.7099 30 30 40 45

AFTER 16000 HzLeft Man Woman 20 38 38.5 37.763 36.785 36.025 40.215 39.501 13.4211 27.9694 3.6635 5.2886 30 15 40 45

AFTER 16000 HzRight Man Woman 20 38 39 38.816 37.559 37.925 40.441 39.707 9.4737 7.3435 3.0779 2.7099 30 30 40 45

BEFORE 18000 HzLeft Man Woman 20 38 39 39.474 37.776 38.637 40.224 40.310 6.8421 6.4723 2.6157 2.5441 30 30 40 45

BEFORE 18000 HzRight Man Woman 20 38 39 39.868 37.776 39.059 40.224 40.678 6.8421 6.0633 2.6157 2.4624 30 30 40 45

AFTER 18000 HzLeft Man Woman 20 38 39 39.474 37.776 38.637 40.224 40.310 6.8421 6.4723 2.6157 2.5441 30 30 40 45

AFTER 18000 HzRight Man Woman 20 38 39 39.868 37.776 39.059 40.224 40.678 6.8421 6.0633 2.6157 2.4624 30 30 40 45

6 Discussion It was determined that hearing levels remain in a mild range at 96.47% of the sample. There was a rate increase of 3.10% among students who have changed their audition level. The students who did not have an audition problem or had a mild hearing loss began to have a moderate hearing loss after their clinical activities. In the same way, an increase of 0.95% of students began to have from a mild to moderate hearing impairment. Also, it is important to point out that 4.05% started to have certain hearing loss. Additionally, an analysis by gender was carried out. In the case of men, the hearing levels remain in a mild range of 98.25% in the sample. It should be noted that there was an increase of 5.00% of students who changed their hearing level. The students did not have hearing impairments or have mild hearing loss, but due to their clinical activities, they reached to have a moderate hearing loss. In contrast, the case of women’s hearing levels remains in a mild range at 95.53% of the sample. There was an increase of 2.11% for students, those of which moved from not having any hearing impairment or having a mild hearing problem, to a moderate hearing loss after their clinical practices. An increase of 1.18% of students moved from having a mild deficiency to a moderate deficiency. Besides, 3.29% of students started to have some kind of hearing loss. Additionally, it is possible to analyze the differences among the measures of the audiometry at the beginning and end of the study (Fig. 5). It was noted that there is a lot of similarity between the masculine and feminine sexes. Both showed a clear tendency of hearing loss in frequencies of 125 Hz, 250 Hz, 1000 Hz, 2000 Hz, 4000 Hz, and 8000 Hz. These results are aligned with the study

Hearing Loss and Its Association with Clinical Practice

15

Fig. 5. Difference of mathematical means of hearing in each ear.

carried out by [23] which sets out that students start their clinic activities with a mild hearing loss. What drew attention is an increase in the liminals within group G5 for the frequencies from 3 kHz to 4 kHz. This can be due to that the group G15 was exposed to higher noise levels for a smaller portion of time compared to the ones who remained within the normal range of noise exposure. According to the student’ hearing levels belonging to the fifth study cycle who finished their clinic dental practices, it was settled down that they present significant changes of audition in frequencies lower at 8000 Hz. Thus, it is verified that the samples are less dispersed at the end of the study. Thus, it is verified that the samples are less dispersed at the end of the study. This is a clear piece of evidence; that all students have become part of the group with a mild and moderate hearing loss. The hearing on the left presents an increase of lower threshold in four different frequencies. Furthermore, the right ear has an increase in the lower threshold in five frequencies. In a similar way, the left ear shows an increase of the superior threshold of audition in four different frequencies. On the other hand, the right ear presents an increase in the upper threshold in two frequencies. The study was carried out in a total sample of 70 students, 48 were women and 22 were male. After ruling out twelve samples, and with a sampling of 38 women and 22 males, a remarkable male predominance was observed that corresponds to [4]. The authors determine that within a sample of 50 students, males had a higher prevalence. It was found that hearing loss level of the students belonging to the fifth

16

J.-C. Cobos-Torres et al.

cycle increased above 10% in the frequencies of 125 Hz, 250 Hz, 1000 Hz, 2000 Hz, and 8000 Hz when their clinical practices ended. In the specific case of 4000 Hz, this increase only happened in the left ear. This coincides with study [16]. The authors state that the scotoma presence in frequency at 4 kHz as at 6 kHz was higher for the left ear in the exposed group. The current study shows that in almost all frequencies within an audible range (less than 2000 Hz) there is higher hearing loss in the left ear. This may be due to the distance and the student’s body posture relative to the patient when using the dental turbine and/or to the student’s bone density. This last statement is observed in the gender analysis. Women have higher hearing loss in the left ear than men.

7 Conclusions This is a highly controversial issue. The present investigation can not affirm that the dental noise exposure levels cause hearing loss; nevertheless, the amount of time exposed plays an important role and this has been proven. The students are exposed to noise from not just one medical equipment during their clinical practice. These practices last ten or fifteen hours per week average. The students can reach noise harmful levels by long period of time. All of this lends itself to may provoke a hearing disability. It would be advisable that under health and safety regulations, all universities require the mandatory use of hearing protectors among students and all staff working in clinics. It would be important to carry out other longitudinal analysis with a cross-sectional study in students that are wearing and not wearing hearing protectors. In any case, we must consider students’ features like laterality and height among other data which could help to identify main causes and implications that the dental equipment about hearing loss. All this is possible thanks to the ICT because as it has been detailed the measurements can be made with a mobile phone application. Acknowledgments. The research leading to these results has received funding from SmartUniverCity 2.0 program, funded by “Modelado y moldura nasoalvear pre-pos-quirúrgica 3D y etiquetado digital odontológico”.

References 1. Ahveninen, J., Jääskeläinen, L.P., Raij, T., Bonmassar, G., Devore, S., Hämäläinen, M., et al.: Task-modulated “what” and “where” pathways in human auditory cortex. Proc. Natl. Acad. Sci. 103, 14608–14613 (2006) 2. Deafness and hearing loss. https://www.who.int/es/news-room/fact-sheets/detail/deafnessand-hearing-loss. Accessed 12 Apr 2019 3. Estadísticas de Discapacidad. https://www.consejodiscapacidades.gob.ec/estadisticas-dediscapacidad/. Accessed 12 Apr 2019 4. Fuentes, E., Rubio, C., Cardemil, F.: Pérdida auditiva inducida por ruido en estudiantes de la carrera de odontología. Rev. de otorrinolaringología y cirugía de cabeza y cuello 73(3), 249–256 (2013) 5. Lin, Y., Abdulla, W.H.: Principles of psychoacoustics. In: Audio Watermark. Springer, Cham, pp. 15–49 (2015)

Hearing Loss and Its Association with Clinical Practice

17

6. Zwicker, E., Fastl, H.: Psychoacoustics: Facts and Models, 2nd edn. Springer, New York (2013) 7. Barrero, M.V., Vega, A.M.G., Valverde Sánchez, F.: Prevención de Riesgos Laborales en Odontoestomatología, 1st edn. Editorial Mad S.L., Madrid (2003) 8. Mojarad, F., Massum, T., Samavat, H.E.: Iran. Front. Dent. 6(4), 181–186 (2009) 9. Lozano, F.E., Díaz, A.M., Payano, J.C., Sánchez, F.I., Ambrocio, E.D., Huapaya, M.C., et al.: Nivel de ruido de los procedimientos clínicos odontológicos. Rev. Estomatológica Herediana 27(1), 13–20 (2017) 10. Medina, Á., Velásquez, G.I., Giraldo, L., Henao, L.M., Vásquez, E.M.: Sordera ocupacional: una revisión de su etiología y estrategias de prevención. Rev. CES Salud Pública 4(2), 116–224 (2013) 11. Peter, S.: Potentiation of chemical ototoxicity by noise. Semin. Hear. 30(1), 38–46 (2009) 12. León, N.: Conocimiento estudiantil de la salud ocupacional en la práctica odontológica I, vol. 55, no. 1, pp. 1–22. Fundación Acta Odontológica Venezolana, Universidad Central de Venezuela (2017) 13. Moncayo, J.P., Zumba, D.V.: Prevalencia de hipoacusia y factores de riesgo asociados en los estudiantes de quinto a décimo ciclo de la Facultad de Odontología de la Universidad de Cuenca, 2015–2016, Universidad de Cuenca, Cuenca (2016) 14. Ferrando, K., Chirife, T., Jacquett, N.: Exposición a ruidos por el ejercicio profesional en docentes odontólogos. Rev. de Odontología Latinoamericana 2(1), 59–67 (2012) 15. Paredes Salcedo, G.M.: Ruido ocupacional y niveles de audición en el personal odontológico del servicio de Estomatología del Centro Médico Naval Cirujano Mayor Santiago Távara, 2013. Universidad Nacional Mayor de San Marcos, Lima (2013) 16. Alabdulwahhab, B.M., Alduraiby, R., Ahmed, M.A., Albatli, L., Alhumain, M.S., Softah, N.A., et al.: Hearing loss and its association with occupational noise exposure among Saudi dentists: a cross-sectional study. BDJ open 2, 16006 (2016) 17. Theodoroff, S.M., Folmer, R.L.: Hearing loss associated with long-term exposure to highspeed dental handpieces. Gen. Dent. 63(3), 71–86 (2015) 18. Myers, J., John, A.B., Kimball, S., Fruits, T.: Prevalence of tinnitus and noise-induced hearing loss in dentists. Noise Health 18(85), 347–354 (2016) 19. Barek, S., Adam, O., Motsch, J.F.: Large band spectral analysis and harmful risks of dental turbines. Clin. Oral Investig. 3(1), 49–54 (1999) 20. Masterson, E.A., Themann, C.L., Calvert, G.M.: Prevalence of hearing loss among noiseexposed workers within the health care and social assistance sector, 2003 to 2012. J. Occup. Environ. Med. 60(4), 350–356 (2018) 21. Gijbels, F., Jacobs, R., Princen, K., Nackaerts, O., Debruyne, F.: Potential occupational health problems for dentists in Flanders, Belgium. Clin. Oral Investig. 10(1), 8–16 (2006) 22. Angulo Jerez, A., Mateos Alvarez, F., Blanco, J.L.: Audioprótesis teoría y práctica. 1st ed. Masson (2004) 23. Limia, C.: Pérdida auditiva en Odontólogos. Escola Superior de Tecnologida Saúde de Coimbra (ESTeSC), Coimbra, Portugal (2017)

Data

A Bayesian Network Approach to Identity Climate Teleconnections Within Homogeneous Precipitation Regions in Ecuador Renato Ávila1

and Daniela Ballari2(B)

1 Carrera de Ingeniería de Sistemas y Telemática, Facultad de Ciencias de la Administración,

Universidad del Azuay, 01.01.981, Cuenca, Ecuador [email protected] 2 Universidad del Azuay, 01.01.981, Cuenca, Ecuador [email protected]

Abstract. Reliable precipitation predictions require an understanding of climate teleconnections over precipitation. In Ecuador, these teleconnections were studied with correlation methods, but multivariate studies with several climatic indexes simultaneously has been less study. The objective of this work is to carry out a multivariate study using Bayesian networks to identify the influence of climate indexes in homogenous precipitation regions in Ecuador. The climate teleconnections, defined as the correlation between precipitation satellite data and climate indexes, as well as the regionalization of seasonality of precipitation were used to learn a Bayesian network in R software. It was characterized the structure and strength of the relationship between the teleconnections and the precipitation. Additionally, three types of belief propagation were used: regions to climate index, climate index to regions, and interactions between indexes. This was useful to determine whether the influence of a climate index is homogeneous throughout the country or varies by region, as well as to identify interactions between different indexes. The results of this study contribute to a better understanding of precipitation in Ecuador, and to promote making evidence-based water resource decisions. Keywords: Climate teleconnections · Bayesian networks · Probability propagation

1 Introduction Understanding rainfall patterns is essential for a country’s economy as they influence its productive sectors such as farming [1, 2], livestock [3], fishing [4], tourism [5] and power generation [6]. The hydroelectric sector is crucial for Ecuador as it produces much of the electricity consumed in the country. Additionally, new hydroelectric projects will come into operation in the coming years, and for which a deeper knowledge of the precipitation patterns is required. Nevertheless, the lack of studies on precipitation and climate in a country-wise scale, has made it difficult to understand and predict climate behavior. This is important for promoting a sustainable water resource management. © Springer Nature Switzerland AG 2020 E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 21–35, 2020. https://doi.org/10.1007/978-3-030-35740-5_2

22

R. Ávila and D. Ballari

In this regard, climate teleconnections are one of main climate variability driver that need to be addressed. Climate teleconnections are defined as a correlation between climatic variables separated by great distances. Different methods are currently used to study climate teleconnections such as: linear regression [7], principal component analysis (PCA) [8], wavelet analysis [9] and point-by-point correlation [8]. These methods have been widely used, however, according to De la Torre-Gea et al. [10], they present the limitation that, techniques such as linear regression, PCA and clustering, fragment information and assume spatial independence to reduce dimensionality. Multivariate approaches have been used such as empirical orthogonal teleconnections and empirical orthogonal functions (EOF), which are much better suited to this type of problem. A review of these methods can be found in [11]. Additionally, Mendoza et al. [12] used dynamic harmonic regressions to detected intra and inter-annual teleconnections, however the study area was localized in the Paute Basin, a small region within Ecuador. In recent years, Bayesian networks (BN) for climate teleconnection analysis has become popular. A BN is defined as a directed acyclic graph representing a set of variables as nodes and their probabilistic conditional independences as the arcs of the network [13]. It is a probabilistic model that allows inferring unknown variables from variables that are known or observed. It is a type of expert system that has application in almost any field of science, e.g., medicine, biology, astrophysics, computer science, chemistry, economy and climate. Regarding climate, they have provided an opportunity, due to the growing number of weather stations, as well as the increased availability of climate data from satellite images. These techniques have already been applied for both climate prediction [14, 15] and teleconnection analysis related to climate variables [10] as well as climate indexes [16, 17]. In Ecuador, climatic teleconnections have been studied [18], however no studies were found for the whole country that followed a multivariate approach to explore the interaction among teleconnections. Thus, the objective of this work is to carry out a multivariate study using BNs to identify the influence of climate indexes on homogenous precipitation regions in Ecuador. Applying a Bayesian approach is useful to determine whether the influence of a climate index is homogeneous throughout the country or varies by region, as well as to identify interactions between different indexes. The results of this study contribute to a better understanding of precipitation in our country, and to make evidence-based water resource decisions.

2 Study Area The study area is continental Ecuador, which has a surface area of 250 000 km2 and three clearly defined areas: the coast, the Andes and the Amazon (Fig. 1). Ecuador’s climate is influenced by a large number of factors whose influence changes according to the area, resulting in a large spatial-temporal variability of precipitation.

A Bayesian Network Approach to Identity Climate Teleconnections

23

Fig. 1. Study area and main climate regions of Ecuador: Coast, Highlands and Amazon.

3 Materials 3.1 Modes of Climate Variability and Climate Indexes Climate variability is the result of a combination of spatial patterns. The most prominent spatial patterns are known as modes of climate variability and have the capacity to influence climate at spatial and temporal scales. Modes of climate variability are identified through the study of spatial teleconnections [19]. Among the most common modes of climate variability are: El Niño-Southern Oscillation (ENSO), Central Pacific El Niño (Modoki), Pacific Decadal and Interdecadal Variability (PDO), North Atlantic Oscillation (NAO), and Pacific/North America Atmospheric Teleconnection (PNA). The modes of variability are defined by climatic indexes, e.g., ENSO is defined by various indexes such as: niño3, niño3.4, soi, among others. The changes in the indexes are associated with large-scale changes in the climatic fields. Because no single index can isolate a phenomenon from all other effects, multiple indexes define the same climatic phenomenon, and therefore each index is affected by different climatic phenomena [19]. The climatic indexes used in this work are: • • • • • •

Related to ENSO: niño1+2, niño3, nino 3.4, niño 4, emi, mei, espi. Related to the Pacific Ocean: tni, oni. Related to the Atlantic Ocean: tna, tsa, car_ersst, nta_ersst, amon, ammsst. Related to atmospheric indexes: ao, aao, qbo, soi, glaam. Related to teleconnections: pdo, ea, wp, ep/np, pna, nao, epo. Others: gmsst.

The indexes were retrieved from: https://www.esrl.noaa.gov/psd/ data/climateindexes/list/, http://www.jamstec.go.jp/frsgc/research/d1/iod/e/elnmodoki/ about_elnm.html, https://www.esrl.noaa.gov/psd/data/timeseries/daily/EPO/. 3.2 Downscaled Satellite Precipitation Data The main dataset consists on the monthly precipitation satellite imagery from TRMM 3b43 V7, for the period of 2001–2011. The original available data was corrected and

24

R. Ávila and D. Ballari

downscaled to 5 km from its original resolution of 27 km, using 117 in-situ rain gauges and satellite images as covariables related to cloud properties, normalized vegetation index and soil moisture [20]. Following the terminology of the R software and its “raster” package, the product is represented as a rasterbrick, where each raster layer shows the precipitation of a specific month. Additionally, each pixel in the rasterbrick contains the precipitation time series at such a location. The rasterbrick contains a total of 9919 pixels for the whole Ecuador.

4 Methods This paper focuses on the learning and belief propagation of a BN to identify the relationship between climate teleconnections and homogenous precipitation regions. Figure 2 summarizes the main methodological steps.

Fig. 2. Methodological steps, based on [21]

4.1 Data Preparation Correlation Maps. It was applied on the rasterbrick, pixel by pixel, the non-parametric correlation of Spearman between the time series of precipitation and the different climatic indexes. Figure 3 shows the correlation maps with blue color the positive correlations and red the negative correlations. The high values of positive and negative correlations depicted the existence of climatic teleconnections. Regionalization Map of Precipitation Seasonality at 5 km. A functional regionalization was performed on the downscaled TRMM images at 5 km, following the previous work of [22]. In our case, by applying the regionalization at 5 km of resolution, it was obtained 9 seasonality regions (Fig. 3), which can be classified within the natural areas of

A Bayesian Network Approach to Identity Climate Teleconnections

25

Ecuador (Coast, Highlands and Amazon). However, there are two special cases because they overlapped more than one natural area in equal proportions. This was the case of region 1 (which included the Coast, Highlands and a small part of the Amazon) and region 9 (which included the Highlands and Amazon). We decided to classify region 1 as Coast and region 9 as Amazon. Relationship of Regionalization and Correlation Maps. Each pixel of the regionalization map was related to the correlation values of the different climatic indexes at the same pixel. Thus, a table was constructed for learning the structure and parameters of the BN. Such table consisted on 9919 records (pixels) and 29 columns (Region and 28 climate indexes). The 80% of the records were randomly selected for the training group (random sampling stratified by seasonality region), while the remaining 20% were used for validation.

Fig. 3. Data preparation. (a) Correlation maps and climate teleconnections; (b) Homogenous precipitation regions at 5 km

4.2 Bayesian Network In this work, the nodes of the BN were the seasonality regions and the climate teleconnections (correlations of precipitation and the climate indexes). The arcs were the probabilistic dependencies between the regions and the climatic teleconnections. Step 1. Discretize Data. Because most learning algorithms use discrete data, our data must be discretized. In our case, the correlation data were continuous and needed to be discretized. The 9 regions of seasonality, however, were discrete. There are several discretization methods [23], however here we focused on unsupervised methods that use the distribution of a continuous variable as the sole source of information [24]. One of these method is the Equal Width Discretization (EWD) that divides the range of an attribute in k intervals of equal size, being k determined by the user. Thus we divided in 6 intervals. For positive correlations: [0; 0,3) Low; [0,3; 0,6) Medium; [0,6; 1] High. For negative correlations: [−1; −0,6] High; (−0,6; −0,3] Medium, (−0,3; 0) Low.

26

R. Ávila and D. Ballari

Step 2. Network Learning. Learning a BN consists on estimating a structure model and the associated parameters from data. In our work, these were done with the bnlearn library [25] and the gRain library in R [26], which has tools to read the structure and parameters previously developed with bnlearn. Structure Learning. It consists on identifying the graphical structure of the network. (i.e. nodes, arcs and arc directions). There are different learning algorithms such as constraint-based, score-based and hybrid [27], however following [28], we selected the method based on Hill-Climbing scores (hc). It is a local search algorithm, consisting of an iteration that moves in the direction that increases its value. It ends when it has reached a peak and no neighbor has a higher value. Additionally, hc does not look beyond its immediate neighbours [29]. Score Metrics. Although BN have gained much prominence in recent years, identifying the structure from the data remains as a challenge, because the number of possible structures increases exponentially with the number of nodes. One way to solve this is through a heuristic search based on punctuation metric optimization. Thus the search is based on a scoring function that evaluates the degree of adequacy of the network [30]. We evaluated different metrics such as Akaike’s information criterion (AIC), Bayesian information criterion (BIC), Bayesian Dirichlet equivalence score (BDe). BDe was selected because it depicted the best network score. Parameter Learning. Considering complete and discrete data, and assuming that the structure has already been obtained, we used the Maximum Likelihood Estimation (MLE) method, under which the probabilities are estimated based on the frequencies of the data. The marginal and conditional probability tables were obtained. Belief Propagation. BN provides an inference system. This allows to modify the probability tables once new evidences are observed about the state of certain nodes, and these new probabilities spread out (propagate) to the rest of nodes. This is known as probabilistic inference. There are 2 types of propagation: forward and backward. In the forward propagation the probability is propagated from a parent node to all child nodes and in backward propagation the probability is propagated from the child nodes to the parents. The belief propagation was implemented using three approaches: from regions to climate indexes (forward), from climate indexes to regions (backward), and interaction between teleconnections. Model Validation. We used a confusion matrix approach, which is an nxn matrix with the rows as the real classes and the columns as the predicted classes by the model. We accounted for precision and kappa metrics.

5 Results 5.1 Bayesian Network Structure and Parameters The learned structure was similar to a simple Bayesian classifier (BAN, Naive Bayesian Network), with region as the class variable (Fig. 4). Although region was the ancestor

A Bayesian Network Approach to Identity Climate Teleconnections

27

of most nodes, it does not mean that it is the parent of these nodes, i.e. it does not imply a causal relationship between the region and the teleconnections. In Fig. 4, the child nodes in beige color represent teleconnections from the Pacific Ocean; and in blue from the Atlantic Ocean. The white nodes represent global and north-south poles teleconnections. The first-level descendants of region were: (1) Pacific Ocean: niño 3, nino 1+2, np, niño 34, qbo; (2) Atlantic Ocean: ammsst, CAR_ersst, tsa, epo, nao; and (3) Others: glaam. The second-level descendants of region were: (1) Pacific Ocean: mei, niño4, oni, espi, pdo, wp; (2) Atlantic Ocean: ao, ea, NTA_ersst, amon. Two nodes were isolated (tna and pna), indicating that the relationship between region and these nodes was very weak. Finally, soi was not directly connected with region, but instead it was indirectly connected through mei and tni.

Fig. 4. Structure of the learned Bayesian network

The thickness of the nodes in Fig. 4 is related to the strength of the relationship between nodes. Therefore, the relationship strength was strong between region and the nodes: niño1+2, niño34, epo, qbo and ammsst; with moderate strength with the nodes: nao, np, CAR_ersst and niño3; and weak with the nodes: glaam and tsa. Parameters were also learned. Thus in each node the marginal a priori probability was obtained. These parameters are not shown here for the sake of space. 5.2 Belief Propagation The three types of propagation were performed: (1) from regions to indexes allowing to know the indexes that influence the precipitation of a certain region; (2) from indexes to regions allowing to know the regions in which a given index has an influence; (3)

28

R. Ávila and D. Ballari

interactions between indexes allowing to analyze relations between them. The results are in Tables 1, 2 and 3 showing only the indexes or regions whose probability increased beyond 1% (minimum of representativeness). In addition, index ranges between −0.3 and 0.3 were not considered, because the correlation in these ranges is weak. In the Tables 1, 2 and 3, the indexes were related to the Pacific Ocean in beige color, the Atlantic Ocean in brown, and others in white color. Similarly, the range of the correlation was colored with red for negative correlation and blue for positive. From Regions to Climate Indexes. Table 1 shows the propagation, highlighting the most relevant results when a posteriori probability increased more than 10%. It contains the following information: “region” as the variable established as evidence, and “index” as the indexes whose probability increased as a result of the evidence. The “a priori probability” and “a posteriori probability” as the probability of indexes without evidence (a priori) and with evidence (a posteriori). A priori probability is identified as the influence of an index in the whole country (i.e. frequency without selecting a region as evidence), while a posteriori probability is the influence within a specific region (i.e. frequency when selecting a region as evidence). The “range” is the range of the index correlation. Coast’s Regions. In region 1 (North Coast) there were 6 indexes affecting precipitation: 4 from the Pacific (mei, espi, oni and pdo) and 2 from the Atlantic (CAR_ersst and tsa). They all did it in a negative moderate range, except for tsa which did it in a positive moderate range. In the case of teleconnections from the Atlantic, the a posteriori probability increased more than 10%, while with the Pacific it increased less than 6%. In region 2 (Central Coast), there were 4 indexes from the Pacific and 2 from the Atlantic. The niño1+2 presented the highest increase in a posteriori probability to (95%) and in a positive correlation range of 0.6 to 1. The indexes amon, np and niño4 influenced in a negative moderate range, while niño3 and tsa in a positive moderate ranges. All indexes showed a change in probability greater than 10% except for niño4. In region 5 (South Coast) all the indexes belonged to the Pacific except for amon. Mei, niño4, espi, amon, oni and np had influences in a negative moderate range, niño3 in a positive moderate range, and niño1+2 in a positive high range. The niño1+2, niño3, niño4, amon and np showed a change of a posteriori probability greater than 20% (54.8% for niño1+2), while mei, spi and oni lower than 5%. Highland’s Regions. All the indexes that affected region 3 (Northen Hihglands) were related to the Pacific. They influenced in a negative moderate range with the exception of niño1+2 which did it in a positive moderate range. The indexes that most increased their a posteriori probabilities were niño1+2 and niño4 (higher than 17% in both cases), meanwhile for the others was less than 5%. In region 7 (Central-east Highlands) the only index that had influence was pdo. It did it in a moderate negative range. The probability increase was very low, barely 2%. All indexes affecting region 8 (South Highlands) belonged to the Pacific (except amon), they all influenced in a negative moderate range except for niño1+2 which registered both a positive moderate and positive high influences. Niño4 and np had the highest changes in their probabilities (23.5%). In other cases, the increase in probability did not exceed 7%.

A Bayesian Network Approach to Identity Climate Teleconnections Table 1. Regions to climate indexes propagation. Region

1 North Coast

2 Central Coast

5 South Coast

3 North Highlands 7 Central-E Highlands

8 South Highlands

4 North Amazon 6 Central Amazon 9 South Amazon

Index tsa mei espi CAR oni pdo nino1 nino3 tsa nino4 amon np nino1 nino3 mei nino4 espi amon oni np nino1 mei nino4 espi oni

A priori probability. 14,83% 3,91% 3,41% 4,29% 4,41% 0,86% 39,44% 71,80% 14,83% 29,11% 41,10% 20,66% 39,44% 71,80% 3,91% 29,11% 3,41% 41,10% 4,41% 20,66% 35,97% 3,91% 29,11% 3,41% 4,41%

A posteriori probability 26,92% 5,59% 4,65% 16,20% 6,53% 2,08% 94,68% 93,80% 67,59% 34,11% 88,33% 30,82% 94,31% 92,42% 7,40% 61,83% 6,06% 88,04% 8,65% 47,79% 53,43% 7,94% 51,43% 6,62% 9,27%

Range [0.3,0.6)

[-0.6,-0.3)

[0.6,1) [0.3,0.6) [-0.6,-0.3) [0.6,1) [0.3,0.6)

[-0.6,-0.3)

[0.3,0.6) [-0.6,-0.3)

pdo

0,86%

2,87%

[-0.6,-0.3)

nino1 nino1 mei nino4 espi amon oni np qbo nino1 nino3 nino34 np nino1 nino3 CAR nino1 nino3 nino34 np ammsst

35,97% 39,44% 3,91% 29,11% 3,41% 41,10% 4,41% 20,66% 2,96% 35,97% 71,80% 6,48% 1,16% 35,97% 71,80% 4,29% 35,97% 71,80% 6,48% 1,16% 0,53%

42,93% 45,66% 6,98% 52,62% 5,88% 47,72% 8,16% 44,21% 19,60% 49,81% 95,64% 45,08% 2,94% 82,75% 85,39% 10,57% 39,33% 80,33% 11,00% 17,00% 7,67%

[0.3,0.6) [0.6,1)

[-0.6,-0.3)

[0.3,0.6)

[0.3,0.6) [-0.6,-0.3) [0.3,0.6) [-0.6,-0.3)

29

30

R. Ávila and D. Ballari

Amazon’s Regions. Of the 5 indexes that affected region 4 (North Amazon), all were from the Pacific (niño1+2, niño3, niño34). They influenced in a positive moderate range and the probability increase was greater than 13%, except for np which increased barely 1.7%. For region 6 (Central Amazon), the indexes that affected were niño1+2, niño3 (Pacific) and CAR_ersst (Atlantic). The first two had a moderate positive influence while the third one a moderate negative influence. The one that registered the highest probability increase was niño1+2 with 46.7%, while the one that registered the lowest change was CAR_ersst with 6.2%. In the case of region 9 (South Amazon) there were 7 indexes, 5 belonging to the Pacific and 2 to the Atlantic. The indexes niño1+2, niño3, niño34 and np had moderate positive influence, while ammsst, CAR_ersst and pdo moderate negative influence. Np was the index with the highest increase in probability (15.8%), while in the case of the others it was less than 9%. There were 3 indexes whose probabilities did not showed changes regardless the evidences. They were: soi, tni and modoki. Additionally, it should be noted that regardless of the magnitude of the increase, it is the a posteriori probability of an index which should be observed as an indicator of the influence of such index. Thus an integrated interpretation of both, a posteriori probability and the increase magnitude, is needed. For instance, in the region 9, the niño1+2 and niño3 indexes increased their probability in 3.36% and 8.53% respectively, which was lower than the 15.84% for the np index. However, the a posteriori probability of niño1+2 was 39.33% and of niño3 was 80.33%, while in the case of np was 17%. Therefore, the probability of the presence of the niño1+2 and niño3 indexes in that region was higher than the probability of the presence of np. From Climate Indexes to Regions. In this case, the established evidences were moderate and high ranges of the indexes. Therefore, probabilities were propagated to the region variable. This propagation is useful in the case that anomalies (extremely high or low values) are observed in the indexes, thus it can be explored the main region than might affect. A total of 50 evidences were set, 10 from Atlantic indexes (ammsst, Table 2. Climate index to region propagation. Index

Range [0.3,0.6)

niño1+2 niño34 ammsst amon CAR_ersst tsa

Region

6 Central Amazon 2 Central Coast [0.6,1) 5 South Coast 4 North Amazon [0.3,0.6) 9 South Amazon 8 South Highlands [-0.6,-0.3) 9 South Amazon 2 Central Coast [-0.6,-0.3) 5 South Highlands 1 North Coast [-0.6,-0.3) 6 Central Amazon 1 North Coast [0.3,0.6) 2 Central Coast

A priori proba- A posteriori probbility ability 17,68% 40,68% 15,87% 38,10% 15,96% 38,16% 13,31% 92,61% 3,78% 6,42% 14,80% 40,48% 3,78% 54,76% 15,87% 34,10% 15,96% 34,18% 6,88% 25,99% 17,68% 43,57% 6,88% 12,49% 15,87% 72,30%

A Bayesian Network Approach to Identity Climate Teleconnections

31

ammon, CAR_ersst, tsa), and 40 from Pacific indexes (np, pdo, qbo, niño1+2, niño3, niño34, niño4, mei, espi, oni). Nevertheless, because of space, we illustrate the results with few examples. Table 2 shows the examples with the a priori probability and a posteriori probability of the region, and Fig. 5 shows the cartographic representation of the a posterior probability. The Maps colored in blue represented a positive range of correlation and those colored in orange a negative range of correlation.

Fig. 5. Examples of a posterior probability maps after climate index to region propagation

Climate Indexes Interaction. Table 3 shows the propagated index (with evidence in a specific range) and the target index (where the probability was propagated) Table 3 illustrates 10 interaction examples that had highest probability change after the propagation. Table 3. Climate index interactions. Propagated index espi modoki soi mei oni CAR_ersst mei oni nino1 espi

Range of propagated index [-0.6,-0.3) [0.3,0.6) [-0.6,-0.3) [0.6,1) [-0.6,-0.3)

Target index

A priori probability

oni tni tni oni espi tni nino4 mei amon mei

4,41% 17,90% 17,90% 4,41% 3,41% 17,90% 29,11% 3,91% 41,10% 3,91%

A posterior probability 86,33% 99,52% 92,58% 70,14% 66,67% 79,62% 87,67% 62,13% 92,59% 53,62%

Range of target index [-0.6,-0.3) [0.3,0.6) [-0.6,-0.3) [0.3,0.6) [-0.6,-0.3)

Main results showed that, for the Pacific indexes, qbo, niño1+2, niño3, niño34 indexes had interaction in positive correlation ranges. Pdo, oni, espi, niño4, mei indexes

32

R. Ávila and D. Ballari

had interaction in negative correlation ranges. Np had both positive and negative interactions. Additionally, np had the most numerous influence with a total of 13 records. Meanwhile qbo and niño34 had the least numerous influence with only 2 records each. For the Atlantic indexes, ammsst, amon, CAR_ersst had interaction in negative correlation ranges and tsa in positive ranges. Amon had the most numerous influence with a total of 8 records, and ammst the least with only 1 record. 5.3 Model Validation Validation was performed using 22 nodes of the BN. Regarding the accuracy metric: 19 nodes registered a value greater than 0.8 (high); 3 nodes registered a value between 0.5 and 0.8 (moderate); and there were no records with an accuracy value less than 0.5. Regarding the kappa coefficient: 1 node registered a value greater than 0.8 (high); 17 nodes registered a value between 0.5 and 0.8 (moderate); and 4 nodes had a value less than 0.5 (low).

6 Discussion and Conclusions In this work was carried out the structure and parameter learning of a Bayesian network. Additionally, by using belief propagation, it was possible to study different the teleconnections that influenced precipitation in the different seasonality regions of Ecuador, as well as to determine if there is interaction between teleconnections. Bayesian networks are useful to set what-if scenarios and analyze their possible effects. In climatology this can be a very complex task due to the large number of climatic indexes that interact. Consider as a hypothetical example that the National Institute of Meteorology reports a high probability of occurrence of the El Niño event to the coasts of Ecuador. Faced with this situation, the National Institute for Risk Management must prepare contingency and mitigation plans, which implies to know on which regions must to act according to the anomalies observed in the climate indexes. To make more concrete this example, let suppose that it was reported that niño1+2 and niño3 indexes present “abnormally high” values, which can be related to the moderate and high positive correlation ranges. Thus, having a Bayesian Network those values can be established as evidences, obtaining that the most affected regions would be the region 6 (Amazon) with a 29.2% probability, the region 2 (Coast) with a 19.8% probability, and region 5 (Coast) with 19.6%. In another example, let suppose “abnormally high” values for niño 4 and np indexes, which can be related to moderate and high negative correlation ranges. When establishing these evidences, it is obtained that the most affected regions are the region 5 (Coast) with a 40.2% probability, region 8 (Highlands) with a 33.5% probability, and region 2 (Coast) with 17.9%. Despite the advantages shown by Bayesian networks for identifying the indexes affecting a region and what if scenarios, there were also some limitations: Discretization. It was necessary to discretize the values since the learning algorithm used in this paper worked with this type of data. However, this can also be a disadvantage since, if the intervals are not chosen properly, it is possible to draw biased or erroneous

A Bayesian Network Approach to Identity Climate Teleconnections

33

conclusions. In addition, the method used here to discretize (EWD) can produce intervals with a lot of data and other intervals with little data which could affect the resulting network. In future work, it would be important to use learning methods that ensure a more appropriate data distribution or to work with continuous data distributions. Causality Interpretation. Unless external evidence is available, no causality between nodes can be concluded from a Bayesian network. For example, between two connected nodes there is no causality relationship, but through learning there is a probability “relationship” between those nodes. It is important to highlight this in order to avoid drawing erroneous conclusions. Comparison with Other Methods. In order to validate the obtained results, a fair comparison is needed to be performed with further methods such as regression, decision trees or random forest. Index Correlation. The data used to obtain the Bayesian network were correlation values and not the original values of the climate indexes. This means that in order to create this network, it was first necessary to obtain the correlation maps. It would be interesting for future work to learn a network from the original precipitation and the indexes. Regionalization. We used regions defined as homogenous precipitation seasonality, however, further regionalization could also be considered, for example by administrative limits (provinces or counties). Finally, this work showed that a Bayesian network approach is useful to determine whether the influence of a climate index is homogeneous throughout the country or varies by region, as well as to identify interactions between different indexes. The results of this study contribute to a better understanding of precipitation in our country, and to promote the evidence-based water resource decisions. Acknowledgements. This study has been financed by the Corporación Ecuatoriana para el Desarrollo de la Investigación y la Academia (CEDIA) through the project CEPRA XII “Spatial representation of climatic teleconnections in the precipitation of Ecuador”.

References 1. Ali, S., Jan, A., Manzoor, et al.: Soil amendments strategies to improve water-use efficiency and productivity of maize under different irrigation conditions. Agric. Water Manag. 210, 88–95 (2018). https://doi.org/10.1016/j.agwat.2018.08.009 2. Sudha, V., Venugopal, K., Ambujam, N.K.: Reservoir operation management through optimization and deficit irrigation, 93–102 (2008). https://doi.org/10.1007/s10795-0079041-3 3. Engler, J., Von Wehrden, H., Baumgärtner, S.: Land use policy determinants of farm size and stocking rate in Namibian commercial cattle farming. Land Use Policy 81, 232–246 (2019). https://doi.org/10.1016/j.landusepol.2018.10.009 4. Pratiwi, R., Sukardjo, S.: Effects of rainfall on the population of Shrimps Penaeus Monodon Fabricius in Segara Anakan lagoon, Central Java, Indonesia. 2(3), 156–169 (2018). https:// doi.org/10.11598/btb.2018.25.3.830

34

R. Ávila and D. Ballari

5. Abd-elhamid, H.F., Fathy, I., Zelen, M.: Flood prediction and mitigation in coastal tourism areas, a case study: Hurghada, Egypt (2018). https://doi.org/10.1007/s11069-018-3316-x 6. Hamududu, B., Killingtveit, A., Engineering, E.: Assessing Climate Change Impacts on Global Hydropower, 305–322 (2012). https://doi.org/10.3390/en5020305 7. Liu, Y.-C., Di, P., Chen, S.-H., DaMassa, J.: Relationships of rainy season precipitation and temperature to climate indexes in California: long-term variability and extreme events. J. Clim. 31(5), 1921–1942 (2018). https://doi.org/10.1175/JCLI-D-17-0376.1 8. Fierro, A.O.: Relationships between California rainfall variability and large-scale climate drivers. Int. J. Climatol. 34(13), 3626–3640 (2014). https://doi.org/10.1002/joc.4112 9. Konapala, G., Valiya, A., Ashok, V.: Teleconnection between low flows and large-scale climate indexes in Texas River basins. Stoch. Environ. Res. Risk Assess. (2017). https://doi.org/10. 1007/s00477-017-1460-6 10. De la Torre-Gea, G., Soto-Zarazua, G.M., Guevara-Gonzalez, R.G., Rico-Garcia, E.: Bayesian networks for defining relationships among climate factors. Int. J. Phys. Sci. 6(18), 4412–4418 (2011). https://doi.org/10.1016/j.jmaa.2015.01.055 11. Lee, J.H., Lee, J., Julien, P.Y.: Global climate teleconnection with rainfall erosivity in South Korea. CATENA 167, 28–43 (2018). https://doi.org/10.1016/j.catena.2018.03.008 12. Mendoza, D.E., Samaniego, E.P., Mora, D.E., Espinoza, M.J., Campozano, L.V.: Finding teleconnections from decomposed rainfall signals using dynamic harmonic regressions: a Tropical Andean case study. Clim. Dyn. 1–28 (2018). https://doi.org/10.1007/s00382-0184400-3 13. Correa, M., Bielza, C., Pamies-teixeira, J.: Comparison of Bayesian networks and artificial neural networks for quality detection in a machining process. Expert Syst. Appl. 36(3), 7270–7279 (2009). https://doi.org/10.1016/j.eswa.2008.09.024 14. Das, M., Ghosh, S.K.: A probabilistic approach for weather forecast using spatio-temporal inter-relationships among climate variables. In: 9th International Conference on Industrial and Information Systems, ICIIS 2014 (2015). https://doi.org/10.1109/ICIINFS.2014.7036528 15. Zeng, Z., Hsieh, W.W., Shabbar, A., Burrows, W.R.: Seasonal prediction of winter extreme precipitation over Canada by support vector regression. Hydrol. Earth Syst. Sci. 15(1), 65–74 (2011). https://doi.org/10.5194/hess-15-65-2011 16. Duc, H.N., Rivett, K., MacSween, K., Le-Anh, L.: Association of climate drivers with rainfall in New South Wales, Australia, using Bayesian model averaging. Theor. Appl. Climatol. 127(1–2), 169–185 (2017). https://doi.org/10.1007/s00704-015-1622-8 17. Ebert-Uphoff, I., Deng, Y.: A new type of climate network based on probabilistic graphical models: results of boreal winter versus summer. Geophys. Res. Lett. 39(19), L197011. 1–7 (2012) 18. Vicente-Serrano, S.M., Aguilar, E., Martínez, R., et al.: The complex influence of ENSO on droughts in Ecuador. Clim. Dyn. 48(1–2), 405–427 (2017). https://doi.org/10.1007/s00382016-3082-y 19. Blunden, J., Arndt, D.S., Baringer, M.O., et al.: State of the climate in 2010. Bull. Am. Meteorol. Soc. 92(6), S1-S236 (2011). https://doi.org/10.1175/1520-0477-92.6.S1 20. Ulloa, J., Ballari, D., Campozano, L., Samaniego, E.: Two-step downscaling of Trmm 3b43 V7 precipitation in contrasting climatic regions with sparse monitoring: the case of Ecuador in Tropical South America. Remote Sens. 9(7), 758 (2017). https://doi.org/10.3390/rs9070758 21. Rodríguez, D., Dolado, J.: Redes Bayesianas en la ingeniería del software. CcUahEs 1–21 (2007). https://doi.org/10.2196/jmir.7.3.e31 22. Ballari, D., Giraldo, R., Campozano, L., Samaniego, E.: Spatial functional data analysis for regionalizing precipitation seasonality and intensity in a sparsely monitored region: unveiling the spatio-temporal dependencies of precipitation in Ecuador. Int. J. Climatol. 38(8), 3337–3354 (2018). https://doi.org/10.1002/joc.5504

A Bayesian Network Approach to Identity Climate Teleconnections

35

23. Das, K., Vyas, O.P.: A suitability study of discretization methods for associative classifiers. Int. J. Comput. Appl. 5(10), 46–51 (2010). https://doi.org/10.5120/944-1322 24. López, D.A.G.: Algoritmo de Discretización de Series de Tiempo Basado en Entropía y su Aplicación en Datos Colposcópicos (2007). http://cdigital.uv.mx/bitstream/123456789/ 32352/1/garcialopezdaniel.pdf 25. Scutari, M.: Package ‘bnlearn’ (2019). https://cran.r-project.org/web/packages/bnlearn/ bnlearn.pdf 26. Højsgaard, S.: Graphical independence networks with the gRain package for R. J. Stat. Softw. 46(10), 37–44 (2012). https://doi.org/10.4324/9780429468872-4 27. Nagarajan, R., Scutari, M., Lèbre, S.: Bayesian Networks in R (2013) https://doi.org/10.1007/ 978-1-4614-6446-4 28. Sachs, K., Perez, O., Pe’er, D., Lauffenburger, D.A., Nolan, G.P.: Causal protein-signaling networks derived from multiparameter single-cell data. Science 308(5721), 523–529 (2005) 29. Russell, S., Norvig, P.: Artificial Intelligence A Modern Approach, 3rd edn (2010). https:// doi.org/10.1017/S0269888900007724 30. Carvalho, A.: Scoring functions for learning Bayesian networks. INESC-ID Technical report 54/2009, pp. 1–27 (2009). https://pdfs.semanticscholar.org/6efe/ f4bacfb14cfe4c1ababae751904431b75cc9.pdf

High Impact Innovation

Design and Implementation of an Automatic System for the Monitoring and Monitoring of a Prototype Refrigeration Plant with Parallel Compressors Elsy del Rocio Villamar Garcés1(B) , Jorge Luis Gonzalez Murillo1 , Jacinto Gabriel Lino Sánchez1 , Monica Karina Jaramillo Infante1 , and Oswaldo Villamar Chele2 1 State University Santa Elena Peninsula (UPSE), Santa Elena, Ecuador

[email protected], [email protected], {evillamar, mjaramillo}@upse.edu.ec 2 Instalaciones y Mantenimiento El Termico S.A. Termidigil, Guayaquil, Ecuador [email protected] http://www.eltermico.com.ec/

Abstract. The present article shows the design and implementation of an automatic system for the controlling and monitoring of a prototype refrigeration plant with parallel compressors. The Importance of carrying out this work is to Maintain the adequate temperature in the showcases and in the cold storage rooms, in this Way to Provide a solution to the demand With Which the supermarkets have a system Whose function is to preserve the quality of the product for the consumer, in Addition to Take Control and monitor Efficiently through a PLC, screen and other HDMI devices. The first resource will be the identification of the system through the acquisition of data Which leads to the problem, with the aim of optimizing the temperature control of the products and extracting the heat from an automated and truthful plant. Keywords: Plant monitoring · Refrigeration · PLC · Temperature · Compressors

1 Introduction Nowadays, Ecuador has several companies and supermarkets that have a cooling system to keep all types of food products, being Ecuador the country with more production and marketing of these systems in Latin America. Products must be in good condition to consume, due to this, cooling devices are used to have storage capacity of various products, but the big difference between having a product in good condition and high quality, is much. To know if a product has a good quality, should review issues such as its color, smell, flavor, texture and the absence of pollutants or other harmful products in the food or drink. In some cases, have witnessed products, that stay in the higher production markers, have © Springer Nature Switzerland AG 2020 E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 39–52, 2020. https://doi.org/10.1007/978-3-030-35740-5_3

40

E. del Rocio Villamar Garcés et al.

a drawback in preserving them for mismanaging temperatures due to machinery and equipment used. The lack of temperature control in products, makes teams working constantly in a sequential manner, producing a high demand for energy that hurt the company. Due to this, the aim is to optimize the problems of these cooling systems, making a design for the system control and monitoring via the HMI (human machine interface) for the automatic cooling. This project will help to have more efficient cooling system to conserve food products and at the same time to create broader future research on the proposed topic and help the local market to provide better service.

2 Methodology 2.1 Automation Control and Monitoring of a Plant In this project, automation is the variety of systems, processes or automatic equipment for control and monitoring of the plant, whose function is to perform certain tasks operate independently without the little human, intervention in the operation or control. These systems and processes, are carried out, due to its elements, such as sensors that are used, the same as allow a control to make a change in the process to develop. 2.2 Cooling System Refrigeration systems were used to transport the heat, that means, lowering the temperature of items or products that are only at room temperature and have a greater food

Fig. 1. Prototype refrigeration plant.

Design and Implementation of an Automatic System

41

preservation. In this project the cooling system function will be to keep all kinds of food, that have automated cooling systems that will serve both companies and for future research in educational institutions [3]. 2.3 Control System Control systems consist of several elements that work contiguously to perform a specific target. With these elements, the values are adjusted according to goals. In the case, cooling system allows temperature variation to have a stable control. In At industry, these control systems had an impact on making a change, leaving the “manual” way and do it mechanically and automatically. Control systems have two types: open loop and closed loop. Open Loop In this system, refers to an input signal runs all the way to its output without affecting their regulation, making any changes to the process control (Fig. 2).

Signal Input Reference Signal

Controller

Compressor/ valves

Process /Plant

Signal output Manipulated Variable

Fig. 2. System open loop control.

In the industry have an open loop system may cause a risk because it does not have a control input signal. The open loop system has two components: the controller and the controlled process. Closed Loop In this case the system has a feedback input signal involved in the regulation of the system. The aim is to compare the desired value of the controlled variable output with minimal error and have a reliable system (Fig. 3). 2.4 PID Controller The basis of the automatic control is to calculate a corrective action based on the difference between the current state of the process and the desired state. This is known as feedback control and consists of three basic blocks: a process, a PID (Proportional, Integral, and Derivative) and a sensor as feedback as shown in Fig. 4. Family members of PID controllers include three actions: proportional (P), integral (I) and derivative (D). These drivers are called P, I, PD and PID [5].

42

E. del Rocio Villamar Garcés et al.

Signal Input Reference Signal

Compressor/

Controller

Error

valves

Process /Plant

Signal output Manipulated Variable

Feedback

Sensors Fig. 3. System closed loop control.

2.5 Design of the Proposal Development of the Control Structure of the Plant From the problem the control structure of the plant is designed. Here the design of the electrical panel where electronic, electrical and more for installation and proper operation which are detailed in the following items elements are mounted components is performed. Calculation to Choose the Correct Component and Design Elements Dashboard Before detailing the design, electrical calculations and consultations took place experienced people in the field to choose the right components for controlling the cooling system. Then, each element was chosen for the system is described: [6]. Power Plant: Current calculations of the whole plant by means of measuring elements are made (Table 1).

Table 1. Current consumption of the entire plant. Element

Amount Starter current [A] Nominal current [A]

Compressor 7/8Hp MX18T R404A 220

2

16 A

8.4 A

Compressor 3/8 110V

1

5A

3.9 A

Motor Fan ELCO 127 A 50 Hz 1300–1500 rpm

3

1.6 A

1.3 A

PCL SIEMENS S7 1200

1

0.08 A

0.08 A

Valve Danfoss

1

0.4 A

0.4 A

Fan

1

0.25 A

0.2 A

29.16 A

14.28 A

Total

Design and Implementation of an Automatic System

43

Because the voltage has a variation starter is multiplied by the safety factor that is added by circuit overload. Circuit Breaker: According to the parameters of the plant a circuit breaker that allows the opening of the circuit is selected, in this case a circuit breaker is chosen with a maximum voltage of 32 A type C. Fuse Holder: There are several models or according to their classification as for motors, capacitors, transformers, utility, among others. He was chosen general purpose (gG) with an intensity of 8 A and a voltage of 500 V. Thermal Relay: For choosing proper thermal relay function that is performed it is determined, in our project opted for an electromagnetic relay brand CHINT with 220 V power, which this type of contactor to adjust the nominal to the compressor intensity. Type of Power Cord: According to the intensity and power plant, size 14 AWG wire for installation of the electronic components that support up to 500 V-20 A was chosen, and for the installation of electrical components, the wire gauge 12 AWG supporting up to 500 V was chosen −25 Â. Dimensions of Rails and Gutters: To adjust the dimensions 7 cm rails components was acquired, similarly gutters are dimensioned 5 cm × 5 cm to conceal the electrical wiring of the whole control system located inside the metal housing. Contactor: As required plant contactors we acquire a power coil 110 V, for igniting 220 V compressors and fan control Solenoid Valves and 110 V. Design and Installation of the Control Panel with Its Components On the first board or main system board electrical and electronic components are described; on the second board distribution VX-950 controller and its components are. Control System Installation • The prototype plant needs a 220 V power due to technical data that contain it, so that a circuit breaker is placed to protect the equipment in case of mishandling. This will start to function as electronic components are mainly the Programmable Logic Controller (PLC), the man-machine interface and VX-950 controller. • The temperature sensors are located inside the chamber and near the evaporator freezing showcase for greater accuracy in measurements. a sensor near the product to have data of temperature and power control with the PLC controller to which you want to have is also installed. • The pressure sensors are placed in the pipes gas passage near the compressor to have a pressure monitoring and is suitable for the cooling system. • Control was used for a Programmable Logic Controller (PLC) 6ES7 2121BE400XB0 S S7 that the PLC will perform the control of the plant.

44

E. del Rocio Villamar Garcés et al.

Installation of the Electrical Elements of the Control Panel 1: • It needs a 220 V power, which allows operation of all electrical and electronic circuit. According to the features mentioned above and 2-pole circuit breaker to be operated by a switch to have control over it and that is not operated directly connected to the outlet to avoid the high stresses that may occur is placed. Another function will cut the intensity if there is a short circuit or high currents and cause damage to the devices. • To supply the PLC, which works with 110 V/220 V, is taken a line neutral connection 220 on the circuit breaker. • The fuse holder 32 A will be fed according to the power that is required for the contactors, that is, two contactors 220 V need to drive two compressors working with one of said voltage and 110 V. In addition to providing general protection to control devices such as PLC, compressors and contactors. • Thermal relays will be responsible for protecting the compressors due to overcurrent and overheating that may exist, as if the compressors have overload by overwork. These are in series contactors, the output of the contactor with the relay input. • In the portion of the control panel door, pushbuttons, indicators, selector and HMI where you can manipulate the general lighting and commissioning the prototype plant is located. • For electrical connection cables 12 and 14 numbers according to the calculations made to determine what type of cable used is needed. In addition, grooved channels, terminal blocks 32 A, rails, terminals, among others (Fig. 5). Topological Plant Diagram An Ethernet network via a router was made because the structure is in a very accessible network that provides the university. Here they communicate the three elements to perform due configurations in: programmable logic controller (PLC), HMI and laptop screen [6] (Fig. 6). Table 2. IP routing components. Component

IP address

Mask

PC PORTÁTIL

192.168.0.100 255.255.255.0

PLC SIEMENS S7 1200 192.168.0.10

255.255.255.0

HMI DELTA

255.255.255.0

192.168.0.53

Blocks Diagram In Fig. 7 we represent the block diagram of the cooling system, where the setpoint or required and the error (this gives the controlled value), is the difference between them to have an input signal to the driver and then operate the actuator (Table 2). General Operation of the Control System For operation of the whole system, it will be controlled from the front of the metal panel (door), besides having a monitoring system with the HMI. For the control system and

Design and Implementation of an Automatic System

45

temperature sensors providing pressure information from both the camera and display case as the product to be preserved in such cooling systems they are used. These sensors are placed in the PLC inputs and this regulates the temperature and pressure operating compressors according to the configuration thereof according to the product required and obtaining desired temperatures in refrigeration systems. Declaration of Variables Before starting the process of implementing the control system must define input variables and output of the cooling system in the following table the variables that help to conduct control system with their respective processes detailed in et shown document (Table 3). Table 3. Definition of variables of the plant cooling system Elements

Statement

TEMPERATURE SENSOR Parameter output SENSOR DE PRESIÓN

Parameter output

EXPANSION VALVE

Parameter input

PLANT CURRENT

Parameter input

With the defined variables proceed to perform the respective processes for project construction. Plant Identification There are several methods for identifying a plant, an experimental project for identification according to the requirement of the plant was conducted. The experimental identification is one of the most used for a control system, with which it was chosen to work, where they are made by collecting experimental data input and output of the plant. For this use of a digital controller for data collection process it was made. Figure 8 shows graphical data of the temperature and pressure of the chamber and in Fig. 9 of the cabinet with the experimental data was obtained, see appendix for data collected. Graphics software obtained Sitrad: Fig. 8 shows the data obtained Sitrad software being green pressure used in both cases: case chamber and by design, red and blue ambient temperature of the evaporator. In this graph the ambient temperature is blue and red evaporator. From the data collected identification with the transfer function for the system is done and see if this behavior is stable in a short period of time. After obtaining the data, how to perform identification using MATLAB software is using the IDENT tool [7].

46

E. del Rocio Villamar Garcés et al.

Obtaining the Transfer Function To obtain the transfer function described below process: We started identifying the MATLAB program for this experimental data obtained in the “Import Data option is imported. The transfer functions obtained are of the following characteristics (Table 4): Table 4. Transfer function in MATLAB per system. Systems

Type of transfer function

Parameters

Percentage more like actual graphs

Plant (compressor)

Discrete

2 polos 1 zeros

73,18%

Camera

Continuo

4 polos 2 zeros

87,83%

Cooling display case Continuo

2 polos 1 zeros 2 relay 87,83%

Configure to perform an autotuning to determine the values for the PID controller of each system, the values of each controller component are presented below (Table 5). Table 5. Controller PID in MATLAB per system. Systems

KP

Compressor (Pressure) 0.02154

KI

KD

0.02154

0.005386

Camera

0.002475 1.141e−08 0.4032

Cooling display case

0

0.1526

0

PID controller values for the different systems involving the plant (Table 6). Table 6. Controller PID in PLC and full gauge per system. Systems

KI

KD

Compressor (Pressure) PLC 1

KP

20

0

Camera

0

0.122

0

Cooling display case

4.662 0.3266 16.64

After the PID controller values were found, simulation and implementation was performed on the plant in Fig. 1.

Design and Implementation of an Automatic System

47

Fig. 4. Block diagram of a PID control.

Fig. 5. Installing the electrical panel.

3 Results 3.1 Comparison of Energy Consumption of the Existing Plant with the PID Controller Cold Storage Based on tests developed in prototype systems implemented cooling energy consumption comparisons between the cooling system and unregulated system with a PID controller in the cold room are performed. According to the obtained results, the system with PID control implemented optimizes the prototype plant, giving a stability time much less than the current plant making this a more efficient system, in addition, result in lower energy costs and therefore a reduction in amounts payable.

48

E. del Rocio Villamar Garcés et al.

Fig. 6. Topological design prototype plant cooling system. Reference Signal Error Signal Input

Controller

Controller

Compressors

VX-950

PLC

220V/110V

Camera/ Cooling display case

Signal output Manipulated Variable

Electro valve

Feedback

Temperature sensors

Fig. 7. Block diagram of the cooling system.

Fig. 8. Temperature data and graphics chamber pressure. (Color figure online)

Design and Implementation of an Automatic System

49

Fig. 9. Graphical data of temperature and pressure of the showcase. (Color figure online)

3.2 Comparison of Energy Consumption of the Existing Plant with the PID Controller Refrigerated Showcase As the cold energy consumption comparisons are made between the cooling system and unregulated system with a PID controller refrigerated showcase (Table 7). Table 7. Data energy consumption of the existing. Description

Values of the plant Values of the cooling display case

Values of the camera

Consumption in KwH 752 KwH/mes

439,72 wH/mes

440,03 KwH/mes

Stabilization time

1800 s

400 s

500 s

Cost per month

$30.08

$17,58

$17.60

According to the obtained results, the system with PID control implemented optimizes the prototype plant, giving a stability time much less than the current plant making this a more efficient system, in addition, result in lower energy costs and therefore a reduction in amounts payable. 3.3 Stabilization Time Comparison Between the Existing Plant and the Plant Control System After performing the tests of stability of the system in the chamber and refrigerated showcase, it is that with the implementation of the PID controller response time in which the system reaches the setpoint value is best compared with the existing plant, providing optimal and efficient system (Table 8).

50

E. del Rocio Villamar Garcés et al. Table 8. Comparison of stability time. Valor setpoint Planta actual Control PID 20 psi

1800 s

1080 s

With the data shown is that the resulting difference in stability between the existing plant and the control system with a time of 720 s. This shows that the implemented control system improves the prototype plant by 67%. 3.4 Compared to Other Jobs Quantitative, qualitative comparison of other controllers and how much energy can be saved with each of them, is the reason that inspired for this project [7]. In supermarkets there are refrigeration equipment with two controllers, which are the most used (PID and on/off). In analyzing the prototype cooling was performed between existing plant without controller and the PID controller plant floor [15]. Optimization was observed in operation, where energy savings of 40% in Fig. 45 also in other documents reflected showed that the fuzzy logic control can further reduce power consumption by similar equipment operating in supermarkets reaching up to 60% without load [14] (Fig. 10).

Porcentaje de consumo eléctrico

Consumo con control difuso 40%

Consumo Planta con PID 60%

Fig. 10. Comparison of percentage of electricity consumption in this plant over two controllers [7].

In Ecuador it does not have companies that manufacture these devices, so investment in equipment and plant to experience different control strategies, also serves to analyze variables such as disturbances (climate) that occur in these systems according to the site where the plant is analyzed. The comparison is obvious and can be checked with different items of reduced consumption, electricity savings according to the type of control used in these systems [13].

Design and Implementation of an Automatic System

51

4 Discussion or Conclusions • With the implementation of the interface HMI could monitor the values in psi providing the pressure sensor, showing that when the pressure exceeds the values of active Setpoint an alarm indicating the unstable behavior of the compressors. • the controller designed for cold storage based on the transfer function obtained, wherein temperature ranges between 4 °C was set at −8 °C, for durability of foodstuffs was introduced. • In identifying the plant acquiring real-time data was performed to obtain the different transfer functions of the camera, display and compressors with estimated percentages of 72.62% (continuous type), 87.83% (continuous type) and 73.18% (discrete type) respectively, obtaining stable for controller design results. • Refrigeration plant is useful for the student who chooses to take the chair of automation and control, with the help of a trained person to identify and manipulate all system installations. • Time comparison between stabilization system implemented control and the current plant has a time difference of 720 s. Thus, it concludes a PID control system is more efficient in the cooling system prototype. • Power consumption with PID controller, shown in Fig. 46, if consumption is 60% the reduction is 40%, even in the thesis can display a savings of up to 60% with other controller as the fuzzy logic.

5 Recommendations – It is recommended to obtain the transfer function of the plant model, perform the experimental acquisition of the collected data with a higher range than this project, in order to have greater samples and better results in the identification of the model of the plant. – The use of measuring tools is recommended when working with electrical wiring since it allows to have current or voltage readings checking that the plant has enough power to the electronics and Electrical. This calculates the energy savings that this demand sits. – For the reading of the temperature of the product that is placed as a reference inside the display case or cold room, it is advisable to place the temperature sensor near the product to have an exact value. – When using the controllers in the project, PLC and VX-950, it is advisable to make the respective measurements independently and have no inconvenience for the readings of the elements to be controlled, but when working independently the Readings made of both temperature and pressure have a difference range of 4% above a value normally accepted as 1%. This excludes the VX-950 controller from the topological diagram after testing, prioritizing the Programmable Logic Controller (PLC) according to the objective set.

52

E. del Rocio Villamar Garcés et al.

References 1. B & Company: Ministry of Industry and Productivity. Ministry of Industry and Productivity (2015) 2. Compostela, S.: jmrivas. School Technical Professions, 1 January 2002 3. EYDC electrical engineering department, department of electrical, electronic engineering and control, 1 January 2011 4. Inductive Automation. Inductive Automation, 10 August 2018 5. Victor Manuel: Victor blog, victor, 28 August 2015 6. EDR Villamar garcés, evaluation of two control strategies to improve energy efficiency of a parallel system for cooling conpresión: fuzzy control and discrete control spaces of states, Guayaquil - Ecuador: Escuela Superior Politecnica del Litoral (2014) 7. Garcia, I.: National University of Loja, National University of Loja, 1 September 2010 8. Free software. Media programs, 15 January 2019 9. Andrew, P.: Universidad de Piura. University of Piura, 1 March 2017 10. Alfaya, J.A., Bejarano, G., Ortega, M.G., Rubio, F.R.: Multi-operating-point robust control of a one-stage refrigeration cycle. In: European Control Conference (ECC), Linz, Austria. IEEE (2015) 11. Shafiei, S.E., Rasmussen, H., Stoustrup, J.: Supermarket refrigeration systems modeling for demand-side management. Department of Electronic Systems, Automation and Control, Aalborg University, Aalborg, 8 February 2013 12. Witt, H., Taylor, S., Lomas, K.J., Liddiard, R.: Simulation of energy use in UK supermarkets using EnergyPlus. Loughborough University’s Institutional (2015) 13. Ge, Y.T., Tassou, S.A.: Mathematical modeling of supermarket refrigeration systems for design, energy prediction and control. Department of Mechanical Engineering, Brunel University, Uxbridge, Middlesex, UK (2000) 14. Bejarano, G., Alfaya, J.A., Rodríguez, D., Morilla, F., Ortega, M.G.: Benchmark for PID control of refrigeration systems based on vapor compression. University of Sevilla (2018)

Design of Emergency Call Record Support System Applying Natural Language Processing Techniques Andrea Trujillo(B)

, Marcos Orellana , and María Inés Acosta

Universidad del Azuay, Av. 24 de Mayo 7-77, Cuenca, Ecuador {atrujillo,marore,macosta}@uazuay.edu.ec

Abstract. Currently Command and Control Centers (C2), such as the Integrated Security Service ECU 911, are managed by Computer Assisted Dispatch Systems (CADS). These systems facilitate the registration of incidents and the distribution of rescue resources. However, the information registration process does not yet have automated methods. Important data such as: name, address, reference and categories are recorded manually, which generates problems of loss of information and inefficiency, in terms of time and attention to the incident. As a solution to these problems, the design of an emergency call record support system is proposed, based on Natural Language Processing (NLP) techniques and algorithms. Taking into consideration an analysis of the processes of the ECU 911, three modules are proposed: (1) transcription of audio to text calls (ASR), (2) extraction of relevant information (NER) such as: address and references; and (3) call classification (TF-IDF/SVM) according to service and priority. Thus obtaining the design of an automated support system for CADS, which provides quality information in a timely manner. Keywords: Emergency call · Natural Language Processing · Transcription · Information extraction · Classification of emergencies

1 Introduction Command and Control Center (C2) such as ECU 911 [1] in Ecuador, manage their processes through Computer Assisted Dispatch Systems (CADS). These systems have a fundamental role to play in the administration of rescue units. In addition, they provide important information in the registration and decision-making processes. Currently, CAD systems used in emergency departments around the world have three main functions: (1) call reception: attention and verification of information, prioritizing location, (2) identification of resources and rescue institutions: allocation of articulated response agencies (National Police, Fire Department, Ministry of Health, Army, etc.); and (3) mobilization and resource management: sending resources and evaluating the performance of rescue planning [2]. The use of CAD systems has indicated that the entry of information occurs during the initial call [3]. For example, when validating information such as: location and site © Springer Nature Switzerland AG 2020 E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 53–65, 2020. https://doi.org/10.1007/978-3-030-35740-5_4

54

A. Trujillo et al.

references, operators try to collect this data through the caller or using a geospatial location system during the call [4]. Location systems generally have inaccuracies and a considerable range of error when positioning the call. This problem commonly occurs in rural sectors, since those are sectors without adequate delimitation in the geographical area of dominance, which produces conflicts in attention and agility. In emergencies, arrival time is crucial, considering the environment that involves the incident. According to the report of the San Francisco Fire Department [5], the time used in the emergency analysis process is between 35% and 45% of the total unit deployment time. Similarly, the monthly statistics of the ECU 911 Ambato [6] report an average response time of 19 min between the call and the analysis of information for the dispatch of resources. Consequently, the emergency response compromises the level of efficiency in terms of time. Emergency departments work in an environment of great pressure and stress, as they are exposed to the attention of sensitive calls. This environment makes registration and dispatch processes difficult, generating unreliable information due to the subjectivity of the operators, typical of the situation. Likewise, these systems tend to be vulnerable if there are points of failure in their structure, that is, if one of the processes has a lag, this will affect the subsequent ones, and consequently a delay in emergency responses [7]. A challenging issue in CAD systems is how to improve the registration and categorization of information, that is, what information should and can be automated. In this way, the information collected requires to be entered according to the description of the caller. Some of the data collected in the registration process are: (1) location of the caller, which allows the operator to know the location of the incident, (2) the categorization by service, which includes the assignment of the responsible rescue institution, (3) classification by priority, which determines the level of speed of according the event. The faster this information is retrieved, the faster the response will be. The literature reflects related works to emergency call location and classification systems where classification criteria are exposed through predictive factors during the call [8, 9]. In addition, there was found research that seeks to determine the location of calls through mobile phones and geographic information systems [4, 10]. However, regarding the location, researchers opt for positioning technologies, which due to their inaccuracy are still relying on the manual registration of the operator. Similarly, in terms of classification, the investigations show call categorization systems supported by studies subsequent to emergency care. Based on the problems evidenced in the emergency call log, a technological solution is proposed that follows the principles of common language, such as Natural Language Processing (NLP). The objective of the investigation is to manage the information of the people during an emergency call, since in this phase the entry of relevant information begins. For this, a call log support system model is designed, based on NLP techniques and algorithms. The present document is structured as follows: Sect. 1 is the introduction, then, Sect. 2 presents the related work in the process of registering calls in CAD systems. Section 3 describes the proposed design for the emergency registration support system based on NLP; and finally, in Sect. 4 the conclusions and benefits that would be obtained with a support system in the registry of emerging calls are presented.

Design of Emergency Call Record Support System Applying NLP Techniques

55

2 Related Work The CADS manage the rescue resources in the C2 and the Medical Priority Dispatch Systems (SDPM). Although these systems have automated processes, there are still manual tasks [4], such as: entering names, registering addresses and place references and categorizing calls. This information makes the operation of the system complex, because problems are generated such as: delays due to the writing speed of the operators, background noise, and incomplete or ambiguous information that the caller could present [8]. Based on these problems, related works involving the use of NLP techniques in the emergency registration process were investigated. NLP techniques are used in tasks of reading and understanding audio and text in large quantities, detecting similarities and differences [11]. The purpose of the application of NLP techniques is to classify and extract information automatically, for the benefit of the attention time of the caller. The main source of information in CADS is the caller, who narrates the event in informal language. Therefore, the incoming call can be processed and transcribed, in order to show information in greater detail and in a format that facilitates its processing. The authors of [12] exposed the transcription of calls in dispatch system algorithms. This study analyzed the interaction between caller and operator to identify the factors that affect the execution of medical maneuvers guided by office. In this work, the Channel Trans software from the University of California was used to audit incoming calls and give an assessment of the medical office service. The transcripts of this study were not executed in real time and the software used has support only for English language. On the other hand, in [13], the authors studied the transcription of emergency calls outside hospitals. In this study, these transcripts were coded as linguistic and interactional variables. The purpose was to answer a series of research questions about the recognition of cardiac arrests outside the hospital. The transcription software was Child Language Analysis (CLAN), a support tool for manual transcriptions, with greater support in English. Finally, in [8] it was presented a machine-learning model based on the collection of recorded calls of type “cardiac arrest”, its purpose was to recognize this kind of emergency in future calls. The tools and software used in previous research did not demonstrate reliability when using in Spanish language and in real-time transcription. However, there are tools with high performance in voice recognition. Among these tools are: Google Cloud Speech-to-text [14–16], IBM Watson Speech-to-text [17], Amazon Transcribe [18] and PocketSphinx [19]. Most of these tools are based on cloud storage and require payment for service. The few free and offline tools such as PocketSphinx do not yet have a stable Spanish model, in addition, it requires training and adaptation of the existing acoustic model [19]. However, the latter is widely used in audio transcripts with a common language and little processing. In the text classification area, CADS require information and descriptive factors in terms of priority level (yellow, orange, red and green) and service (Health management, Citizen Security, Traffic and mobility, etc.). In [9] the authors investigate the predictive factors for the categorization of emergency calls and their effect on response times and mortality risks. This approach is directed towards two types of calls: (1) with relevant information and (2) with ambiguous information. In the same area, in [20] it was study a group of selected characteristics of emerging events, such as: crimes, traffic and public

56

A. Trujillo et al.

intervention. Additionally, they investigate the classification in terms of characteristics of the caller such as: age, time of call of men and intonation, with a classification model based on artificial neural networks. In [21] the authors define a classification model, post-emergency. This study is based on decision trees, whose set of categories to be classified is composed of: first aid, disasters and rescue. Currently, CADS record the location and reference of the event, and have positioning systems with GPS technology [4, 10]. This type of system presents a considerable margin of error in emergency events, mainly in rural areas or sectors outside the geographical coverage area, which is why the address and location reference depends directly on the information issued by the caller. In the NLP area, the Name Entity Recognition (NER) is presented, its objective is to improve the performance of NLP tasks such as: information extraction, query response systems, automatic summary, etc. [22]. However, its use in CADS processes has not been considered in previous research. The NER analysis can be a solution to location models. This technique extracts data from entities such as addresses and names of organizations (banks, schools, colleges, etc.), thus generating more information for operators. Among the most used algorithms in NER, there are probabilistic models such as Conditional Random Fields (CRF) [22, 23]. This model can be applied in different languages, since it is an algorithm that feeds on previous supervised and unsupervised characteristics. Features for supervised analysis require large amounts of tagged data, to achieve good performance, while unsupervised ones do not require large datasets to assess their performance [24]. In contrast to the goal that has been set, the literature review demonstrates a reduced analysis of processes, systems or technologies that improve the performance of CADS. The studies focus on post-emergency care processing, and the tools or software used are better suited for the English language.

3 CAD Support System Model The reports made by the ECU 911 service demonstrate 72% of emergencies are made from landlines and cellphones. The remaining emergencies are received from video surveillance, panic buttons, or other mechanisms of warning [25]. For this reason, the emergency calls registration was chosen as the main process to evaluate and improve in a CAD system. Based on the problems recognized in the registration process of calls in CADS, and the process of registration calls of the C2 ECU 911, it is proposed a support system model in the phase of receiving calls. The model consists of three main modules: transcription of calls to text, information extraction, and call classification (See Fig. 1). The methodology is based in a qualitative research. The information collected of the call registration process was obtained through interviews and visits to the facilities of the ECU 911 service. The purpose was to analyze methods, technologies and information relevant to the current state of the process. A variety of research questions were generated through the data collected. Such questions seek to identify interaction factors that could affect the sending of information and the necessary cooperation between the caller and the operator. Additionally, the proposed system was based on a previous analysis of the main source of information, which is the caller. The information in this analysis was generated

Design of Emergency Call Record Support System Applying NLP Techniques

57

Fig. 1. CADS support system model

through the simulation of calls, based on observations and experiences that are presented in the registration process. The generated calls provided a study of events that occur at this stage, and that should be considered in the system. In addition, a general call reception process was defined (See Fig. 2), grounded on the current state of the process in the ECU 911 service. For the design of the support system, only the information-recording phase was considered, because the following phases are in charge of the rescue institutions associated with ECU 911.

Fig. 2. Call reception process

Subsequently, the modules designed in the support system are presented. Similarly, the processes and algorithms proposed to solve the current needs and problems that arise in CAD systems are described. 3.1 Transcription Module The function of the transcription module is to obtain information in an easily processed format. Likewise, the information will serve as an input source in the subsequent modules. The call transcription through NLP techniques is applied in Automatic Speech Recognition (ASR), a technology that improves human-computer interactions [25, 26]. Based on this technique, an online voice recognition system is proposed, due to its significant characteristics and support, which are offered for audio transcription (See Fig. 3).

58

A. Trujillo et al.

Fig. 3. Transcription module design.

Among the voice recognition tools evaluated according to the needs of the model (See Table 1) Google Cloud Speech-To-Text stands out. This tool has a variety of languages in the voice recognition engine. In addition, the regional division service provides a training corpus typical of the locality. Table 1. Tools automatic speech recognition. Tools IBM watson

Features Free Offline Multiple language

Precision/Noise

No

No

11

No

Google cloud STT

No

No

120

Yes

Amazon transcribe

No

No

7

No

PocketSphinx

Yes

Yes

12

No

The proposal begins with the communication between caller and operator. This information is transmitted on separate channels, that is, the operator and the caller are communicating through different audio tracks; however, the call is recorded in a single file. Google Cloud Speech-to-text, provides multichannel recognition, in other words, recognizes each channel separately and performs transcription sequentially. Additionally, the tool offers a prior training of its language model, improving overall performance. Based on the model training service, the transcription of a considerable number of calls is suggested, in order to train the model with terminology and languages from the region. 3.2 Text Preprocessing Module The audio transcription generates plain text files, expressed in an informal language without processing, that is, the produced text exhibits a large amount of irrelevant content for analysis. Therefore, a text pre-processing is proposed, which generates clean and easily processed transcripts. The text pre-processing techniques considered would reduce the dimensionality and therefore the complexity when applying a NLP analysis or method. These techniques are defined as follows:

Design of Emergency Call Record Support System Applying NLP Techniques

59

Remove Special Characters. Reduce the processing time of information with terms that are not relevant to the text [27]. Lowercase. Reduces the complexity and number of different terms in the texts [27]. Remove Stop Words. Discriminates words that are not relevant in the text such as conjunctions, pronouns, articles, etc. [28]. Taking into account that some of these words are considered as connectors that can intervene in the information extraction process; an own dictionary of stop words for the set of texts is created. Lemmatization. Unification of terms at a single time and person [27].

3.3 Extraction Information Module According to the incident reporting protocol of the ECU 911 service, the caller must provide the following information [1]: • • • • • •

Emergency type. City. Main and Secondary Street. Neighborhood or Sector. Identify reference places (pharmacy, park, church, business, etc.) First Name and Last name.

The manual registration of the exposed information creates the need to automate this process through the extraction of information, based exactly on the NER. The NER analysis as a solution to the extraction of information is evaluated from three approaches: Systems Rule-Based. Search for clues within the structure and grammar of the text to indicate that a named entity exists. In addition, it can be equipped with regular expressions that extract patterns of entities such as dates, currencies, identifications, etc. [29, 30]. Systems Based on Machine Learning. They establish their operation in the search for patterns and relationships in documents, to create models with algorithms, supervised, unsupervised and semi-supervised [30]. Systems Based on Knowledge Basis. They require a previous dictionary, a corpus or an external source, which includes the entities that can be presented within the text [29]. Considering the text features, a model based on supervised machine learning is exposed, in order to verify the accuracy and performance of the model. In addition, to create a corpus of its own, to achieve greater precision, when working with languages and terminology of the place (See Fig. 4).

60

A. Trujillo et al.

Fig. 4. Information extraction module design.

Among the machine learning systems for NER analysis are statistical and probabilistic models. Probabilistic models generally use CRF-based algorithms. This algorithm is present in libraries such as: Stanford [31], Sklearn [32] and CRF++ [33]. One of the features of the CRF algorithm is the adaptability to different languages, and the opportunity to combine a system based on rules and a system based on machine learning. However, the libraries mentioned do not yet have efficient models for Spanish, which cover the needs of the support system. As a solution, the SpaCy library [34] is presented, with a statistical model based on a random training corpus, using the probability to identify named entities. In addition, to have support and adaptability in their models for Spanish language. Alleging that the NER analysis study in Spanish is poorly studied and the scarce research makes use of structured and generic corpus such as AnCora and CoNLL2002, which do not adapt to the model to be developed, a real comparison between libraries cannot be made. However, SpaCy is one of the best performing libraries and is substantially faster than many other libraries [35] and according to the documentation set out in its API, this library has programming adjustment features in its NER model for Spanish language. This makes it the best option for the design of the support system. 3.4 Classification Module The emergency calls classification takes an important role in integrated security services, whose objective is the communication and assignment of emergencies to rescue institutions. The call reception process of ECU 911 has two important categories of incidents: service and priority. The type of service defines the rescue institution responsible for the emergency. The type of priority specifies the level of urgency of the incident. Based on this analysis, two call classification modules are proposed. These modules look for predictive factors, to categorize calls according to the service and the level of priority with which they should be treated.

Design of Emergency Call Record Support System Applying NLP Techniques

61

There are multiple algorithms for automatic document classification such as: Naïve Bayes, Support Vector Machine (SVM), K-Neighbors Neighbors (K-NN), neural networks, decision trees, among others. However, in classification models, the representation of documents play a significant role in accuracy. Machines and algorithms do not include text or characters, it is important to convert them into an understandable format for the machine, to perform any text analysis [36]. This work is done during the representation of documents, where the text to be classified is transformed according to statistical models based on the weight or frequency of the terms. Some of the most used models in the representation of documents are: TF-IDF [37–39], Latent Dirichlet Allocation (LDA) [40], word bag [37] and Paragraph Vector (PV) [41]. TF-IDF is considered a word weighting algorithm, recognized for its accuracy in classification models. This technique considers two approaches to representation: (1) TF (term frequency): indicates the frequency of occurrence of a term in a document, and (2) IDF (inverse document frequency): indicates the frequency of occurrence of the term in the collection of documents [40]. Based on Nakamori [41], in his comparative research between: LDA, TF-IDF and PV, combined with the SVM classification algorithm, he concludes that TF-IDF shows the best results. As for LDA, it is an algorithm that requires longer texts to infer and determine the type of content. On the other hand, PV exhibits low performance results, compared to the previous algorithms. The study relied on metrics such as: speed and performance. According to this data, LDA has a disadvantage when working with transcribed calls because they are short texts that would prevent deducing the type of content. In this way, it is argued that TF-IDF when considering the weight and the burden it takes on the document and the set of documents, works equally in short and long texts. This and other features give it a greater advantage over the LDA technique (See Table 2). Regarding to the classification algorithm, SVM is chosen, according to the characteristics and advantages it has over other algorithms. This conclusion is based on a comparative analysis between classification algorithms, presented in Table 3. Table 2. Documents representation techniques. Techniques Features

Limitations

LDA

- Generative model - Reduce dimensions - Application: grouping and classification of documents

- The documents need very strong trends to identify the topic - Intrusive words involved in the distribution - Word bag for each topic

TF-IDF

• Statistical model - Increase dimensionality in their • Word frequency at the document level vectors and in the complete set of documents • Application: grouping, document classification, SEO, etc.

Finally, the design of the classification phase is presented in Fig. 5. The modules exposed, unlike the literature reviewed, focus on immediate classification, that is, when the call is over, predictive factors are sought that establish a category for the text. The

62

A. Trujillo et al. Table 3. Documents classification algorithms.

Algorithms

Features

Limitations

SVM

- High precision - Linear and nonlinear model - The presence of kernel variations makes it robust to other algorithms

- It requires a training corpus

(C4.5)

- It has an easy interpretation and implementation - Variation in data leads to different decision trees - Robust with noisy data

- Requires a large set of training data - Overfitting

K-NN

- It doesn’t need linearly separable classes - Its learning process has zero costs - Robust with noisy data (relatively)

- Excessive training time - It may be sensitive to noisy

Naïve Bayes

- Computational efficiency and classification - Simple implementation

- The number of records affects the accuracy

Neural networks - It does not need reprogramming - High precision

- High processing time - Complicated adjustment parameters (layers, neurons) - Slow learning

development of the classification modules has the same process for categorization by service and priority, except for some configuration parameters and the labeling of the transcribed calls.

Fig. 5. Classification modules designs.

Design of Emergency Call Record Support System Applying NLP Techniques

63

4 Results The result of this research is the design of emergency call record support system, compatible with the requirements and infrastructure of the Integrated Service ECU 911. As a proposal, its performance would be tested in the development phase; however, it has been made in base to techniques and tools that meet with objective of improving the time for registration emergency calls. The design has limitations and weaknesses that should be considered, an example of this is the informal and varied language that occurs in calls of this type, which causes a difficult grammatical structuring of the corpus, therefore, we should evaluate in different C2’s since other needs can be presented. In addition, a limiting can be its external use, since its design is based on a national reality, that is, this proposal is tied on the processes found in the command and control center ECU 911 of Ecuador.

5 Conclusions and Future Work This work presented the design of a support system that will increase efficiency and response time in the process of registering calls in a CAD system. The proposed design was based on NLP techniques and algorithms, with the aim of receiving quality information and in less time. In this way, the loss of data generated in registration and validation processes that are carried out manually by the operators would be avoided. Among the information that is expected to be extracted is: address, references, categorization by service and categorization by priority, the same that would pass to a second instance, known as the dispatch process. Consequently, the proposed system expedite the attention in emergency incidents. For future works, the development of the system is proposed, a structured corpus that serves as the basis for the NER analysis in Spanish and, an investigation that improves the tools or models of automatic voice recognition in Spanish, attached to the needs of rescue institutions. Acknowledgment. This research was supported by the vice-rectorate of investigations of the Universidad del Azuay. We thank our colleagues from Laboratorio de Investigación y Desarrollo en Informática (LIDI) de la Universidad del Azuay who provided insight and expertise that greatly assisted this research.

References 1. Servicio Integrado de Seguridad ECU 9111. Servicio Integrado de Seguridad ECU911 (2019). http://www.ecu911.gob.ec/. Accessed 2 May 2019 2. Beynon-Davies, P.: Human error and information systems failure: the case of the London ambulance service computer-aided despatch system project. Interact. Comput. 11, 699–720 (1999). https://doi.org/10.1016/S0953-5438(98)00050-2 3. Seattle Fire Department: Surveillance Impact Report: Computer-Aided Dispatch (CAD) (2019) 4. Zhang, J., Zhang, M., Ren, F., et al.: Enable automated emergency responses through an agentbased computer-aided dispatch system. In: Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems, pp. 1844–1846 (2018)

64

A. Trujillo et al.

5. San Francisco Fire Department. Fire Department Calls for Service (2018) 6. Unidad de Estadística y Evaluación ECU 911 Ambato. Estadísticas mensuales de alertas recibidas, incidentes atendidos y despachos realizados por el ECU 911 Ambato (2013) 7. Souza, J., Botega, L.C., Santar, E., et al.: Conceptual framework to enrich situation awareness of emergency dispatchers. In: 17th International Conference, HCI International 2015, Los Angeles, CA, USA, 2–7 August 2015, Proceedings, Part II, pp. 33–44 (2015) 8. Blomberg, S.N., Folke, F., Ersbøll, A.K., et al.: Machine learning as a supportive tool to recognize cardiac arrest in emergency calls. Resuscitation 1–8 (2019). https://doi.org/10. 1016/j.resuscitation.2019.01.015 9. Møller, T.P., Kjærulff, T.M., Viereck, S., et al.: The difficult medical emergency call: a registerbased study of predictors and outcomes. Scand. J. Trauma Resusc. Emerg. Med. 25, 1–9 (2017). https://doi.org/10.1186/s13049-017-0366-0 10. Yeh, L.-Y., Tsaur, W.-J., Huang, H.-H.: Secure IoT-based, incentive-aware emergency personnel dispatching scheme with weighted fine-grained access control. ACM Trans. Intell. Syst. Technol. 9, 1–23 (2017). https://doi.org/10.1145/3063716 11. Nakata, T.: Text-mining on incident reports to find knowledge on industrial safety. In: Proceedings - Annual Reliability and Maintainability Symposium (2017) 12. Clegg, G.R., Lyon, R.M., James, S., et al.: Dispatch-assisted CPR: where are the hold-ups during calls to emergency dispatchers? A preliminary analysis of caller-dispatcher interactions during out-of-hospital cardiac arrest using a novel call transcription technique. Resuscitation 85, 49–52 (2014). https://doi.org/10.1016/j.resuscitation.2013.08.018 13. Riou, M., Ball, S., Williams, T.A., et al.: The linguistic and interactional factors impacting recognition and dispatch in emergency calls for out-of-hospital cardiac arrest: a mixed-method linguistic analysis study protocol. BMJ Open 7, 1–8 (2017). https://doi.org/10.1136/bmjopen2017-016510 14. Ashwell, T., Elam, J.R.: How accurately can the Google web speech API recognize and transcribe Japanese L2 English learners’ oral production? Jalt Call J. 13, 59–76 (2017) 15. Iancu, B.: Evaluating Google speech-to-text api’s performance for romanian e-learning resources. Inform. Econ. 23, 17–25 (2019). https://doi.org/10.12948/issn14531305/23.1. 2019.02 16. Stefanovic, M., Cetic, N., Kovacevic, M., et al.: Voice control system with advanced recognition. In: 2012 20th Telecommunications Forum, TELFOR 2012 – Proceedings, pp. 1601–1604 (2012) 17. IBM Corporation. Watson Speech to Text (2019). https://www.ibm.com/es-es/cloud/watsonspeech-to-text. Accessed 2 June 2019 18. Amazon Web Service. Amazon Transcribe: Automatic speech recognition (2019). https:// aws.amazon.com/transcribe/. Accessed 2 June 2019 19. Principi, E., Squartini, S., Bonfigli, R., et al.: An integrated system for voice command recognition and emergency detection based on audio signals. Expert Syst. Appl. 42, 5668–5683 (2015). https://doi.org/10.1016/j.eswa.2015.02.036 20. Balcerek, J., Pawlowski, P., Dabrowski, A.: Classification of emergency phone conversations with artificial neural network. In: Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA, pp. 343–348 (2017) 21. Lee, K., Kim, J.K., Park, M.W., et al.: A situation-based dialogue classification model for emergency calls. In: 2017 International Conference on Platform Technology and Service, PlatCon 2017 – Proceedings, pp. 1–4 (2017) 22. Gutiérrez, R., Castillo, A., Bucheli, V., Solarte, O.: Named Entity Recognition for Spanish language and applications in technology forecasting Reconocimiento de entidades nombradas para el idioma Español y su aplicación en la vigilancia tecnológica. Rev. Antioqueña las Ciencias Comput. y la Ing Softw. 5, 43–47 (2015)

Design of Emergency Call Record Support System Applying NLP Techniques

65

23. Molina, C.A.C., Gutierrez, R.E., Solarte, O.: Prototipo para el reconocimiento de entidades nombradas en el idioma Español. In: 2015 10th Colombian Computing Conference, 10CCC 2015, pp. 364–371 (2015) 24. Copara, J., Ochoa, J., Thorne, C., Glavas, G.: Exploring unsupervised features in conditional random fields for spanish named entity recognition. In: Proceedings - 2016 5th Brazilian Conference on Intelligent Systems, BRACIS 2016, pp. 283–288 (2017) 25. Ziman, K., Heusser, A.C., Fitzpatrick, P.C., et al.: Is automatic speech-to-text transcription ready for use in psychological experiments? Behav. Res. Methods 50, 2597–2605 (2018). https://doi.org/10.3758/s13428-018-1037-4 26. Yu, D., Deng, L.: Automatic Speech Recognition. Springer, London (2015) 27. Symeonidis, S., Effrosynidis, D., Arampatzis, A.: A comparative evaluation of pre-processing techniques and their interactions for Twitter sentiment analysis. Expert Syst. Appl. 110, 298–310 (2018). https://doi.org/10.1016/J.ESWA.2018.06.022 28. Krouska, A., Troussas, C., Virvou, M.: The effect of preprocessing techniques on Twitter sentiment analysis. In: 2016 7th International Conference on Information, Intelligence, Systems & Applications (IISA), pp. 1–5. IEEE (2016) 29. Moreno, E.I., Dra, A.M., Dra, C., Azc, P.: Reconocimiento y clasificación automatizada de Entidades Nombradas en documentos medievales (s. XIV): Libro Becerro de las Behetrías. Universidad III de Madrid (2017) 30. Alicia Pérez, M., Carolina Cardoso, A.: Técnicas de extracción de entidades con nombre. Intel. Artif. 17, 3–12 (2014) 31. Stanford NLP Group. Stanford Named Entity Recognizer (NER) (2019). https://nlp.stanford. edu/software/CRF-NER.html 32. Korobov, M.: Sklearn Crfsuite (2015). https://sklearn-crfsuite.readthedocs.io/en/latest/ tutorial.html. Accessed 3 June 2019 33. Kudo, T.: CRF++: Yet Another CRF toolkit (2003) 34. Explosion AI. Industrial-Strength Natural Language Processing. In: Train. named entity recognizer (2015). https://spacy.io/. Accessed 3 June 2019 35. Al Omran, F.N.A., Treude, C.: Choosing an NLP library for analyzing software documentation: a systematic literature review and a series of experiments. In: IEEE International Working Conference on Mining Software Repositories, pp. 187–197 (2017) 36. King, B.E., Reinold, K.: Natural language processing Recipes (2019) 37. Nursalman, M., Kusnendar, J., Fadhila, U.F.: Implementation of k-nearest neighbor with cosine similarity for classification abstract international journal of computer science. In: 2018 International Conference on Information Technology Systems and Innovation, ICITSI 2018 - Proceedings, pp. 43–48. IEEE (2019) 38. Yang, Z., Yang, D., Dyer, C., et al.: Hierarchical attention networks for document classificatio. In: Proceedings of NAACL-HLT, pp. 1480–1489 (2016) 39. Kim, D., Seo, D., Cho, S., Kang, P.: Multi-co-training for document classification using various document representations: TF–IDF, LDA, and Doc2Vec. Inf. Sci. (Ny) 477, 15–29 (2019). https://doi.org/10.1016/j.ins.2018.10.006 40. Hakim, A.A., Erwin, A., Eng, K.I., et al.: Automated document classification for news article in Bahasa Indonesia based on term frequency inverse document frequency (TF-IDF) approach. In: Proceedings - 2014 6th International Conference on Information Technology and Electrical Engineering: Leveraging Research and Technology Through University-Industry Collaboration, ICITEE 2014, pp. 0–3 (2015) 41. Nakamori, Y.: Performance comparison of TF*IDF, LDA and paragraph vector for document classification. Knowl. Syst. Sci. 2, 225–235 (2016). https://doi.org/10.1201/b15155

Integrating ISA-95 and IEC-61499 for Distributed Control System Monitoring Jairo D. Llamuca1 , Carlos A. Garcia1 , Jose E. Naranjo1 , Cesar Rosero1 , Edison Alvarez-M1 , and Marcelo V. Garcia1,2(B) 1 Universidad Tecnica de Ambato, UTA, 180103 Ambato, Ecuador {jllamuca6533,ca.garcia,jnaranjo0463,cesararosero,ealvarez, mv.garcia}@uta.edu.ec 2 University of Basque Country, UPV/EHU, 48013 Bilbao, Spain [email protected]

Abstract. The automation solutions over time have allowed relating and integrating different applications; this is complicated when involving the organization functional levels, so they need systems better flexibility, interoperability, and greater integrity and scalability in information management. This has been a trigger for the startup of Industry 4.0, which presents any solution with representative challenges, so it is of great interest to propose design methods that facilitate the portability of control systems and the exchange of information between heterogeneous systems. For this context, for the development of this project is selected; the ISA95 standard, to adopt flexibility with an optimal exchange of information; the IEC-61499 standard, to provide portability and interoperability; and SOA, integration way in Industry 4.0. This paper proposes the development of a modular communication architecture of efficient distributed systems (low cost) under the IEC-61499 standard, whose parameters are monitored framed to the data flow of the equipment model of the ISA95 standard, using an OPC-UA’s SOA framework with MQTT for its integration. In this paper, the result shows a flexible lightweight communication architecture with enhancing scalability and interoperability.

Keywords: Distributed systems MQTT communication protocol

1

· IEC-61499 · ISA-95 · OPC-UA ·

Introduction

The current need to optimize production processes, in order to avoid losing competitiveness, has required organizations to incorporate new solutions into their processes. This leads to technological advancement in the industry in order to drive companies to produce customized products, but which leads, according to the analysis of [2], the increased use of information in the processes, together with the need to develop techniques oriented at distributed control and integrated monitoring systems. c Springer Nature Switzerland AG 2020  E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 66–80, 2020. https://doi.org/10.1007/978-3-030-35740-5_5

ISA-95 and IEC-61499 for Distributed Control System Monitoring

67

Control and monitoring actions are the center of the need for the exchange of information, they adopt a hierarchical disposition, which implies according to [20], traffic problems and low information flow rates, by the transfer of frequently incompatible data. A perfect communication conduit for data flow control, between hierarchical levels, is vertical integration, which according to [10], today it is an ambition to gain competitiveness against modern systems, due to its approach of Industry 4.0 (I4.0) with links to Industrial Internet of Things (IIoT). The architecture defined in FB types (Function Blocks) by IEC-61499 provides portability and interoperability benefits at plant levels, which allow to optimize resources in flexible systems, and enables the integration of distributed applications from design documentation and data from other standards [15]. IoT Communication Protocols ensure interoperability between incompatible systems, certain protocols can be combined without losing interoperability, according to the analysis [3] and the study [1], MQTT is a lightweight and mature protocol that enables high-frequency data aggregation from the plant for vertical connectivity, it also enables to be perfectly combined with Middleware protocols such as OPC-UA at high levels; this has prompted the OPC committee to currently work on an OPC-UA PUB/SUB solution on MQTT protocol. In view of all the above. This project proposes a lightweight IoT-based communication architecture, for network monitoring of equipment parameters. This work presents a novel method based on the combination of OPC-UA and MQTT protocols and IEC-61499 and ISA-95 standards, to extract information of interest from equipment and physical assets. Architecture interoperability allows information to be transferred in real time from the plant to the server in charge of making it available for integration with compatible management level tools. Finally, the scalability of the architecture allows you to add and publish more information parameters to the service. In this way, the end client is allowed to monitor in real time the information of the parameters of equipment and physical assets that they want, from a browser connected to the industrial network, to evaluate their status and facilitate decision making. The content of the article is structured as follows: Sect. 2 presents a series of studies that are related and incentivized the development of this work; Sect. 3 provides brief concepts of: ISA-95, IEC-61499, MQTT and OPC-UA, which will allow a better understanding of the following sections; Then, the description of the communication architecture, and the rest of the proposed methodology, is presented in Sect. 4; In Sect. 5, the idea is illustrated in a case study example; Finally, the work is completed and possible future work is presented in Sect. 6.

2

Related Works

In this section, studies are released, which, in addition to having to do with the present project, contribute to the context of the project. In this reason, a series of work related to monitoring/supervision platforms, using I4.0 and IoT-enabled communication protocols are presented, as well as using IEC-61499 and ISA-95 standards.

68

J. D. Llamuca et al.

H¨astbacka et al. [6] propose a scalable monitoring architecture that provides plant device status information to enterprise-level applications, in this study they use OPC-UA as an intermediary to facilitate the heterogeneous exchange of information between devices from different vendors reliably. In this way, they achieve the discovery and dynamic monitoring of the devices inherited from the workshop, however, by not using an optimal information model instead of OPC-UA advanced information modeling, it causes increased data traffic, so it is necessary to store data locally temporarily until the network is available. A similar study is that of Hoffmann et al. [7], which instead use semantic models of intelligent multi-agent systems (MAS) in the exchange of information, through an OPC-UA address space, for a more agile execution of the system, but due to the learning capacity of the agents, high performance equipment is needed within the workshop. In common, these studies set aside their integration with standards suitable for automated processes such as IEC-61499 and ISA-95 and communication protocols that would enhance the monitoring system as MQTT. Mizuya et al. [13], proposes a case study of interest, in which they analyze the data acquisition capacity of the OPC-UA and MQTT protocols from the field devices, using an RPi card as a gateway. In this work, the effectiveness of OPC-UA in control, and MQTT in monitoring control variables is demonstrated. However, the limitations of OPC-UA performance on controller equipment with reduced features are evident. Also, MQTT’s agility is corroborated when publishing and subscribing messages from field devices. Another interesting work is that presented by Wenger et al. [16], in which they propose a novel agile cloud-based monitoring infrastructure, based on IEC61499 standard. In the monitoring scenario they propose, the domain-specific language determines the type of behavior of FBs, to collect, in a register, events and timestamps, from specific variables, without affecting the control functions. These values are published to the topics of the broker located in the cloud, from the MQTT client supported by FORTE; In this way, its monitoring service subscribes to the Broker’s logging topics, and manages monitoring of the properties of control devices from the cloud. However, an information reference model that facilitates the development of automated interfaces, such as ISA-95/IEC-62264, is not covered in this research, as well as the integration of new communication standards at the enterprise level such as OPC-UA. According to these studies, monitoring solutions only with OPC-UA, are a very effective option for the interoperability of systems, but because of, its increased consumption of resource and performance on low-level intermediary devices, it needs greater effort in low-cost systems with limited bandwidth. This is quite the opposite with monitoring systems using only MQTT, but still, according to the analysis of Ye et al. [17], OPC-UA has a better reach in high-level systems, because of this, the OPC-UA Client is used as an end user in enterprise-layer IoT applications, facilitating the integration of graphical interfaces, for the monitoring and management of configuration parameters and real-time information of field systems. This motivates this project in the use of

ISA-95 and IEC-61499 for Distributed Control System Monitoring

69

MQTT at the workshop level to obtain the parameters of the field devices, and OPC-UA for the presentation of the parameters at the enterprise level. In summary, these works seek to use techniques involving the design and use of new communication infrastructures to ensure the monitoring of field device variables focused on specific standards and protocols, such as: IEC-61499, ISA95, OPC-UA, and MQTT, but don’t focus on designing systems that advantage the integration of these standards to achieve a more useful system in smart factories, taking advantage of the strength of each of these. The point of departure for the present project, is given by the Study of Garc´ıa et al. [4,5], which incentivizes finding solutions that reduce efforts in the integration of CPPSs by incorporating compatible standards with OPC-UA.

3

State of the Art

This section introduces brief concepts and features for understanding the monitoring architecture and case study proposed in this paper. 3.1

ISA-95/IEC-62264

The ISA-95/IEC-62264 standard provides a reference model that prepares in categories, the type of information that can be exchanged to develop automated interfaces between business and control systems. ISA-95 hierarchically defines computers, these contain parameter and their values can be mapped through data type properties or object properties [9]. The hierarchical arrangement of the equipment is flexible and allow to extend levels below according to the depth of the system analysis, so too, the standard allows you to use levels of other specifications. Role-based team information represents the abstract description of the automation system’s installations, each equipment is composed of physical parts that are represented by the physical asset. Both can be related through assignments. 3.2

IEC-61499

IEC-61499 is a modeling language that facilitates the representation of distributed control systems. The main element of the standard architecture is the FB, it contains features that allow you to model complex applications by decentralizing your control logic to make it feasible to implement holistic controls that make the system more flexible and more interoperable [8,14]. The IEC-61499 standard provides three types of Fbs: (1) Basic Function Blocks (BFB), internal execution is conditioned by execution control graphs (CCPs) consisting of states, transitions and execution control actions; (2) Composite Function Blocks (CFB), whose configuration allows to encapsulate an FB network in another FB or other interconnected and synchronized CFBs; (3) Service Interface Function Blocks (SIFB), whose function is used for device control and allows you to associate the application of the FB with a specific hardware target.

70

J. D. Llamuca et al.

The FBs design process makes it possible to carry out control and supervision tasks on distributed industrial processes, this reduces the engineering effort to deal with data from flexible distributed systems, the contribution of [19], provides a more complete description. There are several software tools compatible with IEC-61499, especially for this project 4DIAC has been used this software is composed of 4DIAC-IDE and its Runtime FORTE environment, which allow the design and execution of control programs and distributed monitoring in small integrated devices [18]. 3.3

OPC-UA

OPC-UA extends process control level interoperability to information analysis, for: data acquisition, information modeling, and communication between applications, with security and reliability. The OPC-UA platform increases visibility and connectivity range, facilitating the communication of legacy systems with Management systems in IoT applications, integrating, more easily, graphical interfaces for monitoring and managing configuration parameters and real-time information of field systems [11]. OPC-UA allows you to develop and design communication systems that distribute the location of data in the plant using Client/Server communication models. The Object-Oriented Approach of OPC-UA enables the generation of Au models from an ISA-95 common object information model, such as rolebased team information and physical asset information; the resulting maximized model includes ISA-95 and OPC-UA terms that can be easily distinguished in information models NameSpace within each AddressSpace [4]. 3.4

MQTT

The MQTT protocol works on the TCP/IP transport protocol for message transfer, this protocol is very lightweight, making it suitable for limited resources and restricted environments with minimal network bandwidth. MQTT ensures message delivery asynchronously using Client-Server Publish-Subscribe protocols. Any IoT object can be an MQTT Subscriber-Publisher client, in a complementary way, the MQTT Broker can handle the communication of thousands of MQTT clients bound to a specific Topic. The message delivered by MQTT is filtered by Topic, which is a text-based hierarchical structure, and is positioned by separating each hierarchy with a slash (/) [12]. MQTT is considered a safe path in IoT, allows for own, external authentication, or take advantage of operating system mechanisms. An application can protect or restrict access to the content of its messages both over the network and at rest. There is no mechanism that allows the Client to authenticate the Server.

4

Methodology

c Modular Production System (MPS) is considered for this work, A FESTO  this represents a scale model of a real industrial process. The MPS is composed

ISA-95 and IEC-61499 for Distributed Control System Monitoring

71

of a distribution station and a sorting station, each station has two modules; feed and transfer, conveyor and storage, respectively. This section presents the hardware composition required by the system and the proposed system architecture for IoT-based monitoring of the MPS.

Fig. 1. Overall communication architecture of the proposed monitoring system.

4.1

Hardware Architecture

The Hardware Architecture is composed of the low-cost boards Raspberry Pi 3 model B+ (RPi3) and BeagleBone Black (BBB). Both are small board computers that allow you to work with open source operating systems (OSs) developed for embedded devices, these boards have installed distributions based on Linux Debian. On the one hand, RPi3 with Raspbian Stretch, and BBB with Debian Stretch pre-installed. As for the hardware characteristics of the boards. The RPi3 features: one 1.4 GHz quad-core-64bits processor with 1 GB of RAM, one 40-pin GPIO interaction terminal, Ethernet communication port, four USB 2.0 communication ports, Wireless LAN and Bluetooth connection capability, and one micro SD expansion port for loading the OS and storing the data. The BBB has: one AM335x 1 GHz processor with 512 MB of RAM, two GPIO general purpose terminals with 46 pins each, Wireless LAN connection capability, one USB mini-port for communication and power, one USB communication port 2.0, and one micro SD expansion port for storage. To enable boards to communicate with industrial signals, it is necessary to use a circuit designed to read from 3.3 v to 5 v sensors, supported by the I/O terminals of the boards.

72

4.2

J. D. Llamuca et al.

Overall Communication Architecture

The general communication architecture of the proposed monitoring system is presented in Fig. 1. This system is based on the combination of MQTT and OPC-UA IoT protocols, assigning an application for each protocol. Attributing the transfer of information to different protocols, and in applications with conditions where these best work, allows to maximize the capacity of the system. In this reason, the system is divided into two specific applications; integration application to obtain the status of field-level devices and/or components and the visual interface application in order to monitor the information obtained. MQTT is used for the acquisition of information, and OPC-UA is used for the taking of significant data from the graphic interface. Integration Application. In this application the devices connect to the boards to acquire their status information, that is, the physical process is integrated with the system, using MQTT and FBs. The use of FBs in the reading of signals from the GPIO pins of the boards facilitates the distribution of the execution logic in the hardware, adding flexibility to the system. The types of required FBs are contained in IEC-61499 applications.

Fig. 2. Configuration of the SIFB for communication with the Broker.

4DIAC uses a library of device description for each card, in this way it allows to map the IEC-61499 applications to each device through its resources. On the network of FBs, the information obtained is transmitted to designed FBs (See Fig. 3) to store the equipment specifications. This information is then published to the broker within topics determined by the provider, through the address of SIFBs configured as MQTT clients, as shown in Fig. 2. The broker distributes the information to the subscribing clients at the higher level. 4.3

Designed FBs

For the design of the FBs shown in Fig. 3, the information required by the proposed ISA-95 models is taken as reference.

ISA-95 and IEC-61499 for Distributed Control System Monitoring

73

Equipment Model FBs. The FB EQUIPMENT is a BFB, this is the main FB, it contains the required information of all the equipment within the control algorithm in its ECC. Unlike information on the status of the equipment, the rest of the information contained remains constant, and is extracted according to the identification of the equipment. The FB EQUIPMENT interface is shown in Fig. 3, this has a flow of events common to a typical FB. In the data flow: (1) the QI data is linked to the INIT event and initializes the BFB every time an INIT event is requested; (2) the ID data contains the identification of the equipment; (3) the IN data contains the current value of the pin assigned to the equipment; (4) the QO data informs about the state of the last event executed in the BFB; (5) the data Enterprise, Site, Area, WorkCenters, WorkUnits, correspond to the information of the hierarchy of each team based on the role, according to ISA95; (6) the rest of the output data correspond to the attributes proposed by the information model of ISA-95. The FB EQUIPMENT MODEL is a CFB, it contains a network of dedicated FBs, some for the conversion of data, and others for the construction of the string type chain with all the attributes of the BFB EQUIPMENT, each one separated by a character of the alphabet. The interface of the FB EQUIPMENT MODEL is also shown in Fig. 3, in this the events INIT, REQ, INITO, CNF, and the data QI, ID, QO, Enterprise, Site, Area, WorkCenters, WorkUnits, are directly connected to its corresponding of the FB EQUIPMENT. On the other hand: (1) the IN data goes to string type before connecting to the IN data of the FB EQUIPMENT; (2) Equipment data offers the string of all the attributes with their separation character.

Fig. 3. Interface of FBs for the transmission of data of Equipment and Physical Assets.

Physical Asset Model FBs. The FB PHYSICAL ASSET is a BFB and the FB PHYSICAL ASSET MODEL is a CFB, these FBs presented in Fig. 3, have design characteristics similar to the FBs EQUIPMENT and EQUIPMENT MODEL, but with reference to the model of physical assets. Its difference is

74

J. D. Llamuca et al.

noticeable in the interface of the FBs, differences such as: (1) increases input data because physical assets are considered by ISA-95 as physical equipment that can be replaced, therefore most of its attributes change periodically; (2) the output data Enterprise, Site, Area, WorkCenters and WorkUnits, from the equipment hierarchy, are not considered for the physical assets model; (3) the attributes of the physical assets are different from the attributes of the equipment. Due to the raw[ ] condition in the communication parameters (Fig. 2) admits only the string data type, all outputs of the CBFs are string type to avoid communication problems. Visual Interface Application. For an optimal visualization, this application allows to graphically present an interface with the information obtained from the field devices. For this reason, the internal interconnection environment of this application perform tasks for the debugging of information and the presentation of information to the user. Debugging carries all information processing, this is composed of stages of reception, structuring and transmitting information, these stages communicate with each other, through a message flow in the NodeRED work environment. The MQTT subscriber client node acts for the information reception stage, which acquires raw string information from the broker. The structuring stage is essential, in the first instance, it formats the message, decomposing the long string with the attributes that are acquired from the broker, using configuration nodes as shown in Fig. 4; the ISA-95 node set fulfills the key role of this stage. Finally, the OPC-UA Server typical of the ISA-95 node set saves the information to NameSpaces for the information transmission stage. The result of the debug task is information structured according to ISA-95 so that it is suitable for I4.0, In addition, the user can configure which information will be transmitted to the visualization interface. Note that because the ISA-95 structuring node does not support string data, it is necessary to do the conversion before entering it.

Fig. 4. Message format in information processing.

The presentation of information allows the structured monitoring of the equipment and physical assets of the process. For the presentation of information the working environment in NodeRED, has an OPC-UA Client node and the set of graphical interface nodes for the dashboard web server. Here the user

ISA-95 and IEC-61499 for Distributed Control System Monitoring

75

can determine the final interface distribution, because, you can create multiple frames or display groups on different tabs according to your needs. 4.4

System Information Flow

The information flow of the proposed monitoring system can be seen in the sequence diagram in Fig. 5. In which, to enable the information flow, MQTT and OPC-UA components must be started. When the MQTT broker and the OPC-UA server are started, they enable their IP addresses for MQTT and OPCUA clients respectively, can connect and initiate the information flow. The information flow loop is concise. The information that enters the smart boards is constantly obtained at the beginning of the information flow by the FBs network, this network adds information of parameters proposed by ISA-95 to the message, then this information is published through an MQTT client, supported by FORTE, towards a topic address space of the MQTT broker, then the MQTT subscriber client gets the simple message and sends it immediately, for debugging and structuring according to ISA-95 specifications, in sequence the structured information within the message, is written in a namespace of the OPC-UA server, after an address the OPC-UA client reads the information structure, breaks down the message and finally sends it to a web server with IoT technology of dashboard graphical interface for its visual presentation and subsequent monitoring.

Fig. 5. Diagram of workflow sequence and messaging of the monitoring system.

76

J. D. Llamuca et al.

The MQTT session is constituted when the MQTT publisher client supported by FORTE and the MQTT subscriber client node supported by NodeRED establish communication by adding the broker’s IP address and configuring the communication port. Similarly to establish the OPC-UA session, the OPC-UA server must accept the connection requirement from the OPC-UA client once it locates the IP address of the OPC-UA server on the communication port.

5

Case Study

This section presents a sample configuration of the proposed monitoring system. The two MPS stations are considered for the case study flow diagram shown in Fig. 6, but finally only the feeding module, specifically the stacking equipment is considered for monitoring.

Fig. 6. Workflow diagram of the system for the case study

Enter the hardware composition required by the system; the RPi 3 board is used as a means of acquiring parameters from the distribution station and the BBB board as a means of acquiring parameters of the sorting station; both distributed periphery boards run the FORTE Runtime, and work as MQTT clients for the publication of parameters. The system is communicated with the boards via Wi-Fi connection to the industrial network, in this way the main PC can communicate to the boards by means of an Ethernet connection to the industrial network. The main PC configures the broker; Also, the NodeRED working environment with MQTT subscriber clients, the structuring information according to ISA-95 and the OPC-UA server. OPC-UA clients are configured in another NodeRED flow.

ISA-95 and IEC-61499 for Distributed Control System Monitoring

77

Fig. 7. Flow of FBs in 4DIAC to acquire and transmit the parameters of the stacker equipment.

The integration application for obtaining the stacker parameters is shown in Fig. 7, in which observed the obtaining of the status of the connected equipment in the GPIO2 pin. The stages of reception, structuring and transmission for the debugging task, as well as the application for the presentation of data in the visual interface are shown in Fig. 8.

Fig. 8. Message flow between communication and function nodes in NodeRED.

Figure 9 shows the resulting monitoring interface of the stacking equipment. To give the graphical interface a structure similar to the hierarchy of ISA-95 role based equipment, the MQTT subscriber client nodes transmit directly to the given space in the graphical interface.

78

J. D. Llamuca et al.

Fig. 9. Structure of the graphic interface resulting from the monitoring system for the equipment.

6

Conclusions and Recommendations

A communication architecture is presented that allows acquiring the parameters of the field equipment and physical assets, in an agile way, encapsulating the relevant information in types of FBs, to then relate it efficiently, due to the common structure of the types of FBs, and exchange it more lightly using the MQTT communication protocol supported by FORTE. On a low-cost smartboard architecture. The architecture approach allows the growth of the system from the plant level, to expand the coverage capacity of the system. Monitoring system infrastructure supports reliable integration into the production process of high-level applications consumers of pre-processed bulk information, using OPC-UA communications. A set of FBs is proposed, to store the equipment and physical assets parameters, with well-defined interfaces, in order to obtain greater flexibility in the design of future applications of the system through IEC-61499. Flexibility, based on information flow, allows you to add ISA-95 specifications required without addressing problems, being compatible with other ISA-95 products. The information structure, based on ISA-95, provided from NodeRED, reduces data load to the high performance of communication architectures in IoT environments. So, the communication architecture that is presented, improves the interoperability between the components of the monitoring system, reducing efforts in the exchange of information and enabling communication to other top-level applications such as MES and ERP systems. The resulting monitoring system, on IoT’s NodeRED platform, features a refined interface with information structured in categories, which facilitates realtime monitoring of equipment and physical asset parameters, of distribution and sorting stations of the FESTO training plant, integrated in a distributed manner, simulating, a distributed remote system of production plants, modeled according to the paradigm of industries 4.0.

ISA-95 and IEC-61499 for Distributed Control System Monitoring

79

Finally, a complete scope system is implemented, helping to introduce new concepts of integration and compatibility between current upgrade standards within the Industry I4.0 paradigm, as well as IEC-61499, ISA-95, OPC-UA, and MQTT. Future research will focus on integrating different communication protocols, such as AMQP and STOMP, with OPC-UA, for an assessment and analysis of the characteristics and properties of the transmission of information.

References 1. Al-Fuqaha, A., Guizani, M., Mohammadi, M., Aledhari, M., Ayyash, M.: Internet of things: a survey on enabling technologies, protocols, and applications. IEEE Commun. Surv. Tutorials 17(4), 2347–2376 (2015). https://doi.org/10.1109/ COMST.2015.2444095. https://ieeexplore.ieee.org/document/7123563/ 2. Brettel, M., Friederichsen, N., Keller, M., Rosenberg, M.: How Virtualization, Decentralization and Network Building Change the Manufacturing Landscape: An Industry 4.0 Perspective. Technical Report RWTH-2015-01642, WASET (2014). https://publications.rwth-aachen.de/record/465283/export/hx?ln=en 3. Fysarakis, K., Askoxylakis, I., Soultatos, O., Papaefstathiou, I., Manifavas, C., Katos, V.: Which IoT protocol? comparing standardized approaches over a common M2m application. In: 2016 IEEE Global Communications Conference (GLOBECOM), pp. 1–7. IEEE, Washington, DC, USA, December 2016. https://doi.org/10. 1109/GLOCOM.2016.7842383. http://ieeexplore.ieee.org/document/7842383/ 4. Garcia, M.V., Irisarri, E., Perez, F., Estevez, E., Marcos, M.: OPC-UA communications integration using a CPPS architecture. In: 2016 IEEE Ecuador Technical Chapters Meeting (ETCM), pp. 1–6 (2016). https://doi.org/10.1109/ETCM.2016. 7750838 5. Garcia, M.V., Perez, F., Calvo, I., Moran, G.: Developing CPPS within IEC-61499 based on low cost devices. In: IEEE International Workshop on Factory Communication Systems - Proceedings, WFCS, pp. 1–4, July 2015. https://doi.org/10. 1109/WFCS.2015.7160574 6. Hastbacka, D., Barna, L., Karaila, M., Liang, Y., Tuominen, P., Kuikka, S.: Device status information service architecture for condition monitoring using OPC UA. In: Proceedings of the 2014 IEEE Emerging Technology and Factory Automation (ETFA), pp. 1–7. IEEE, Barcelona, Spain, September 2014. https://doi.org/10. 1109/ETFA.2014.7005141. http://ieeexplore.ieee.org/document/7005141/ 7. Hoffmann, M., Meisen, T., Jeschke, S.: OPC UA based ERP agents: enabling scalable communication solutions in heterogeneous automation environments. In: Demazeau, Y., Davidsson, P., Bajo, J., Vale, Z. (eds.) Advances in Practical Applications of Cyber-Physical Multi-Agent Systems: The PAAMS Collection, vol. 10349, pp. 120–131. Springer, Cham (2017). https://doi.org/10.1007/978-3-31959930-4 10 8. IEC: IEC 62769-1 CDV: Field Device Integration (FDI), Technical Specification Part 1: Overview (2013) 9. ISA: ISA95, Enterprise-Control System Integration (2017). https://www.isa.org/ isa95/ 10. Ismail, A., Kastner, W.: A middleware architecture for vertical integration. In: 2016 1st International Workshop on Cyber-Physical Production Systems (CPPS), pp. 1–4. IEEE, April 2016. https://doi.org/10.1109/CPPS.2016.7483915. http:// ieeexplore.ieee.org/document/7483915/

80

J. D. Llamuca et al.

11. Mahnke, W., Leitner, S.H., Damm, M.: OPC Unified Architecture. Springer, Heidelberg (2009). https://doi.org/10.1007/978-3-540-68899-0 12. Manandhar, S.: MQTT based communication in IoT (2017) 13. Mizuya, T., Okuda, M., Nagao, T.: A case study of data acquisition from field devices using OPC UA and MQTT. In: 2017 56th Annual Conference of the Society of Instrument and Control Engineers of Japan (SICE), pp. 611–614. IEEE, Kanazawa, September 2017. https://doi.org/10.23919/SICE.2017.8105594. http:// ieeexplore.ieee.org/document/8105594/ 14. Trejo-Hernandez, M., Lopez-Mellado, E.: Specification of manufacturing systems controllers using the standard IEC61499. In: CONIELECOMP 2013, 23rd International Conference on Electronics, Communications and Computing, pp. 179– 184. IEEE, Cholula, March 2013. https://doi.org/10.1109/CONIELECOMP.2013. 6525782. http://ieeexplore.ieee.org/document/6525782/ 15. Vyatkin, V.: IEC 61499 as enabler of distributed and intelligent automation: stateof-the-art review. IEEE Trans. Ind. Inform. 7(4), 768–781 (2011). https://doi.org/ 10.1109/TII.2011.2166785. http://ieeexplore.ieee.org/document/6021366/ 16. Wenger, M., Zoitl, A., Blech, J.O., Peake, I., Fernando, L.: Cloud based monitoring of timed events for industrial automation. In: 2015 IEEE 21st International Conference on Parallel and Distributed Systems (ICPADS), pp. 827–830. IEEE, Melbourne, VIC, December 2015. https://doi.org/10.1109/ICPADS.2015.111. http:// ieeexplore.ieee.org/document/7384374/ 17. Ye, X., Hong, S.H.: An automationML/OPC UA-based industry 4.0 solution for a manufacturing system. In: 2018 IEEE 23rd International Conference on Emerging Technologies and Factory Automation (ETFA), pp. 543–550. IEEE, Turin, September 2018. https://doi.org/10.1109/ETFA.2018.8502637. https://ieeexplore. ieee.org/document/8502637/ 18. Zoitl, A., Grabmair, G., Auinger, F., Sunder, C.: Executing real-time constrained control applications modelled in IEC 61499 with respect to dynamic reconfiguration. In: 2005 3rd IEEE International Conference on Industrial Informatics INDIN 2005, pp. 62–67. IEEE (2005). https://doi.org/10.1109/INDIN.2005. 1560353. http://ieeexplore.ieee.org/document/1560353/ 19. Zoitl, A., Strasser, T., Ebenhofer, G.: Developing modular reusable IEC 61499 control applications with 4diac. In: 2013 11th IEEE International Conference on Industrial Informatics (INDIN), pp. 358–363. IEEE, July 2013. https://doi.org/ 10.1109/INDIN.2013.6622910. http://ieeexplore.ieee.org/document/6622910/ 20. Zurawski, R. (ed.): Integration Technologies for Industrial Automated Systems, vol. 1, 1st edn. CRC Taylor & Francis, Boca Raton (2007). oCLC: ocm64592270

Knowledge Management and Intellectual Property

Genomic Databases Exploration Using Conceptual Models C. Vanessa Solis(B) , P. Ana León, and Oscar Pastor Lopez Research Center on Software Production Methods (PROS), Universitat Politècnica de València, Valencia, Spain {vsolis,aleon,opastor}@pros.upv.es

Abstract. The modeling of the human genome is a fundamental part that allows us to consider the involved entities and their relationships. For this reason, the present work incorporates a conceptual model under a mapping with different existing genomic databases, establishing links between the information genomic databases contrasted with each of the elements of the Conceptual Schema of the Human Genome (CSHG). This work presents the development of the exploration of genomic databases found in lists endorsed by research institutes in the genomic area, as a basis for a later construction of an information system oriented to the genomic field. It states the verification process of the found sites, since some have suffered changes in the servers or have simply stopped working. Also, exposes generated depuration tasks, because each of the genomic databases have different structures, information organization, or even in some cases unusual nomenclature was used. Subsequently, the mapping of each genomic database with the elements of the CSHG is presented. Finally, the results obtained are shown with statistics established in the exploration of the genomic databases. Keywords: Conceptual schema of the human genome (CSHG) · Genomic data bases · Human genome · Genomic information system · Conceptual schema (CM)

1 Introduction From the discovery of the genome, an information storage site was required. The number of biological databases available for public consultation is growing rapidly both publicly and privately [1, 2]. These databases cover different parts of human biology, from the genetic sequence to the pharmacotherapeutic one. However, many of these repositories were created with very specific purposes, and no considering the connection between them. This situation has caused problems such as data inconsistencies, redundancy and even discrepancies when representing the same concept. For example, Tao and Embley have collected this concept as “genomic chaos” [3]. For this case the use of a conceptual scheme (oriented to the human genome) makes it possible to unify all the knowledge generated in different fields of genetics study under a single perspective [4, 5]. Likewise, © Springer Nature Switzerland AG 2020 E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 83–96, 2020. https://doi.org/10.1007/978-3-030-35740-5_6

84

C. V. Solis et al.

the Conceptual Human Genome Scheme (CSHG) [6] also serves as a basis for the creation of a Genomic Information System (SIGe) that allows the analysis and exploitation of stored information for diagnostic purposes. But in order to populate this SIGe with relevant information for clinical diagnosis, it is important to solve the integration problems that arise when joining information from different repositories. Thus, the first step will be the identification of genomic databases that have information and were related to the CSHG. With the purpose of that the later union of all this knowledge under a single global perspective allows the emergence and support of new prevention paradigms, diagnosis and treatment such as Precision Medicine [7, 8].

2 Conceptual Model Conceptual Model (CM) is used in different work areas since it allows the abstraction of different elements to express its functioning or its performance [9, 10]. Mylopoulos in 1992 defines conceptual modeling as the activity of formally describing some aspects of the physical and social world that surrounds us in order to understand and communicate [11]. Thalheim for its part states that modeling is governed by its purpose [12], for example, the construction of a system, simulation of real world situations, construction of theories, explanation of phenomena or documentation of an existing system. Thus, modeling is also an engineering activity with engineering steps and engineering results [12]. The conceptual model is a tool that allows a versatile visual description of the relationships between the elements that intervene in a task, activity or process. Many authors come to consider the conceptual model as an abstract process to develop an alternative view of the situation of the problem, this model will later allow that when returning to the real world it is possible to evaluate and effectively test the functioning of the created model [13]. Although there is no precise description where each of the modeling artifacts is framed in a rigorous manner [14], the scientific community is in the position to present improvements or discuss its elaboration [15]. In the exploration of genomic databases oriented to the human genome, we considered the conceptual model presented by Reyes Román [16] (See Fig. 1). In this scheme

STRUCTURAL

VARIATIONS

TRANSCRIPTIONS

PATHWAYS DATA SOURCES AND BIBLOGRAPHY

Fig. 1. Views of holistic conceptual model.

Genomic Databases Exploration Using Conceptual Models

85

you can see 5 views grouping each of the elements of the CSHG. The CSHG has been developed and proposed in a holistic way, so in this work we have the same approach, considering all the views. Each of the different views presented in the CSHG consist of a group of classes which collect different attributes of each of them, then a brief description of these is given: • Structural view: Presents information related to the structure of the genome associated with a species. • View of transcription: Contains information related to the synthesis of proteins. • View of variations: Represents each of the possible variations that can be presented in the DNA according to what the scientific community has been able to collect up to now. • View of metabolic pathways or pathways: Contains information on the chemical reactions that take place within a cell. • View of data sources and bibliography: Contains information related to the repositories that check or review, give scientific validity and above all that allow obtaining information on chromosomal elements, variations, populations, proteins, among others.

3 Genomic Databases Genomic databases were conceived from the relational databases focused on the biological field in which the design, development and long-term management have come to form the central point in the bioinformatics field [17]. The content of said databases is the result of DNA sequencing from a sample of a living being (See Fig. 2). Later, the sequencing machines allow to obtain a genomic report, which is processed by the experts (geneticists in its great majority). Experts, who treat genomic information based on their technique, area, species of study, among others, are those who generate genomic databases with different types of information including nucleotide sequences, gene expressions, antibiotic resistance genes, taxonomy, genomes, mutations, genetic variations, secondary structure of proteins, families, domains, index of publications of scientific articles, interaction routes, protein and enzyme reactions, among others [18–21]. At present, the management of databases has become very common, so that information gathered in the different work areas is easily found. Thus, in the genomic domain there is a wide variety of genetic databases, which is why with the appearance of NGS, or also known as Next Generation Sequencing [1], is one of the promoters that handled information should be helpful for medical reports, as well as for research. As a starting point, we considered the databases found in the lists proposed by Oxford Academic [18], National Center for Biotechnology Information (NCBI) [19], Human Genome Variation Society (HGVS) [22] and Health Sciences Library System (HSLS) [21]. The lists appear from initiatives of research groups; these groups have tried to group various genomic databases that have been appearing over time. However, here are weaknesses:

86

C. V. Solis et al.

Fig. 2. Genomic databases origins.

• The bases presented in the different lists are not the same. • Each list presents databases that differ in their content. • Elements present in each database are different, in some databases the information can be more detailed, and in others have more attributes. • The creation of each database depends on the criteria of the authors or administrators of the research group that manages the information.

4 Methodology It has followed Design Sciences that has allowed to identify each of the elements to be developed based on the context of the human genome, as well as the elaboration of each one of the activities to be carried out. 1. Problem formulation: A delimitation and approach of the problem will be carried out in order to have a central and clear point to focus on the process of finding the solution. 2. Knowledge of basic concepts: The description of the concepts of conceptual modeling as well as genomic databases will be performed to de-terminate existing relationships with respect to the CSHG. 3. Research of genomic databases: An investigation of each database will be carried out to review the information stored and the different forms of access it provides. 4. Structural analysis of the genomic databases: Description of data structures provided by each repository, according to the form of access. 5. Information mapping: Mapping between the information required by the CSHG and the information provided by each database.

Genomic Databases Exploration Using Conceptual Models

87

6. Identification of problems: The identification of the different problems originated at the time of integrating the information, which was found in each of the genomic databases, with the genomic conceptual model. 7. Generation and presentation of results: After the mapping and identification of problems, a tool will be developed to visualize the results of the mapping between the CSHG and each of the genomic databases. 8. Conclusions: The main conclusions will be drawn from each of the chapters, as well as those related to the objectives set out in this work.

5 Mapping Genomic Databases Based on the Conceptual Model It is very important in this section to first establish different criteria for the selection of the databases since, the approach to be considered to select the existing genomic databases must be defined, for this reason criteria will be established that lead to an easy alignment of the concepts to be obtained based on the CSHG. Likewise, from this section you will obtain the necessary information for the generation of the application for presenting the results of the exploration of genomic databases. On the other hand, the problems that have been found or could exist for this entire process will be identified. 5.1 Selection Criteria Compliance criteria have been defined to gather information that is used in the Conceptual Schema Human Genome, which are detailed below: 1. Considering the different views that the conceptual scheme of the human genome poses, the databases must have information of: a. Genes: Databases that show information about genes, either nomenclatures, descriptions, expressions, among others. b. Genomes: The databases in which information on the genomes of different species or populations will be available, which is in accordance with the CSHG, will be considered. c. Species: They are considered for the selection of the databases that have information about species, in this case when dealing with the human genome, all the databases that present human information will be filtered. However, the databases that indicate information of other species will be presented in a list indicating the species to which they correspond. d. Proteins: The databases that show protein information, within which you can see description, nomenclatures, location, among others. e. Chemicals: Databases that indicate information on chemicals that are related to alterations or variations corresponding to the genome will be considered. f. Pathways: The information that is framed in the metabolic routes will also be considered as a selection criterion in the databases. g. Variations: The information in the databases that is related to the different variations present in humans is considered.

88

C. V. Solis et al.

h. Populations: Genomic databases that contain information on the populations in which the studies have been conducted or where the information comes from. i. Bibliography: All databases that have information on scientific references about the studies carried out. 2. Credibility of the sources: The sources must be considered the pages published or referenced by entities whether educational or dedicated to the field of health. Likewise, it will be indicated if the information in the database has been reviewed or there is an author (person/people/organization) that endorses the veracity of the published information.

5.2 Selection of Genomic Databases For the selection of the genomic databases, those presented in the 4 collections were first considered, the first collection of bases is those associated with NBCI [19], second the online collection of bioinformatics resources of the University of Pittsburgh [20], which presents online a set of annotations and links for bioinformatics databases, as well as software tools. Likewise, the collection presented by the Oxford Academy [18] has also been considered in which there are scientific articles on the different investigations both at the biological level, as well as chemical, physical and biochemical, its access is totally open what generates great ease in accessing content. It is important to mention that despite taking all the information in July 2018, the websites present a copyright of the year 2014. Additionally, the list proposed by the Human Genome Variation Society (HGVS) [22] was considered, whose objective is to promote the discovery and characterization of genomic variations, including population distribution and phenotypic associations. In the case of the generation of a single list based on the accessed sources, firstly, the websites in which the text of the databases are simply presented by means of an HTML were reviewed. None of the sites allows downloading csv or txt files which contains, for example, names, description, URL. That information is useful for forming an initial database, are already available, for this reason the extraction of information became slow when you have to copy each of these attributes in a spreadsheet or go manually entering a database. To streamline this process, different scripts were generated in Python where information was filtered for said extraction. Subsequently, it is essential to perform a data cleansing to obtain a final list of the same genomic databases that will allow us to map between each of the databases and the conceptual model of the human genome: • Duplication was verified in the names of each of the databases included in the general list, so it is of vital importance since at the time of accessing them, double work in said activity will be avoided. It should be noted that sometimes the database is indicated with an acronym by name and in other cases it is with an extended name, so this case is also considered. In this way, 4552 databases were initially presented (See Table 1).

Genomic Databases Exploration Using Conceptual Models

89

Table 1. Genomic databases preselection Source No. NCBI

43

Oxford 1610 HSLS

2401

HGVS

498

• A revision of the URLs was made and those that presented duplication were eliminated, this is since in several occasions the genomic databases had different names, but their links pointed to the same site. • Elimination of special characters, and invalid syntax. • By using a script made in Python, which allowed to determine which databases have access (Upload to a website) and those that do not (No website is loaded), this determination is made based on the access code that the script returned. Based on the codes obtained, the databases that obtained code 200 were selected, since they are those that do not have any problem in access and connection. For the case of the bases that obtained codes 400 (client errors, the page is available, but the resource is not) and 500 (server errors) were discarded. • Finally, we considered the selection criteria where each of the databases that presented information for the human species was filtered and presented information related to the conceptual model, thus leaving a total of 761 genomic databases.

5.3 Traceability Between the CSHG and Genomic Databases Once the databases were obtained, the mapping between the CSHG and each of the databases under the flow indicated in Fig. 3 was carried out. As a first check, access to the website is verified, in some cases the result was “Not found”, “No access”, “Not allowed”, “Discontinued”, “In maintenance”, “Server not available”, “Server error”. If a website is accessed correctly, the basic information that was not available has been extracted, such as: “Author/Curators”, “Description”, “Last update”, “Download type”. This information allows the person who is interested in the details of the database to be able to obtain it easily. Subsequently, the identification of each of the elements presented by the genomic database is made (See Fig. 4), in contrast to the CSHG. This information is very useful because in some cases it must be considered that when a text is represented behind it, it contains a URL that links to other websites, so this information is also extracted. Finally, the mapping of the identified elements is done, for which it can be seen in Table 2 that each of the attributes of the CSHG belongs to an element presented in the website of the genomic database. Additionally, to link the section of the cases that the information requested were presented only as text, it was considered a html structure used in the website of the genomic database. In Fig. 6 an example is presented in which within the PubMed database

90

C. V. Solis et al.

Start

Open link

No

Site error cataloging

Did you upload the website?

Yes

AƩribute Mapping of the CSHG

Review of the website

ExtracƟon of basic informaƟon from the DB

Storage of informaƟon

End

Is there more BD?

Fig. 3. Mapping flowchart of DBs and CSHG

is the authors of the scientific article for which in the html structure of the site this information is within the “auths” section. In the same way, for the case of the abstract, the section in which it is found inside the html has been found, so it is called the “abstr” section. So, all this information will be added to the csv file as referential data of the found attribute.

Mapping elements Fig. 4. Elements identification for mapping

Genomic Databases Exploration Using Conceptual Models

91

Table 2. Mapping elements of CSHG and Genomic DB DB elements

Class

Attribute

Phenotype

Phenotype

name

Mutation

Mutation

Exon

Exon

Id_symbol

Mutation database

Data Bank

name description url

OMIM

Variation

omim

Reference

Bibliography reference

bibliography_reference_id

Reference

Bibliography DB

URL

5.4 Results and Discussion The exploration of the different genomic databases (See Fig. 5) has allowed us to assess that 3542 were found, of which 21.49% correspond to the human species, 14.12% to animals, 10.53% to plants or fungi, 12.73% to cells or bacterium and finally 41.13% is dispersed in different subcategories. With the mapping of each one of the databases, it has been possible to verify that the initially provided URLs actually become the correct ones, since at the moment of reviewing the contents of these it was found that many of them were not accessible or their link was broken. Of the 100% exploration of genomic databases, 32% represents 1148 genomic databases that have not been accessed.

Fig. 5. Genomic DB by specie

92

C. V. Solis et al.

Fig. 6. Traceability through web page elements

However, it is essential to indicate that in many sites explored pages were found that did not correspond to the indicated, or was in maintenance, this can be seen in Table 3, which presents the cataloging that has been given to the sites that have presented certain irregularities, after the normal load of the web page of the genomic database. Table 3. Errors found on sites with connection Irregularities

No. %

Authorization required 6

16,67

Database error

2

5,56

Discontinued

1

2,78

Internal error

6

16,67

Internal server error

3

8,33

Maintenance

7

19,44

Redirection

1

2,78

Retired

1

2,78

Unavailable

8

22,22

Website closed

1

2,78

Of the 761 genomic databases that could finally be found both in access and in information, it could be seen that some of them allow downloading files or resources in different formats in their websites, which are detailed in Table 4, having 242 genomic DBs that allow downloading. Similarly, some websites have incorporated tools that contain a set of functions, which allow the creation of applications to access the data it contains. These tools have been cataloged as APIs, Table 5 shows that 9.33% allows the use of them, while 90.67% does not have these tools or services incorporated.

Genomic Databases Exploration Using Conceptual Models

93

Table 4. Genomic DB allows download resources (Human species) Allows download No. % Yes

242 31.80

No

519 68.20

Total

761 100

Table 5. Genomic DB with APIs (Human species) Allows API No. % Yes

71 9.33

No

690 90.67

Total

761 100

Finally, we wanted to make visible the counting of the databases that are grouped by the 5 genomic views considered in the CSHG. Figure 7 indicates that the largest group of DB is in the transcription view, since there are 502 DB. Being the least presence of DB for the sight of metabolic pathways or pathways with 62 genomic DB.

Fig. 7. DB by views

For a better visualization of the information, a website was created in Django that allows queries regarding which databases present information based on the views, classes and attributes respectively (See Fig. 8) thus facilitating the location of the elements to be sought.

94

C. V. Solis et al.

Home Views

All

Classes

All

Search

About Contact

AƩributes All Review Search

Name

Authors

DescripƟon

Source

Download type

API

Fig. 8. Web created for searches

6 Conclusion One of the main problems encountered is the so-called “Genomic data chaos”. This is considered very relevant since the information presented in the different sources maintains a degree of dispersion, heterogeneity, redundancy and, in many cases, inconsistency, making it difficult to process. Thus, the domain of the subject is fundamental, with the knowledge of the data structure presented by each of the websites corresponding to the databases oriented to the human genome. Although each repository has a different structure, it is also the case of the nomenclatures that are available and presented at the time of consultation of the information, so the researcher or person who collects such information must have basic knowledge that can guide and Drive in the research. For this reason, the collection and mapping of each of the genomic databases with the CSHG was done manually, not discarding important data by changing nomenclatures or abbreviations that could lead to confusion. Because of these “synonyms” in genetic terminology, the chaos of genomic data becomes visible due to the heterogeneity of the information. At this point, thanks to the CSHG versatility, it allows the direct mapping of each attribute and, if necessary, to add any element that may be useful and relevant to the researcher or doctor. The exploration of each of the genomic databases allowed us to specify the correspondence between the different attributes required by the CSHG. The mapping of each element generated a traceability between the CSHG and the genomic DB, in this way the different areas of knowledge that are handled in the genomic domain were identified. The use of a conceptual scheme provides a versatility in the handling of information in this case focused on the human genome. This is because the attributes required by researchers or doctors are tailored to the new knowledge acquired in base of scientific research. If it is necessary to incorporate new information, the conceptual model allows it to be carried out without problems, as well as if it is necessary to omit or eliminate it.

Genomic Databases Exploration Using Conceptual Models

95

Acknowledges. The authors wish to thank the members of the Genome Group of the PROS Research Center for the fruitful discussions on the application of conceptual modeling in the field of medicine. This work has been supported by the Ministry of Science and Innovation of Spain through the DataME project (ref: TIN2016-80811-P) and the Research and Development Assistance Program (PAID-01-16) of the Universitat Politècnica de València under the FPI 2137 grant.

References 1. Cook, C.E., Bergman, M.T.C.G., Apweiler, R., Birney, E.: The European bioinformatics institute in 2017: data coordination and integration. Nucleic Acids Res. 46, D21–D29 (2017) 2. UK government, Strategy for UK life sciences: one year on (2012). https://www.gov.uk/ government/publications/strategy-for-uk-life-sciences-one-year-on 3. Tao, C., Embley, D.: Seed-based generation of personalized bio-ontologies for information extraction. In: Advances in Conceptual Modeling–Foundations and Applications, pp. 74–84 (2007) 4. Olivé, A.: Conceptual Modeling of Information Systems, 1st edn. Springer, Heidelberg (2007) 5. Aguilera, D., Gómez, C., Olivé, A.: Enforcement of Conceptual Schema Quality Issues in Current Integrated Development Environments, pp. 626–640. Springer, Heidelberg (2013) 6. Reyes Román, J.F., Pastor, Ó., Casamayor, J.C., Valverde, F.: Applying conceptual modeling to better understand the human genome. In: ER 2016 Concept Model, pp. 404–412. Springer, Gifu (2016). https://doi.org/10.1007/978-3-319-46397-1_31 7. Mirnezami, R., Nicholson, J., Darzi, A.: Preparing for precision medicine. N. Engl. J. Med. 6(366), 489–491 (2012) 8. Middleton, A.: Society and personal genome data. Hum. Mol. Genet. 27(R1), R8–R13 (2018) 9. Coll, V.B.: Gestión de mutaciones en ambientes genómicos: una perspectiva basada en Modelos Conceptuales (2012) 10. Cabot, J., Gómez, C., Pastor, O., Sancho, M.R., Teniente, E. (Eds.): Conceptual Modeling Perspectives. Springer, Heidelberg (2017) 11. Mylopoulos, J., Chung, L., Nixon, B.: Representing and using nonfunctional requirements: a process-oriented approach. IEEE Trans. Softw. Eng. 18(6), 483–497 (1992) 12. Thalheim, B.: The Theory of Conceptual Models, The Theory of Conceptual Modelling and Foundations of Conceptual Modelling, pp. 543–577. Springer, Heidelberg (2011) 13. Cabot, J., Gómez, C., Sancho, M.R., Teniente, E.: 30 years of contributions to conceptual. In: Conceptual Modeling Perspectives, pp. 7–20. Springer, Heidelberg (2017) 14. Muller, G.: System and context modeling - the role of time-boxing and multi-view interaction. Syst. Res. Forum 3, 139–152 (2009) 15. Ludewig, J.: Models in software engineering–an introduction. Softw. Syst. Model. 2(1), 5–14 (2003) 16. Reyes Román, J.F.: Diseño y Desarrollo de un Sistema de Información Genómica Basado en un Modelo Conceptual Holístico del Genoma Humano (Doctoral dissertation), Valencia: Tesis Doctoral (2018) 17. Bourne, P.: Will a biological database be different from a biological journal? PLoS Comput. Biol. 1(3), 179–181 (2005) 18. Oxford University: NAR Database Summary Paper. Oxford University Press (2014). www. oxfordjournals.org 19. National Center for Biotechnology Information, “All resources,” National Center for Biotechnology Information. https://www.ncbi.nlm.nih.gov/

96

C. V. Solis et al.

20. University of Pittsburgh, “search.HSLS.OBRC,” The Health Sciences Library System 2014 (2018). https://www.hsls.pitt.edu/obrc/ 21. Health Sciences Library Systems, “OBRC: Online Bioinformatics Resources Collections,” University of Pittsburgh (2014). https://www.hsls.pitt.edu/obrc/ 22. Human Genome Variation Society, “Databases & Tools,” HGVS (2018). http://www.hgvs. org/content/databases-tools

Management of Humanitarian Logistics in the Stages Prior to Natural Disasters in Canton Ambato, Ecuador Santiago Velastegui1 , Rosa Galleguillos-Pozo2 , Cesar Rosero1 , and Marcelo V. Garcia1,3(B) 1

2

Universidad Tecnica de Ambato, UTA, 180103 Ambato, Ecuador {rs.velastegui,cesararosero,mv.garcia}@uta.edu.ec Universidad Politecnica de Catalunya, UPC, 08028 Barcelona, Spain [email protected] 3 University of Basque Country, UPV/EHU, 48013 Bilbao, Spain [email protected]

Abstract. In recent years in Ecuador’s country, an increasing number of domestic natural disasters has caused losses of lives and material. This study aims to generate Humanitarian Logistics for the canton Ambato in Ecuador. The safe zones are identified for defining the collection centers and the public resources to be available for each parish based on the Analytic hierarchy process (AHP) and Geographic Information System (GIS) methods. As a result of this technique the main collection center for the canton Ambato is determined, allowing the movement of humanitarian assistance in the shortest time possible due to the accessibility of transport routes, available space, organization, own resources and travel time to every location. Keywords: Humanitarian Logistics · Catastrophes · Analytic hierarchy process (AHP) · Center of Gravity · ArcGIS

1

Introduction

Natural events can be characterized as natural disasters when they occur in populated areas causing the local infrastructure destruction, people’s suffering and deprivation. In the last three decades, the occurrence of natural disasters has increased significantly [2,6]. The disasters caused by natural hazards have the potential of causing interruptions in basic services and a significant cost in human lives. Recent examples of disasters are the earthquakes in Japan, Chili, Haiti and New Zealand in 2010 and 2011; the devastation of Vanuatu due to the cyclone Pam in 2015; floods in Thailand in 2011; the eruption of Eyjafjallaj¨ okull in Island in 2010 and the current drought in California. The most vulnerable regions are those with the highest concentrations of population and economic activities [18,20]. c Springer Nature Switzerland AG 2020  E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 97–108, 2020. https://doi.org/10.1007/978-3-030-35740-5_7

98

S. Velastegui et al.

Immediately after the occurrence of the disaster, the humanitarian operations start with the purpose of providing quick assistance to the victims in different ways: rescuing injured people, collecting and disposing the dead bodies, allocating resources, supplying food, shelter, medical aid and restoring the access to remote places. Delays in the delivery of emergency kits or in granting aid may cost human lives [11]. Therefore, the efficiency in logistics is a key factor for success because it ensures a fluid flow of goods and services in a complex supply chain. Logistics plays a key role on the response operations in disasters; it represents the link connecting disaster preparedness factor, public procurement, distribution, main site and land characteristics being crucial for an effective capability of response concerning the principal humanitarian programs such as health, food and sanitation [14]. One of the last disasters occurred in Ecuador was the earthquake of April 16th, 2016 in the province of Manab´ı, where it was clearly proven the lack of culture of the citizenship in the prevention and response for this kind of phenomenon, and also the absence of humanitarian logistics and institutional organization [9]. On August 5th, 1949 another disaster took place in Ecuador, specifically in Ambato, in the province of Tungurahua. This earthquake left 6000 victims, 100.000 affected people and 1920 Km2 of devastated area. Approximately 80% of the canton Ambato was destroyed. Ambato canton was also affected in 1999 by Tungurahua’s volcanic eruption leaving 32 deaths, 25 000 evacuees and agriculture losses of 17,600,000 USD. In the eruptive processes, the ash fall affects people’s and animal’s health, it generates loss of crops, livestock and products, making negative impact on the economy of the region [7]. Humanitarian assistance in a disaster area represents 80% of logistics, in other words, it highly depends on an adequate management of the Supply Chain [3,5]. Given the circumstances of canton Ambato and focusing, above all, on the well-being of the population, the aim of this study is to propose a model of humanitarian logistics allowing to optimize the available resources and assistance for the canton. This document relies on 4 sections, including the introduction. Section 2 explains the methodology based on the Analytic hierarchy process (AHP) to determine humanitarian assistance in the shortest time, Sect. 3 explains about the study case used in this research and Sect. 4 show the results obtained and conclusions of the work carried out.

2

Proposed Methodology

The methodology used in this study is explained below, it is structured in four steps.

Management of Humanitarian Logistics to Natural Disasters

2.1

99

Step 1: Determining Safe Zones of the Canton Ambato

The safe zones of the canton Ambato were determined according to the Seismic Code of Costa Rica of 1986, which establishes that it is forbidden to build new structures at 50 m or less from seismically active geologic fault lines. For this reason, in the definition of the open spaces, resources that cross the fault lines or that are under the regulated distance, were excluded [13]. In this step the ArcGIS 10.1 software, which is a Geographic Information System (GIS), is also used for a landscape evaluation. The GIS software allows do the cartography of the canton Ambato and to represent the different elements located in the region [17]. 2.2

Step 2: Selecting the Collection Center for Each Parish with AHP

The decision making or selection processes become more difficult when the problems gain in complexity or when the situations are less structured and many scenarios coexist, actors and factors, being necessary to have more information and reasoning [11]. This situation is simplified with the AHP methodology that uses comparisons between pairs of elements, building matrices from these comparisons and using Matrix algebra elements for setting priorities among the elements of a level with respect to an element of the next level up. When the priorities of the elements of each level are defined, they are aggregated to obtain the global priorities in order to accomplish the main objective. The results next to the alternatives, become an important support element for the persons responsible of making the decisions. The notation used is the following: For i given objectives, i = 1, 2, ..., m; respective weights wi are calculated. For each i objective, the j = 1, 2, ..., n alternatives are compared and the weights wij are determined with respect to the objective i. The final weight for the alternative wj is calculated with respect to all the objectives having that: wj = w1 jw1 + w2 jw2 + ... + wm jwm The alternatives are ordered according to the wj value in descendent order, where the greater value indicates the most preferred alternative [19]. 2.3

Step 3: Using Center of Gravity Method for Determining the Main Collection Center with ArcGIS

This mathematical method is frequently used by many organizations when the problem is simply placing only one facility, thus, it is mathematically classified as a static method of continuous localization [10]. In the case study the coordinates for the general collection center are resultant from the Eqs. 1 and 2. n V i ∗ Xi (1) Xc = i=1 Vi

100

S. Velastegui et al.

n Yc =

Vi∗Yi Vi

i=1

(2)

Where: – Xc , Yc = Coordinate points of the General Collection Center. – Vi = Volume of the amount of persons per parish. – Xi , Yi = Coordinate points of the collection centers of each parish determined by AHP [1,12]. In this stage, the ArcGIS software facilitates the coordinates of each of the civil structures established as collection centers per parish. 2.4

Step 4: Determining Costs of Humanitarian Assistance Kits

The complementary humanitarian assistance kits are defined as inputs to cover the immediate needs of shelter, food and hygiene for the affected population [16]. The preparation of the kits is based on the information provided by the American Red Cross where some instructions are given about the type of food necessary in emergency situations [15]. It is established that a person needs minimum 2 l of water and has to eat a balanced meal at least once a day; a balanced meal must contain portions from the different food groups as follows, from large to small portions: bread, cereal, rice and pasta group; fruit group; vegetables group; meat, poultry, fish, beans, eggs and nuts group; milk, yogurt and cheese group; and finally fat, oils and desserts [4]. The costs of the kits will be set according to the budgets made for three different supermarkets of canton Ambato.

3

Case Study

The application of the abovementioned methods is developed in the following steps: 3.1

Step 1: Determining Safe Zones of the Canton Ambato

The canton Ambato is politically divided in 9 urban parishes that depend directly from the GADMA (Decentralized Autonomous Government of Ambato Municipality) administration, and 18 rural parishes represented by the Decentralized Autonomous Governments of each one of them. The safe zones are established according to the regulations of the Seismic Code of Costa Rica in the guide for management of temporary shelters in predefined buildings. The safe zones are the regions with no interceptions with the micro fault lines, and satisfying the risk management criteria of providing assistance in the shortest time possible. So, the public infrastructure can be considered as safe places if they are not close to a geological fault line, that is, educational units, health centers, parks among others. The exceptions are shown in Fig. 1, where the spaces located

Management of Humanitarian Logistics to Natural Disasters

101

over geologic faults are represented by red lines and are excluded from this study, for example: the parks Casigana, Chasinato, La Cantera, Los Sauces, Nicol´as Mart´ınez, Samanga, La Cantera Paseo Ecol´ogico, the sports fields from the neighborhood Los Girasoles, Urbanization Los Ceibos, Sports and Leisure Center Los Sauces.

Fig. 1. Unsafe zones in the canton Ambato.

3.2

Step 2: Selecting the Collection Center for Each Parish with AHP

From the previous step it can be established that the total of civil works (health center, hospitals, educational units, open spaces as parks, sports fields, stadiums and coliseums) are located in safe zones, so it is defined which civil work is the optimal in each parish. Based on Risk Management criteria, the places of higher importance are prioritized through the AHP method, defining 3 levels: the first level is the Objective: finding a safe place for the collection center in each parish. In the second level 5 decision criteria are considered: 1. Availability of routes, nowadays the majority of ways are of first and second category; 2. Available Space, due to the fact that in the urban sector there are few open spaces with large areas for accommodating high amounts of people; 3. Organization, is a very important criteria because culturally, especially in the rural sector, there are ancient traditions and it is difficult dealing with or organizing groups for agreeing on certain issues; 4. Own Resources, this is related with the natural resources of each sector, as it is possible to lose vital supply lines, it must be considered that in the rural sector the natural hydric sources can provide safe drinking water after a basic treatment, in contrast, the urban sectors have no natural resources to be consumed. 5. Travel Time, when an emergency occurs the response times are decisive for guaranteeing an effective assistance, that is why, the shortest

102

S. Velastegui et al.

distances are taken as reference, giving them the highest score when they must be covered on foot. In the third level, the alternatives related to civil works (health centers, hospitals, educational units and open spaces as parks, sports fields, stadiums and coliseums) can be found. They vary in number and type from one parish to another. Once the objective, the criteria and the alternatives of each parish are defined, the Risk Management Technical team of the Risk Management Secretariat (RMS) with 5 years of experience, qualify by pairs each one of the criteria and the alternatives with respect to each criteria applying AHP. In Fig. 2 the hierarchic structure for Cunchibamba parish made with the Super Decisions software is shown.

Fig. 2. Hierarchic structure for Cunchibamba parish.

The results from the previous analysis, obtained with the Super Decisions software, is shown in Table 1, where the total percentages of each alternative of the Cunchibamba parish are listed, allowing to determine the most suitable spot for the collection center in the parish according to the experts. In the inconsistency analysis of each parish, the resultant value was no more than 10% [8]. Table 1. Results of the alternatives for Cunchibamba parish with AHP. Civil Work, Parish Ambato Prioritization U.E. H´ector Lara

18.17%

U.E. Dar´ıo Guevara

24.42%

Central Park

24.94%

Sub Center Cunchibamba

32.47%

Finally, the safe zone for the Cunchibamba parish is located on the Sub Center Cuchibamba, with a value of 32.47%. The previous process was made for each of the 27 parishes, evaluating a total of 581 Alternatives as resources of the canton Ambato. Among these alternatives

Management of Humanitarian Logistics to Natural Disasters

103

there are: 35 Health Centers and Sub centers, 182 educational units, 128 Sport Facilities (sports fields, stadiums, others), 127 Community Police Units (CPU), 109 Squares and parks. These results are presented in Table 2, where the Civil Works chosen as collection centers for each parish are listed. For La Matriz parish, due to its vast area, two collection centers were defined. Table 2. Selected locations by AHP as collection centers in the parishes of the canton Ambato. Parish

Collection center

Atocha-Ficoa U.E. Ruminahui

Parish

Collection center

Celiano Monge

U.E. Maria Natalia Vaca

Huachi Chico Centro de Salud N.- 3

Huachi Loreto

Estadio Bellavista

La Matriz

U.E. Augusto Nicolas Martinez

Martinez

U.E. Bolivar

La Merced

U.E. canton Ambato

La Peninsula

Sub Centro La Peninsula

Pishilata

Polideportivo Ivan Vallejo

San francisco

U.E. Liceo Juan Montalvo

Ambatillo

Sub Centro Ambatillo

Atahualpa (Chisalata)

U.E. Atahualpa

Augusto N. Martinez

Sub Centro Augusto N. Martinez

Constantino Fernandez

Estadio Central

Cunchibamba Sub Centro Cunchibamba

Huachi Grande

U.E. Huachi Grande

Izamba

Sub Centro Izamba

Juan Benigno Vela

U.E. Chibuleo

Montalvo

Sub Centro Montalvo

Pasa

Sub Centro Pasa

Picaigua

Estadio Central Pilaguin (Pilahuin)Cancha del Barrio La Matriz

Quisapincha

Estado Cashauco

San Bartolome de Pinllog

Estadio Santa Elena

San Fernando Sub Centro San Fernando

Santa Rosa

Sub Centro Misquilli

Totoras

Unamuncho

Parque Central

3.3

Sub Centro Totoras

Step 3: Using Center of Gravity Method for Determining the Main Collection Center with ArcGIS

It is necessary to define a general collection center to guarantee that the estimated average travel time for heavy machinery to any place of the canton Ambato, will not exceed 6 h in extreme conditions and by third category ways. For defining the location of the general collection center in canton Ambato, the resultant places of the previous step are considered, each one has coordinates X and Y obtained with the ArcGIS software and for calculating the center of gravity among all the locations, the volume value used is the number of inhabitants of each parish. For illustrating this let’s analyze Quisapincha parish example: it has the coordinates X: 757954.62; Y:9863146.70; and a population of 13.001 inhabitants, so this volume is multiplied by the coordinates X and Y as shown below in Eqs. 3 and 4. X1 =

757954, 62 (U T M ) ∗ 13001 (personas) = 757954, 62(U T M ) 13001 (personas)

(3)

Y1 =

9863146, 7(U T M ) ∗ 13001 (personas) = 9863146, 7(U T M ) 13001 (personas)

(4)

104

S. Velastegui et al.

This procedure will be repeated for each prioritized spot and for all of the parishes, as shown in Eqs. 1 and 2, obtaining the following results Xc = 764440, 51(U T M ) and Y c = 9862268, 45(U T M ) for the coordinates of the general collection center of the canton Ambato. The resultant coordinates locate the general collection center for the canton Ambato in the Educational Unit Bolivar, which is in the urban parish La Matriz, in the Huachi Chico sector. 3.4

Step 4: Determining Costs of Humanitarian Assistance Kits

The humanitarian assistance kits and the delivery criteria are defined in Table 3. Table 3. Humanitarian Assistance Kits and delivery criteria. Complementary Kit name Delivery Criteria. Food

It will be delivered according to the number of family members: from 1 to 4 members, every 15 days up to twice; from 5 to 8, two deliveries every 15 days; from 9 members and more, 3 deliveries every 15 days, up to twice

Kitchen/Family crockery

1 kit for a four-member family and for just once

Sleeping

Delivery per family according to the number of members and only once

Personal Hygiene

1 kit for a four-member family, for just once

Cleaning Kit for family 1 kit for a four-member family and for just once Temporary Emergency Provisioning the active shelters for common use Shelters of the accommodated families. Average of 10 families, 40 people Community kitchenware for Shelters

For common use of the accommodated families (in correspondence with the number of families in the shelter, the standard for a shelter is 10 families)

Cleaning and shelter

For common use of the accommodated families (the amount is in correspondence with the number of families in the shelter)

Volcano

1 kit per person and for just once

Calculating the Number of Humanitarian Assistance Kits. The population of each parish was determined by summing up the total of persons of the zones and subzones indicated in the Census Plan of the province’s capital (Ambato), provided by the INEC (National Institute of Statistics and Censuses,

Management of Humanitarian Logistics to Natural Disasters

105

Fig. 3. Parish population of Ambato canton.

2016) which provided also the total amount of persons in canton Ambato which is 326834 inhabitants. (See Fig. 3). For calculating the amount of kits let’s observe the following example for the Food Kit in the urban parish of Pishilata where the given scenario of seismic risk has a damage value of 80% (percentage based on the damage caused by the earthquake occurred in canton Ambato, 1949). The estimation is done based on the number of inhabitants per sector, percentage of damage and delivery criteria as shown in the following Eq. 5: numberF oodKits =

P arish population ∗ % damage Delivery criteria

(5)

The urban parish Pishilata, according to the INEC, has approximately 12.993 inhabitants; it is considered that there is a damage of 80% and the delivery criteria for the food kits is 1kit for 4 persons, every 15 days obtaining the following Eq. 6: 12933 persons ∗ 80% numberF oodKits = (6) 4 persons/kit numberF oodKits = 2586.6 kits = 2587 kits

(7)

Therefore, for emergency situations the Pishilata parish must have approximately 2587 food kits every 15 days.

106

S. Velastegui et al.

Emergency Budget. Once the amount of emergency kits is calculated per sectors, considering 80% of damage, the budget in the main supermarkets and in the web sites for products sale, was determined for the city of Ambato, obtaining the cost of the complementary Kit in American dollars. Table 4 shows, for each type of kit, the amount of kits needed, the price and the total budget in American dollars. Table 4. Total budget for Complementary Kits. Type of complementary kit

Amount of kits Price (US$) Total

Food

65366

Kitchen/Family crockery Sleeping

58.58

3,829,140.3

65064

92.70

6,031,432.8

65366

213.00

13,922,958.0

Personal Hygiene

65366

45.93

3,002,260.4

Cleaning Kit for family

65366

27.99

1,829,594.3

6537

361.95

2,366,067.2

326834

15.00

Community kitchenware for Shelter Volcano (personal) Total

4,902,510.0 35,883,963.0

With an 80% of damage, in a similar scenario to the earthquake of 1949, a total budget of approximately $35,883,963.00 will be needed for facing an emergency.

4

Conclusions

A Humanitarian Logistics proposal was generated for the canton Ambato with the aim of improving the decision making process and reducing the time of response in emergency situations. For each parish, different alternatives of temporary shelters were defined. Also, the amount of emergency kits per sector was determined and the emergency budget for granting aid to the affected citizens was calculated. This way, the competent authorities have a clearer view of the scenario and can fully intervene with better results. From the 581 alternatives for safe zones defined as collection centers, 27 locations were defined as the most convenient using georeferentiation with the ArcGIS software and the AHP method. In addition, the Super Decisions software allowed to reduce calculation times for resources prioritization and also allowed to obtain the inconsistency index, which resulted in a value under 10% for all the evaluations of each parish. Regarding the definition of the general collection center, determined with the Center of Gravity method, storage and/or temporary shelters, considering the location of the main spots, the Educational Unit Bolivar, in the urban parish La Matriz, of Huachi Chico sector, was the best result obtained.

Management of Humanitarian Logistics to Natural Disasters

107

Finally, it can be stated that for facing a natural catastrophe in the canton Ambato, with 80% of damage, an emergency budget of approximately USD. 35,883,963.0 will be needed for covering the cost of the emergency kits. As future work it is recommended to carry out studies where optimal alternate routes of delivery of humanitarian aid are defined, from the collection center to the prioritized points, since if an adverse event occurs it may cause several damages in road infrastructure, preventing the transfer and making delivery more difficult. Conduct seismic micro-zoning studies in the canton Ambato to deepen the study of humanitarian logistics in each of the parishes of the canton Ambato, through obtaining more precise topographic charts and shapefiles, which will increase the accuracy of the distribution of optimal routes and delivery of kits, when applying the ANP decision-making tool.

References 1. Ratna, S., et al.: A hybrid MCDM model for supplier selection in supply chain. Int. J. Mech. Prod. Eng. Res. Dev. 9(3), 143–150 (2019). https://doi.org/ 10.24247/ijmperdjun201915. http://tjprc.org/publishpapers/2-67-1555150838-15. IJMPERDJUN201915.pdf 2. Anjomshoae, A., Hassan, A., Wong, K.Y.: An integrated AHP-based scheme for performance measurement in humanitarian supply chains. Int. J. Productivity Perform. Manag. 68(5), 938–957 (2019). https://doi.org/10.1108/IJPPM-04-20180132. https://www.emeraldinsight.com/doi/10.1108/IJPPM-04-2018-0132 3. Badina, S.V.: Socio-economic potential of municipalities in the context of natural risk (case study – Southern Siberian regions). In: IOP Conference Series: Earth and Environmental Science, vol. 190, October 2018. https://doi. org/10.1088/1755-1315/190/1/012001. http://stacks.iop.org/1755-1315/190/i=1/ a=012001?key=crossref.fd4b96081a9491c64fe74f90941cf771. 012001 4. Beatty, T.K.M., Shimshack, J.P., Volpe, R.J.: Disaster preparedness and disaster response: evidence from sales of emergency supplies before and after hurricanes. J. Assoc. Environ. Resour. Economists 6(4), 633–668 (2019). https://doi.org/10. 1086/703379. https://www.journals.uchicago.edu/doi/10.1086/703379 5. Borensztein, E., Cavallo, E., Jeanne, O.: The welfare gains from macro-insurance against natural disasters. J. Dev. Econ. 124, 142–156 (2017). https://doi.org/10. 1016/j.jdeveco.2016.08.004. https://linkinghub.elsevier.com/retrieve/pii/S03043 87816300621 6. Caruso, G.D.: The legacy of natural disasters: the intergenerational impact of 100 years of disasters in Latin America. J. Dev. Econ. 127, 209–233 (2017). https://doi. org/10.1016/j.jdeveco.2017.03.007. https://linkinghub.elsevier.com/retrieve/pii/ S0304387817300317 7. d’Ercole, R., Trujillo, M., Zucchelli, M., Portaluppi, C.: Amenazas, vulnerabilidad, capacidades y riesgo en el Ecuador: los desastres, un reto para el desarrollo. Cooperazione Internazionale (COOPI); Institut de Recherche pour le D´eveloppement (IRD); OXFAM GB (2003) 8. Ergu, D., Kou, G., Shi, Y., Shi, Y.: Analytic network process in risk assessment and decision analysis. Comput. Oper. Res. 42, 58–74 (2014). https://doi.org/10.1016/ j.cor.2011.03.005. http://linkinghub.elsevier.com/retrieve/pii/S0305054811000785

108

S. Velastegui et al.

9. Geofisico, I.: Informe Sismico Especial. Technical report 13 (2016). https://www. igepn.edu.ec/servicios/noticias/1317-informe-sismico-especial-n-13-2016 10. Giannakis, M., Papadopoulos, T.: Supply chain sustainability: a risk management approach. Int. J. Production Econ. 171, 455–470 (2016). https://doi.org/10.1016/j. ijpe.2015.06.032. https://linkinghub.elsevier.com/retrieve/pii/S0925527315002704 11. Hoque, M., Tasfia, S., Ahmed, N., Pradhan, B.: Assessing spatial flood vulnerability at Kalapara Upazila in Bangladesh using an analytic hierarchy process. Sensors 19(6), 1302 (2019). https://doi.org/10.3390/s19061302. https://www.mdpi.com/1424-8220/19/6/1302 12. Hussain, M., Khan, M., Ajmal, M., Sheikh, K.S., Ahamat, A.: A multi-stakeholders view of the barriers of social sustainability in healthcare supply chains: analytic hierarchy process approach. Sustain. Acc. Manag. Policy J. 10(2), 290–313 (2019). https://doi.org/10.1108/SAMPJ-05-2018-0140/full/html 13. de Ingenieros, C.F.: Comision Permanente de Estudio y Revision del Codigo Sismico de Costa Rica (2005). Codigo Sismico de Costa Rica (2002) 14. Kovacs, G., Moshtari, M.: A roadmap for higher research quality in humanitarian operations: a methodological perspective. Eur. J. Oper. Res. 276(2), 395–408 (2019). https://doi.org/10.1016/j.ejor.2018.07.052. https://linkinghub. elsevier.com/retrieve/pii/S037722171830674X 15. Montagnese, C., Santarpia, L., Buonifacio, M., Nardelli, A., Caldara, A.R., Silvestri, E., Contaldo, F., Pasanisi, F.: European food-based dietary guidelines: a comparison and update. Nutrition 31(7–8), 908–915 (2015). https:// doi.org/10.1016/j.nut.2015.01.002. http://linkinghub.elsevier.com/retrieve/pii/ S0899900715000076 16. Puri, J., Aladysheva, A., Iversen, V., Ghorpade, Y., Br¨ uck, T.: What methods may be used in impact evaluations of humanitarian assistance? (2014) 17. Seyedmohammadi, J., Sarmadian, F., Jafarzadeh, A.A., McDowell, R.W.: Development of a model using matter element, AHP and GIS techniques to assess the suitability of land for agriculture. Geoderma 352, 80–95 (2019). https://doi.org/ 10.1016/j.geoderma.2019.05.046. https://linkinghub.elsevier.com/retrieve/pii/ S0016706118302805 18. Skilodimou, H.D., Bathrellos, G.D., Chousianitis, K., Youssef, A.M., Pradhan, B.: Multi-hazard assessment modeling via multi-criteria analysis and GIS: a case study. Environ. Earth Sci. 78(2) (2019). https://doi.org/10.1007/s12665-018-8003-4 19. Urango-Licona, O.D., P´erez-Ortega, G., Romo-Morales, G.: Aplicacion de las tecnicas de centro de gravedad y AHP para la localizacion de un centro de distribucion de productos industriales en Colombia. Revista CEA 1(2), 79 (2015). https:// doi.org/10.22430/24223182.132. http://revistas.itm.edu.co/ojs/index.php/revistacea/article/view/132 20. Zhang, X., Song, J., Peng, J., Wu, J.: Landslides-oriented urban disaster resilience assessment-a case study in ShenZhen. China. Sci. Total Environ. 661, 95– 106 (2019). https://doi.org/10.1016/j.scitotenv.2018.12.074. https://linkinghub. elsevier.com/retrieve/pii/S0048969718349143

Variability Features in Building Approaches for Context-Aware Mobile Applications Estevan Gómez-Torres1(B)

, Cecilia Challiol2,3

, and Silvia E. Gordillo2,4

1 Carrera de Ingeniería en Informática, Universidad UTE,

Av. Mariscal Sucre y Mariana de Jesús, Quito, Ecuador [email protected] 2 LIFIA, Facultad de Informática, UNLP, La Plata, Buenos Aires, Argentina {ceciliac,gordillo}@lifia.info.unlp.edu.ar 3 CONICET, Buenos Aires, Argentina 4 CICPBA, Buenos Aires, Argentina

Abstract. The growth of mobile applications has been exponential in the last couple of years and it has come with technological advances, such as embedded sensors in mobile devices. This brings out greater challenges in the development of context-aware mobile applications, according to the demands of the current market. Currently, there are building approaches for this kind of applications, but these do not have flexibility in the generated applications. Until now, there is not a unified solution for this kind of applications, so, this is an open research area. This paper presents a taxonomy of variability concepts (Relevance, Combination, Precision and Accuracy’s Margins, Configuration Type, and Execution Type) to be taken into account when designing building approaches for context-aware mobile application. When these approaches are designed from scratch considering these variability concepts, this allows generating a wide variety of applications. The contribution of our taxonomy is to help the designer to identify the potential variability points in order to obtain more flexible approaches. The aim is to generate a discussion in relation to the variability concepts of the proposed taxonomy, this provides guidelines to be able to achieve variability in this kind of approaches. We hope this will enrich the discussion in relation to this kind of approaches in order to the unification of features that should be handled by these building approaches to obtain variability. Keywords: Building approaches · Context-aware mobile applications · Variability · Mobile computing

1 Introduction In recent years, the growth of mobile applications has been exponential and it has come with technological advances such as embedded sensors (GPS, accelerometer, etc.) in mobile devices. This kind of progress has allowed context-aware mobile applications

© Springer Nature Switzerland AG 2020 E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 109–123, 2020. https://doi.org/10.1007/978-3-030-35740-5_8

110

E. Gómez-Torres et al.

[1] to penetrate the market, becoming more and more commons. Until a couple of years ago, this was only a research topic and the applications that have been built did not go beyond the prototypical stage. This generates a new challenge when having to think about how to support the creation of this kind of applications to adapt to the demand. The concept of context has been explored in different areas of Computer Science [2] (e.g.: Artificial Intelligence, Home Automation, etc.). However, this is carried out from different perspectives depending on each author [1], for example, modeling solutions, building approaches for this kind of applications, etc. Although, this is a topic that has been investigated in the last twenty years, there is not yet unified solution for this kind of applications as mentioned in [1] and [3]. Therefore, it remains an open area of research. Moreover, it has emerged a new issue, which consists of how to build contextaware mobile applications that are really useful for users [4], something vital in today’s market. A feature of this kind of applications is the variability, which can be managed from different levels of abstraction. For example, in [5] a way is proposed to provide variability support at the modeling level at both designing and execution time. In [6], a taxonomy is detailed understanding the differences and similarities between various ways of handling the variability in context-aware software. When should be define building approaches for context-aware mobile applications, there are not clear guidelines to take into account to ensure that these support variability, particularly in the kind of applications that can be generated. Besides, there are building approaches for both non-expert users [7] and experts [8] which require some technical knowledge as modeling features. However, these approaches are not designed to have variability features in the applications that derive; for example, they only provide GPS as a location mechanism. The aim of this paper is to present a taxonomy that specifies variability concepts to be taken into account when designing approaches for the creation of context-aware mobile applications. When these approaches are designed from scratch considering these variability concepts, this allows generating a wide variety of applications. The contribution of our taxonomy is to help the designer to identify the potential variability points in order to obtain more flexible approaches. To do that, each concept of the taxonomy is described using a pattern-based format, describing how and what could be considered in the designing phase of this kind of approaches. The aim is to generate a discussion in relation to the variability concepts of the proposed taxonomy, this provides guidelines to be able to achieve variability in this kind of approaches. We hope this will enrich the discussion in relation to this kind of approaches in order to the unification of features that should be handled by these building approaches to obtain variability. The paper is structured as follows. In Sect. 2, related works are presented which are related to the variability of context-aware applications. The proposed variability taxonomy is detailed in Sect. 3. In Sect. 4 a discussion of the topic is generated. Conclusions and Future works are detailed in Sect. 5.

Variability Features in Building Approaches

111

2 Related Works The concept of variability in context-aware applications has been handled from different perspectives. In [9], 36 different ways of modeling context-aware are analyzed, which implies variability in the form of representation of this kind of applications. Based on these identified models, in [9] 10,498 elements of context are identified which more than half do not have a clear classification of how to categorize them. In some cases, context-aware models allow for handler variability not only at the design level, but also at execution time [5]. This allows, for example, dynamically add context while the application is running without requiring to be compiled again. Therefore, to have this support, the modeling approach should be designed to consider variability. In [6], a taxonomy is presented understanding the differences and similarities in how variability is handled in context-aware software. This taxonomy focuses on analyzing three axes: mechanisms to support variability in binding time (post-deployment and runtime), context-feature’s types and dependencies between contextual and no-contextual features. In [6] the authors are focus on variability in context-aware software not on building approaches. There are currently several building approaches for context-aware mobile applications. For example, the App Inventor [7] is an “online” program, which allows users to create Android applications without having any technical knowledge. The generated applications can include only GPS as a location-sensing mechanism. That is, users can only create applications for outdoor spaces. The App Inventor allows configuring precision and accuracy of the location sensor (in this case, GPS). However, it is not possible to combine accelerometer sensor to orientation sensor, in that the configuration reacts to these three sensors separately. In this case, the App Inventor only focuses on some contexts of the device and the user’s location. On the other hand, WebRatio Mobile [8] allows users to create context-aware mobile web applications. WebRatio Mobile is oriented to expert users who should have knowledge of databases and hypermedia design. The generated applications are packaged in PhoneGap, facilitating their use on both Android and iOS platforms. The users could define contexts related to Device, Network Connectivity and Location. In this case, location setting is also limited to GPS, allowing only applications for outdoor spaces to be generated. It can define precision and accuracy related to GPS. Considering the description mentioned above, these approaches [7] and [8] are limited. In particular, they focus only of building applications for outdoor spaces where certain contexts are also considered, most of them related to the available APIs. Therefore, these approaches are not designed to support variability in the kind of generated applications. We hope to contribute in this aspect with the proposed taxonomy in this paper.

112

E. Gómez-Torres et al.

3 The Proposed Variability Taxonomy In [10] we present an initial version of a taxonomy for context-aware mobile applications building approaches. Only a brief description of each concept of the taxonomy is detailed in [10], without going into about how should be handled in building approaches. We have been working in the area of context-aware mobile applications from more than ten years, this allows us to know and identify points of variability in this kind of applications. According to that, we think that the format used for design patterns in different areas of Software Engineering is a good way to describe our taxonomy. Thus, in this paper will expand each concept by describing them with a pattern-based format. Note that each concept is described not only considered the existing literature (such as [1–5]) but also based on our expertise in the area. Each pattern of variability has the following structure: • Name of the Variability Concept • Scenario: Describe situations in which the concept be handled by building approaches. This allows understanding situations that could be presented to the user of a building approach, and how this concept would provide flexibility if it should be handled by building approaches. • Purpose: Define the specific objective or motivation for handler the concept from a building approach. • Applicability: Every concept has challenges that should be faced when it wants to be implemented as part of a building approach. These challenges could be occur at both levels conceptual approach and tool. Therefore, how it is feasible to handle these challenges be shown. – Applicability at Conceptual Approach Level: Represented the conceptual way to handle these challenges. This is the first step that the challenges be handled by a tool. – Applicability at Tool Level: The ability of a tool to handle these challenges. In other words, how the challenges can be treated from a tool. Sometimes happen that at the conceptual level could be achieved, but at a tool level demands a lot of cost or it is not viable according to the current technology. Using the structure described above, each variability concept of the proposed taxonomy is described below in Tables 1, 2, 3, 4, 5 and 6.

Variability Features in Building Approaches

113

Table 1. Variability concept: relevance. Name of the variability concept

Relevance

Scenario

Classify, in some way, the level of importance of a context would allow working with contexts at different levels. For example, to take into account only the most important contexts to generate reduced versions of applications which only provide services related to some selected contexts. Having reduced versions allows, for example, generating applications that are less heavy or that consume fewer resources

Purpose

Being able to specify the importance of each context, from a building approach, would allow making different decisions

Applicability at conceptual approach level

A way of indicating the relevance of each defined context should be provided, for example, using a numerical or descriptive ranking. This could do it by associating a certain value with each defined context. The scale of relevance should be clearly established

Applicability at tool level

Using the scale of relevance defined at the conceptual level, from a tool could be implemented using, such as, a combo-box with the possible values. In addition, some actions could be taken related to this values. For example, this could be implemented in the tool to be considered when applications are derived, indicate which contexts are considered in them Although specifying the relevance is technically feasible to implement in a tool, handling how this impacts, for example, in the derivation of applications and also involves defining heuristics of what actions to take based on the indicated values Note that the flexibility of the tool could be affected by the heuristics that are defined in relation to relevance

Table 2. Variability concept: combination. Name of the variability concept

Combination

Scenario

Allowing indicating which contexts could be combined, for example, to provide services that are more complex. This implies defining which context they want to combine but also how they will behave together to provide, for example, services

Purpose

Being able to specify how contexts can be combined, from a building approach, would allow providing complex services that depend on several contexts (continued)

114

E. Gómez-Torres et al. Table 2. (continued)

Name of the variability concept

Combination

Applicability at A way of indicating how to combine contexts should be provided; conceptual approach level for example, grouping them in some way. Moreover, each group should define how the different values of each contexts are working together to provide, for example services. This could be implemented using rules as if-them in which the conditions indicated what is the specific value of each considered context, in order to apply the rule Applicability at tool level

In a tool could be used the specification of combination defined at the conceptual level. For example, visually represent the group of contexts that want to combine, and then it associates services to this group Specify the combination of contexts is technically feasible to implement in a tool; however, the specification of how to combine services can be complex since there should be clear rules of how to react to each value that these contexts could take It is worth mentioning that sometimes the contexts take values from sensors; this brings as a consequence the fact should be considered the margin of error of these sensors when defining how the contexts behave when they are combined

Table 3. Variability concept: precision and accuracy’s margins. Name of the variability Precision and accuracy’s margins concept Scenario

Defining values of precision and accuracy’s margin related to physical sensors allow having what is the error respect to the sensed value. This helps to interpret better the context’s values that depend on these sensors. For example, the accuracy of the GPS will allow knowing how much the user’s location is accurate

Purpose

Being able to specify the precision and accuracy in relation to physical sensors, from a building approach, would allow interpreting better the context’s values. According to these values, different decisions could be made (continued)

Variability Features in Building Approaches

115

Table 3. (continued) Name of the variability Precision and accuracy’s margins concept Applicability at conceptual approach level

A way to indicate the precision and accuracy related to physical sensors should be provided. Different actions could be specified according to these values, for example, to adjust the sensed value according to the precision or accuracy which has. This could be represented using value ranges

Applicability at tool level

A tool could use how precision and accuracy’s margin have been defined at the conceptual level. For example, entering numerical values related to physical sensors, to indicate both the precision and the accuracy of them. In addition, it should consider what services (associated with these contexts) can be affected by how these sensors behave. For example, if the GPS has a certain precision and accuracy, then in the generated application it will be possible to reflect the user’s location more accurately. Specifying precision and accuracy’s margin related to physical sensors is technically feasible to implement in a tool; however, the complexity is associated with how these values affect services related to the contexts. So, this is not only involved to define how to react to each sensed value but also how to consider the precision and accuracy of them

Table 4. Variability concept: categorization. Name of the variability Categorization concept Scenario

Classify the types of context would allow working with them in different ways. For example, the derivation of the contexts of the user, environment or mobile objects does not have the same treatment. That is, it is not the same to take location from the user that a mobile object; this impacts on which sensors are used in each case to take the location

Purpose

Being able to specify to which category a context belongs, from a building approach, would allow being able to make different decisions or to enable different options (continued)

116

E. Gómez-Torres et al. Table 4. (continued)

Name of the variability Categorization concept Applicability at conceptual approach level

A way of indicating the context’s category to which it belongs should be provided; for example, by a description. It should be clearly established what concept is defined and what is associated with each category, for example, sensors available for each category

Applicability at tool level

Using the categories defined at the conceptual level, from a tool could be implemented using, such as, combo-box with the possible values. In addition, it should specify in the tool which restrictions are associated with each category, for example, what sensors are available or how these are derived in the generated applications Although specifying the categories is technically feasible to implement in a tool; however, handling how each of these impacts the generated applications requires defining different heuristics. For example, user’s contexts should be specified differently from mobile objects’ context, but also be derived differently

Table 5. Variability concept: configuration type. Name of the variability Configuration type concept Scenario

Specifying the configuration type associated with each context would allow providing more flexibility in the generated applications. For example, the configuration type could be passive or active, according to [1]; in which these types are defined in relation to if it requires user’s intervention or not. The passive configuration requires specifying what data the user should enter, meanwhile the active configuration is automatic, for example, using automatic learning. It could also represent the default configuration

Purpose

Being able to specify the configuration type, from a building approach, would allow establishing in which way the contexts are set to take their values (continued)

Variability Features in Building Approaches

117

Table 5. (continued) Name of the variability Configuration type concept Applicability at conceptual approach level

When the configuration type is chosen, more information should be specified in relation to what each type requires. For example, the passive configuration requires defining what the user should specify. Meanwhile, the active configuration should consider with what automatic mechanism, it will be bound in order to make the configurations dynamically in runtime. In the default configuration, the values should be set. The configuration type impacts on the generated application since it is necessary to incorporate as part of it what each one requires

Applicability at tool level

Using the types of configuration defined at the conceptual level, from a tool could be implemented using, such as, a combo-box with the possible values. In addition, what each type requires should be implemented as a part of building approach in order to generate applications with the corresponding configuration Meanwhile, specifying the configuration type is technically feasible to implement in a tool, handling what each implies is not trivial. For example, the active configuration type requires automatic learning which it would not be simple to have as part of the generated application The passive configuration type should be a little less complex to implement since it could be defined by a form with the data that the user could configure in relation to each context (or sensor associated with it). The default configuration could be set from the tool with, for example, selected options

Table 6. Variability concept: execution type. Name of the variability Execution type concept Scenario

Specifying the execution type associated with each context would allow providing more flexibility in the generated applications. For example, execution type could be passive or active, according to [1]; in which these types are defined in relation to if it requires user’s intervention or not. Passive execution requires user intervention (for example, QR code reading); meanwhile active execution is automatic (for example, GPS)

Purpose

Being able to specify execution type, from a building approach, would allow establishing information about how the contexts behave (continued)

118

E. Gómez-Torres et al. Table 6. (continued)

Name of the variability Execution type concept Applicability at Conceptual approach level

When the execution type is chosen, more information should be specified in relation to what each type requires. For example, passive execution requires defining how the user interact with it, meanwhile active execution is an automatic mechanism without intervention. The execution type impacts on the generated application since it is necessary to incorporate as part of it what each one requires

Applicability at tool level

Using the types of execution defined at the conceptual level, form a tool could be implemented using, such as, a combo-box with the possible values. In addition, what each type requires should be implemented as a part of building approach in order to generate applications with the corresponding mechanisms Specifying the execution type is technically feasible to implement in a tool; however, handling what each implies is not trivial. The passive execution requires detail of how the application will interact with the user; as well as, how it will react for each possible interaction. For example, in the case of having a passive execution mechanism, such as reading QR codes to take the user’s location, it should be indicated how the generated application behaves when the user reads an incorrect code. Active execution is simpler to implement because it only requires the available APIs. For example, if GPS is used to take the user’s location, it is enough to make the appropriate connection to the location’s API as part of the generated application. This execution type is transparent to the users and their intervention is not required It could be possible that a context, such as user’s location could involve a mixed execution type, for example, reading of QR codes and GPS. In this case, the two previous solutions are combined, but in addition heuristics should be defined of which mechanism has more priority given that the GPS works constantly. This add complexity in the tools

Each variability concept of the proposed taxonomy is specified above. It could observe that each concept could involve a value to be established, but the complexity arises in relation to how each value impacts or what has to be defined based on it. The biggest challenge comes from being handler from a tool and how to derive applications considering what implies each variability concept. It should be mentioned that the taxonomy is not closed, but is an initial proposal to achieve flexibility in the context-aware mobile applications building approaches.

Variability Features in Building Approaches

119

4 Discussion In this section, it discusses different aspects of the taxonomy presented in Sect. 3; in order to help the reader understand how each concept can affect the variability supported by the building approaches. In [5], a model is proposed which considers the separation between the aware-objects concept of its context-feature and in addition from the sensors (which assign values to these features). When this decoupling of concepts is considered from a building approach, it allows the different layers could combine and thus support variability. For example, the location-feature could set its values from different associated sensors, and each of them could has different implementations. This decoupling allows reuse and extensibility. This is a possible way to represent these concepts in a building approach to handler a first level of variability; beyond that incorporates the variability concepts of the proposed taxonomy in this paper. According to what is detailed in [11], the sensors can be of different types, for example, physical, virtual (using applications and services), direct user input, etc. Each of them has its own configuration and way of execution. Considering what has been analyzed previously, Fig. 1 is shown an example of two aware-objects (User 1 and Package), each one with different context-features according to its nature.

Fig. 1. Examples of aware-objects with its context-features which are related to abstract and concrete sensors.

120

E. Gómez-Torres et al.

In addition, in Fig. 1 there is a separation between the concepts of abstract and concrete sensors. This allows the flexibility that the same concept of an abstract sensor is implemented differently, such as occurs with the GPS example; that could be implemented using the Google Places API [12] or the Android Location API [13]. The separation between the concepts of abstract and concrete sensors when it is considered from a building approach allows having more level of variability in relation to the combination that can be handled. It can also be observed in Fig. 1 that two sensors could set values to the same context-feature (in this case, the location-feature). This decoupling allows sensors to be added or removed without affecting the context concepts. This provides flexibility when considered as part of a building approach. Based on the example in Fig. 1, it is analyzed with more details how the variability concepts of the proposed taxonomy can be considered. The aware-objects could have its category associated. The range’s values of the category are used by the building approaches when deriving context-aware mobile applications. For example, deriving code that represents the user concept is not the same as referring to a mobile object. The relevance could be associated with the context-features. The relevance range could be used by the building approaches to generate reduced applications without all the functionality. For example, in the example presented in Fig. 1, the location could be more relevant than the activity; this depends on the services of the application that are being modeled. The combination could occur observing the values of different context-feature. From building approaches should be able to allow the specification of rules, as well as the actions that it triggers. Each time that a context-feature change its value the rules are evaluated. In the case of triggering services, they should be integrated into the building approach; in order to later derive these as part of the generated applications. Other triggers could update some context-features values. In the example of Fig. 1, it could happen that User 1 is waiting for the Package to arrive; a rule could be that each time the object’s state changes, the user is notified as long as she/he is not in a meeting. In this case, the rule observes the object’s state and the user’s activity, and based on its values, triggers the warning to the user as a service. Based on the analysis carried out, in Fig. 2 is shown a possible generalization of the variability concepts: categorization, relevance and combination. Following the analysis of the variability concepts of the proposed taxonomy, each of the specific sensors could involve different types of configuration and execution. That is, at this level these variability concepts could be handled from building approaches. The configuration type is useful for a building approach since it allows identifying if user’s intervention would be required or not. In the case of it is requiring it is important to define how this would be carried out. In the case “Active” configuration [1], it should be taken into account that the building approach should have, at least, one way of including the monitoring program as part of the generated application. “Passive” configuration [1] when is supported by building approach should have, at least, one way of indicating how the user will perform such a configuration; for example, designing the form which would

Variability Features in Building Approaches

121

Fig. 2. Generalization of the variability concepts: relevance, combination and categorization.

then be embedded in the generated application. In the case of “default” configuration, the generated application defines the fixed values which have been defined using the building approach. In the example of Fig. 1, the APIs related to GPS and Estimote Beacons [14], they could be set, for example, as the default configuration. The monitoring program could be set as “Active” configuration, so, it could learn to configure itself as it monitors. Whereas the manual entry could have, for example, a “Passive” configuration, being able to design from the building approach the form in which the user could enter new possible values. For the execution type, some specific sensors determine automatically how they work; this facilitates the auto-completion of this value when it is handled by a building approach. For example, if the option is GPS or Beacons, they have an “Active” execution, meanwhile manual entry is “Passive” and it should define how the user could interact with the generated application, for example, using a form. In the case of the monitoring program could require or not user’s intervention, and this defines the execution type associated with it. In this last case, the building approach could not autocompleted the execution type associated with the sensor. The specification of precision and accuracy is usually associated with concrete physical sensors. Building approaches should consider the range’s values that precision’s margin can take. However, this is used at runtime in the derived applications; that is, the margin of error is specified in the building approach but allows decisions to be handled in the generated applications. In the example of the Fig. 1, the APIs related to GPS and Beacons Estimote are those that could define precision and accuracy’s margins. Based on the analysis carried out, in Fig. 3 is shown a possible generalization of the variability concepts: configuration type, execution type and the precision and accuracy’s margins. Figures 2 and 3 allow to observe a way of representing the variability concepts which should be associated with aware-objects, context-features or specific sensors. This is important when designing building approaches to consider these variability concepts. In this session, it has been discussed as a possible way to represent the variability concepts of the proposed taxonomy. This representation could vary according to how

122

E. Gómez-Torres et al.

Fig. 3. Representation of the variability concepts: configuration type, execution type and precision and accuracy’s margins.

the basic concepts to be represented by each building approach. The examples presented in this session are based on the concepts defined in [5].

5 Conclusions and Future Work In this paper, a taxonomy was presented that specifies variability concepts to be taken into account when designing approaches for the creation of context-aware mobile applications. This taxonomy is focused on providing support to generate a wide range in the kind of applications from these approaches. Each variability concept of the taxonomy is described using a pattern-based format, focusing on the challenges involved in considering each of them as part of an approach to build this kind of applications. The proposed taxonomy is not complete or closed, but it is the first definition in order to have variability in the building approaches. In addition, a discussion has been generated in relation to the variability concepts of the proposed taxonomy. It has been used the concepts defined in [5] in which separates the concepts of aware-objects, context-features and sensors; based on this, a possible way of handler each variability concept was described. Using this paper, designers of building approaches for context-aware mobile applications have a guidelines to be able to achieve variability. We hope the presented discussion will enrich how could be handled variability in these building approaches. Moreover, we wish to contribute to a unified solution to this kind of approaches. As future work, a concrete building approach will be designed to put into practice the proposed taxonomy. It is desirable that by designing this approach, the taxonomy could be enriched by incorporating new variability concepts. In addition, we will analyze how to enrich our taxonomy with other aspects of variability such as those proposed in [6].

Variability Features in Building Approaches

123

References 1. Alegre, U., Augusto, J.C., Clark, T.: Engineering context-aware systems and applications: a survey. J. Syst. Softw. 117, 55–83 (2016) 2. Augusto, J., Aztiria, A., Kramer, D., Alegre, U.: A survey on the evolution of the notion of context-awareness. Appl. Artif. Intell. 31(7–8), 613–642 (2017) 3. Bauer, C., Dey, A.K.: Considering context in the design of intelligent systems: current practices and suggestions for improvement. J. Syst. Softw. 112, 26–47 (2016) 4. Alegre-Ibarra, U., Augusto, J.C., Evans, C.: Perspectives on engineering more usable contextaware systems. J. Ambient Intell. Humanized Comput. 9(5), 1593–1609 (2018) 5. Fortier, A., Rossi, G., Gordillo, S.E., Challiol, C.: Dealing with variability in context-aware mobile software. J. Syst. Softw. 83(6), 915–936 (2010) 6. Mens, K., Cardozo, N., Duhoux, B.: A context-oriented software architecture. In: 8th International Workshop on Context-Oriented Programming, pp. 7–12. ACM, New York (2016) 7. Bales, S.: Build Android Apps Without Coding: Get Started with Android Apps Using Thunkable-MIT app Inventor. Independently published. ACM, New York (2018) 8. Hamdani, M., Butt, W.H., Anwar, M.W., Azam, F.: A systematic literature review on interaction flow modeling language (IFML). In: 2nd International Conference on Management Engineering, Software Engineering and Service Sciences, pp. 134–138. ACM, New York (2018) 9. Bauer, C., Novotny, A.: A consolidated view of context for intelligent systems. J. Ambient Intell. Smart Environ. 9(4), 377–393 (2017) 10. Gómez-Torres, E.R., Challiol, C., Gordillo, S.E.: Context-aware mobile applications: taxonomy of factors for building approaches. In: XXV International Conference on Electronics, Electrical Engineering and Computing, pp. 1–4. IEEE (2018) 11. Rivero-Rodriguez, A., Pileggi, P., Nykänen, O.A.: Mobile context-aware systems: technologies, resources and applications. Int. J. Interact. Mobile Technol. 10(2), 25–32 (2016) 12. Google Place API. https://developers.google.com/places/web-service/intro. Accessed 28 June 2019 13. Android Location API. https://developer.android.com/training/location. Accessed 28 June 2019 14. Estimote Beacons. https://estimote.com. Accessed 28 June 2019

Wheelchair Controlled by Eye Movement Using Raspberry Pi for ALS Patients Jorge Buele1,2,4(B)

, José Varela-Aldás1 , Franklin W. Salazar2 and Víctor H. Andaluz4

, Angel Soria3

,

1 SISAu Research Group, Universidad Tecnológica Indoamérica, Ambato, Ecuador

{jorgebuele,josevarela}@uti.edu.ec 2 Universidad Técnica de Ambato, Ambato, Ecuador

[email protected] 3 Purdue University, Lafayette, USA

[email protected] 4 Universidad de Las Fuerzas Armadas ESPE, Sangolquí, Ecuador

[email protected]

Abstract. The mobility of people who have suffered a degenerative disease or an accident is partially or totally reduced, which limits their locomotive independence. Therefore, this paper presents a proposal that facilitates the mobility of people suffering from moderate levels of amyotrophic lateral sclerosis (ALS). A control system has been adapted to an electric wheelchair to provide it with a certain degree of intelligence. The acquisition of multimedia data is done with a small camera adapted to a glasses frame that the person must use. For eye patterns tracking, a recognition system is performed using the LabVIEW software environment. The control system that regulates the movement of the wheelchair was designed on the Raspberry Pi embedded board as a low-cost proposal. Experimental tests and user surveys validate the correct operation of this device. Keywords: Amyotrophic lateral sclerosis · Eye movement · Image processing · Smart wheelchair

1 Introduction Wheelchairs are medical devices that adapt a chair to four wheels (or six), and enabling free motion for people who have problems in their lower extremities [1–3]. They are mainly used in medical, rehabilitation centers and in homes where there are people with this type of disability [4, 5]. Technology advancement have turned this device from being a tool to a complex means of mobility [6, 7]. Especially for the users that wish to obtain greater autonomy despite their physical limitations, thus transforming them into active entities of society [8]. They can be classified into three large groups that include: manual, electric and intelligent wheelchairs, the former being the most common and the basis for the following. Manual type possesses a constraint requiring the accompaniment of a person or the physical effort of the user who often cannot exercise and is in a great limitation [9]. The electric wheelchairs already have batteries that provide electrical power © Springer Nature Switzerland AG 2020 E. Fonseca C et al. (Eds.): TICEC 2019, AISC 1099, pp. 124–136, 2020. https://doi.org/10.1007/978-3-030-35740-5_9

Wheelchair Controlled by Eye Movement

125

supply for the wheel motion, and incorporate electronic elements that provide greater benefits, making the company of another person not necessary [10]. Moreover, smart wheelchairs adhere to more advanced technological elements, and robotic techniques such as computer vision, trajectory control, pattern recognition for obstacle avoidance, etc. [11–13]. These devices are distributed for a greater economic value, which for most families represents a latent problem being unable to acquire them. The integration of hardware and software to the devices, allows the extension or improvement of the mobility of patients who are victims of motor diseases. These systems are increasingly adaptable to new technologies through power control systems [14]. Based in control systems, solutions for wheelchair navigation of patients with severe degenerative motor diseases, and with a severe degree of movement disability can be presented. Conditions related to diseases and/or physical traumas resulting from accidents, developed by medical conditions such as paraplegia, amputations caused by diabetes, arthritis, total or partial paralysis due to damage in one of the cerebral hemispheres, among others [15]. Amyotrophic Lateral Sclerosis (ALS) is a disorder that affects the nervous and muscular system that causes premature degeneration of motor neurons [16]. It occurs mainly in people between 40 and 60 years of age, 50 being the average age according to the diagnoses. However, there have been cases at earlier ages, between 20 and 40 years [17]. Half of the population that is diagnosed with ALS lives three or more years after diagnosis and up to ten percent manage to live for more than ten years. Therefore, solutions like this are required during the lifespan of patients, so they can perform various activities with autonomy, safety and comfort.

2 Related Works People with visual, motor and strength disabilities find it difficult to use a manual or an electric wheelchair as shown in [18]. To solve this problem, in [19] the main characteristics of different prototypes of intelligent wheelchairs proposed by several authors are presented. With the objective of becoming a bibliographic reference for future research. Due to the continuous advancement of technology in the field of automation and artificial intelligence. As a result, remarkable research has been carried out on intelligent wheelchairs, with various methods. In the work of [20], a preliminary study on people’s emotional perception of emotions and behavior that patients have, this as a first step in the construction of smart wheelchairs. By the Affect Control Theory (ACT), emotions are modeled based on social interactions that have proven suitable for the design of emotionally intelligent systems. Some of the respondents associated a greater amount of power and response of the wheelchair as obedience to them and the opposite effect as a resistance (as if it had a life of its own). An intelligent wheelchair that can be driven with different movements of the user’s head is presented in [21], in this work the user gestures can be detected through an accelerometer. The control algorithm was executed with the Arduino board, and for the electrical power supply a solar panel was used, which is intended to integrate clean energy solution into the system design. Similarly [22] describes the design of an intelligent wheelchair that uses an IMU (Inertial Measurement Unit) to capture the head movement and based on the captures the system establishes the movements that the user

126

J. Buele et al.

wishes to make. As a limitation of these two proposals, it should be noted that the movement of the head, over longer periods of time, can cause pain or discomfort in the patient’s neck and that is why this paper did not use a similar approach on this occasion. As a basis for this research, the proposal presented by [23], an intelligent wheelchair prototype based on eye tracking, designed for people with locomotive disabilities using OpenCV was used. In this work the image processing module consists of a webcam installed on the user’s glasses frame and a customized C++ image processing software. From the context of previous articles, this paper presents the design of a system that provides some intelligence to an electric wheelchair with the incorporation of electronic elements. For the acquisition of images, a webcam is used and an interface was developed in the LabVIEW software. The Raspberry Pi 3 embedded board is the central processing unit (CPU) that provides control actions for the motors that move each wheel respectively. This project is divided into 5 sections, including the introduction in Sect. 1 and related works in Sect. 2. Section 3 presents the methodology used and Sect. 4 describes the tests and results. Finally, the conclusions are presented in Sect. 5.

3 Methodology For the development of this prototype it is necessary to integrate technological elements that allow to repower a conventional wheelchair, whose description is presented in Fig. 1.

Fig. 1. General diagram of the control system implemented.

3.1 Hardware To achieve the correct eyeball tracking and monitoring of the necessary area, it is fixed on the right side of the frame of a pair of glasses, as shown in Fig. 2. These images are

Wheelchair Controlled by Eye Movement

127

sent to the Raspberry pi 3 board (controller system), a single-board computer (SBC) that operates at a processing speed of 1.2 GHz and has 1 GB of random-access memory (RAM). This embedded device not only performs image processing, but also establishes control signals for the movement of an electric wheelchair, which has an adjustable seat, backrest, armrests and footrests. The RoboteQ AX2550 controller is used to control the DC motors, which allows the precise control of the rotation degree, and speed of the motors, and for the required power supply Fuli batteries of 12 V and 35 Ah where used.

Fig. 2. Glasses with adapted webcam.

The transmission of the data obtained is performed using USB (Universal Serial Bus) protocol, this data bus allows a point-to-point connection to a personal computer. The Raspberry Pi 3 board sends two voltage signals received by the RoboteQ AX2550 controller, which provides the voltage necessary to start and maintain the movement of the wheels. To complete this prototype system, a printed circuit board was designed, with the purpose of filtering the noise, amplifying the signal and regulating the power of the motors. It is presented in Fig. 3.

Fig. 3. Electronic board for motor control.

128

J. Buele et al.

3.2 Software The interface design was developed in the National Instruments LabVIEW software, using the Vision Development Module for the image acquisition coming from a real time video capture, as can be seen in Fig. 4. Digital image processing is performed using the NI Vision OpenCV Utilities package, which facilitates the integration of OpenCV algorithms within the LabVIEW environment for execution on computers, as well as deployment to embedded boards (Raspberry Pi 3). To track the eyeball movement, a coherence algorithm has been developed as shown in Table 1, operating in the frames extracted from the real-time video of the user’s right eye.

Fig. 4. Design of the interface.

Table 1. Coherence algorithm conditions. Eye movement Value Wheelchair’s movement Closed eyes

0

Stop

Up

1

Start

Front

2

Forward

Left

3

Left

Right

4

Right

The original capture obtained is converted into a grayscale image and this in turn becomes a binary image through a simple threshold. The sclera is identified as the white area and the iris as a black area that is centered on the graphic indicator, and depending on its displacement, numerical values are assigned, which will then be used in the control stage. To start the algorithm for detecting the position of the eyeball: (i) The relaxed

Wheelchair Controlled by Eye Movement

129

position (forward vision) is determined and its mean is established at the position c (x, y); (ii) Detects the variation from the reference point to the new location; (iii) The direction of iris movement is calculated based on the following conditions, as described in Fig. 5 [22]:

Fig. 5. Pupil detection and movement considerations.

• • • •

Turning left if −65