ICAF 2019 – Structural Integrity in the Age of Additive Manufacturing: Proceedings of the 30th Symposium of the International Committee on Aeronautical Fatigue, June 2-7, 2019, Krakow, Poland [1st ed.] 978-3-030-21502-6;978-3-030-21503-3

Table of contents :
Front Matter ....Pages i-xix
Front Matter ....Pages 1-1
Fatigue Characteristic of Linear Friction Welded Ti-6Al-4V Joints (Hiroshi Kuroki, Yukihiro Kondo, Tsukasa Wakabayashi, Kenji Nakamura, Kikuo Takamatsu, Koji Nezaki et al.)....Pages 3-15
Fatigue in Additive Manufactured Aircraft: The Long Way to Make It Fly (Ivan Meneghin, Goran Ivetic, Matthias Stiller, Gianluca Molinari, Vjola Ristori, Sara Della Ratta et al.)....Pages 16-30
High Cycle Fatigue and Fatigue Crack Growth Rate in Additive Manufactured Titanium Alloys (Xiang Zhang, Abdul Khadar Syed, Romali Biswal, Filomeno Martina, Jialuo Ding, Stewart Williams)....Pages 31-42
Strain Controlled Fatigue Testing of Additive Manufactured Titanium Alloy Ti-6Al-4V (Rob Plaskitt, Andrew Halfpenny, Michelle Hill)....Pages 43-55
The Optimization and Design of Complicated-Surface Panel Based on Automate Fiber Placement (Tieliang Zhang, Liyang Liu, Hao Cui)....Pages 56-70
Front Matter ....Pages 71-71
A Modeling Approach for the Fatigue Behavior of Laser Drilled Micro Perforated Structural Panels (Dort Daandels, Stefan Riekehr, Nikolai Kashaev, Jon Mardaras, Sammy Zein El Dine, Christian Heck)....Pages 73-87
Assessment of Fatigue Behavior of Advanced Aluminum Alloys Under Complex Variable-Amplitude Loading (Kevin Stonaker, David Stanley, John G. Bakuckas Jr., Mike Kulak, Po-Yu Chang, Gongyao Wang et al.)....Pages 88-103
Damage Mechanisms and Mechanical Properties of Directly Bonded CFRTP and Aluminium with Nano-Structured Surface (Kristine Munk Jespersen, Hikaru Abe, Hiroki Ota, Kei Saito, Keita Wada, Atsushi Hosoi et al.)....Pages 104-112
Interaction Between High- and Low-Cycle Thermo-Mechanical Fatigue Crack Propagation Around Cooling Hole in a Ni-Based Superalloy (Masakazu Okazaki, Yuuki Yonaguni)....Pages 113-123
Ply Curving Termination to Suppress Delamination in Composite Ply Drop-Off (Shu Minakuchi, Nobuo Takeda)....Pages 124-132
Studies on the Fatigue Damage Behavior of Active Jet Engine Chevron (Bingfei Liu, Shangyang Jin, Shaozhe Dong, Zhenyu Feng)....Pages 133-142
Front Matter ....Pages 143-143
An Ultrafast Crack Growth Lifing Model to Support Digital Twin, Virtual Testing, and Probabilistic Damage Tolerance Applications (Juan Ocampo, Harry Millwater, Nathan Crosby, Beth Gamble, Christopher Hurst, Michael Reyer et al.)....Pages 145-158
Analytical and Numerical Investigation of the Effect of Secondary Bending in Hard-Point Joints (Yuval Freed, Lior Sagi Machnes, Orel Magidish)....Pages 159-175
Demonstration of an Airframe Digital Twin Framework Using a CF-188 Full-Scale Component Test (Guillaume Renaud, Min Liao, Yan Bombardier)....Pages 176-186
Development of Efficient High-Fidelity Solutions for Virtual Fatigue Testing (Javier Gomez-Escalonilla, Diego Garijo, Oscar Valencia, Ismael Rivero)....Pages 187-200
Effective Durability and Damage Tolerance Training: New Methods for Modern Learners (Brandon D. Chapman)....Pages 201-214
Fatigue Considerations in the Development and Implementation of Mechanical Joining Processes for Commercial Airplane Structures (Robert Jochum, Antonio Rufin, Tanni Sisco, Frederick Swanstrom)....Pages 215-227
Rapid Calculation of Safe Acceleration Values for Aircraft Structures Under Flight Test (Stephen Dosman, Jonathan Gorman)....Pages 228-244
Reliability Approach Applied on Fatigue Safety Factors for Fleet Monitoring (Vincent Montlahuc)....Pages 245-255
Research on the Airworthiness Compliance Strategy of Composite Structure (Weiping Li, Xiaoling Zheng)....Pages 256-264
Risks of Initial Assumptions in Fatigue and Damage Tolerance of Small Aircraft Development Programs (Dejan Romančuk, Juan Ocampo)....Pages 265-278
Russian Practice to Provide Safe Operation of Airplane Structures with Long-Term Operation (Boris G. Nesterenko, Grigory I. Nesterenko, Victor V. Konovalov, Vitaly Ya. Senik)....Pages 279-291
Smarter Testing Through Simulation for Efficient Design and Attainment of Regulatory Compliance (Steven A. Chisholm, Jack F. Castro, Brandon D. Chapman, Kazbek Z. Karayev, Andrea J. Gunther, Mohammed H. Kabir)....Pages 292-307
Widespread Fatigue Damage Evaluation for Multiple Elements Based on Probabilistic Approach (Fabiano Hernandes)....Pages 308-318
Front Matter ....Pages 319-319
A Framework to Implement Probabilistic Fatigue Design of Safe-Life Components (Joshua Hoole, Pia Sartor, Julian Booker, Jonathan Cooper, Xenofon V. Gogouvitis, Amine Ghouali et al.)....Pages 321-335
A Multiaxial Fatigue Damage Model for Isotropic Materials (Mauricio V. Donadon, Mariano A. Arbelo, Paulo Rizzi, Carlos V. Montestruque, Lucas Amaro, Saullo Castro et al.)....Pages 336-348
A Specimen to Evaluate Susceptibility of Aluminium Alloys to L-S Crack Deviation (Erembert Nizery, Jean-Christophe Ehrström, Guillaume Delgrange, Bruno Wusyk)....Pages 349-359
A Numerical Approach to the Disbonding Mechanism of Adhesive Joints (Nicola Zavatta, Enrico Troiani)....Pages 360-371
An Engineering Calculation Method of Probability Distribution of Crack Initiation Life for Widespread Fatigue Damage (Xi Wei, Li Qiang, Shen Peiliang, Yang Gang, Huang Fu, Zhao Jianjun)....Pages 372-383
Assessment of Aircraft Structural Service Life Using Generalized Correction Methodology (Hongna Dui, Xiaodong Liu, Jiang Dong, Lixin Zhang)....Pages 384-398
Examination of the KAWAI CLD Method for Fatigue Life Prediction of Composites (Yael Buimovich, Dvir Elmalich)....Pages 399-409
Fatigue Crack Growth Prediction and Verification of Aircraft Fuselage Panels with Multiple Site Damage (Shaopu Su, Jianghai Liao, Wendong Zhang, Dengke Dong)....Pages 410-422
Fatigue Life Prediction of CFRP Laminate Under Quasi-Random Loading (Vitaly E. Strizhius)....Pages 423-431
Fatigue Life Simulation and Experiment of 2024 Aluminum Joints with Multi-Fasteners Interference-Fit (Qingyun Zhao, Yunliang Wang, Hong Huang, Sirui Cheng, Fenglei Liu)....Pages 432-443
Influence of Heat Treatment on Near-Threshold Fatigue Crack Growth Behavior of High Strength Aluminum Alloy 7010 (M. S. Nandana, Bhat K. Udaya, C. M. Manjunatha)....Pages 444-451
Multiaxial Fatigue Behavior of 30HGSA Steel Under Cyclic Tension-Compression and Reversed Torsion (Daniel Dębski, Krzysztof Gołoś, Marek Dębski, Andrzej Misztela)....Pages 452-460
Novel Methods for Measuring the Mode I and Mixed Modes I/II Interlaminar Fracture Toughnesses of Composite (W. Xu, Z. Z. Guo, Y. Yu, X. J. Zhang)....Pages 461-476
Numerical Investigations on the Three-Dimensional I/II Mixed-Mode Elasto-Plastic Fracture for Through-Thickness Cracked Bodies (Fang-li Wang, Ming-bo Tong, Shu-wei Bai, Nan Jiang, Chong-min She, Jun-ling Fan)....Pages 477-487
Probabilistic Reliability Assessment of a Component in the Presence of Internal Defects (Fedor Fomin, Nikolai Kashaev)....Pages 488-502
Stress-Intensity Factor Solutions for Tapered Lugs with Oblique Pin Loads (James C. Sobotka, Yi-Der Lee, R. Craig McClung, Joseph W. Cardinal)....Pages 503-517
Summary of Recent Round Robin Life Prediction Efforts for Crack Shape and Residual Stress Effects (Alexander V. Litvinov, James A. Harter, Robert Pilarczyk)....Pages 518-527
The Influence of Low and High-Cycle Fatigue on Dislocations Density and Residual Stresses in Inconel 718 (Elżbieta Gadalińska, Maciej Malicki, Bartosz Madejski, Grzegorz Socha)....Pages 528-538
Effect of Crack Length and Reference Stress on Variable Amplitude Fatigue Crack Growth Rate (E. Amsterdam)....Pages 539-550
Weibull or Log-Normal Distribution to Characterize Fatigue Life Scatter – Which Is More Suitable? (Abraham Brot)....Pages 551-561
Front Matter ....Pages 563-563
Bonded Repairs of Composite Panels Representative of Wing Structure (John G. Bakuckas Jr., Reewanshu Chadha, Paul Swindell, Michael Fleming, John Z. Lin, J. B. Ihn et al.)....Pages 565-580
Comparison of Rivet Hole Expansion for Protruding Rivets; Universal and with Compensator (Wojciech Wronicz)....Pages 581-588
Effect of Alternative Paint Stripping Processes on the Fatigue Performance of Aluminium Alloys - Atmospheric Plasma De-painting (Ali Merati, Marko Yanishevsky, Yan Bombardier)....Pages 589-599
Effect of Strengthened Hole on the Fatigue Life of 2024-T3 Aluminum Alloy (Hong Huang, Qingyun Zhao, Fenglei Liu)....Pages 600-605
Fatigue Crack Growth in Pin Loaded Cold-Worked Holes (Luisa Boni, Daniele Fanteria, Domenico Furfari, Luigi Lazzeri)....Pages 606-616
Fatigue Crack Propagation Influenced by Laser Shock Peening Introduced Residual Stress Fields in Aluminium Specimens (Sören Keller, Manfred Horstmann, Nikolai Kashaev, Benjamin Klusemann)....Pages 617-631
Influence of Bonded Crack Retarders on Damage Tolerance Performance of Fuselage Panel (Haiying Zhang, Dengke Dong, Yulong Wei, Weifeng Zang, Wenwei Yan)....Pages 632-642
Is the Civil Aerospace Industry Ready to Implement Laser Shock Peening into Maintenance Environment? Questions to Be Answered and Minimum Requirements from Aircraft Manufacturer’s Perspective (D. Furfari, U. C. Heckenberger, V. Holzinger, E. Hombergsmeier, J. Vignot, N. Ohrloff)....Pages 643-657
Fatigue Life Prediction at Cold Expanded Fastener Holes with ForceMate Bushings (Yan Bombardier, Gang Li, Guillaume Renaud)....Pages 658-673
Why Should We Encourage Usage of Interference-Fit Fasteners at Airframe Structural Joints? (Carmel Matias, Ekaterina Katsav)....Pages 674-691
Front Matter ....Pages 693-693
Analysis Prediction and Correlation of Fiber Metal Laminate Crack Growth in Semi-Wing Full-Scale Test (Willy R. P. Mendonça, Danielle F. N. R. da Silva)....Pages 695-707
Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning to ¼ Life (C. André Beltempo, Alexandre Beaudoin, Robert Pothier)....Pages 708-722
Changing the Philosophy of Full-Scale-Fatigue-Tests Derived from 50 Years of IABG Experience Towards a Virtual Environment (Gerhard Hilfer, Olaf Tusch, Don Wu, Michael Stodt)....Pages 723-735
Combined Static and Fatigue Tests of the Full-Scale Structure of a Transport Aircraft (K. S. Shcherban, A. A. Surnachev, M. V. Limonin, A. G. Kalish, O. V. Chuvilin)....Pages 736-746
Conception of Modular Test Stand for Fatigue Testing of Aeronautical Structures (Andrzej Leski, Wojciech Wronicz, Piotr Kowalczyk, Michał Szmidt)....Pages 747-761
Full Scale Fatigue Testing for Mitsubishi Regional Jet (Koji Setta, Toshiyasu Fukuoka, Kasumi Nagao, Keisuke Kumagai)....Pages 762-770
Full-Scale Fatigue and Residual Strength Tests of the Composite Wing Box of a Passenger Aircraft (K. S. Scherban, A. Yu. Zakharenkova, V. V. Konovalov, S. V. Kulikov, V. E. Strizhius)....Pages 771-787
Full-Scale Fatigue Testing from a Structural Analysis Perspective (Derk Daverschot, Paul Mattheij, Mathias Renner, Yudi Ardianto, Manuel De Araujo, Kyle Graham)....Pages 788-800
Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests (Risto Laakso, Jussi Kettunen, Juha Lähteenmäki)....Pages 801-815
Progress on the Pathway to a Virtual Fatigue Test (Ben Dixon, Madeleine Burchill, Ben Main, Thierry Stehlin, Raphaël Rigoli)....Pages 816-830
Testing Approach for Over Wing Doors Using Curved Fuselage Panel Testing Technology (Mirko Sachse, Matthias Götze, Silvio Nebel, Sven Berssin, Christian Göpel)....Pages 831-837
Very High-Cycle Fatigue Characteristics of Cross-Ply CFRP Laminates in Transverse Crack Initiation (Atsushi Hosoi, Takuro Suzuki, Kensuke Kosugi, Takeru Atsumi, Yoshinobu Shimamura, Terumasa Tsuda et al.)....Pages 838-846
Application of Optical Fiber-Based Strain Sensing for the Full-Scale Static and Fatigue Tests of Aircraft Structure (U. Ben-Simon, S. Shoham, R. Davidi, N. Goldstein, I. Kressel, M. Tur)....Pages 847-852
Front Matter ....Pages 853-853
Analysis of Adhesive Disbond Occurrences in Rotor Blades of Mi-2 Helicopters (Piotr Synaszko, Krzysztof Dragan, Michał Sałaciński, Mirosław Wrona)....Pages 855-864
Approach to Evaluation of Delamination on the MiG-29’s Vertical Stabilizers Composite Skin (Michał Sałaciński, Piotr Synaszko, Dawid Olesiński, Piotr Samoraj)....Pages 865-873
Evaluation of a PC-9/A Wing Main Spar with Misdrills Using Enhanced Teardown at Resonance (Ben Main, Keith Muller, Michael Konak, Michael Jones, Sudeep Sudhakar, Simon Barter)....Pages 874-888
Holistic Approach for Determining a Helicopter’s Airframe Interval for Depot Induction (Craig L. Brooks, Samuel Benavides)....Pages 889-903
Front Matter ....Pages 905-905
A Guidance to Derive Statistical Data for Asymmetrical Maneuvers on Transport Operation (Juliana Diniz Mattos, Diego Silva Peixoto, Frank Machado)....Pages 907-921
A Machine Learning Approach to Load Tracking and Usage Monitoring for Legacy Fleets (Catherine Cheung, Srishti Sehgal, Julio J. Valdés)....Pages 922-937
Structural Integrity Control Technology Based on Structural Damage Monitoring (Yuting He, Teng Zhang, Binlin Ma, Xianghong Fan, Xu Du)....Pages 938-955
Application of RLC Filters and Analog Circuits for Increasing Information Bandwidth of Channels of Data Acquisition Units (Kamil Kowalczyk, Michal Dziendzikowski, Artur Kurnyta, Patryk Niedbala, Krzysztof Dragan)....Pages 956-965
Effect of Plate Thickness and Paint on Lightning Strike Damage of Aluminum Alloy Sheet (Takao Okada, Hiromitsu Miyaki, Yoshiyasu Hirano)....Pages 966-975
Evaluating the Influence of SHM on Damage Tolerant Aircraft Structures Considering Fatigue (Dominik M. Steinweg, Mirko Hornung)....Pages 976-993
Fatigue Crack Growth Approach for Fleet Monitoring (Olivier Gillet, Bastien Bayart)....Pages 994-1009
Flight Testing of an Ultrasonic Based SHM System (Hideki Soejima, Takuya Nakano, Makoto Yokozuka, Yoji Okabe, Nobuo Takeda, Noriyuki Sawai)....Pages 1010-1021
Lightning Strike Damage of CF/Epoxy Composite Laminates with Conductive Polymer Layers (Tomohiro Yokozeki, Vipin Kumar, Yu Zhou, Takao Okada, Teruya Goto, Tatsuhiro Takahashi)....Pages 1022-1030
Machine Learning Application on Aircraft Fatigue Stress Predictions (Eugene O’Higgins, Kyle Graham, Derk Daverschot, Julien Baris)....Pages 1031-1042
Modernizing the A-10 Loading Spectrum Development Process (Luciano Smith, Mark Thomsen, Devin Butts, Kurt Schrader)....Pages 1043-1053
Nondestructive Evaluation for Damage Tolerance Life Management of Composite Structures (Eric A. Lindgren, John C. Aldrin, David H. Mollenhauer, Mark D. Flores)....Pages 1054-1064
Perspective of Structural Health Monitoring for Military Aviation in Poland (Krzysztof Dragan, Michał Dziendzikowski, Artur Kurnyta, Kamil Kowalczyk)....Pages 1065-1081
Real-Time Stress Concentration Monitoring of Aircraft Structure During Flights Using Optical Fiber Distributed Sensor with High Spatial Resolution (Daichi Wada, Hirotaka Igawa, Masato Tamayama, Tokio Kasai, Hitoshi Arizono, Hideaki Murayama)....Pages 1082-1090
Research on the Scatter of Structural Load-Time History in a Fleet (Tang Li, Yongjun Wang, Hongna Dui, Jiang Dong)....Pages 1091-1100
Study of Composite Impact Dent Visual Detectability and Damage Relaxation Phenomena (Stanislav Dubinskii, Vitaliy Senik, Yuri Feygenbaum)....Pages 1101-1111
Study of Load Spectrum Occurring in the Course of Photogrammetric Missions of the UAV (Miroslaw Rodzewicz, Dominik Glowacki)....Pages 1112-1127
Substitute Models for Structural Components Loads Estimation Based on Flight Parameters and Statistical Inference Methods (Michal Dziendzikowski, Wojciech Zielinski, Piotr Reymer, Marcin Kurdelski, Piotr Synaszko, Witold Klimczyk et al.)....Pages 1128-1137
Technical Justification for an Ultrasonic Inspection Procedure Applied to a Helicopter Component (Muzibur Khan)....Pages 1138-1149
The Research of Aircraft Structure Health Monitoring System Based on Big Data Analysis (Zhinan Zhang, Yu Ning, Xinbo Wang, Bintuan Wang)....Pages 1150-1159
Back Matter ....Pages 1161-1164

Citation preview

Lecture Notes in Mechanical Engineering

Antoni Niepokolczycki Jerzy Komorowski Editors

ICAF 2019 – Structural Integrity in the Age of Additive Manufacturing Proceedings of the 30th Symposium of the International Committee on Aeronautical Fatigue, June 2–7, 2019, Krakow, Poland

Lecture Notes in Mechanical Engineering

Lecture Notes in Mechanical Engineering (LNME) publishes the latest developments in Mechanical Engineering - quickly, informally and with high quality. Original research reported in proceedings and post-proceedings represents the core of LNME. Volumes published in LNME embrace all aspects, subfields and new challenges of mechanical engineering. Topics in the series include: • • • • • • • • • • • • • • • • •

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

To submit a proposal or request further information, please contact the Springer Editor in your country: China: Li Shen at [email protected] India: Dr. Akash Chakraborty at [email protected] Rest of Asia, Australia, New Zealand: Swati Meherishi at [email protected] All other countries: Dr. Leontina Di Cecco at [email protected] To submit a proposal for a monograph, please check our Springer Tracts in Mechanical Engineering at http://www.springer.com/series/11693 or contact [email protected] Indexed by SCOPUS. The books of the series are submitted for indexing to Web of Science.

More information about this series at http://www.springer.com/series/11236

Antoni Niepokolczycki Jerzy Komorowski

•

Editors

ICAF 2019 – Structural Integrity in the Age of Additive Manufacturing Proceedings of the 30th Symposium of the International Committee on Aeronautical Fatigue, June 2–7, 2019, Krakow, Poland

123

Editors Antoni Niepokolczycki Institute of Aviation Warsaw, Poland

Jerzy Komorowski JPWK Aerospace Ottawa, ON, Canada

ISSN 2195-4356 ISSN 2195-4364 (electronic) Lecture Notes in Mechanical Engineering ISBN 978-3-030-21502-6 ISBN 978-3-030-21503-3 (eBook) https://doi.org/10.1007/978-3-030-21503-3 © 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 to ICAF 30th Symposium Proceedings

The International Committee on Aeronautical Fatigue and Structural Integrity (ICAF) Conference and Symposium are coming for the first time to Poland, reflecting successful expansion of ICAF since the early 2000s. Poland has fairly rich aviation history going back to 1918 when the country regained its independence. Over the last 100 years, many airplanes and helicopters have been designed, built and certified in Poland. Today, the industry has been integrated well into the global aerospace enterprise. The industry is supported by several universities and research institutes with the Institute of Aviation as the oldest and largest. ICAF was formed in 1951 in response to the growing concerns regarding fatigue problems in metal aircraft structures. In keeping with the technological changes that have been taking place since then, ICAF encompasses today broader issues of aircraft structural integrity. The Committee has been keeping with the growth of the global aircraft industry, and the membership now consists of 17 countries. The stated aims of ICAF are to exchange information concerning aircraft structural fatigue and to encourage contacts between people active in this field. To this end, a Conference and Symposium are organized every two years for attendance by representatives of industry, universities and institutes, regulatory agencies and operators throughout the world. The two-day Conference consists of reviews of aeronautical fatigue activities presented by the National Delegates of ICAF member nations. It is followed by the three-day Symposium that consists of specialist papers presented by authors with backgrounds and expertise in design, manufacturing, airworthiness regulations, operations and research. The Conference and Symposium theme this year is “Structural Integrity in the Age of Additive Manufacturing”, reflecting one of the new challenges that these manufacturing technologies present to designers and regulators. Each ICAF Symposium starts with a lecture honouring the memory of Dr. Fredrick J. Plantema, the founder of the International Committee on Aeronautical Fatigue. Dr. Plantema took the initiative of forming the International Committee on Aeronautical Fatigue in 1951 with the stated objectives of forming closer cooperation with various institutes carrying out non-classified work. This year Mr. Steven Swift from Australia, a long-term contributor to ICAF, will follow a long v

vi

Preface to ICAF 30th Symposium Proceedings

list of distinguished Plantema lectures. The title of his lecture is “Last Diamond: An Appeal for Holistic Regulatory Leadership”. The lecture is timely following recent events. Steve Swift will draw on his long and diverse regulatory experience, civil and military, to give us a rare insight into the qualities of good regulatory “craftsmanship”, and why this “craft” is not only as important to structural integrity as good engineering, it is also every bit as interesting, challenging and satisfying. Dr. Tommaso Ghidini of European Space Agency offered to present the state of the art of fatigue and damage tolerance for spacecraft and rocket structures. His talk is entitled “Fatigue and Damage Tolerance from Aeronautics to Astronautics”. This is probably first time since 2003 Plantema Lecture that the space community structural integrity concerns will be discussed at ICAF and the talk will no doubt be synergistic and inspiring. A biennial award for young and talented academics in the field of aeronautical fatigue has been established in 2007 by Delft University and the National Aerospace Laboratory (NLR) in the Netherlands. The award is named after Prof. Jaap Schijve, to celebrate his 80th birthday. Recognizing its promotional value, ICAF has offered the opportunity to present the award on the last day of the ICAF Symposium. This year Dr. Adam L. Pilchak from the Materials and Manufacturing Directorate of United States Air Force Research Laboratory has been selected to present the Schijve Lecture. Many thanks to Anna Jędrocha, Małgorzata Małek and their enthusiastic team of Symposium Cracoviense for their tremendous work to help organize the conference from venue selection, webpage design, abstract and paper organization, help with the proceedings to running of the Conference and Symposium. The ICAF 2019 would never have been possible without the financial support of the sponsors. Their contribution and support is gratefully acknowledged. Last but not least, we would like to thank the Institute of Aviation for years of support to ICAF, providing home to the Polish National Delegate and his efforts to organize the 2019 ICAF Conference and Symposium. We would like to recognize the National Delegates and the Technical Committee for initial review of abstracts and papers. The papers included in this volume should be of interest to the broad aerospace community concerned with structural integrity and to the newcomers to the field. A particular expression of gratitude is in order to the authors and their organizations. Without their contribution, the proceedings would have never been published. Kraków (as Cracow is called in Polish) is the second largest and one of the oldest cities in Poland dating back to the seventh century. It was the official capital of Poland until 1596 and has traditionally been one of the leading centres of Polish academic, economic, cultural and artistic life. The Old Town was declared a UNESCO World Heritage Site. We hope that all will enjoy rich history of Cracow and the region. Antoni Niepokolczycki Jerzy Komorowski Co-chair ICAF 2019

International Committee on Aeronautical Fatigue and Structural Integrity

ICAF General Secretary

Marcel J. Bos Netherlands Aerospace Centre (NLR) P.O. Box 153 NL-8300 AD Emmeloord, The Netherlands [email protected]

National Delegates Australia Phil Jackson Airworthiness and Life Evaluation Group Defense Science and Technology 506 Lorimer Street Fishermans Bend VIC 3207, Australia [email protected] Brazil Carlos E. Chaves Senior Structures Engineer Chief Engineering Office Embraer - São José dos Campos, Brazil [email protected] Canada Min Liao Group Leader-Structural Integrity National Research Council Canada

vii

viii

International Committee on Aeronautical Fatigue and Structural Integrity

1200 Montreal Rd, Ottawa, Ontario Canada K1A 0R6 [email protected] China Degang Cui Scientific Consultant Chinese Aeronautical Establishment No.2 Anwai Beiyuan Beijing 100012 P. R. China [email protected] Finland Tomi Viitanen Senior Scientist VTT Technical Research Centre of Finland Lifetime Management, Machine Health P. O. Box 1000, FI-02044 VTT, Finland Tomi.Viitanen@vtt.fi France Thierry Ansart Head of Structures Division DGA Techniques aéronautiques - BP 93123 31131 BALMA, France [email protected] Germany Elke Hombergsmeier Airbus Corporate Technology Office 81663 Munich, Germany [email protected] Israel Yuval Freed Manager, Structural Analysis Department Engineering & Development Group Israel Aerospace Industries Tel Aviv, Israel [email protected] Italy Luigi Lazzeri Universita di Pisa Department of Civil and Industrial Engineering - Aerospace Division Via G. Caruso, 8 56122 Pisa, Italy [email protected]

International Committee on Aeronautical Fatigue and Structural Integrity

Japan Shigeru Machida Aeronautical Technology Directorate Japan Aerospace Exploration Agency (JAXA), Japan [email protected] Poland Antoni Niepokólczycki Materials and Structures Research Center (MSRC) Institute of Aviation Al. Krakowska 110/114 02-256 Warsaw, Poland [email protected] Russia Boris Nesterenko Head of Aircraft Division National Research Center “Zhukovsky Institute” Viktorenko str., 7, Moscow, Russia, 125319 [email protected] Sweden Zlatan Kapidzic Saab Aeronautics Saab AB SE-581 88 Linköping, Sweden [email protected] Switzerland Michel Guillaume ZHAW Zurich University of Applied Sciences Head of the Centre for Aviation Technikumstrasse 9 8400 Winterthur, Switzerland [email protected] The Netherlands René Alderliesten Delft University of Technology, The Netherlands [email protected] UK David Hallam Senior Ageing Platforms Engineer [dstl] Porton Down Salisbury, Wilts, UK SP4 0JQ [email protected]

ix

x

International Committee on Aeronautical Fatigue and Structural Integrity

USA Ravi Chona USAF Senior Scientist - Structural Integrity Air Force Research Laboratory Aerospace Systems Directorate, AFRL/RQ Wright-Patterson Air Force Base, OHIO 45433-7402, USA [email protected]

Contents

Additive Manufacturing Fatigue Characteristic of Linear Friction Welded Ti-6Al-4V Joints . . . . Hiroshi Kuroki, Yukihiro Kondo, Tsukasa Wakabayashi, Kenji Nakamura, Kikuo Takamatsu, Koji Nezaki, and Mitsuyoshi Tsunori Fatigue in Additive Manufactured Aircraft: The Long Way to Make It Fly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ivan Meneghin, Goran Ivetic, Matthias Stiller, Gianluca Molinari, Vjola Ristori, Sara Della Ratta, and François Dumont High Cycle Fatigue and Fatigue Crack Growth Rate in Additive Manufactured Titanium Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Xiang Zhang, Abdul Khadar Syed, Romali Biswal, Filomeno Martina, Jialuo Ding, and Stewart Williams

3

16

31

Strain Controlled Fatigue Testing of Additive Manufactured Titanium Alloy Ti-6Al-4V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rob Plaskitt, Andrew Halfpenny, and Michelle Hill

43

The Optimization and Design of Complicated-Surface Panel Based on Automate Fiber Placement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tieliang Zhang, Liyang Liu, and Hao Cui

56

Advanced Materials and Innovative Structural Concepts A Modeling Approach for the Fatigue Behavior of Laser Drilled Micro Perforated Structural Panels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dort Daandels, Stefan Riekehr, Nikolai Kashaev, Jon Mardaras, Sammy Zein El Dine, and Christian Heck

73

xi

xii

Contents

Assessment of Fatigue Behavior of Advanced Aluminum Alloys Under Complex Variable-Amplitude Loading . . . . . . . . . . . . . . . . . . . . Kevin Stonaker, David Stanley, John G. Bakuckas Jr., Mike Kulak, Po-Yu Chang, Gongyao Wang, and Mark Freisthler

88

Damage Mechanisms and Mechanical Properties of Directly Bonded CFRTP and Aluminium with Nano-Structured Surface . . . . . . . . . . . . . 104 Kristine Munk Jespersen, Hikaru Abe, Hiroki Ota, Kei Saito, Keita Wada, Atsushi Hosoi, and Hiroyuki Kawada Interaction Between High- and Low-Cycle Thermo-Mechanical Fatigue Crack Propagation Around Cooling Hole in a Ni-Based Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 Masakazu Okazaki and Yuuki Yonaguni Ply Curving Termination to Suppress Delamination in Composite Ply Drop-Off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 Shu Minakuchi and Nobuo Takeda Studies on the Fatigue Damage Behavior of Active Jet Engine Chevron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Bingfei Liu, Shangyang Jin, Shaozhe Dong, and Zhenyu Feng Airworthiness and Other Considerations An Ultrafast Crack Growth Lifing Model to Support Digital Twin, Virtual Testing, and Probabilistic Damage Tolerance Applications . . . . 145 Juan Ocampo, Harry Millwater, Nathan Crosby, Beth Gamble, Christopher Hurst, Michael Reyer, Sohrob Mottaghi, and Marv Nuss Analytical and Numerical Investigation of the Effect of Secondary Bending in Hard-Point Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Yuval Freed, Lior Sagi Machnes, and Orel Magidish Demonstration of an Airframe Digital Twin Framework Using a CF-188 Full-Scale Component Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 Guillaume Renaud, Min Liao, and Yan Bombardier Development of Efficient High-Fidelity Solutions for Virtual Fatigue Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 Javier Gomez-Escalonilla, Diego Garijo, Oscar Valencia, and Ismael Rivero Effective Durability and Damage Tolerance Training: New Methods for Modern Learners . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Brandon D. Chapman

Contents

xiii

Fatigue Considerations in the Development and Implementation of Mechanical Joining Processes for Commercial Airplane Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 Robert Jochum, Antonio Rufin, Tanni Sisco, and Frederick Swanstrom Rapid Calculation of Safe Acceleration Values for Aircraft Structures Under Flight Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228 Stephen Dosman and Jonathan Gorman Reliability Approach Applied on Fatigue Safety Factors for Fleet Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 Vincent Montlahuc Research on the Airworthiness Compliance Strategy of Composite Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256 Weiping Li and Xiaoling Zheng Risks of Initial Assumptions in Fatigue and Damage Tolerance of Small Aircraft Development Programs . . . . . . . . . . . . . . . . . . . . . . . . 265 Dejan Romančuk and Juan Ocampo Russian Practice to Provide Safe Operation of Airplane Structures with Long-Term Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279 Boris G. Nesterenko, Grigory I. Nesterenko, Victor V. Konovalov, and Vitaly Ya. Senik Smarter Testing Through Simulation for Efficient Design and Attainment of Regulatory Compliance . . . . . . . . . . . . . . . . . . . . . . 292 Steven A. Chisholm, Jack F. Castro, Brandon D. Chapman, Kazbek Z. Karayev, Andrea J. Gunther, and Mohammed H. Kabir Widespread Fatigue Damage Evaluation for Multiple Elements Based on Probabilistic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308 Fabiano Hernandes Fatigue Crack Growth and Life Prediction Methods A Framework to Implement Probabilistic Fatigue Design of Safe-Life Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 Joshua Hoole, Pia Sartor, Julian Booker, Jonathan Cooper, Xenofon V. Gogouvitis, Amine Ghouali, and R. Kyle Schmidt A Multiaxial Fatigue Damage Model for Isotropic Materials . . . . . . . . . 336 Mauricio V. Donadon, Mariano A. Arbelo, Paulo Rizzi, Carlos V. Montestruque, Lucas Amaro, Saullo Castro, and Marcos Shiino

xiv

Contents

A Specimen to Evaluate Susceptibility of Aluminium Alloys to L-S Crack Deviation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 Erembert Nizery, Jean-Christophe Ehrström, Guillaume Delgrange, and Bruno Wusyk A Numerical Approach to the Disbonding Mechanism of Adhesive Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360 Nicola Zavatta and Enrico Troiani An Engineering Calculation Method of Probability Distribution of Crack Initiation Life for Widespread Fatigue Damage . . . . . . . . . . . 372 Xi Wei, Li Qiang, Shen Peiliang, Yang Gang, Huang Fu, and Zhao Jianjun Assessment of Aircraft Structural Service Life Using Generalized Correction Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 Hongna Dui, Xiaodong Liu, Jiang Dong, and Lixin Zhang Examination of the KAWAI CLD Method for Fatigue Life Prediction of Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399 Yael Buimovich and Dvir Elmalich Fatigue Crack Growth Prediction and Verification of Aircraft Fuselage Panels with Multiple Site Damage . . . . . . . . . . . . . 410 Shaopu Su, Jianghai Liao, Wendong Zhang, and Dengke Dong Fatigue Life Prediction of CFRP Laminate Under Quasi-Random Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423 Vitaly E. Strizhius Fatigue Life Simulation and Experiment of 2024 Aluminum Joints with Multi-Fasteners Interference-Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . 432 Qingyun Zhao, Yunliang Wang, Hong Huang, Sirui Cheng, and Fenglei Liu Influence of Heat Treatment on Near-Threshold Fatigue Crack Growth Behavior of High Strength Aluminum Alloy 7010 . . . . . . . . . . . 444 M. S. Nandana, Bhat K. Udaya, and C. M. Manjunatha Multiaxial Fatigue Behavior of 30HGSA Steel Under Cyclic Tension-Compression and Reversed Torsion . . . . . . . . . . . . . . . . 452 Daniel Dębski, Krzysztof Gołoś, Marek Dębski, and Andrzej Misztela Novel Methods for Measuring the Mode I and Mixed Modes I/II Interlaminar Fracture Toughnesses of Composite . . . . . . . . . . . . . . . . . 461 W. Xu, Z. Z. Guo, Y. Yu, and X. J. Zhang Numerical Investigations on the Three-Dimensional I/II Mixed-Mode Elasto-Plastic Fracture for Through-Thickness Cracked Bodies . . . . . . . 477 Fang-li Wang, Ming-bo Tong, Shu-wei Bai, Nan Jiang, Chong-min She, and Jun-ling Fan

Contents

xv

Probabilistic Reliability Assessment of a Component in the Presence of Internal Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488 Fedor Fomin and Nikolai Kashaev Stress-Intensity Factor Solutions for Tapered Lugs with Oblique Pin Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503 James C. Sobotka, Yi-Der Lee, R. Craig McClung, and Joseph W. Cardinal Summary of Recent Round Robin Life Prediction Efforts for Crack Shape and Residual Stress Effects . . . . . . . . . . . . . . . . . . . . . 518 Alexander V. Litvinov, James A. Harter, and Robert Pilarczyk The Influence of Low and High-Cycle Fatigue on Dislocations Density and Residual Stresses in Inconel 718 . . . . . . . . . . . . . . . . . . . . . . . . . . . 528 Elżbieta Gadalińska, Maciej Malicki, Bartosz Madejski, and Grzegorz Socha Effect of Crack Length and Reference Stress on Variable Amplitude Fatigue Crack Growth Rate . . . . . . . . . . . . . . . 539 E. Amsterdam Weibull or Log-Normal Distribution to Characterize Fatigue Life Scatter – Which Is More Suitable? . . . . . . . . . . . . . . . . . . . . . . . . . 551 Abraham Brot Fatigue Life Enhancement Methods and Repair Solutions Bonded Repairs of Composite Panels Representative of Wing Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565 John G. Bakuckas Jr., Reewanshu Chadha, Paul Swindell, Michael Fleming, John Z. Lin, J. B. Ihn, Nihar Desai, Erick Espinar-Mick, and Mark Freisthler Comparison of Rivet Hole Expansion for Protruding Rivets; Universal and with Compensator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581 Wojciech Wronicz Effect of Alternative Paint Stripping Processes on the Fatigue Performance of Aluminium Alloys - Atmospheric Plasma De-painting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589 Ali Merati, Marko Yanishevsky, and Yan Bombardier Effect of Strengthened Hole on the Fatigue Life of 2024-T3 Aluminum Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 600 Hong Huang, Qingyun Zhao, and Fenglei Liu Fatigue Crack Growth in Pin Loaded Cold-Worked Holes . . . . . . . . . . 606 Luisa Boni, Daniele Fanteria, Domenico Furfari, and Luigi Lazzeri

xvi

Contents

Fatigue Crack Propagation Influenced by Laser Shock Peening Introduced Residual Stress Fields in Aluminium Specimens . . . . . . . . . 617 Sören Keller, Manfred Horstmann, Nikolai Kashaev, and Benjamin Klusemann Influence of Bonded Crack Retarders on Damage Tolerance Performance of Fuselage Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632 Haiying Zhang, Dengke Dong, Yulong Wei, Weifeng Zang, and Wenwei Yan Is the Civil Aerospace Industry Ready to Implement Laser Shock Peening into Maintenance Environment? Questions to Be Answered and Minimum Requirements from Aircraft Manufacturer’s Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643 D. Furfari, U. C. Heckenberger, V. Holzinger, E. Hombergsmeier, J. Vignot, and N. Ohrloff Fatigue Life Prediction at Cold Expanded Fastener Holes with ForceMate Bushings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658 Yan Bombardier, Gang Li, and Guillaume Renaud Why Should We Encourage Usage of Interference-Fit Fasteners at Airframe Structural Joints? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674 Carmel Matias and Ekaterina Katsav Full Scale Fatigue Testing of Aircraft and Aircraft Components Analysis Prediction and Correlation of Fiber Metal Laminate Crack Growth in Semi-Wing Full-Scale Test . . . . . . . . . . . . . . . . . . . . . 695 Willy R. P. Mendonça and Danielle F. N. R. da Silva Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning to ¼ Life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 708 C. André Beltempo, Alexandre Beaudoin, and Robert Pothier Changing the Philosophy of Full-Scale-Fatigue-Tests Derived from 50 Years of IABG Experience Towards a Virtual Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723 Gerhard Hilfer, Olaf Tusch, Don Wu, and Michael Stodt Combined Static and Fatigue Tests of the Full-Scale Structure of a Transport Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 736 K. S. Shcherban, A. A. Surnachev, M. V. Limonin, A. G. Kalish, and O. V. Chuvilin Conception of Modular Test Stand for Fatigue Testing of Aeronautical Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747 Andrzej Leski, Wojciech Wronicz, Piotr Kowalczyk, and Michał Szmidt

Contents

xvii

Full Scale Fatigue Testing for Mitsubishi Regional Jet . . . . . . . . . . . . . . 762 Koji Setta, Toshiyasu Fukuoka, Kasumi Nagao, and Keisuke Kumagai Full-Scale Fatigue and Residual Strength Tests of the Composite Wing Box of a Passenger Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 771 K. S. Scherban, A. Yu. Zakharenkova, V. V. Konovalov, S. V. Kulikov, and V. E. Strizhius Full-Scale Fatigue Testing from a Structural Analysis Perspective . . . . . 788 Derk Daverschot, Paul Mattheij, Mathias Renner, Yudi Ardianto, Manuel De Araujo, and Kyle Graham Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests . . . . . . . . . . . . 801 Risto Laakso, Jussi Kettunen, and Juha Lähteenmäki Progress on the Pathway to a Virtual Fatigue Test . . . . . . . . . . . . . . . . 816 Ben Dixon, Madeleine Burchill, Ben Main, Thierry Stehlin, and Raphaël Rigoli Testing Approach for Over Wing Doors Using Curved Fuselage Panel Testing Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 831 Mirko Sachse, Matthias Götze, Silvio Nebel, Sven Berssin, and Christian Göpel Very High-Cycle Fatigue Characteristics of Cross-Ply CFRP Laminates in Transverse Crack Initiation . . . . . . . . . . . . . . . . . . . . . . . 838 Atsushi Hosoi, Takuro Suzuki, Kensuke Kosugi, Takeru Atsumi, Yoshinobu Shimamura, Terumasa Tsuda, and Hiroyuki Kawada Application of Optical Fiber-Based Strain Sensing for the Full-Scale Static and Fatigue Tests of Aircraft Structure . . . . . . 847 U. Ben-Simon, S. Shoham, R. Davidi, N. Goldstein, I. Kressel, and M. Tur In-Service Experience, Life Extension and Management of Aging Fleets Analysis of Adhesive Disbond Occurrences in Rotor Blades of Mi-2 Helicopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855 Piotr Synaszko, Krzysztof Dragan, Michał Sałaciński, and Mirosław Wrona Approach to Evaluation of Delamination on the MiG-29’s Vertical Stabilizers Composite Skin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865 Michał Sałaciński, Piotr Synaszko, Dawid Olesiński, and Piotr Samoraj Evaluation of a PC-9/A Wing Main Spar with Misdrills Using Enhanced Teardown at Resonance . . . . . . . . . . . . . . . . . . . . . . . . 874 Ben Main, Keith Muller, Michael Konak, Michael Jones, Sudeep Sudhakar, and Simon Barter

xviii

Contents

Holistic Approach for Determining a Helicopter’s Airframe Interval for Depot Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 889 Craig L. Brooks and Samuel Benavides Structural Health and Structural Loads Monitoring A Guidance to Derive Statistical Data for Asymmetrical Maneuvers on Transport Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 907 Juliana Diniz Mattos, Diego Silva Peixoto, and Frank Machado A Machine Learning Approach to Load Tracking and Usage Monitoring for Legacy Fleets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 922 Catherine Cheung, Srishti Sehgal, and Julio J. Valdés Structural Integrity Control Technology Based on Structural Damage Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 938 Yuting He, Teng Zhang, Binlin Ma, Xianghong Fan, and Xu Du Application of RLC Filters and Analog Circuits for Increasing Information Bandwidth of Channels of Data Acquisition Units . . . . . . . 956 Kamil Kowalczyk, Michal Dziendzikowski, Artur Kurnyta, Patryk Niedbala, and Krzysztof Dragan Effect of Plate Thickness and Paint on Lightning Strike Damage of Aluminum Alloy Sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 966 Takao Okada, Hiromitsu Miyaki, and Yoshiyasu Hirano Evaluating the Influence of SHM on Damage Tolerant Aircraft Structures Considering Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 976 Dominik M. Steinweg and Mirko Hornung Fatigue Crack Growth Approach for Fleet Monitoring . . . . . . . . . . . . . 994 Olivier Gillet and Bastien Bayart Flight Testing of an Ultrasonic Based SHM System . . . . . . . . . . . . . . . . 1010 Hideki Soejima, Takuya Nakano, Makoto Yokozuka, Yoji Okabe, Nobuo Takeda, and Noriyuki Sawai Lightning Strike Damage of CF/Epoxy Composite Laminates with Conductive Polymer Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022 Tomohiro Yokozeki, Vipin Kumar, Yu Zhou, Takao Okada, Teruya Goto, and Tatsuhiro Takahashi Machine Learning Application on Aircraft Fatigue Stress Predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1031 Eugene O’Higgins, Kyle Graham, Derk Daverschot, and Julien Baris Modernizing the A-10 Loading Spectrum Development Process . . . . . . . 1043 Luciano Smith, Mark Thomsen, Devin Butts, and Kurt Schrader

Contents

xix

Nondestructive Evaluation for Damage Tolerance Life Management of Composite Structures . . . . . . . . . . . . . . . . . . . . . . 1054 Eric A. Lindgren, John C. Aldrin, David H. Mollenhauer, and Mark D. Flores Perspective of Structural Health Monitoring for Military Aviation in Poland . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1065 Krzysztof Dragan, Michał Dziendzikowski, Artur Kurnyta, and Kamil Kowalczyk Real-Time Stress Concentration Monitoring of Aircraft Structure During Flights Using Optical Fiber Distributed Sensor with High Spatial Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1082 Daichi Wada, Hirotaka Igawa, Masato Tamayama, Tokio Kasai, Hitoshi Arizono, and Hideaki Murayama Research on the Scatter of Structural Load-Time History in a Fleet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1091 Tang Li, Yongjun Wang, Hongna Dui, and Jiang Dong Study of Composite Impact Dent Visual Detectability and Damage Relaxation Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . 1101 Stanislav Dubinskii, Vitaliy Senik, and Yuri Feygenbaum Study of Load Spectrum Occurring in the Course of Photogrammetric Missions of the UAV . . . . . . . . . . . . . . . . . . . . . . . 1112 Miroslaw Rodzewicz and Dominik Glowacki Substitute Models for Structural Components Loads Estimation Based on Flight Parameters and Statistical Inference Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1128 Michal Dziendzikowski, Wojciech Zielinski, Piotr Reymer, Marcin Kurdelski, Piotr Synaszko, Witold Klimczyk, Andrzej Leski, and Krzysztof Dragan Technical Justification for an Ultrasonic Inspection Procedure Applied to a Helicopter Component . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1138 Muzibur Khan The Research of Aircraft Structure Health Monitoring System Based on Big Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1150 Zhinan Zhang, Yu Ning, Xinbo Wang, and Bintuan Wang Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1161

Additive Manufacturing

Fatigue Characteristic of Linear Friction Welded Ti-6Al-4V Joints Hiroshi Kuroki1(&), Yukihiro Kondo1, Tsukasa Wakabayashi2, Kenji Nakamura2, Kikuo Takamatsu3, Koji Nezaki4, and Mitsuyoshi Tsunori5 1

IHI Corporation, Research & Engineering Division, Aero-Engine, Space & Defense Business Area, Tokyo, Japan [email protected] 2 Production Engineering Development Department Manufacturing Division, Aero-Engine, Space & Defense Business Area, Tokyo, Japan 3 Quality System Department Aero-Engine, Space & Defense Business Area, Tokyo, Japan 4 Welding Technology Department, Production Engineering Center, Yokohama, Japan 5 Structural Strength Department, Research Laboratory, Yokohama, Japan

Abstract. A Blisk is the integrated part of rotating blades and a disk, which is recently being adopted for fan and compressor modules of jet engines for the purpose of the weight reduction and the performance improvement. The blisk is generally manufactured by milling from the large forged material. Therefore large amount of material is wasted as cutting chips. Linear friction welding (LFW) is expected to save the wasted material, which is a kind of the solid state joining. The purpose of this study was to investigate the reduction of the fatigue strength due to LFW process. The fatigue test specimens of LFW joints and the base material were made from the same forged material to minimize the effect of scatter. Fatigue tests and analysis of covariance were conducted, and it was made clear that the fatigue strength of as-welded joint was slightly lower than that of base material, and that that of post weld heat treated (PWHTed) LFW joint was the same as that of base material. Since it was thought that the fatigue strength reduction of as-welded LFW joint was due to the effect of residual stress, the residual stress was measured before and after PWHT by Center Hole Drilling method. It was confirmed that the residual stress of as-welded LFW joint remained approximately 3 mm from the weld line where the fracture occurred in the fatigue tests, and that the residual stress of PWHTed LFW joint was reduced drastically. As a conclusion, it was made clear that the fatigue strength of PWHTed LFW joint was equivalent to that of base material with the statistical data. Keywords: Linear friction welding
Titanium alloy Post weld heat treatment
Analysis of covariance


Fatigue


© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 3–15, 2020. https://doi.org/10.1007/978-3-030-21503-3_1

4

H. Kuroki et al.

1 Introduction Blisk is the component that consists of “blade” part and “disk” part integrally. This structure enables to eliminate the blade roots and disk slots which are used for installation, and to reduce the leak air between blade platforms (see Fig. 1). In recent aircraft jet engines, the applications of blisks are increasing so as to achieve weight reduction and performance improvement.

Blade and disk

Blisk

Fig. 1. Comparison between bladed disk and blisk (Turner et al. [1])

LFW is a solid state joining process using frictional heat generated by a linear reciprocating motion between the weld surfaces of work pieces. Because of small heat input and high accuracy of blade positioning, LFW is known as the suitable manufacturing technique of blisks Figure 2 shows a schematic view of LFW process. One of the materials is slid at high speed parallel to the weld interface while the other material is pressed in a direction orthogonal to the weld interface. Vairis and Frost [2, 3] reported that the welding process of LFW can be divided into four phases: the initial phase (Phase I), in which the stationary work piece starts contacting the sliding work piece and frictional heat is generated. The asperities soften and locally deform so as to broaden true area of contact between work pieces; the transition phase (Phase II), in which the materials start to form flash at the interface of the two pieces (burn-off begins); the equilibrium phase (Phase III), in which the materials continue to soften and the burn-off rate becomes constant; and the deceleration phase (Phase IV), in which the sliding motion ends and the materials cool down and become fixed.

Fatigue Characteristic of Linear Friction Welded Ti-6Al-4V Joints

5

Fig. 2. Schematic view of LFW process

Using this technology, the blisk is fabricated from a disk part and blade parts as shown in Fig. 3. Blades parts are slid in reciprocal motion, and the disk is pressed against weld surface. Flash and the blocks to hold blade parts need to be milled off by machining after welding process. Forge Pressure Reciprocation

LFW Disk Part

+

Blade Part Blisk

Fig. 3. Schematic view of blisk fabrication using LFW

Because a blisk is a rotating part, it is very important to assure the integrity of welding joint. LFW is the joining method in which the welding defects seldom remain as far as the sufficient burn off distance is maintained. Static strength of the welding joint is also equivalent to that of base material. However the effect of welding on the fatigue capability needs to be carefully studied because of the existence of residual stress.

6

H. Kuroki et al.

Kuroki et al. [4] conducted HCF test using airfoil parts fabricated by LFW. Since the weld location was carefully selected to avoid the peak stress location, HCF test of airfoil parts were successfully completed. But they pointed out the necessity of careful evaluation of fatigue characteristics due to the existence of the many influence parameters. According to Wen et al. [5], the fatigue crack initiated at base material, and the fatigue strength of LFW joints was within the experimental scatter of base material. Fillipo et al. [6, 7] mentioned that the fatigue strength of as-welded LFW joints exceeded the minimum value of design requirements, and that the fatigue strength of post weld heat treated (PWHTed) LFW joints exceeded that of base material. However they didn’t address the effect of the scatter of the material strength. According to Stinville et al. [8], LFW reduced LCF properties even if the fracture occurred at base material. In this way, the fatigue characteristics of LFW joints were reported differently. The objective of this work is to evaluate the fatigue characteristics of the linear friction welded Ti-6Al-4V joints and to make it clear the effect of PWHT with the statistical method.

2 Test and Analysis Procedures 2.1

Linear Friction Welding

The material used in this study was Ti-6Al-4V, which is widely used for fan and compressor blisks. The welding work pieces were milled from the forged material, which was actually used for the production of the jet engine rotor parts. In order to reduce the effect of the scatter, all the test pieces were milled from the identical forged material. The composition of the material is shown in Table 1. Microstructure of the base material is shown in Fig. 4, which is the typical bimodal structure composed of alpha grains and beta grains.

Table 1. Composition of material Fe V Al C O N B Y Element wt% Top 0.158 4.04 6.42 0.020 0.18 0.001 0.0006 F0 (1, 19, a = 0.05) = 4.32. That is to say there is significance difference between 2 SN curves.

12

H. Kuroki et al.

Based on these analyses, it became clear that SN curve of base material was parallel to that of as-welded LFW joint, but fatigue strength of as-welded LFW joints was lower than that of base material. In the same way, analysis of covariance between base material and PWHTed LFW joint was conducted. At first, The hypothesis of the parallelism is not rejected because VNP/VEʹ = 1.51 < F0 (1, 30, a = 0.05) = 4.17. Next, the hypothesis that the slope is 0 is rejected because VR/VE = 198.69 > F0 (1, 31, a = 0.05) = 4.16. Finally the hypothesis that the constant terms are equal is rejected because VA/VE = 1.31 < F0 (1, 31, a = 0.05) = 4.16. That is to say, there is no significance difference between base material and LFW joints with stress relief. Based on these results, it became clear that the fatigue strength of LFW joints was recovered by applying PWHT. 3.3

Residual Stress Measurement Results

The reason why the difference between as-welded LFW joints and PWHTed LFW joints occurs was thought to be the effect of residual stress. The fracture locations of the test specimens were measured as shown in Fig. 9.

500

Applied Alternative Stress

Base Material as-welded LFW joint PWHTed LFW jointf

Oscillation Side

Forge Side

0

-10

-8

-6

-4

-2

0

2

4

6

Distance between Weld Line and Fracture Surface [mm] Fig. 9. Fracture locations of fatigue test specimens

8

10

Fatigue Characteristic of Linear Friction Welded Ti-6Al-4V Joints

13

Fracture locations of as-welded LFW joints were within 3.25 mm from the weld line. But those of PWHTed LFW joints scatter widely, and the scatter of fracture locations was similar to that of base material. Fracture locations scattered equally from oscillation side to forge side. Residual stress was measured for 2 weld work pieces (specimen A and B). Residual stress after PWHT was measured on specimen B. The results are shown in Fig. 10. The residual stress due to LFW remains from approximately 3 mm from the weld line. Since the fracture locations of as-welded LFW joints were within 3.25 mm from the weld line, it was thought that the fatigue strength reduction is due to the effect of the residual stress. Wen et al. [5] also reported that the fracture locations of as-welded joint was approximately 3 mm from weld line. And also residual stress measurement shows that residual stress of PWHTed LFW joint was decreased drastically. It is understandable that fatigue strength of PWHTed LFW joint is equivalent to that of base material.

600

Oscillation Direction / As Welded (Specimen A) Forge Direction / As Welded (Specimen A) Oscillation Direction / As Welded (Specimen B) Forge Direction / As Welded (Specimen B) Oscillation Direction / PWHT (Specimen B) Forge Direction / PWHT (Specimen B)

Residual Stress [MPa]

500 400 300 200 100 0 -100 -200 0

0.5

1

1.5

2

2.5

3

3.5

Distance from Weld Line [mm] Fig. 10. Result of residual stress measurement

4 Conclusion In order to investigate the fatigue property of Ti-6Al-4V LFW joints, fatigue tests and analysis of covariance were conducted. The fatigue test specimens of base material and LFW joints were made from the identical forged material to minimize the effect of scatter of material strength. It was made clear that the fatigue strength of as-welded LFW joints was slightly reduced from base material. Based on the residual stress measurement it was found that residual stress remains at least 3 mm from the weld line.

14

H. Kuroki et al.

Since the fracture locations of fatigue test specimens of as-welded LFW joints were within 3.25 mm from the weld line, it was concluded that the fatigue strength reduction was due to the effect of residual stress. On the other hand, it was made clear that the fatigue strength of PWHTed LFW joints was equivalent to that of base material based on analysis of covariance. It was also confirmed that residual stress after PWHT was drastically reduced.

Appendix Formula of Analysis of Covariance Total Variation

STy ¼

Ni a X X

P P a Ni i¼1

Pa

y2ij 

i¼1

i¼1 j¼1

STx ¼

Ni a X X

STyx ¼

Ni

P P a Ni i¼1

Pa

x2ij 

P

a i¼1

yij xij 

2

j¼1 xij

i¼1

i¼1 j¼1 Ni a X X

2

j¼1 yij

Ni

P P  a Ni y x ij ij i¼1 j¼1 j¼1 Pa N i¼1 i

P Ni

i¼1 j¼1

Inter-level variation P SAy ¼

N1 j¼1 x1j

2

N1 j¼1 y1j

P 

P Na

N1 j¼1 x1j

N1 a P Ni i¼1

j¼1 yij

i¼1

Pa

 P

þ


þ P P a i¼1

Ni

Ni

P P a Ni

Ni j¼1 xij

Na j¼1 yaj



Ni

P Na

2

j¼1 xij

i¼1



i¼1

Pa

2

Na

Pa

i¼1

2

j¼1 yij

i¼1

j¼1 xaj

þ


þ

P P a Ni 

Na P Na

P

2

j¼1 yaj

þ


þ

N1 P

SAyx ¼

2

N1 P

SAx ¼

N1 j¼1 y1j

Na j¼1 xaj



Fatigue Characteristic of Linear Friction Welded Ti-6Al-4V Joints

15

Variation within level SEy ¼ STy  SAy SEx ¼ STx  SAx SEyx ¼ STyx  SAyx

SBx

 P  P P 2 N1 N1 N1 N1 j¼1 y1j x1j  j¼1 y1j j¼1 x1j   ¼ þ


2  P N1 2   P N1 N1 N1 x x  j¼1 1j j¼1 1j  P  P P 2 Na Na Na Na y x y x  aj aj aj aj j¼1 j¼1 j¼1   þ 2  P Na 2   P Na Na Na j¼1 xaj  j¼1 xaj


SNP

SEyx ¼ SBx  SEx

2

SE0 ¼ SEy  SBx SNP is variation of non-parallelism. SEʹ is residual variation.

References 1. Turner, R., Gebelin, J.-C., Ward, R.M., Reed, R.C.: Linear friction welding of Ti-6Al-4V: modelling and validation. Acta Mater. 59(2011), 3792–3803 (2011) 2. Vairis, A., Frost, M.: High frequency linear friction welding of a titanium alloy. Wear 217, 117–131 (1998) 3. Vairis, A., Frost, M.: Modeling the linear friction welding of titanium blocks. Mater. Sci. Eng., A 292, 8–17 (2000) 4. Kuroki, H., Nezaki, K., Wakabayashi, T., Nakamura, K.: Application of linear friction welding technique to aircraft engine parts. IHI Eng. Rev. 2014(47), 40–43 (2014) 5. Wen, G.D., Ma, T.J., Li, W.Y., Guo, H.Z., Chen, D.L.: Cyclic deformation behavior of linear friction welded Ti6Al4V joints. Mater. Sci. Eng. A 597(2014), 408–414 (2014) 6. Flipo, B., Beamish, K., Humphreys, B.: Wood M, Linear friction welding of Ti-6Al-4V for aerostructure application. In: Proceedings of 10th International Conference on Trends in Welding Research. Tokyo, Japan, 11–14 Oct. 2016 (2016) 7. Flipo, B.: Ti-6Al-4V LFW for aero structures. In: 4th Linear Friction Weld Symposium. Granta Park, Cambridge, UK (2017) 8. Stinville, J.C., Brdier, F., Ponsen, D., Wanjara, P., Bocher, P.: High and low cycle fatigue behavior of linear friction welded Ti-6Al-4V. Int. J Fatigue 70, 278–288 (2015)

Fatigue in Additive Manufactured Aircraft: The Long Way to Make It Fly Ivan Meneghin(&), Goran Ivetic, Matthias Stiller, Gianluca Molinari, Vjola Ristori, Sara Della Ratta, and François Dumont Premium Aerotec GmbH, Haunstetter Strasse 225, 86179 Augsburg, Germany [email protected]

Abstract. Manufacturing of metallic products is currently facing a historic revolution driven by a group of innovative technologies clustered under the name of Additive Manufacturing (AM) or, more colloquially, 3D-printing. As opposed to the conventional subtractive manufacturing methodologies, AM is able to create net-shaped products by means of the addition, layer by layer, of the needed material. The biggest benefit of the AM in respect to the conventional manufacturing technologies is the substantial freedom the technology gives to the product designers in frame of the product development. Due to the relaxed geometrical constraints, complex parts characterized by optimized shapes can be realized. This can be translated in development of lighter and more cost-efficient parts and assemblies hitting two of the most important targets of the aeronautical industry. Furthermore, as a consequence of the significant weight benefits, the application of topologically optimized AM aircraft structures can provide with an important contribution in reducing the aircrafts fuel consumption and the relative emissions of CO2 & NOX gases, in line with the environmental targets of current aviation research programs. Despite the high investments, the AM still finds a limited applicability for the manufacturing of aircraft components. This is also due the challenges relative to the fulfillment of the Fatigue and Damage Tolerance (F&DT) requirements, particularly sensitive when they have to cope with innovative materials and technologies. This paper presents the state of art of the AM technology in one of the leading companies specialized in manufacturing of large aircraft components and offers a potential way forward toward the wider exploitation of AM technology potentials. Keywords: Additive Manufacturing
Fatigue
Damage tolerance Primary and secondary structures
Powder bed fusion




1 Introduction Additive Manufacturing (AM) is a common name for a group of technologies, able to create net-shaped products by means of the addition, layer by layer, of the needed material (ASTM 2012). Opposed to classical subtractive technologies, they give © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 16–30, 2020. https://doi.org/10.1007/978-3-030-21503-3_2

Fatigue in Additive Manufactured Aircraft: The Long Way

17

substantial freedom to the product designers in frame of the product development, since the constraints relative to e.g. tool accessibility are not present anymore. Due to the relaxed geometrical constraints given by this technology, it is possible to produce complex parts characterized by optimized shapes (Herzog et al. 2016). This capacity allows in turn the development of lighter and more cost-efficient parts and assemblies, hitting two important targets of the aeronautical industry (the third being of course the airworthiness and safety of flight). Furthermore, given the potential of having significant weight benefits, the application of topologically optimized AM aircraft structures could contribute to reduce the fuel consumption and the relative emissions of CO2 & NOX gases, in line with the environmental targets of current aviation research programs (ACARE 2020 and FlightPath 2050). In Fig. 1, a secondary structure (bracket) manufactured by means of conventional technologies, milling and riveting, is shown together with its AM evolution. In this case thanks to the topology optimization of the AM bracket a weight reduction equal to half of the conventional bracket was achieved.

Milling Part Weight 0,4kg

AM Part Weight

0,2kg

Fig. 1. Conventional vs. AM manufactured bracket structure (© Premium Aerotec GmbH)

Despite the high investments, the AM still finds a limited applicability for the manufacturing of aircraft structures. In Thomas and Gilbert (2014) is reported that in 2014 only 0.02% of goods produced by the aerospace industry were manufactured by means of AM technologies. One of the reasons of the limited applicability of the AM in manufacturing aircraft components is the challenge relative to the fulfillment of the Fatigue and Damage Tolerance (F&DT) requirements. F&DT has always been considered one of the most critical design criteria for the aeronautical structures and the results reported in Table 1 can only confirm its criticality: despite the employment of well-known conventional manufacturing

18

I. Meneghin et al.

technologies, fatigue remains the first reason of mechanical failure of aircraft metallic components (Gorelik 2017). Table 1. Failure mechanisms in aircraft components (Gorelik 2017) Failure mechanism Fatigue Corrosion Overload Stress corrosion cracking Wear/abrasion/erosion High temperature corrosion

% Failure 55 16 14 7 6 2

The main reason is the difficulty of predicting the long-term behavior of complex structures, such as an aircraft, during its entire service life due to the number and complexity of interdependent phenomena to be taken into account. In this context, it is clear that the challenges of the F&DT requirements are exacerbated whenever they have to cope with innovative materials and technologies, such as AM parts, by the current lack of experience in serial production, tests and inservice application: • AM is a complex process constituted of millions of localized melting and solidification cycles that produces metallic components with non-homogeneous macroand micro- structural characteristics and thus inconsistent mechanical proprieties. The incomplete understanding of the relation between the process parameters, the induced thermal cycles, the obtained structure and the mechanical performances of the produced AM parts is currently the main obstacle toward the standardization of the AM process (Yadollahi and Shamsaei 2017). A robust and standard manufacturing process capable of producing components with well-known and repeatable mechanical proprieties is a must in order to provide reliable F&DT assessments. • Yadollahi and Shamsaei (2017) argues that the representativeness of the standard laboratory AM specimens, in respect to final AM parts, may not be ensured, mostly due to differences in their geometry and size. These differences influence the thermal histories experienced during fabrication, and consequently, the macro- and microstructural features of the produces AM specimens and of the final AM parts. As a consequence, not only there is a lack of in-service data for AM parts to support their F&DT justification but also the standards for mechanical testing methods may need to be revised in order to provide the requested experimental evidences. • Inspections plans performed with reliable Non Destructive Inspection (NDI) techniques are considered mandatory to guarantee the continued airworthiness of aeronautical structures. Gorelik (2017) discusses that the combination of the nonhomogeneous macro- and micro- structural characteristics of the AM parts with the

Fatigue in Additive Manufactured Aircraft: The Long Way

19

complex shapes of topologically optimized AM parts (see AM bracket shown in Fig. 1) may challenge the conventional NDI capabilities and lead to their redefinition on the basis of AM needs.

2 Additive Manufacturing of Metallic Aircraft Components Several techniques have an industrial relevance for the manufacturing of metal AM components which differs each other on the basis of the feed stock form (e.g. powder or wire), feeding system (powder bed or blown powder) and heat source (e.g. laser beam or electron beam). The most relevant techniques are powder bed fusion (PBF) and direct energy deposition (DED) both with wire or blown powder (Frazier 2014). On the process side, the PBF techniques are the most used for the manufacturing of aeronautical metallic components. On the materials side, Aluminum is by far the most common metal used in aircrafts. Titanium however, is becoming very popular especially for CFRP-metallic hybrid structures due to its galvanic compatibility with CFRP parts and its high fatigue and static strength. Because the high specific strength of Ti6Al-4V AM can maximize the weight benefits of topologically optimized structures. Therefore, this paper is focused on the AM of Titanium Ti-6Al-4V parts through PBF techniques. The PBF is a cyclic process where the following steps take place (see Fig. 2): – Step 1: deposition of a thin layer of metallic powder on a build plate – Step 2: selective melting of the metallic powder creating the cross-section of the part – Step 3: lowering of the build plate in order to allow the deposition of a new layer of metallic powder. The component is thus built layer after layer (Herzog et al. 2016). The powder can be melted using either a laser beam (Selective Laser Melting, SLM) or an electron beam (Selective Electron Beam Melting, SEBM). To avoid oxidation phenomena during the melting phase, the AM process takes place in an inert atmosphere, i.e. in a chamber flooded with argon or in vacuum (Fig. 3). Due to the peculiarity of the process, a number of internal and external defects characterize the as-built AM parts (intended as just after printing). Typical internal defects associated to the AM process are: lack of fusion, gas porosity and voids. Moreover, as a result of the repetitive and localized melting-solidification cycles, high internal residual stresses, even higher than the material yield stress, may develop inside the bulk material impacting significantly the fatigue performances. Typical external defects associated to AM process are: layer defects, swelling and bobbling (Mardaras et al. 2017). It is also well known that the surface quality of the as-built AM could be very coarse and morphologically heterogeneous (Bagehorn et al. 2017), similarly to sand castings (Qian et al. 2016).

20

I. Meneghin et al.

Fig. 2. PBF process (Herzog et al. 2016)

Fig. 3. Selective Laser Melting of the Titanium component (© Premium Aerotec GmbH)

Fatigue in Additive Manufactured Aircraft: The Long Way

21

In particular, partially molten powder particles stick on the surface, as shown in Fig. 4, increasing the roughness (Ra) up to 20 µm, being 3.2 µm the minimum standard for aerospace applications.

Fig. 4. Surface status of an AM part in as-built condition (© Premium Aerotec GmbH)

For all these reasons, additional processes are required to maximize the mechanical performances of the as-built AM parts (Uhlmann et al. 2015) (see Fig. 6). Annealing is widely used as post-AM process because it is very effective in relieving the intense residual stresses introduced into the part (Mukherjee et al. 2017). Annealing, to relieve the residual stresses, is carried out at about 700 °C for 1 h (Uhlmann et al. 2015).

Pre - HIP

Post-HIP

Fig. 5. Effect of HIP on internal defects (© Premium Aerotec GmbH)

22

I. Meneghin et al.

Hot Isostatic Pressing (HIP) is performed to reduce the dimensions and numbers of defects present into the bulk material of the AM parts (see Fig. 4) and to produce a more homogeneous structure under the combined effect of temperature and pressure. HIP is typically carried out over 900 °C at 100 MPa for minimum 2 h in an inert atmosphere (Uhlmann et al. 2015) (Fig. 5). A final surface finishing process is performed to bring the surface quality to acceptable levels. In order to guarantee the maximum quality standards of the produced AM parts, an intensive inspection plan is currently required to support the AM process. Detailed visual inspection (DVI) represents the first check of the surface quality of the produced parts. Dye penetrant inspection is performed in order to detect external defects as cracks and voids on the surface of the AM part. X-ray and Computer Tomography (CT) are performed in order to detect internal defects as cracks, voids and inclusions.

Fig. 6. AM process with integrate inspection plan

As a result of the PBF process described above, combined with a meticulous inspection plan of the produced AM metallic parts, the final product has the following characteristics:

Fatigue in Additive Manufactured Aircraft: The Long Way

23

• no significant surface defects • controlled roughness • internal defects similar to the defects of conventionally produced parts. The following Figure illustrates some of the parts produced at Premium Aerotec GmbH (Fig. 7).

Fig. 7. Titanium AM structures produced by PBF (© Premium Aerotec GmbH)

In Fig. 8 two cross-sections, extracted in longitudinal and transverse directions, are shown. The structure is well homogenous in both the directions and without evident macro-defects. A uniform and fine microstructure is also visible in Fig. 9.

Fig. 8. Macrostructure, left: (10x) transverse direction; right: (10x) longitu-dinal direction (© Premium Aerotec GmbH)

24

I. Meneghin et al.

Fig. 9. Microstructure, left: (200x); right: (200x) (© Premium Aerotec GmbH)

In Fig. 10 the fatigue life of machined (mechanically milled) HIP-annealed-AM SLM specimens is plotted (Ra  1.6 µm - EN6072 Mini-T-Type, Kt 2.3); the AM specimens performed better than the reference material (Ti-6Al-4V forged, machined and annealed, plotted as a dotted line) that represents the minimum requirement for the part qualification. A similar comparison for the fatigue crack propagation rate is presented in Fig. 11 (ASTM E647 CTW40 B6), and again the AM specimens exceeded the minimum requirement. The fatigue and crack propagation performances were also

Fig. 10. Fatigue behavior of Ti-6Al-4V AM specimens (© Premium Aerotec GmbH)

Fatigue in Additive Manufactured Aircraft: The Long Way

25

compared to the design values (continuous line) used during the aircraft justification for certification phase. The behavior of the AM specimens are in line with the expected behavior of the baseline (forged, machined and annealed) material used during the standard manufacturing phase, confirming the general good quality and robustness of the employed AM process.

Fig. 11. Crack propagation rate of Ti-6Al-4V AM specimens (© Premium Aerotec GmbH)

3 Limits of the AM Process and Future Developments The AM process as described in Fig. 6 permits to produce metallic components characterized by mechanical performance which are comparable with conventional counterparts. Nevertheless the aforementioned AM process, although robust, can hardly make a business case especially when it comes to aircraft primary structures, which require very high F&DT performances. In fact, to achieve these performances the process requires post-AM heat treatments (e.g. HIP) followed by mechanical milling (to obtain a surface roughness below 3.2 µm). These post AM treatments are de-facto limiting the AM applicability to secondary structure only, where F&DT

26

I. Meneghin et al.

requirement are relaxed. To overcome these limitations the following considerations need to be addressed. 3.1

Techniques of Surface Finishing

As mentioned in previous text, a surface finishing process is necessary after the AM due to the low and inconsistent quality of the surfaces obtained with the AM in order to bring the surface quality to an acceptable level. At the moment, the mechanical milling is the most viable solution for the surface finishing of AM metallic aircraft components and for this reason the only one considered for the manufacturing of primary structures (Mardaras et al. 2017). As a consequence, given the milling tool accessibility limitation, it becomes impossible to exploit the benefits AM technology to manufacture topologically optimized fatigue critical structures. For secondary structures the surface quality requirements are relaxed and they can be installed in “as built” conditions. Nevertheless, the design of secondary structure is anyway penalized by severe Knock Down Factors (KDF) used to take into account the general bad quality of the surface, which brings to a wide scatter of the fatigue performances. In order to overcome this limitations, efforts are currently in place to: • Develop tool-less surface finishing techniques that would permit to improve the quality of the AM metallic part despite of the complexity of its surface (Bagehorn et al. 2017). • Develop a consolidated relationship between fatigue performances and surface finishing techniques. 3.2

Damage Tolerance Approach

The conventional damage tolerance approach is a deterministic one which assumes the presence of a determined initial rogue flaw, i.e. a 1.27 mm crack at the most stressed location (Mardaras et al. 2017). When exploiting the advantages of AM producing topologically optimized, typically constituted by slender elements (bionic design) as shown in Fig. 13, this kind of approach could be too penalizing. One method that could be used to overcome this limitation is the use of a zone based (probabilistic) approach, already used in the past for casted parts (Gorelik 2017). Instead of considering a standard initial defect, in this approach the component is divided in several zones. For each of these zones (Fig. 12), different attributes are defined (material properties, anomalies distribution, NDI requirements, etc.) and associated with fatigue and crack propagation performances.

Fatigue in Additive Manufactured Aircraft: The Long Way

27

Fig. 12. Schematics of zoning approach

In order to apply this type of approach to AM though, a sound link between the process parameters and the performance of the produced parts needs to be established. This link can only be completely defined by correlating the process parameters to the deriving material macro- and microstructure and the relative zone dependent material properties. After the definition of this link coupled with an extensive investigation relative to the defect population and its spatial distribution, it will be possible to tailor the damage tolerance approach for topologically optimized parts.

4 AM Parts with Intrinsically Fail Safe Design Another opportunity that could be exploited in the future once the problems with topological optimization of AM parts are overcome (e.g. surface quality), is the possibility to design and build intrinsically file-safe structures. Structural fail-safety is the capability of a structure to withstand the failure of a load carrying member for an amount of time sufficient to detect, and repair, the damage during the normal maintenance inspection plan (Swift 2003). Usually fail-safety is obtained through redundancy, designing structures with many separated load paths that can bear the structural overload resulting from the failure of one of them. Standard integral structures, such as skin panel with welded or machined stringers, are however not intrinsically file-safe because the cracks can propagate continuously through each load carrying member, reducing the damage tolerance capability of the whole integral structure (Swift 2003). For this reason, the usage of standard metallic integral structure is limited to areas where the fatigue is not a concern, e.g. structures mainly subjected to compression. The capability of AM to produce extremely complex structures, together with a careful topological optimization, could overcome the current limitations of integral structures,

28

I. Meneghin et al.

leading to the possibility to design and build intrinsically fail-safe structures. It is theoretically possible to build a part with several load paths reducing at the same time the weight, an example is shown in Fig. 13.

Fig. 13. 3D printed multiple load path component (© Premium Aerotec GmbH)

Moreover, the complete freedom given by AM to deploy different damage containment features on the same part could slow down efficiently the crack propagation, or even point the cracks from the critical areas towards unloaded areas, as sometime happens in integral structures where the cracks could kink and turn resulting in a reduced crack propagation rate (Petitt 2000; Swift 2003). Obviously this scenario is at the moment purely hypothetical; however it opens to new and exciting research opportunities for the aerospace industry.

5 Conclusions At the moment AM can only be used as an alternative technology for the production of the primary structures with a similar design of the parts currently manufactured with conventional technologies. As a consequence, it is very difficult to obtain a business case for the AM primary structures.

Fatigue in Additive Manufactured Aircraft: The Long Way

29

On the other hand, secondary AM structures can be used in “as-built” conditions but with the application of extremely penalizing KDFs, which have a detrimental impact on the weight optimization. For these reasons AM is not yet able to exploit all its potentiality, especially when it comes to primary structures of an aircraft, de-facto limiting its applicability to “oversized” secondary structures only. Nevertheless, the AM technology still withholds a great potential that can be exploited as long as: • a well-established correlation among process parameters, micro- and macrostructure, surface roughness and the resulting mechanical properties of the AM parts, is defined; • DT analysis methods more suitable for AM parts (probabilistic in place of deterministic approaches) are developed; • reliable NDI techniques for geometrically complex AM parts are qualified. It is still a long way to make it fly but the potential gains of topological optimized AM parts is worth the trouble.

References ASTM (American Society for Testing and Materials): Standard terminology for additive manufacturing technologies. Standard F2792-12a (2012) Bagehorn, S., Wehr, J., Maier, H.J.: Application of mechanical surface finishing processes for roughness reduction and fatigue improvement of additively manufactured. Int. J. Fatigue 102 (2017), 135–142 (2017) Ti-6Al-4V Parts European Commission: European Aeronautics: A Vision for 2020. Luxemburg, European Communities (2001) European Commission: Flightpath 2050. EU, Luxemburg (2011) Frazier, W.E.: Metal additive manufacturing: a review. J. Mater. Eng. Perform. 23(6), 1917–1928 (2014) Gorelik, M.: Additive manufacturing in the context of structural integrity. Int. J. Fatigue 94 (2017), 168–177 (2017) Herzog, D., Seyda, V., Wycisk, E., Emmelmann, C.: Additive manufacturing of metals. Acta Mater. 117(2016), 371–392 (2016) Kahlin, M., Ansell, H., Moverare, J.J.: Fatigue behaviour of notched additive manufactured Ti6Al4V with as-built surfaces. Int. J. Fatigue 101(2017), 51–60 (2017) Murr, L.E., Gaytan, S.M., Ceylan, A., Martinez, E., Martinez, J.L., Hernandez, D.H., Machado, B.I., Ramirez, D.A., Medina, F., Collins, S., Wicker, R.B.: Characterization of titanium aluminide alloy components fabricated by additive manufacturing using electron beam melting. Acta Mater. 58(2010), 1887–1894 (2010) Mardaras, J., Emile, P., Santgerma, A.: Airbus approach for F&DT stress justification of Additive Manufacturing parts. Procedia Struct. Integrity 7(2017), 109–115 (2017) Mukherjee, T., Zhang, W., DebRoy, T.: An improved prediction of residual stresses and distortion in additive Manufacturing. Comput. Mater. Sci. 126(2017), 360–372 (2017) Uhlmann, E., Kersting, R., Borsoi, Klein T., Cruz, M.F., Borille, A.V.: Additive manufacturing of titanium alloy for aircraft components. Procedia CIRP 35(2015), 55–60 (2015)

30

I. Meneghin et al.

Thomas, D.S., Gilbert, S.W.: Costs and Cost Effectiveness of Additive Manufacturing. National Institute of Standards and Technology (NIST), NIST Special Publication 1176, December 2014 Petitt, R.G.: Crack Turning in Integrally Stiffened Aircraft Structures. NASA Langley Research Center, August 2000 Swift, T.: Fail-safe design requirements and features, regulatory requirements. In: AIAA/ ICAS International Air and Space Symposium and Exposition. Dayton, OH (2003) Yadollahi, A., Shamsaei, N.: Additive manufacturing of fatigue resistant materials: Challenges and opportunities. Int. J. Fatigue 98(2017), 14–31 (2017) Qian, M., Xu, W., Brandt, M., Tang, H.P.: Additive manufacturing and postprocessing of Ti-6Al4V for superior mechanical properties. Mater. Res. Soc. Bull. 41(10), 775–784 (2016)

High Cycle Fatigue and Fatigue Crack Growth Rate in Additive Manufactured Titanium Alloys Xiang Zhang1(&), Abdul Khadar Syed1, Romali Biswal1, Filomeno Martina2, Jialuo Ding2, and Stewart Williams2 1

2

Research Centre for Manufacturing and Materials Engineering, Coventry University, Coventry CV1 5FB, UK [email protected] Welding Engineering and Laser Processing Centre, Cranfield University, Cranfield MK43 0AL, UK

Abstract. The Wire + Arc Additive Manufacture (WAAM) process can produce large metal parts in the metre scale, at much higher deposition rate and more efficient material usage compared to the powder bed fusion additive manufacturing (AM) processes. WAAM process also offers lead time reduction and much lower buy-to-fly ratio compared to traditional process methods, e.g. forgings. Research is much needed in the areas of fatigue and fracture performance for qualification and certification of additive manufactured aircraft components. In this study, specimens made of WAAM Ti-6Al-4V alloy were tested and analysed focusing on two key areas of structural integrity and durability: (1) High cycle fatigue and effect of defects: crack initiation at porosity defects was investigated via fatigue and interrupted fatigue-tomography testing performed on specimens with porosity defects purposely embedded in the specimen gauge section. Key findings are as follows. Presence of porosity did not affect the tensile strengths, however both ductility and fatigue strength were significantly reduced. Fatigue life could not be correlated by the applied stress, e.g. in terms of the S-N curves, owing to the different pore sizes. Using the fracture mechanics approach and Murakami’s stress intensity factor equation for pores, good correlation was found between the fatigue life and stress intensity factor range of the crack initiating defects. Predictive methods for fatigue strength reduction were developed taking account of the defect size, location, and distribution. (2) Fatigue crack growth rate: effect of heterogeneous microstructure was investigated via two different material deposition methods and testing two crack orientations. Fatigue crack growth rates were measured for damage tolerance design considerations. Unique microstructure features and their effect on the property anisotropy are discussed. Keywords: Additive manufacturing
Porosity defects Fatigue crack initiation
Fatigue crack growth rate




© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 31–42, 2020. https://doi.org/10.1007/978-3-030-21503-3_3

32

X. Zhang et al.

1 Introduction Wire + Arc Additive Manufacturing (WAAM) is a novel process where a wire is fed through an electric or plasma arc at a constant rate to build a near net shape part. WAAM titanium alloy Ti-6Al-4V has found its applications in the aerospace industry to produce large parts at high deposition rate, affordable cost, and much reduced lead time comparing to the traditional manufacturing and the powder bed fusion AM processes (Williams et al. 2016). WAAM Ti-6Al-4V parts are practically defect-free, i.e. material density is 99.99%. However, it is recognised that feedstock contamination may occur during the wire production and/or part building process that can lead to process induced porosity defects (Wang et al. 2013). In addition, complex thermal cycles and directional solidification during the WAAM process result in nonconventional microstructure and texture. High cycle fatigue properties of WAAM Ti-6Al-4V produced by single bead deposition method showed a 10% longer fatigue life compared to the mill annealed Ti6Al-4V (Wang et al. 2013). Vertical build samples (loading axis parallel to the AM build direction) showed better fatigue properties (Wang et al. 2013). The lower fatigue strength of the horizontal samples (loading axis perpendicular to the build direction can be explained by the orientation of the prior b grains that are perpendicular to the loading axis in case of horizontal specimens, which makes it easier for the micro-cracks to grow rapidly resulting in fatigue life reduction. Presence of defects and nonconventional microstructure will also influence the fatigue performance and may limit the industrial adoption process of WAAM processed materials. So far, few studies have been done on the effects of porosity and non-conventional microstructure on the fatigue properties of WAAM Ti-6Al-4V. Published studies on WAAM Ti-6Al-4V have shown that defects are the preferred fatigue crack initiation sites that can lead to fatigue strength reduction (Wang et al. 2013; Biswal et al. 2019). Biswal et al. investigated the effect of designed porosity on fatigue performance of WAAM Ti-6Al-4V built by the oscillation strategy. Fatigue strength of porosity specimens was reduced by a factor of 1.5 compared to specimens without porosity (Biswal et al. 2019). Fatigue crack propagation rate is mainly affected by the microstructure features and process induced residual stresses; defects and surface roughness have no or negligible effect. Our previous study on single bead build WAAM Ti-6Al-4V (Zhang et al. 2017) found that crack growth rate was lower when crack propagated across the build layers comparing to crack growing parallel to layers, albeit the small difference. In contrast, Xie et al. (2018) showed crack growth rate being 5% lower when crack grew in parallel with the build layers, and related this to the continuous interaction of the crack tip with the columnar b grains, which resulted in greater resistance to crack growth rate repeatedly. On the other hand, Lorant (2010) did not find any anisotropy in the crack growth rate property in WAAM Ti-6Al-4V. Studies on this front so far (Zhang et al. 2017; Xie et al. 2018; Lorant 2010) are inconclusive in terms of crack growth rate anisotropy in WAAM Ti-6Al-4V. The difference in these studies is likely caused by the competing mechanisms from microstructure and residual stress. Furthermore, crack growth rate properties in Zhang et al. (2017) were only investigated with the material by single bead deposition method, which is limited to the maximum build thickness of

High Cycle Fatigue and Fatigue Crack Growth Rate

33

7 mm. For thicker structures, parallel and oscillation build methods are currently available. Hence, it is important to study the effect of different deposition methods on fatigue crack growth rate and the property anisotropy. This paper is focused on the material properties crucial to structural integrity and durability of WAAM Ti-6Al-4V produced by different deposition methods, which are the influence of defects on fatigue crack initiation and the effect of microstructure on fatigue crack growth rate. The key objectives are: (a) to develop predictive methods for porosity defect and its effect on fatigue life taking into account of the defect size, location, distribution and current NDT capabilities; (b) to study fatigue crack propagation behaviour in two build methods and two crack orientations, in order to provide crack growth rate properties for damage tolerance design for this alloy.

2 Methods 2.1

Experimental Design

WAAM Ti-6Al-4V material was deposited using a grade-5 Ti-6Al-4V wire with a wire diameter 1.2 mm, using a HiVE machine that had a Fronius Plasma 10 module attached to a rotator, wire feeder and the wire spool around the plasma torch. The base plate was a hot-rolled Ti-6Al-4V of 12 mm thickness and was clamped to a rigid steel backing block during deposition. Heat from the substrate was extracted by a water-cooled backing plate. Argon gas of 99.99% purity was used as shielding gas (placed ahead of the torch and as well as the trailing end of the torch), directed precisely at the melt pool to avoid oxidation. The test program to study the effect of porosity on high cycle fatigue consisted of two batches of WAAM Ti-6Al-4V specimens, (1) material was deposited using the standard processing route as shown in Fig. 1(a), referred to as reference group, (2) using contaminated wires to deposit the specimen gauge section as shown in Fig. 1 (b), referred to as porosity group. The wire was contaminated using water displacement 40th formula (WD-40®), details can be found in (Biswal et al. 2019). The chemical composition (% by weight) of all the influential elements in both the batches was within the permissible limits laid out by ASTM F3302-18.

Fig. 1. Specimen extraction: (a) reference specimens, (b) porosity specimens.

34

X. Zhang et al.

The test programme to study fatigue crack growth behavior was performed using two different material deposition methods, i.e. parallel path and oscillation builds. Schematic of the two deposition methods is shown in Fig. 2(a). Two WAAM Ti-6Al4V walls were deposited, one by the parallel and one by oscillation method, both are of 450 mm length, 150 mm height and 22 mm thick as shown in Fig. 2(b). After the deposition, the walls were separated from the substrate plate and compact tension samples were extracted. Figure 2(b) shows the schematic of the two sample orientations with respect to the build layers.

Fig. 2. (a) Two different deposition methods for building two WAAM Ti-6Al-4V walls, (b) schematic of compact tension sample extraction and crack orientations

2.2

Specimen Design and Testing Method

2.2.1 Tensile and Fatigue Testing Flat tensile specimens were machined according to ASTM E8 as shown in Fig. 3(a). Load controlled fatigue test specimens were extracted from the porosity wall in vertical orientation with respect to the build direction and machined according to ASTM E466 standard as shown in Fig. 3(b). Three specimens were subjected to interrupted fatiguetomography testing which have been referred to as Fatigue-Tomography (FT) specimens FT-1, FT-2 and FT-3 while the remaining specimens were used to develop the SN curves with and without porosity defects. Specimen FT-1 was subjected to interrupted fatigue test with six X-ray computed tomography (CT) measurements taken at the test start and at suitable intervals during the fatigue test. The remaining two specimens, i.e. FT-2 and FT-3, were scanned before the start of the test and fatigue tested till failure. The interrupted fatigue-tomography test set-up consisted of two units: a 50 kN load controlled fatigue test unit (Shimadzu) and an X-ray CT scanning unit (X TH 160 system, Nikon Metrology). The entire gauge section (5 mm diameter and 10 mm length) of the specimen was scanned by the X-ray CT to map the initial porosity distribution. Specimen FT-1 was then fatigue tested at an applied stress amplitude of 315 MPa, stress ratio of 0.1 and frequency of 20 Hz for 2  104 cycles which is approximately 65% of the expected fatigue life of the specimen based on published test results (Biswal et al. 2019). At this point the fatigue test was interrupted to perform Xray CT scanning, following which the fatigue test was resumed. Subsequent fatigue tests were interrupted at every few thousand cycles to perform X-ray CT scans. All the

High Cycle Fatigue and Fatigue Crack Growth Rate

35

X-ray CT scans were conducted at an accelerating voltage of 120 kV, current of 58 µA and voxel size of 20 µm (effective pixel size of 10 µm). Exposure time was set at 500 ms while capturing 1583 projections (rotation step size 0.22°), which resulted in each interrupted scan time of approximately 50 min. The scanned files were analysed using Volume graphics® software (VG Studio Max 2.2). 2.2.2 Fatigue Crack Growth Testing For each deposition strategy, five compact tension samples were extracted, two for crack parallel to the layers and three for crack perpendicular to the layers. The compact tension samples were extracted from the mid thickness of the wall. Figure 3(d) shows the geometry and dimension of the compact tension sample. Crack growth testing was performed according to the ASTM E647 standard, at a maximum load of 3 kN, load ratio 0.1 and 10 Hz frequency.

Fig. 3. Specimens for (a) tensile test, (b) load-controlled fatigue test; (c) X-ray computer tomography set-up for porosity mapping (test facility at TU Dortmund, Germany); (d) compact tension specimen and dimensions.

3 Results and Discussion 3.1

Effect of Porosity Defects on High Cycle Fatigue

3.1.1 Porosity Analysis The size distribution of gas pores was studied using the X-ray CT scans and it was evident that the reference group specimens can consist of micron sized pores (62 ± 23 µm). In contrast, the size of gas pores in the porosity group specimens (206 ± 80 µm) were found to be much larger. Further, the X-ray CT scans showed that

36

X. Zhang et al.

the gas pore morphology in both the groups was near spherical and the spatial distribution was isolated from each other, such that material density was found to be 99.99% and 99.96% in the reference and porosity groups respectively. 3.1.2 Tensile and Fatigue Tests Tensile test results (Fig. 4a) showed that the tensile strength and yield strength of both the groups were comparable (tensile strength 859 ± 4 MPa and yield strength 802 ± 7 MPa). However, the uniform elongation measured in the two groups was widely separated (porosity group 4% and reference group 10%). Similarly, the S-N test data (Fig. 4b) showed that the fatigue limit i.e. applied stress range for a life of 107 cycles, was 540 MPa for the reference group and 360 MPa for the porosity group at an applied stress ratio of 0.1. This indicates that WAAM Ti-6Al-4V has a notch fatigue factor of 1.5 and notch sensitivity of 0.5 for spherical gas pores. It is worth mentioning that the fatigue performance of as-built (i.e. non-heat treated, polished) WAAM Ti-6Al4V specimens was close to the wrought and HIPed AM Ti-6Al-4V alloy.

Fig. 4. Results of (a) static tensile test, (b) fatigue test at applied stress ratio 0.1; Reference data and sources: wrought (MMPDS 2008); powder fusion AM - hot isostatically pressed (HIP) AM: selective laser melting (Bagehorn et al. 2017; Kahlin et al. 2017); electron beam melting (Svensson and Ackelid 2009; Brandl et al. 2011; Hrabe et al. 2015; Kahlin et al. 2017; Shui et al. 2017).

3.1.3 Interrupted Fatigue-Tomography Test Interrupted fatigue-tomography test (Fig. 5) showed tortuous cracks originated preferentially at sub-surface pores that grew towards the nearest free surface. This can be explained by the higher values of the stress concentration factor (SCF) for sub-surface pores, which is at least 25% higher than that of an internal pore and depends on the distance of the pore to the specimen free surface. The location of the cracks were found to be at the mid-riff section of the pores as shown in Fig. 5(c), so the critical subsurface pores were expanding under cyclic loads.

High Cycle Fatigue and Fatigue Crack Growth Rate

37

Fig. 5. Images of interrupted fatigue-tomography test showing the overall cross-sectional (or top view) of the X-ray CT scan taken at: (a) zero cycle, just before the fatigue test, (b) 3.2  104 cycles (specimen failed at 32380 cycles), (c) three dimensional view of the scanned section at 3.2  104 cycles.

Fig. 6. Correlation between fatigue life and stress intensity factor range (SIF range) for porosity defects tested in this study (SIF range is calculated using the Murakami’s equation, Eq. 1). Note: encircled data points denote crack initiation at internal pores; the majority failed by cracks initiating from sub-surface pores.

3.1.4 Stress Intensity Factor Range Applied to Gas Pore Defects Since the applied stress range and the porosity defect size are both play an important role in determining the fatigue performance, the stress intensity factor (SIF) range was used to correlate the fatigue test results. According to Murakami’s model (Murakami

38

X. Zhang et al.

and Endo 1986), a spherical gas pore can be treated as a planar crack of size equal to the square root of the projected area of the pore. Stress intensity factor range was calculated by Murakami’s equation as shown here by Eq. (1). qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffi p area

DK ¼ C  Dr

ð1Þ

where ΔK is the stress intensity factor range, Δr the applied stress range, √area the square root of the projected area of the pore, and parameter C is 0.5 for internal defects and 0.65 for surface defects. Fatigue test result presented in Fig. 4(b) have been re-plotted as a relation of DK vs. N in Fig. 6 showing a much better correlation between the fatigue life and the fracture mechanics parameter. 3.2

Fatigue Crack Growth Behaviour

Measured fatigue crack growth rates (FCGR) for the parallel path and oscillation wave builds are shown in Fig. 7, for two different crack orientations, and compared with traditional method processed Ti-6Al-4V, i.e. cast, wrought (mill annealed) and wrought (beta annealed). Numbers in the figure legends represent repeated fatigue tests for each case. 3.2.1 Crack Orientation When crack propagated in parallel to the WAAM build layers, the influence of different deposition method on FCGR is negligible, as shown in Fig. 7a. In this crack orientation, crack growth rate was lower than that of the wrought material (mill annealed), but greater than that of casting and beta annealed before DK reached 18 MPa√m; beyond this value, crack growth rate in WAAM and beta annealed conditions are similar. When crack growth was across the WAAM layers, Fig. 7b, WAAM Ti-6Al-4V showed lower crack growth rate compared to the wrought and similar crack growth rates as the cast and the beta annealed. The oscillation build had much lower crack growth rate in two tests, but another test had much higher growth rate, which is similar to that from the parallel build. Therefore, it is not yet conclusive for the oscillation wave build; further research is required on the competing mechanisms of microstructure and residual stress; the latter depends on specimen extraction location. If we analyse the crack propagation with respect to the columnar prior-b grain orientation, the crack in this orientation propagated in parallel with the columnar prior-b grain orientation. The role of the macro- and micro-structure and morphology of the microstructural constituents can play an important role on the fatigue crack behaviour of a material. In Ti-6Al-4V, size and morphology of a lath are the primary microstructural parameters that control the plastic zone ahead of the crack tip, therefore the fatigue crack growth rate (Han et al. 2016; Tan et al. 2018). If we analyse the crack growth rate with respect to the microstructure, the low crack growth rate in the cast and beta annealed materials is owing to the presence of fully lamellar and lamellar plus basket weave microstructure

High Cycle Fatigue and Fatigue Crack Growth Rate

39

respectively. On the other hand, the mill annealed wrought material consists of duplex microstructure with near equiaxed grains, and consequently leads to faster crack growth rate.

Fig. 7. Fatigue crack growth rates and comparison of two different build methods: (a) crack propagating parallel to layers, (b) crack across the layers

3.2.2 Build Methods and Effect on Property Anisotropy Degrees of anisotropy in fatigue crack growth rate for the parallel path and oscillation wave built samples are presented in Fig. 8. The parallel build samples did not show considerable anisotropy in crack growth rate, with only a small difference between the two crack orientations at DK values of 12-15 MPa√m, Fig. 8a. The oscillation build samples showed significant anisotropy in Fig. 8b, where samples with crack crossing the layers showed much lower crack growth rate, except the Test No. 2. The lower crack growth rate in the across layer samples is thought to be associated with the crack propagating through the layer banding along the build direction. Ho et al. (2019) presented detailed microstructure analysis of single path build WAAM Ti6Al-4V showing that the banding appearance regions are either heat affected zones (HAZ) or segregation areas that are formed due to repeated thermal cycling during the deposition process. The study also showed that the top of each HAZ zone is associated with very fine a packets with very fine spacing along the build direction, resulting in more resistance to crack propagation across the layers comparing to crack growth in parallel to the layers. For the crack across the layer specimens, the crack tip passes through each layer band, so it continuously encounters fine a packets in the HAZ and segregation zones that will slow down the crack growth rate. Detailed microstructure analysis of layer banding can be found in (Ho et al. 2019). It remains to be studied whether the parallel and oscillation deposition methods have also influenced the HAZ and segregation zones. The absence of anisotropic crack growth rate in the parallel path build samples might be associated with the very low or absence of HAZ and segregation zones. The anisotropic crack growth rate observed in the oscillation build samples might be due to the presence of pronounced HAZ and segregation zones and the crack propagating through these regions in crack across the layer samples. Hence, a lower crack growth rate is observed.

40

X. Zhang et al.

Fig. 8. Crack growth rates in both crack orientations in (a) parallel and (b) oscillation builds.

4 Conclusions • Porosity defects were embedded to specimen gauge section by using contaminated wires during the additive manufacturing. X-ray computed tomography showed spherical pores distributed scarcely (i.e. isolated from each other). Interrupted fatigue-tomography tests showed preferential crack initiation from the near surface pores; the subsequent crack growth was directed towards the nearest free surface. • Porosity defects did not affect the tensile strengths, however the ductility was reduced from 10% to 4% and fatigue strength (in terms of cyclic stress range at stress ratio 0.1) was reduced from 540 MPa (reference, defect-free) to 360 MPa. • Good correlation was found between the fatigue life and the stress intensity factor (SIF) range of the crack initiating defect, where the SIF for pore geometry was calculated using Murakami’s equation. • Fatigue crack growth rate in WAAM Ti-6Al-4V is lower than the mill annealed wrought but higher than the b annealed wrought and casting materials. • The parallel path deposition method showed higher crack growth rate than the oscillation build when crack is across the layers. However, there is virtually no difference between the two build methods when crack is parallel to the layers. • Anisotropic fatigue crack growth rate was observed only for the oscillation build, whereas the parallel path build had almost isotropic crack growth rates between the two major crack orientations. Acknowledgement. The work was supported by the industrial partners through the WAAMMat programme. We also acknowledge UK Engineering and Physical Sciences Research Council (EPSRC) for funding through grants EP/K029010/1 and EP/R027218/1.

High Cycle Fatigue and Fatigue Crack Growth Rate

41

References Åkerfeldt, P., Antti, M.L., Pederson, R.: Influence of microstructure on mechanical properties of laser metal wire-deposited Ti-6Al-4V. Mater. Sci. Eng. A 674, 428–437 (2016) Bagehorn, S., Wehr, J., Maier, H.J.: Application of mechanical surface finishing processes for roughness reduction and fatigue improvement of additively manufactured Ti-6Al-4V parts. Int. J. Fatigue 102, 135–142 (2017) Bermingham, M.J., Kent, D., et al.: Controlling the microstructure and properties of wire arc additive manufactured Ti-6Al-4V with trace boron additions. Acta Mater. 91, 289–303 (2015) Biswal, R., Zhang, X., et al.: Criticality of porosity defects on the fatigue performance of wire + arc additive manufactured titanium alloy. Int. J. Fatigue 122, 208–217 (2019) Brandl, E., Palm, F., et al.: Mechanical properties of additive manufactured titanium (Ti-6Al-4V) blocks deposited by a solid-state laser and wire. Mater. Des. 32(10), 4665–4675 (2011) Colegrove, P.A., Donoghue, J., et al.: Application of bulk deformation methods for microstructural and material property improvement and residual stress and distortion control in additively manufactured components. Scripta Mater. 135, 111–118 (2017) Günther, J., Krewerth, D., et al.: Fatigue life of additively manufactured Ti-6Al-4V in the very high cycle fatigue regime. Int. J. Fatigue 94, 236–245 (2016) Han, F., Tang, B., et al.: Cyclic softening behavior of Ti–6Al–4V alloy at macro and micro-scale. Mater. Lett. 185, 115–118 (2016) Ho, A., Zhao, H., et al.: On the origin of microstructural banding in Ti-6Al4V wire-arc based high deposition rate additive manufacturing. Acta. Mat. 166, 306–323 (2019) Hönnige, J.R., Colegrove, P.A., et al.: Residual stress and texture control in Ti-6Al-4V wire + arc additively manufactured intersections by stress relief and rolling. Mater. Des. 150 (2017), 193–205 (2018) Hrabe, N., Gnaupel-Herold, T., Quinn, T.: Fatigue properties of a titanium alloy (Ti-6Al-4V) fabricated via electron beam melting (EBM): Effects of internal defects and residual stress. Int. J. Fatigue 94, 202–210 (2015) Kahlin, M., Ansell, H., Moverare, J.J.: Fatigue behaviour of notched additive manufactured Ti6Al4V with as-built surfaces. Int. J. Fatigue 101, 51–60 (2017) Leung, C.L.A., Marussi, S., et al.: In situ X-ray imaging of defect and molten pool dynamics in laser additive manufacturing. Nature Communications. 9(1), 1–9 (2018) Lorant, E.: Effect of microstructure on mechanical properties of Ti-6Al-4V structures made by additive layer manufacturing. MSc thesis, Cranfield University (2010) Lu, S.L., Tang, H.P., et al.: Microstructure and mechanical properties of long Ti-6Al-4V rods additively manufactured by selective electron beam melting out of a deep powder bed and the effect of subsequent hot isostatic pressing. Metall. Mater. Trans. A 46(9), 3824–3834 (2015) Metallic Materials Properties Development and Standardization (MMPDS-04), Battelle Memorial Institute (2010). http://app.knovel.com/hotlink/toc/id:kpMMPDSM11/metallic-materialsproperties/metallic-materials-properties Murakami, Y., Endo, M.: Effects of Hardness and Crack Geometries on DKth of Small Cracks Emanating from Small Defects. Mechanical Engineering Publications (1986) Seifi, M., Salem, A., et al.: Defect distribution and microstructure heterogeneity effects on fracture resistance and fatigue behavior of EBM Ti-6Al-4V. Int. J. Fatigue 94, 263–287 (2017) Shui, X., Yamanaka, K., et al.: Effects of post processing on cyclic fatigue response of a titanium alloy additively manufactured by electron beam melting. Mater. Sci. Eng. A 680, 239–248 (2017)

42

X. Zhang et al.

Sterling, A.J., Torries, B., et al.: Fatigue behavior and failure mechanisms of direct laser deposited Ti-6Al-4V. Mater. Sci. Eng. A 655, 100–112 (2016) Svensson, M., Ackelid, U.: Additive manufacturing of dense metal parts by electron beam melting. In: Material Science and Technology Conference, pp. 2711–2719 (2009) Tammas-Williams, S., Withers, P.J., et al.: The influence of porosity on fatigue crack initiation in additively manufactured titanium components. Scientific Reports 7(1), 1–13 (2017) Tan, C., Sun, Q., et al.: Cyclic deformation and microcrack initiation during stress controlled high cycle fatigue of a titanium alloy. Mater. Sci. Eng. A 711, 212–222 (2018) Wang, F., Williams, S., et al.: Microstructure and mechanical properties of wire and arc additive manufactured Ti-6Al-4V. Metall. Mater. Trans. A 44(2), 968–977 (2013) Williams, S.W., Martina, F., et al.: Wire + Arc Additive Manufacturing. Mater. Sci. Technol. 32 (7), 641–647 (2016) Wycisk, E., Solbach, A., et al.: Effects of defects in laser additive manufactured Ti-6Al-4V on fatigue properties. Phys. Procedia 56, 371–378 (2014) Xie, Y., Gao, M., et al.: Anisotropy of fatigue crack growth in wire arc additive manufactured Ti6Al-4V. Mater. Sci. Eng. A 709, 265–269 (2018) Zhang, J., Wang, X., et al.: Fatigue crack propagation behaviour in wire + arc additive manufactured Ti-6Al-4V: effects of microstructure and residual stress. Mater. Des. 90, 551– 561 (2016) Zhang, X., Martina, F., et al. Fatigue crack growth in additive manufactured titanium: residual stress control and life evaluation method development. In: 29th ICAF Symposium. Nagoya, 7–9 June 2017

Strain Controlled Fatigue Testing of Additive Manufactured Titanium Alloy Ti-6Al-4V Rob Plaskitt(&), Andrew Halfpenny, and Michelle Hill HBM Prenscia, Rotherham, UK [email protected]

Abstract. This paper describes strain controlled fatigue testing of a titanium Ti6Al-4V alloy, additive manufactured by “electron beam melting” (EBM). The EBM material is manufactured in two conditions; with no post-manufacture heat treatment (“As-Built”) and after a hot isostatic pressing (HIP) treatment. The EBM HIP treatment condition is manufactured in three build orientations; vertical, horizontal and at 45°. The fatigue test results for these EBM material conditions are compared with those for similar titanium Ti-6Al-4V alloy powder, manufactured by powder metallurgy hot-isostatic pressing (PM HIP), and for similar titanium Ti-6Al-4V alloy manufactured by traditional wrought mill into bar and sheet material. The strain-life fatigue damage model and fatigue characterisation method used to fit fatigue test results from traditional manufacturing methods (wrought and PM HIP) appears to be applicable to the additive layer manufacturing method (EBM) for this titanium Ti-6Al-4V alloy material. The EBM As-Built and HIP conditions in the low-cycle region all show similar fatigue performance. This is expected given their similarity in tensile strength. The effect of the HIP on the EBM additive manufactured material is seen in the high-cycle region with much better fatigue performance. This is expected as the HIP treatment reduces porosity in the material and improves the fatigue life. The three EBM HIP build orientations all show very similar fatigue performance, though the vertical has slightly longer lives than the corresponding horizontal and 45° build orientations. It is not possible to identify whether these slightly longer lives are because of a build orientation difference, a build-tobuild difference, or an effect of powder recycling. In conclusion, fatigue tests on additive manufactured material, including both manufacturing process and any post manufacturing treatment, is considered essential because the fatigue performance of additive manufactured material cannot be inferred from tensile tests or from comparable wrought material. Keywords: Strain controlled fatigue testing Titanium


Additive manufacturing


1 Material Conditions Table 1 details the material and manufacturing conditions of the source “blank” titanium Ti-6Al-4V alloy material. All powder-based material conditions were manufactured from plasma atomized Ti-6Al-4V (grade 5). The same powder batch was used for © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 43–55, 2020. https://doi.org/10.1007/978-3-030-21503-3_4

44

R. Plaskitt et al.

all EBM conditions. The SLM and PM HIP material conditions were manufactured by different suppliers and so each used different powder batches. Where: • “EBM” – electron beam melting, an electron-beam powder bed fusion method • “SLM” – selective laser melting, a laser powder bed fusion method • “PM HIP” – powder metallurgy hot isostatic pressing. Most additive manufactured blanks (EBM and SLM in Table 1) were manufactured in vertical (V) orientation, with the long axis perpendicular to the deposition layers. To assess material property anisotropy, some EBM HIP treatment conditions were manufactured in horizontal (H) and 45 degrees (45°) build orientation. The vertical EBM blanks were manufactured together in one build, and the horizontal and 45 degrees blanks manufactured in a following build. The EBM material conditions were manufactured by Arcam EBM® using an Arcam Q20plus, in the UK National Centre for Additive Manufacturing, at the Manufacturing Technology Centre, and further described in (Plaskitt et al. 2018). The supplier manufacturing details for the SLM, PM HIP and wrought material conditions are confidential. Table 1. Description of the material conditions. Material condition EBM As-Built (V)

EBM HIP (V)

EBM HIP (H)

EBM HIP (45°)

SLM SR (V)

PM HIP Wrought bar annealed Wrought sheet annealed

Description Cylindrical blanks additive manufactured by EBM, with deposition layers perpendicular to the long axis (built vertically), with no postmanufacturing heat treatment Cylindrical blanks additive manufactured by EBM, with deposition layers perpendicular to the long axis (built vertically), with postmanufacturing HIP; 2 h at 920 °C, with 100 MPa pressure and cooled in an inert argon atmosphere to below 425 °C Cylindrical blanks additive manufactured by EBM, with deposition layers parallel to the long axis (built horizontally), with postmanufacturing HIP; 2 h at 920 °C, with 100 MPa pressure and cooled in an inert argon atmosphere to below 425 °C Cylindrical blanks additive manufactured by EBM, with deposition layers 45° to the long axis (built at a 45° angle), with postmanufacturing HIP; 2 h at 920 °C, with 100 MPa pressure and cooled in an inert argon atmosphere to below 425 °C Cylindrical blanks additive manufactured by SLM, with deposition layers perpendicular to the long axis (built vertically), with postmanufacturing stress relief (SR) heat treatment at 800 °C gradually lowered over 4 h Rectangular blanks, saw cut from a larger block manufactured from plasma atomized powder by HIP’ing at 900 °C for 2 h Cylindrical blanks saw cut and machined from 65 mm square bar (hot rolled then machined to bar). Bar annealed by heat treatment at 700 °C for 2 h and air cooled. From two separate batches Flat sheet blanks saw cut and machined from longitudinal and transverse (hot rolled) 4 mm sheets. Sheets annealed by heat treatment at 770 °C for 50 min and air cooled

Strain Controlled Fatigue Testing of Additive Manufactured

45

2 Specimen Preparation For all material conditions except wrought sheet, the blanks were machined into cylindrical tensile and fatigue specimens. The tensile specimens were machined to 5.75 mm gauge diameter and 31 mm gauge length and tested with their as-machined surface finish. The fatigue specimens were machined to 5.0 mm gauge diameter and 10 mm gauge length, and polished to have a surface finish of Ra = 0.2 or better.

3 Tensile Tests Tensile tests were performed to an internal procedure combining best practice from ASTM E8 and BS 6892 standards. Quasi-static tensile tests were performed with the material in an “as-received” and a “cyclically-stabilised” condition. Specimens in the cyclically-stabilised condition were subjected to a number of fatigue cycles under a given strain level in the plastic regime for long enough to reach cyclic-stability before being statically loaded to failure. Whilst specimens in the as-received condition were subjected to no fatigue cycles before tensile testing. These cyclically-stabilised tensile tests were performed to understand any hardening or softening behaviour that may occur within the material. These tests confirm that the mechanical properties are within expected values for the given material type and heat treatment, and were used to help set initial strain levels for fatigue tests. Table 2 shows the average ultimate tensile strength (UTS) for the tensile tests of these material conditions. The cyclically-stabilised tensile test results are used in the subsequent fatigue characterisation as they are more representative of material behaviour during in-service use than material in an unused as-received condition. These UTS averages are for only 2 or 3 test results in each tensile condition. This is too few tests to assess UTS variability but is sufficient to demonstrate that the material is manufactured with close to expected UTS values. These UTS are within the ranges reported by (Lewandowski and Seifi 2016).

Table 2. UTS from as-received and cyclically stabilised tensile tests. Material condition EBM As-Built (V) EBM HIP (V) EBM HIP (H) EBM HIP (45°) SLM SR (V) PM HIP Wrought bar annealed Wrought sheet annealed

As-received (MPa) Cyclically-stabilised (MPa) 1133 1133 1073 1064 984 1002 1000 1001 1025 (not available) 986 982 981 989 1110 1104

46

R. Plaskitt et al.

4 Fatigue Tests Fatigue tests were performed to an internal procedure combining best practice from ASTM E606, BS 7270 and ISO 12106 standards. All fatigue tests were conducted in strain control under fully reversed axial loading conditions (R = −1) for the full duration of each fatigue test. Strain controlled fatigue testing is preferred to load controlled to enable study of the plastic behaviour of the material at load levels above the yield strength. Whereas load controlled fatigue testing is only applicable to load levels below the yield strength where the material responds elastically. A triangular waveform was used for all EBM fatigue tests and a sinusoidal waveform for all others. Strain control was achieved using a standard clip on extensometer. The measured strain, applied load and displacement, cycle number and time within each cycle were recorded for the full duration of each fatigue test. Strain levels were applied to produce failures targeting fatigue life from around 500 cycles to between 1E6 or 2E7 cycles depending on the run-out cycles limit for the material condition. Strain levels were chosen to achieve an even distribution of fatigue life rather than grouped at particular strain levels. This is done to improve the statistical behaviour of the results, to provide a better basis for regression analysis during fatigue characterisation, particularly in the plastic region. The test frequency for each fatigue test was varied according to the strain level of the test, and the material response to maintain the waveform, from 0.25 Hz at the highest strain range to 10 Hz at the lowest. Fatigue tests were stopped and classed as “run-outs” when they exceeded the run-out cycles limit for their material condition without failure. All fatigue tests were conducted in air at ambient laboratory conditions. Table 3 shows the number of fatigue tests, the run out and waveform for these material conditions.

Table 3. Number of fatigue tests, run-out cycles limit and waveform. Material condition Number of tests Run out Waveform EBM As-Built (V) 15 1.0E6 Triangular EBM HIP (V) 27 2.0E7 Triangular EBM HIP (H) 16 1.5E7 Triangular EBM HIP (45°) 15 1.0E6 Triangular not applicable Sinusoidal SLM SR (V) 2a PM HIP 25 1.0E7 Sinusoidal Wrought bar annealed 40 1.0E7 Sinusoidal Wrought sheet annealed 25 1.0E7 Sinusoidal a The SLM fatigue testing programme was delayed, so there are less SLM results available to report than originally expected.

Strain Controlled Fatigue Testing of Additive Manufactured

47

5 Fatigue Failure Criteria In a load controlled fatigue test the applied load is varied to maintain the load waveform but this type of fatigue test has no knowledge of the actual strain in the gauge length. Failure can only be identified as the number of cycles to separation. In a strain controlled fatigue test the applied load is varied to maintain the strain waveform with the required strain amplitude in the gauge length. The material is considered to have failed when the load required to achieve this strain drops by a given percentage. This load drop is explained by the presence of a crack, or by a material which has become entirely plastic in its behaviour. The number of cycles to a given percentage load drop failure can be identified by post-test analysis of the measured strain in the gauge length, the applied load to achieve the strain and the number of cycles measured during the fatigue test. During fatigue testing the number of cycles to separation failure was automatically recorded by the test rig controller. However, cycles to separation is not the most appropriate failure criteria to use for the strain-life fatigue damage model. For strain-life fatigue characterisation it is better to use a “percentage load drop” failure criteria. A load drop failure criteria considers that components which have lost a percentage of their strength are unlikely to perform as intended and are at risk of imminent failure. Figure 1 shows an example from one of the PM HIP fatigue tests where separation occurred at 19136 cycles but the load required to maintain the constant strain began to reduce much earlier, from 14625 cycles for a 5% load drop. A 20% load drop failure criteria was applied in the post-test analysis to all fatigue tests reported in this paper.

Failure Criteria

Cycles

5% Load Drop

14625

10% Load Drop

15675

20% Load Drop

16620

50% Load Drop

18195

Separation

19136

Fig. 1. The number of cycles to failure for different fatigue failure criteria.

6 Fatigue Characterisation Fatigue characterisation is the process and analyses required to convert the raw fatigue test results into parametric strain-life fatigue curves for subsequent use in a strain-life fatigue damage model. This damage model contains both elastic and plastic terms to represent material behaviour, so is appropriate for both low-cycle fatigue (LCF) dominated by plasticity and high-cycle fatigue (HCF) dominated by elasticity. Each fatigue

48

R. Plaskitt et al.

test result was examined to determine whether the fatigue failure was predominantly plastic material behaviour or predominantly elastic; the fracture surfaces of the broken test specimen were examined under a microscope, and the strain, load and fatigue cycles measured during testing were post-processed into stress-strain hysteresis loops. The width of these stress-strain hysteresis loops show the degree of plasticity at the fatigue test strain amplitude. Figure 2 shows a predominantly plastic fatigue failure with a wide stress-strain hysteresis loop and a rough and lumpy fracture surface, similar to, but less pronounced than, the cup-and-cone shear surface and necking that occurs during a tensile test. Figure 3 shows a predominantly elastic fatigue failure with a very narrow stress-strain hysteresis loop, resulting from strains mostly within the elastic region of the stressstrain curve, and a fracture surface with an obvious crack nucleation location and growth region. There are no obvious semi-circular “beach marks” often associated with a growing fatigue crack because this is a constant amplitude test and such marks typically result from occasional overloads.

Fig. 2. Plastic fatigue failure.

Fig. 3. Elastic fatigue failure.

Figures 2 and 3 both show a pair of similar stress-strain hysteresis loops one after initial stabilisation, and a second mid-life hysteresis loop. In addition, Fig. 2 shows a third “Load Drop” hysteresis loop, clearly showing the reduced ability for the material to support the tensile load.

Strain Controlled Fatigue Testing of Additive Manufactured

49

The elastic strain (ee) plastic strain (ep), total strain (et = ee + ep) and total stress (r) are determined from the mid-cycle hysteresis loops for each fatigue test. These are used to obtain the cyclic stress-strain curve, shown in Fig. 4, by regression analysis of the elastic and plastic terms of the Ramberg-Osgood cyclic-stress strain equation, calculating the parameters: • • • •

cyclic cyclic cyclic cyclic

elastic modulus (E) strength coefficient (kʹ) strain-hardening exponent (nʹ) standard error (SEc).

Fig. 4. Determination of the cyclic stress-strain curve from regression analysis of the elastic and plastic terms of the Ramberg-Osgood equation.

The strain-life curve, shown in Fig. 5, is obtained by regression analysis of the elastic (Basquin) and plastic (Coffin-Manson) terms of the strain-life equation, calculating the parameters: • Elastic (Basquin): fatigue strength coefficient (rʹf), fatigue strength exponent (b) and elastic standard error (SEe). • Plastic (Coffin-Manson): fatigue ductility coefficient (eʹf), fatigue ductility exponent (c) and plastic standard error (SEp).

50

R. Plaskitt et al.

Figure 5 also shows the mean strain-life curve through the fatigue test results and a “design” curve for 97.7% Certainty of Survival (2 standard deviations below the mean curve), with 95% Confidence. Design curves can be calculated for a specific reliability target (or certainty of survival) and confidence interval using the standard errors SEe and SEp. Further fatigue characterisation details are described in (Plaskitt et al. 2018), and presented by (Halfpenny 2018).

Fig. 5. Determination of the strain-life curve from regression analysis of the elastic (Basquin) and plastic (Coffin-Manson) terms of the strain-life equation.

7 Fatigue Test Results and Fatigue Curves The fatigue test results for the previously tested wrought bar and sheet material conditions were combined for use as a “wrought baseline”. Figure 6 shows these fatigue test results, their mean regression curve and ±3 standard deviations (sd) from this mean curve. These results show higher than normal scatter because they have been combined from two batches of wrought bar and one batch of wrought sheet tested in both longitudinal and transverse rolling directions. The mean and ±3sd curves are retained in the following plots for comparison purposes.

Strain Controlled Fatigue Testing of Additive Manufactured

51

Fig. 6. Baseline fatigue test results for wrought bar and sheet combined.

Figure 7 compares the EBM As-Built (V) and EBM HIP (V) additive manufactured conditions with the wrought baseline. In the low-cycle region, below 1E4 reversals (5000 cycles), with high strain amplitudes and where plasticity dominates, there is very little difference in the As-Built and HIP conditions. This is expected given the similarity in ultimate tensile strength and existing knowledge that tensile additive manufactured properties are comparable to or even slightly better than wrought materials. It is noted that this result would not have been captured if a load controlled fatigue testing method had been used because it cannot control for the plasticity occurring in this lowcycle region. Conversely in the high-cycle region the HIP condition shows much better fatigue performance, with longer fatigue lives though with a notable increase in scatter. The HIP treatment is expected to be reducing porosity in the material and improving the fatigue life, whilst any remaining porosity is potentially contributing to the increased scatter. When comparing with the wrought baseline in the low-cycle region, where plasticity dominates, both the As-Built and HIP condition have a notable steeper slope. The reason for this steeper slope is not known.

52

R. Plaskitt et al.

Fig. 7. Compare EBM As-Built (V) and EBM HIP (V) with wrought baseline.

Figure 8 compares the three EBM HIP build orientations; vertical, 45° and horizontal. All three conditions show similar characteristics; low scatter in the low-cycle region, and a notable increase in life and scatter in the high-cycle region as a result of the HIP treatment. The vertical build orientation generally has longer fatigue lives at the same strain amplitude than both the 45° and horizontal orientations which are very similar to each other. It is not known whether this is a build orientation difference, a build-to-build difference, or an effect of powder recycling, as the verticals were all built together in one build and the 45° and horizontals all built together in a following build. (In this figure the build orientation mean curves are not shown to improve clarity of the very similar fatigue test results.)

Strain Controlled Fatigue Testing of Additive Manufactured

53

Fig. 8. Compare EBM HIP build orientations (V-45-H) with wrought baseline.

Figure 9 compares the additive manufactured EBM As-Built (V) and HIP (V) conditions with the powder metallurgy manufactured PM HIP condition. Considering the two HIP conditions, the EBM HIP has much better fatigue performance than the PM HIP condition, with longer fatigue lives. Post-test microscopic inspection observed inclusions on some PM HIP fracture surfaces, and it is likely that these had a detrimental effect on fatigue performance. Though only two SLM SR (V) fatigue tests results are available, these are included in this figure because it is notable that they are both below the wrought -3sd curve. This result is not surprising as this material condition has only had a heat treatment and not a HIP, so they are more likely to still contain some porosity. Their fracture surfaces also show the presence of unmelted or partially fused powder.

54

R. Plaskitt et al.

Fig. 9. Compare EBM As-Built (V), HIP (V), PM HIP and SLM SR (V).

8 Conclusions This work has led to useful knowledge and comparison of fatigue performance for titanium Ti-6Al-4V alloy material manufactured by different methods. The strain-life fatigue damage model and fatigue characterisation method used to fit fatigue test results from traditional manufacturing methods (wrought and PM HIP) appears to be applicable to the new additive layer manufacturing method (EBM) for this material. The EBM As-Built and HIP conditions in the low-cycle region all show similar fatigue performance. This is expected given the similarity in ultimate tensile strength and existing knowledge that tensile additive manufactured properties are comparable to or even slightly better than wrought materials. It is noted that this result would not have been captured if a load controlled fatigue testing method had been used because it cannot control for the plasticity occurring in this low-cycle region. The effect of the HIP on the EBM additive manufactured material is seen in the high-cycle region with much better fatigue performance, with longer fatigue lives though with a notable increase in scatter. This is expected as the HIP treatment closes pores to reduce porosity in the material and improves the fatigue life, whilst any remaining porosity is potentially contributing to the increased scatter. The three EBM HIP build orientations all show very similar fatigue test results, though the vertical (V) build orientation has slightly longer fatigue lives than the

Strain Controlled Fatigue Testing of Additive Manufactured

55

corresponding horizontal (H) and 45 degrees (45°) build orientations. These slightly longer lives in the vertical build orientation should be treated with caution as it is not possible to identify whether they are because of a build orientation difference, a buildto-build difference, or an effect of powder recycling. All of the vertical cylindrical blanks were manufactured first within the same build, and then the powder was re-used in a following build to manufacture the horizontal and 45 degrees cylindrical blanks. The PM HIP fatigue performance lies towards the lower -3sd bound of the wrought baseline, with fatigue cracks often propagating from inclusions observed on the fracture surfaces. No conclusions are drawn from the two SLM fatigue tests. A full study of fatigue tests for titanium Ti-6Al-4V alloy material manufactured by SLM is planned for 2019 and 2020. In conclusion it is imperative that fatigue data are obtained from specimens that are made with the same manufacturing and post-manufacturing treatment processes as the intended in-service component, rather than from a generic additive manufacturing process. The additive manufacturing build parameters and powder are known to have a direct effect on porosity levels and therefore fatigue performance. The fatigue performance of additive manufactured material cannot be inferred either from tensile testing the same additive manufactured material or from the fatigue performance of the comparable wrought material. HBM Prenscia thank and acknowledge the support of the UK National Centre for Additive Manufacturing, at the Manufacturing Technology Centre in this work.

References Halfpenny, A.: Fatigue Characterization and Testing of Materials (2018). https://www.ncode. com/videos/fatigue-characterisation-and-testing-of-materials Lewandowski, J.J., Seifi, M.: Metal additive manufacturing: a review of mechanical properties. Mater. Issues Add. Manuf. 46, 151–186 (2016) Plaskitt, R., Hill, M., Halfpenny, A., Lavelle, P.: Strain-life fatigue testing of additive layer manufactured titanium alloy Ti-6Al-4V by electron beam powder bed fusion. In: Proceedings of 6th Aircraft Structural Design Conference, Royal Aeronautical Society. Bristol, UK (2018)

The Optimization and Design of Complicated-Surface Panel Based on Automate Fiber Placement Tieliang Zhang(&), Liyang Liu, and Hao Cui Shenyang Aircraft Design and Research Institute, Aviation Industry Corporation of China, No. 40 Tawan Street, Shenyang 110035, People’s Republic of China [email protected], [email protected], [email protected] Abstract. Aiming at the special configurations of the automated fiber placement composite panel and the complex loads, the applicable analysis criteria and failure criteria are screened and perfected. Iso-parametric shell element is used to simulate skin and stiffener. A specific interface element is used to simulate between skin and stiffener. During the analysis process of the buckling, postbuckling and failure of the structure under complex loads, the modified Hashin criterion is used to simulate the longitudinal and transverse failure of the material. The tensile and shear failure criteria are used to simulate the interface between skin and stiffener. With the increase of the number and height of stiffeners, the buckling and post-buckling capacity of double-curvature stiffened panel increases. But the buckling and post-buckling capacity of longitudinal and transverse stiffeners are very close. Keywords: Double-curvature stiffened panel
Automated fiber placement Hashin failure criteria
Buckling
Post-buckling




1 Introduction Due to the developments of high-performance composite materials, multi-constraint optimization designing methodologies and automate fiber placement (AFP) technologies, various advantages of applying AFP to fabricate composite panels are obtained (Niu 1992). However, in terms of the composite panels subjected to the out-of-plane pressures, there are still many problems to be solved. According to the requirements of structure’s stiffness, strength, stability, durability, damage tolerance, maintenance capability, the structure is not allowed to buckle under the limit load, while local buckling can occur before the ultimate load (Niu 1997). Besides, when the overpressure is applied, some local damages of the materials are permitted in such a structure which is designed based on constraints of the loads, temperature, sealant condition, corrosion resistance and so on. The post-buckling behavior of stiffened curved plates under axial compression has been studied. It is found that the load-carrying capacity of stiffened curved plates is greater than that of stiffened plates under the same size. The failure mode is mainly the © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 56–70, 2020. https://doi.org/10.1007/978-3-030-21503-3_5

The Optimization and Design of Complicated-Surface Panel

57

debonding of the bond layer between stiffeners and curved skin. For the purpose of minimizing the stiffened panel’s weight and optimizing the best manufacturing technology, this paper investigates the design of composite double-curvature stiffened panel considering the buckling, post-buckling, failure and stiffeners details.

2 Theory For the special characters, such as the irregularly double curvature and largely negative curvature, the panel under the complicated loading conditions, the analysis method and failure criteria are proposed. Failure Criterion. Strength characteristics of material are the important part of their mechanical properties. Failure criterion is the important part of material research. Material failure are based on the mechanical analysis and numerical simulation. A series of composite materials failure criteria are established. By comparing the simulation results with the experimental data, the different failure criterion for AFP composites are studied, and the optimal failure criterion are obtained. The finite element model is set up in the software of Abaqus (see Fig. 1), which geometrical parameters are the same with the open-hole tensile test pieces (see Table 1).

Fig. 1. The finite element model of the open-hole tensile piece.

Table 1. Geometrical parameters of the open-hole tensile test piece. Geometrical parameters Stacking sequence Length Width Diameter of hole

Value [45/0/-45/90]3s 300 mm 36 mm 6 mm

The AFP material is CCF300/BA9916-II, whose properties is below (see Table 2). Table 2. Properties of CCF300/BA9916-II. Properties E1 Value

E2

m

G12

XT

XC

YT

YC

S12

S23

129 GPa 9.83 GPa 0.3 5.38 GPa 1773 MPa 1264 MPa 68.8 MPa 225 MPa 132 MPa 115 MPa

58

T. Zhang et al. Table 3. Ultimate strength and maximum displacement of different failure criteria.

Failure criteria Chang-Chang Hashin Modified Hashin Tsai-Wu Experimental value

Ultimate strength (MPa) 300.3 298 308.9 312.2 305.46

Error 1.69% 2.44% 1.12%

Maximum displacement (mm) 0.587 0.616 0.548

Error 46.8% 54% 37%

2.2% -

0.614 0.4

-

53.5%

Fig. 2. Load-displacement curve of different failure criterion.

Where XT, Xc, YT, Yc, S12, S23 are the tensile strength in fiber direction, the compressive strength in fiber direction, the tensile strength transverse to fiber direction, the compressive strength transverse to fiber direction, the axial shear strength and the transverse shear strength, respectively. E1, E2, m, G12 are the Young’s modulus in fiber direction, the Young’s modulus transverse to fiber direction, Poisson’s ratio and the transverse shear modulus. One end of the model is fixed, and the other end is axially loaded. Four failure criteria, including Chang-Chang (Chang and Chang 1987), Hashin (Hashin 1980), modified Hashin and Tsai-Wu (Tsai and Wu 1971), are used to write into the UMAT

The Optimization and Design of Complicated-Surface Panel

59

program of Abaqus. The data of stiffness and strength are compared by each failure criterion (see Table 3). The load-displacement curve is shown in Fig. 2. According to the above analysis data and the figure, it can be seen that the ultimate strength and the maximum displacement of the modified Hashin failure criterion is the closest to the experiment data. Therefore, the modified Hashin failure criterion is used to simulate the damage process of AFP composite structure. Hashin criterion has been widely used in the current research, especially in the analysis of composite laminates. It takes into account the failure modes under different load cases, and reflects various failure modes of composite materials. Based on the analysis of the experimental phenomena of the test pieces by AFP, it is found that a large number of shear failure occur. Therefore, the Hashin failure criterion is modified in the finite element analysis. The failure modes are: (1) Tensile Fiber Mode ðr11  0Þ


r11 XT

2 1

ð1Þ

(2) Compressive Fiber Mode ðr11 \0Þ


r11 XC

2 1

ð2Þ

(3) Tensile Matrix Mode ðr22  0Þ
2
2 r222 s23 s12 þ þ 1 S23 S12 YT2

ð3Þ

(4) Compressive Matrix Mode ðr22 \0Þ !



2 r22 2 r22 YC 2 s12 2 s23 þ 1 þ þ 1 2S23 YC 2S23 S12 S23

ð4Þ

(5) Shear Mode


r11 XC

2







s12 þ S12

s12 S12

2

2







s13 þ S13

s13 þ S13

2  1 r11 \0

ð5Þ

r11  0

ð6Þ

2 1

60

T. Zhang et al.

When one type of failure mode occurs, the elastic properties of material will be reduced accordingly. The specific reduction scheme is as follows: (1) Fiber tensile failure: E11 ; E22 ; G12 ; G13 are reduced to 7% of the original value. (2) Fiber compressive failure: E11 ; E22 ; G12 ; G13 are reduced to 14% of the original value. (3) Matrix tensile failure: E22 ; G12 ; G23 are reduced to 20% of the original value. (4) Matrix compressive failure: E22 ; G12 ; G23 are reduced to 40% of the original value. (5) Shear failure: G12 ; G13 are reduced to zero. Iso-Parametric Shell Element. For the negative curvature surface, the overlap and distortion between layers of AFP composite structure are serious and the transverse shear deformation is very big. The traditional shell element analysis can not accurately reflect the effect of cross-section warpage caused by transverse shear deformation, so the analysis error is very large. The exponential displacement function is established according to the laying mode and the bonding condition of the interface layer, and the iso-parametric shell element (see Fig. 3) is utilized to simulate the skin and stiffeners. On this basis, the finite element program UEL is compiled through the user-defined interface of Abaqus. The exponential iso-parametric shell element is embedded in the commercial finite element analysis software to running the finite element analysis of the negative curvature AFP structure.

Fig. 3. The iso-parametric shell element.

In order to verify the accuracy of the exponential iso-parametric shell element, a cylindrical bending laminate with uniform load on the surface is considered. The ratio of length to thickness is 8. Table 4 shows the comparison results of different analysis elements under two boundary conditions. It can be seen that the exponential isoparametric shell element can obtain more accurate results.

The Optimization and Design of Complicated-Surface Panel

61

Table 4. Comparison results of different analysis elements. Analysis element

Two ends clamped Displacement(mm) Error Stress (MPa) Error Exact solution −1.734 0.227 Traditional shell element −1.456 16% 0.179 −21% Reddy element −1.481 14.6% 0.27 19% First order shear element −1.456 16% 0.179 −21% Iso-parametric shell element −1.764 −1.7% 0.225 −0.9% Analysis element One end clamped, the other end simply supported Displacement(mm) Error Stress (MPa) Error Exact solution −2.941 −1.147 Traditional shell element −2.185 25.7% −1.076 6.2% Reddy element −2.697 8.3% −1.191 −3.8% First order shear element −2.829 3.8% −1.108 3.4% Iso-parametric shell element −2.93 0.4% −1.149 −0.17%

Interface Element. Under in-plane loads, the composite stiffened panels are deboned between the skin and the stiffeners. The traditional method is that the stiffeners are simply equivalent to a bar or beam with certain bending stiffness. Without considering the interface effect between the skin and the stiffeners, it is impossible to accurately simulate the debonding failure between the skin and the stiffeners, which affects the accuracy of structural analysis (In most cases, the calculation results are higher than the actual results. The structure designed is more dangerous). Therefore, it is necessary to consider the interfacial effect between the skin and the stiffeners in the analysis of the composite stiffened panels. At present, virtual crack closure technique (VCCT) and interface element method are feasible. But VCCT requires fine mesh, and it can not predict the crack initiation location. In addition, VCCT is only suitable for small-scale structure. Therefore, its further application is limited. The interface element method based on cohesive zone model is widely used to simulate the whole damage process, including crack initiation, propagation and failure (Rybicki and Kanninen 1977). So, the interface element is introduced to simulate the interlaminar characteristics between the skin and the stiffeners (see Fig. 4). The debonding between the above two components is simulated by the stress analysis and the quadratic failure criteria (Cui et al.1992) of the interface element.

62

T. Zhang et al.

Fig. 4. The interface element.

Fig. 5. The finite element model.

3 Example Figure 5 is the finite element model of double-curvature stiffened panel. Its geometric parameters are: the quadrilateral arc length is 200 mm, and the curvature radiuses (Rx and Ry) in both directions are 800 mm. The number of T-shaped stiffener is one. The double curvature skin and T-shaped stiffener are all laid with 45o layer, and the stacking sequence is [45/-45/45/-45]s. The boundary condition of the model is fixed on four sides and the load is uniform pressure load. The mesh density of the model is: skin 40  40, stiffener flange 4  40, stiffener web 4  40. The interface element meshing is the same with the stiffener flange. The linear analysis results of structural stability problems are far from the actual results, but the buckling load and mode can be used as a reference for the nonlinear analysis. In many studies of nonlinear buckling and post-buckling considering the structural defects, the first few modes of linear buckling are replaced by the initial defects because the actual geometric defect are difficult to determine. In order to verify the rationality and validity of the finite element model, buckling loads and modes are provided for the nonlinear post-buckling analysis. The first-order buckling load and the second-order buckling load of double-curvature stiffened panels are 21756 N and 23390 N respectively (see Fig. 6).

The Optimization and Design of Complicated-Surface Panel

63

(a) The first-order buckling mode

(b) The second-order buckling mode Fig. 6. The first-order and second-order buckling mode of a double-curvature stiffened panel.

The constitutive equation, failure criterion and stiffness reduced scheme of composite materials and interface materials are established by writing subroutine UMAT. The Newton-Raphson method is used to solve the nonlinear equations. The detailed analysis process is shown in Fig. 7. Step 1: The linear buckling analysis is carried out, and the first-order buckling mode is added to the subsequent nonlinear analysis as the initial disturbance. Step 2: In the process of analysis, the displacement step-by-step loading method is adopted for external loads. In each load increment step, the total stiffness matrix is established. The equilibrium equation is solved. The stress of the interface element and each layer of the composite structure is obtained. Step 3: According to the failure criterion, the damage of interface elements and isoparametric shell element can be judged. If there is a damage, the corresponding damage variables can be calculated according to the current stress. The finite element equilibrium equation is re-established according to the reduced material properties. Repeat the above analysis until no new damage occurs in the structure. Step 4: Increase the load and return to Step 2 for next step analysis until the structural is destroyed.

64

T. Zhang et al.

Finite element model

Buckling analysis Initial disturbance Nonlinear buckling analysis

Initial load P0

First step iteration

Stress analysis

Failure criterion Stiffness reduction Interface element failure

Failure judgement

Iso-parameter shell element failure

Material degeneration

No Pi+1=Pi+ ΔP

Damage judgement

No

Yes

Finish

Fig. 7. Process of model failure analysis.

Considering the interface debonding and composite material failure, the nonlinear post-buckling analysis of double-curvature stiffened panel is carried out. Figure 8 is the load-center point displacement curve.

The Optimization and Design of Complicated-Surface Panel

65

Fig. 8. Load-center point displacement curve.

In the Fig. 8, point A is the maxima point, the structure buckles. Point B is the minima point, and the web of T-shaped stiffener is failure at point C (see Fig. 9). The failure mode is fiber cracking.

Fig. 9. Web of T-shaped stiffener failure.

As the load increasing, the failure of interface elements occurs between the skin and the stiffener. Debonding occurs at both ends of the interface (see Fig. 10). Then the interface element crushing occurs at the center of the interface (see Fig. 11).

66

T. Zhang et al.

Fig. 10. Debonding at both ends of the interface.

Fig. 11. Crushing at the center of the interface.

It can be seen that after buckling, the bearing capacity of the double-curvature stiffened panel does not decrease dramatically due to the role of stiffener. But it decreases dramatically after the web of T-shaped stiffener damaged. It can be seen that the dangerous location of the double-curvature stiffened panel is in the middle of the web.

4 Effect of Stiffener Parameters on Double-Curvature Stiffened Panel The effects of different stiffener parameters on the buckling and post-buckling capacity of composite double-curvature stiffened panel (see Fig. 12) are studied. The thickness of single layer is 0.125 mm. The stacking sequence of orthogonal layers is [0/90]2s. The thickness of skin t is 1 mm. x is 1 direction, corresponding to 0o layers. The xdirectional and y-directional spans are a and b, and the curvature radius are Rx and Ry. a = b = 100 h, Rx = 1000 h, Ry = 0.8 Rx. The height of the ribs is d and the thickness is t. The ribs are laid symmetrically along the x and y directions, using the same material and thickness as the skin. The boundary conditions of the finite element model are free at both ends of the stiffeners and hinged at all sides of the shell.

The Optimization and Design of Complicated-Surface Panel

67

Fig. 12. Geometric model of composite double-curvature stiffened panel.

Effect of Stiffener Arrangement Direction on Buckling and Post-buckling Load. The effects of stiffener arrangement different directions on the buckling and post-buckling load of bi-curvature panel are different due to their different longitudinal and transverse curvatures. The height of the stiffener d is 2 mm, and the stacking sequence of orthogonal layers is [0/90]2s. At the midpoint, one stiffener is respectively arranged along the x and y directions, and the stiffeners are intersected along the x and y directions. The load-displacement curves of stiffener with different directions are shown in Fig. 13. The load-displacement curves of longitudinal and transverse stiffeners are very close. The load-displacement curve of along y direction is slightly smaller than along x direction. That is to say, stiffener along the direction with smaller curvature is more stiffness and stability. Compared with single stiffener, cross stiffener has a certain increase in buckling and post-buckling loads, which is due to the increase of the number of stiffeners and the increase of the bending stiffness of the panel.

Fig. 13. Load-displacement curves of stiffener with different directions.

68

T. Zhang et al.

Effect of Stiffener Height on Buckling and Post-buckling Load. A stiffener is laid along the y direction, and the stacking sequence of orthogonal layers is [0/90]2s. The height of the stiffener is d = 2 mm, 3 mm and 4 mm, respectively. The load-displacement curves of stiffener with different heights are shown in Fig. 14. With the increase of stiffener height, the buckling loads and post-buckling loads increase steadily, and the jump instability phenomenon decreases gradually. This is because that the bending stiffness increases with the increase of stiffener height, so the buckling and the post-buckling carrying capacity increase.

Fig. 14. Load-displacement curves of stiffener with different heights.

Effect of the Number of Stiffeners on Buckling and Post-buckling Load. In the finite element model, the stiffener height h is 3 mm and the sequence of orthogonal layers is [0/90]2s. The numbers of stiffener are 1, 3 and 5, respectively. The stiffeners are laid along the y direction. The load-displacement curves of different numbers of stiffener are shown in Fig. 15. With the increase of the number of stiffeners, the buckling and post-buckling loads increase, but the load amplitude is small. This is because that the bending stiffness increase with the increase of the number of stiffeners. The load-displacement curves of three and five stiffeners are very close, because the stiffeners arranged on the panel edge do not contribute much to the bending stiffness of the double-curvature stiffened panel under four ends hinged.

The Optimization and Design of Complicated-Surface Panel

69

Fig. 15. Load-displacement curves of different numbers of stiffener

5 Conclusion (1) The iso-parametric shell element is used to simulate the skin and stiffener of complex curved surface. The interface between the above two components is simulated by interface element. (2) The process of buckling, post-buckling and damaging can be analyzed, which means the interlaminar damage of both skin and stiffeners is detailed through the modified Hashin criteria. Furthermore, the interlaminar damages at the normal and tangential directions are controlled by the tension and shear criteria respectively. The analysis methodologies of the stiffness, strength and load-carrying capability of the double-curvature stiffened panel are established. (3) With the increase of the number and height of stiffeners, the buckling and postbuckling capacity of double-curvature stiffened panel increases. But the buckling and post-buckling capacity of longitudinal and transverse stiffeners are very close.

References Chang, F.K., Chang, K.Y.: A progressive damage model for laminated composites containing stress concentrations. J. Composite. Matel. 21(9), 834–855 (1987) Cui, W.C., Wisnom, M.R., Jones, M.: A comparison of failure criteria to predict delamination of unidirectional glass/epoxy specimens waisted through the thickness. Composites 23(3), 158– 166 (1992) Hashin, Z.: Failure criteria for unidirectional fiber composites. J. Appl. Mech. 47(2), 329–335 (1980)

70

T. Zhang et al.

Niu, C.Y.: Composite Airframe Structures. Conmilit Press LTD, Hong Kong (1992) Niu, C.Y.: Airframe Stress Analysis and Sizing. Conmilit Press LTD, Hong Kong (1997) Rybicki, E.F., Kanninen, M.F.: A finite element calculation of stress intensity factors by a modified crack closure integral. Eng. Frac. Mech. 9, 931–938 (1977) Tsai, S.W., Wu, E.M.: A general theory of strength for anisotropic materials. J. Composite. Matel. 5, 58–80 (1971)

Advanced Materials and Innovative Structural Concepts

A Modeling Approach for the Fatigue Behavior of Laser Drilled Micro Perforated Structural Panels Dort Daandels1(&), Stefan Riekehr2, Nikolai Kashaev2, Jon Mardaras3, Sammy Zein El Dine1, and Christian Heck1 1

Airbus Operations GmbH, Bremen, Germany [email protected] 2 Institute of Materials Research, Materials Mechanics, Helmholtz-Zentrum Geesthacht, Geesthacht, Germany {stefan.riekehr,nikolai.kashaev}@hzg.de 3 Airbus S.A.S, Toulouse, France [email protected] Abstract. With hybrid laminar flow control the drag can be reduced for airfoils. This is done by boundary layer suction through millions of small holes in the skin panels. The goal of this paper is to assess the impact on fatigue properties and to describe a modeling approach to account for the laser drilled holes in the skin panels. Two fatigue test programs were performed to obtain inputs to model the fatigue behavior of micro perforated titanium panels. From the test data several SN curves were generated. The all the SN curves have a similar slope as the basic material. This allows for a factorization approach to model the fatigue behavior. Three factors are applied to the basic material properties. The first factor takes into account the hole geometry, which be captured analytically with a stress concentration factor. The pitch between the holes is large enough that there is no interference for fatigue initiation. Secondly multiple site damage was visually observed by multiple crack plateaus on the crack surface. The crack jumps between the rows of holes. This can be seen as shift in the SN curve which is in line with the knock down factor found by an analytical approach. And finally a technology factor to account for the manufacturing process was obtained. The manufacturing process for the coupons of the two test programs was different, resulting in a different technology factor. Therefore setting specific requirements for the manufacturing process can reduce the impact on the fatigue properties. By multiplying the basic material properties with these three factors according to Eq. (5) it is possible to do standard fatigue assessments in commonly used fatigue tools and with detailed spectra. Keywords: Micro perforation
Laser drilling
Multiple site damage Hybrid laminar flow control
HLFC
Durability




1 Introduction into Hybrid Laminar Flow Control Hybrid laminar flow control (HLFC) is a technology approach to reduce the drag of civil jet transport aircrafts. The aim of HLFC is to create a longer laminar flow along the airfoil on an aircraft. This is done by boundary layer suction through small holes in © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 73–87, 2020. https://doi.org/10.1007/978-3-030-21503-3_6

74

D. Daandels et al.

the skin panels. The holes are small ( r r v m m = < UTS
rm
rm > ra  rva UTSrvm ð1Þ  ¼
 2wv v > rva v > ; : rm rmv UCS  r  r m m UCSr m

where, rav and rmv are the alternating and mean stress components of the fatigue stress for a given constant value of life, N, under fatigue loading at the critical stress ratio, v = UCS/UTS. The variable wv denotes the fatigue strength ratio and it is defined as wv ¼

rvmax rb

ð2Þ

where, rb (>0) is the reference strength to define the peak of the static failure envelope in the (rm- ra) plane and rvmax is the maximum stress extracted from the S-N curve of the critical R-ratio, v, for a given number of fatigue cycles. Therefore, this normalization guarantees that wv always varies in the range [0, 1] and the exponents (2-wv) in Eq. (1) are always greater than unity. Subsequently, linear (2-wv = 1) or parabolic (2wv > 1) curves can be obtained from Eq. (1). The critical fatigue strength ratio represents the normalized cyclic stress, and its relation to the number of loading cycles defines the normalized critical S–N curve: a

2Nf ¼

2 ð1  wv Þ K wnv

ð3Þ

where the constants a, n and K* are material constants determined iteratively. The relation between the normalized stress and the number of loading cycles presented in Kawai et al. and shown in Eq. 3 is given implicitly. For most sets of a, n and K* the normalized stress cannot be extracted directly from Eq. (3). In these cases, the use of Eq. (2) is more convenient for the determination of wv. After determining the critical S–N curve by fitting to the available fatigue data, the CFL diagram can be constructed based on the static strengths, UTS and UCS, and the reference S–N relationship.

3 Test Description Static and fatigue tests were carried out for open-hole coupon specimens made of unidirectional carbon/epoxy tapes for examination of the applicability of the KAWAI constant life model. A quasi-isotopic lay-up of intermediate modulus unidirectional

402

Y. Buimovich and D. Elmalich

(UD) Carbon/epoxy tapes was examined. The lamination sequence, [(+45°, 90°, −45°, 0°)s]s, is balanced and symmetrically stacked. The geometry of the specimens was in accordance to ASTM standard for open-hole tests (ASTM D5766, ASTM D6484). The specimen is illustrated in Fig. 1.

Fig. 1. IM carbon UD/ER450 open-hole tension and compression specimen for quasi-isotropic lamination

A total of 56 composite specimens were tested. The tests included static compression and tension strength to obtain the critical R ratio. Fatigue tests included five R ratios; i.e., 0.5, 0.1, v, −1 and −10. Most R-levels consisted of five levels of stresses (with two tests at each stress level). A picture of the tested specimen at the loading apparatus is given in Fig. 2.

Fig. 2. The tested specimen at the loading apparatus

Examination of the KAWAI CLD Method

403

4 Analysis of Test Results The critical R-ratio obtained from the static tests is v = −0.5. Hence, the Kawai CLD was built according to fatigue tests results for R = v. The predicted Kawai CFL and test results of fatigue tests are shown in Fig. 3. Note that the values of the mean and amplitude stress in Fig. 3 are normalized with respect to the reference stress rb. A relatively good agreement was obtained between the predicted and experimental results, with the exception of R = −10. The prediction of the Kawai model was slightly conservative for R-ratios
−1 and un-conservative for R-ratio that equals 10. The Goodman CLD was also compared to test results and is illustrated in Fig. 4. Note that the values of the mean and amplitude stress in Fig. 4 are normalized with respect to the reference stress rb. It may be observed in Fig. 4, that the Goodman CLD highly overestimated fatigue life for purely tension and tension-compression areas and underestimated fatigue life for R-ratios lower than R = −1. SN curves were built for each R-ratio for analyzing the precision of the Kawai and Goodman constant life diagrams. The SN curves built for R = 0.1, 0.5, −0.5, and −1 are given in Fig. 5 through Fig. 8, respectively, in terms of the normalized maximum stress. The SN curve built for R = 10 is given in Fig. 9 in terms of the normalized minimum stress. In the curves at Fig. 5 through Fig. 9, both Kawai and Goodman curves are shown, as well as the experimental results for the R-ratio. Note that in Fig. 8, experimental results are compared only to the Kawai curve, since the Goodman curve is built according to test results of R = −1. In Fig. 7, experimental results are compared

Fig. 3. Kawai Constant Fatigue Life Diagram for unidirectional carbon/epoxy laminates

404

Y. Buimovich and D. Elmalich

Fig. 4. Goodman Constant Fatigue Life Diagram for unidirectional carbon/epoxy laminates

Fig. 5. SN curve for R = 0.1 for unidirectional carbon/epoxy laminates

Examination of the KAWAI CLD Method

405

Fig. 6. SN curve for R = 0.5 for unidirectional carbon/epoxy laminates

Fig. 7. SN curve for R = −0.5 for unidirectional carbon/epoxy laminates

only to the Goodman curve, since the Kawai curve is built according to test results of R = −0.5 (v). It may be obtained in Figs. 5 and 6 that both SN curves are below the experimental results. However, the Goodman curve is much more conservative than the

406

Y. Buimovich and D. Elmalich

Fig. 8. SN curve for R = −1 for unidirectional carbon/epoxy laminates

Fig. 9. SN curve for R = 10 for unidirectional carbon/epoxy laminates

Examination of the KAWAI CLD Method

407

Kawai method in most of the cases. The great conservatism of the Goodman model is also observed in Fig. 7. It may be seen in Fig. 8, that there is a relatively good agreement between the experimental results and the curve built according to the Kawai method for R = −1. As opposed to the conservative trend of the Kawai method for the pure tension zone and the good agreement obtained for R = −1, the Kawai SN curve for R = 10 is below the experimental results and hence yields un-conservative prediction (see Fig. 9). It may also be seen in Fig. 9 that the Goodman curve is unconservative, as compared to the experimental results.

5 Modified KAWAI CLD A modified model was suggested to overcome the un-conservatism of the Kawai model at the pure compression zone. Thus, the constant life curve is built from two parts; i.e. for UTS
rm
rvm, constant life diagrams are built according to KAWAI method and for UCS  rm  rvm, constant life diagrams are built by a linear line between rva and UCS. Hence, CFL formulation is determined as follows v

a ¼  rarr v a

ra ¼




rm rvm UTSrvm

ðrm rc Þrva rvm rc

2wv

UTS
rm
rvm

:

ð4Þ

UCS  rm  rvm

Fig. 10. Modified Kawai Constant Fatigue Life Diagram for unidirectional carbon/epoxy laminates

408

Y. Buimovich and D. Elmalich

The predicted modified Kawai CFL and test results of fatigue tests are shown in Fig. 3. Note that the values of the mean and amplitude stress in Fig. 10 are normalized with respect to the reference stress rb. One may observe in Fig. 10 that the experimental results of R = 10 are within the range of the CLD, as opposed to these values in Fig. 3. SN curve was built for R = 10, based on the modified Kawai methodology. It is presented in Fig. 11. It may be observed in Fig. 11 that the experimental results are slightly below the predicted curve according to the modified Kawai method for fatigue life above 105 cycles.

Fig. 11. SN curve for R = 10 for unidirectional carbon/epoxy laminates

6 Summary and Conclusions The applicability of KAWAI CLD method for fatigue life prediction of composites was examined. The Goodman methodology was examined as well. Static tests were carried out first for open-hole coupon specimens made of unidirectional carbon/epoxy tapes for obtaining the critical R-ratio, v = UCS/UTS. This value was determined as v = −0.5. Fatigue tests were conducted at the critical R-ratio for modeling the constant life diagrams, per KAWAI method. In addition, fatigue tests were carried out at different Rratios for examining the Kawai model. The prediction of the Kawai model was slightly conservative for R-ratios
−1 and un-conservative for R-ratio that equals 10. Whereas, the Goodman CLD highly overestimated fatigue life for purely tension and tension-compression areas and underestimated fatigue life for R-ratios lower than R = −1.

Examination of the KAWAI CLD Method

409

A modified model was suggested to overcome the un-conservatism of the Kawai model at the pure compression zone. Thus, the constant life curve is built from two parts; i.e. for UTS
rm
rvm, constant life diagrams are built according to KAWAI method and for UCS  rm  rvm, constant life diagrams are built by a linear line between rva and UCS. The predicted curve of the suggested model was compared to the experimental results and was shown as conservative for fatigue life greater than 105. Hence, with application of the appropriate scatter factor, the use of the modified Kawai model is considered relatively accurate.

References Boerstra, G.K.: The multislope model: a new description for the fatigue strength of glass reinforced plastic. Int. J. Fatigue 29(8), 1571–1576 (2007) FAA Doc: DOT/FAA/AR-10/6: Determining the fatigue life of composites aircraft structures using life and load-enhancement factors, June 2011 Forecast International, Analysis 1 (2013), the market for UAV reconnaissance systems 2013– 2022 Harris. B.: A parametric constant-life model for prediction of the fatigue lives of fibre-reinforced plastics. In: Harris, B. (ed.) Fatigue in Composites, pp. 546–568. Woodhead Publishing Limited, United Kingdom (2003) Kawai, M., Koizumi, M.: Nonlinear constant fatigue life diagrams for carbon/epoxy laminates at room temperature. Compos.: Part A 38(11), 2342–2353 (2007) Kawai, M., Teranuma, T.: A multiaxial fatigue criterion based on the principal constant life diagrams for unidirectional carbon/epoxy laminates. Compos.: Part A 43, 1252–1266 (2012) Philippidis, T.P., Vassilopouloss, A.P.: Life prediction methodology for GFRP laminates under spectrum loading. Compos.: Part A 35(6), 657–666 (2004) Raiter, L.: How to test hybrid aircraft in fatigue. In: 27th ICAF Symposium, Jerusalem, June 2013 Rouchon, J.: Fatigue and damage tolerance evaluation of structures, the composite materials response. In: 25th ICAF Symposium (2009)

Fatigue Crack Growth Prediction and Verification of Aircraft Fuselage Panels with Multiple Site Damage Shaopu Su(&), Jianghai Liao, Wendong Zhang, and Dengke Dong Department of Metallic Structure Strength Research, Avic Aircraft Strength Research Institute, Xi’an 710065, Shaanxi, China {shaopu_su,dengke623}@sina.com, [email protected] Abstract. A numerical algorithm of fatigue crack growth prediction on aircraft fuselage panels with multiple site damage is discussed in this paper based on Finite Element Method. The objective of this research is to realize the automatic crack propagation and crack growth life calculation. The mixed-mode stress state of structures at crack tip is considered through the published form of effective stress intensity factor, and the incremental fatigue crack growth model is studied to coordinate the propagation between multiple cracks. Maximum tangential stress criterion is applied to determine the crack growth direction at a growth step. Moreover, the new crack tip position in flat plates and curved plates are numerically determined to automatically update the crack tip. Fatigue test of curved panel subjected to inner pressure was conducted by horizontal selfbalanced test facility, and the fatigue tests of plates with two internal collinear cracks or with seven collinear cracks under tensile stress were also analyzed to validate the feasibility of algorithm. It is shown that the efficient fatigue crack prediction algorithm is able to predict various crack growth behaviors observed in tests, and the predicted crack propagation path and lives are in a good agreement with test results and data available in the literature. Keywords: Multiple site damage
Fatigue crack growth life Aircraft fuselage panel
Stress intensity factor




1 Introduction Widespread Fatigue Damage (WFD) is recognized as one of the biggest threats on structural integrity, especially for aging aircraft. Because of many structural detail changes in stress levels, local loading and assembly forms in the process of crack growth, its evaluation provides a tough challenge for typical aircraft damage tolerance design and analysis [Jones et al. 2008]. The paper focuses on the analysis of multiple site damage (MSD) in aircraft structures. As one form of WFD, it starts when the fuselage pressure cycling fatigue loads lead to crack initiation and propagation at multiple rivet locations. MSD is characterized by the interaction of a major crack with several short cracks located at various sites of the same structural element. MSD may become critical, when cracks become of sufficient size of density to exceed the residual strength of the structural element [Labeas and Diamantakos 2005]. Three main factors should be considered in MSD problems: crack interference, inner-stress redistribution © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 410–422, 2020. https://doi.org/10.1007/978-3-030-21503-3_32

Fatigue Crack Growth Prediction and Verification of Aircraft

411

and overlap between cracks. Independent cracks growth analysis will induce incorrect results in MSD prediction, and the traditional engineering methods face the difficulties to obtain stress intensity factors and interference effects of multiple cracks. Comparatively numerical simulation methods are favored by aeronautical engineers owing to its independence on crack geometries. Engineers prefer to apply Finite element method to evaluate the airworthiness of metallic structure in Civil Aircraft, yet how to build a numerical modeling to coordinate the growth of multiple cracks and to realize automatic propagation of MSD in aircraft panels is the key point to accurately and efficiently predict MSD life based on commercial software. In this paper, the fatigue cracks growth behaviors of MSD in aircraft panels are studied to understand the phenomenon of MSD and to obtain useful datum for maintaining continued airworthiness. Effective Stress Intensity Factor (SIF) is researched based on the developed earlier formulas to evaluate the stress state of crack tips. According to strong ability of SIF in ABAQUS, the numerical FEM algorithm is investigated to automatically predict incremental crack propagation length, failure mode and life of MSD. Meanwhile, the sealing internal pressure load is applied on the fuselage panel through a horizontal self-balanced test rig, and the MSD in plate and in curved panels are analyzed by tests to validate the numerical prediction through comparison with test results.

2 Description of the Model Fatigue Crack Growth Model. Here Constant amplitude spectrum is only considered in this paper. Paris type equation is applied to analyze the fatigue crack growth rate. Due to interference between multiple cracks, the crack tip may be in the mixed mode stress state. So we express the fatigue crack growth rate as a function of the effective stress intensity factor Keff as following:
n da ¼ c DKeff dN

ð1Þ

Where c,n are material parameters, DKeff is the amplitude of Keff under maximum loading and minimum loading. The key point is how to describe the effective factor in the form of I-type, II-type and III-type SIF (noted as KI, KII and KIII). Some work has been done in this research field. Yan proposed the following MTS formula [Yan et al. 1992] to describe Keff : h h h Keq ¼ KI cos3  3KII cos2 sin 2 2 2 Where h is the crack growth direction.

ð2Þ

412

S. Su et al.

Tanaka proposed a corrected Keff through comparison with experimental results, as shown in Formula (3) [Tanaka 1974]: Keq ¼ ½KI4 þ 8KII4 0:25

ð3Þ

In addition, Tanaka suggests that this criterion can be extended to the combination of three mode loadings. Based on the experimental results of homogeneous and isotropic materials in mixed mode state, Richard proposed a new equivalent SIF depending on KI and KII as following [Richard et al. 2014]: Keq ¼

KI 1 þ 2 2

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi KI2 þ 4ða1 KII Þ2

ð4Þ

Where a1 is the ratio of fracture toughness KIC to KIIC. Meanwhile, most commercial software applies the simple square root formula to consider the coupling effect of KI and KII (see formula (5)). Keq ¼

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 KI2 þ KII2 þ ð1  tÞKIII

ð5Þ

Where v is the passion ratio. For the sake of investigating the feasibility of different forms of Keff in mixed-mode fracture, we use fatigue tests of CT structure under different loading attitudes to analyze the life cycles of fatigue crack deflection. Table 1 is the comparisons of test results and predictions by formulas (2) * (5) regarding single crack of CTS structures propagating certain distances. We records the fatigue lives of cracks in CTS specimen at 15o, 30o and 45o loading angle, and obtained that the prediction errors based on Formula (5), Formula (3), Formula (4) and Formula (2) are respectively 7.13%, 6.59%, 10.52% and more than 200%, Tanaka formula is more adequate to calculate the equivalent SIF for CTS specimens in mixed-mode I-II loading. Consequently, we apply Tanaka formula in Formula (1) to estimate the fatigue lives of mixed-mode fracture and fatigue problems. Table 1. Prediction of fatigue lives of CTS structure at different loading angles Loading angle 15o 30o 45o Relative error

Experimental results

Formula (5) 36820

31459 (38 mm * 43.20 mm) 12575 12316 (38 mm * 39.15 mm) 33203 32444 (38 mm * 43.22 mm) – 7.13%

Formula (3) 36927

Formula (4) 36655

Formula (2) 41075

KI 36934

12780

11556

29177

12954

32953

30903

311942

34259

6.59%

10.52%

–

7.87%

Fatigue Crack Growth Prediction and Verification of Aircraft

413

Based on the effective SIF described as Formula (3), some authors [Ma and Hu 2016] amended the parameters forms and proposed different fatigue crack growth rate models to evaluate the influences of loading mode on fatigue life. In view of fatigue tests results of CTS specimens on 2024-T3 aluminum, the following crack growth model is proposed:  b da ¼ ce dN

      1

KI 2 p arctan KII

ðDKeq Þn

ð6Þ

Where c, n, b are material parameters, and b = 1 when material is 2024-T3 aluminum. Incremental Fatigue Crack Growth Model. Due to different stress state of crack tips in MSD structures, every crack has a different growth rate with same number of cycles, how to determine the crack increment of all cracks is the key point to evaluate the fatigue life. Here we limit the crack growth length Da for MSD structures as Da  Damin , where Damin is a presumed value. Considering the application of contour integrity in numerical simulation, we generally choose Damin  1 mm depending on the geometry of simulated structure. Yet too high value of Damin will induce large prediction error, we should balance between calculation efficiency and accuracy. Firstly we calculate the fatigue lives DN of every cracks tips required to extend Damin , and then set the minimum cycles DNmin as the next incremental cycles, finally the crack increments of every crack tip are obtained at DNmin from DK and DN. The propagation length of each crack is written as following: n

Da ¼ f ðmin DNÞ i¼1

ð7Þ

Where f ðxÞ is the crack growth rate model; n is the number of crack tips. The SIF of structures under constant spectrum loading can be considered as constant in small propagation length; Conversely, K depends on loading value for the structures under random spectrum. In order to decrease the numerical calculation amount, K can be obtained as following: Kj þ 1 ¼

pffiffiffiffi r j þ 1 aj pffiffiffiffi Ki ri a i

ð8Þ

Where rj þ 1 is the current loading stress; ri is the loading stress at step i (when FE crack modelling created); aj is the total crack length after loading stress rj ; ai is total crack length at step i; Ki is SIF at step i. This algorithm can efficiently obtain the SIF at random loading stress. Fatigue Crack Growth Direction. Crack direction is one of major aspects in mixedmode fatigue crack growth. Various criteria for the crack growth direction have been discussed for decades, such as Maximum tangential stress criterion (MTS criterion),

414

S. Su et al.

Minimum strain energy density criterion(S-criterion), J-criterion, Dilatational strain energy density criterion (T-criterion) and so on [Qian and Fatemi 1996]. MTS criterion is applied in this paper to determine the crack growth direction, expressed as: KI sin h þ KII ð3 cos h  1Þ ¼ 0

ð9Þ

Where h is crack deflection angle along the direction of crack extension. Most works have been done to verify this criterion provides satisfactory predictions. Determination of New Crack Tip Position. If the crack growth length and crack path direction at step i are respectively Da and hc , the new crack tip location at step i + 1, written as (xi þ 1;1 ; yi þ 1;1 ) in Cartesian coordinate system, can be obtained as following: For the cracks in plane [SU et al. 2015]: 

xi þ 1;1 ¼ xi;1 þ signðxi;1  xi;2 ÞDa cos hc yi þ 1;1 ¼ yi;1 þ signðxi;1  xi;2 ÞDa sin hc

ð10Þ

For the cracks in curve (see in Fig. 1): 8 hc Þ xi þ 1;1 ¼ xi;1 þ signðxi;1  xi;2 ÞDa cos hc cosðDa cos > R > qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi < L 2 yi þ 1;1 ¼ R2  xi þ 1;1  R cosð Þ R > > : zi þ 1;1 ¼ zi;1 þ signðxi;1  xi;2 ÞDa sin hc

ð11Þ

Where xi;1 ; yi;1 are the coordinates of crack tip at step i; xi;2 is the coordinate of the point with the minimum distance with crack tip; L is half arc length of fuselage curve panel; R is the curvature of panel.

Fig. 1. Description of crack coordinates in curved panel

Fatigue Crack Growth Prediction and Verification of Aircraft

415

3 Material and Specimens As mentioned above, the attention is focused on the structures with MSD on the panel surface. The interference between cracks, the growing paths and lives of cracks on flat and curved panels are interesting for this research. Three types of specimens with MSD were considered in the work. The first type of specimen was a rectangular plate with two internal, parallel, noncollinear cracks (length = 10 mm for both), as shown in Fig. 2. A cyclic tension (rmax ¼ 160 N=mm, rmin ¼ 0) was loaded in a constant spectrum at both ends of plate. The material was chosen as Al 2024-T3 the same as in Reference. The second type of specimen was a rectangular plate (thickness was 4 mm) with seven internal collinear cracks. We set the distances between cracks as 50 mm, and each crack length as 8 mm. Meanwhile, the plate with material Al LY12 was subjected to a constant spectrum tensile loading (Pmax ¼ 82 kN,Pmin ¼ 8:2 kN). The third type of specimen was a curved fuselage panel with curvature radius as 1430 mm subjected on cyclic internal pressure (Pmax = 55000 Pa, R = 0.06). The panel included six frames with a 530 mm spacing and seven stringers with a 162.2 mm spacing. Two internal cracks were pre-set on the skin between 4# stringer and 5# stringer to evaluate the crack growth of MSD on the key skin region. The initial lengths of cracks 1# and 2# were respectively 21 mm and 19 mm. The thickness without frames and stringer joints was 1.1 mm, and the thickness on the other connection part was 1.8 mm. The skin material was Al 2524-T3 (Figs. 3 and 4).

A B

Fig. 2. Plate with two non-collinear cracks

416

S. Su et al.

Fig. 3. Plate with seven collinear cracks

Fig. 4. Curved panel with two internal cracks

4 Experimental Technique In view of the published experimental results in reference [Tu and Cai 1993], we evaluated the crack interference effect of two non-collinear cracks in the first type of specimen. Consequently, the experimental technique research is focused on the fatigue crack growth evaluation of MSD in the second and third types of specimens. For the plate with seven internal collinear cracks, the uniform tensile stress was performed on a 1000 kN Instron8801 hydraulic servo fatigue testing machine, and the crack increment was measured by numerical reading microscope. Consider the little buckling effect of specimen in the test, we did not install anti-buckling device on specimen. Three plates with same types were tested to record the fatigue lives.

Fatigue Crack Growth Prediction and Verification of Aircraft

417

Internal pressure is one of the major loading type to simulate the accurate loading state of airplane fuselage panel. How to suppose the uniform sealing pressure and to simulate the hoop stress created by pressure are the key techniques to fatigue test implement of fuselage panel. A horizontal self-balanced test facility was designed to realize the internal pressure loading for curved panel on Al 2524-T3, as shown in Fig. 5. The facility included pressure box, load regulator, load plates, transverse girder and base frames. The test specimen was bonded with an airbag and was put on the pressure box. The pressure load was supposed by pressurizing the airbag. The load plates and load regulator are to balance the hoop stress created by pressure. The skin force can be approximately adjusted by inner pressure according to thin walled cylinder theory. We measured the crack length along the inner surface of the skin using an optical microscope.

Specimen Load plates Load regulator

Transverse girder Pressure box

Base plate Fig. 5. Horizontal self-balanced test facility

5 Crack Growth Analysis Procedure Based on the secondary code development of ABAQUS software, the progressive failure analysis of MSD in fuselage panel structures are carried out to obtain the failure mode and fatigue lives through automatic crack growth code. The procedure is described in Fig. 6. Combined with contour integral technique and MTS criteria in ABAQUS, the SIF and crack growth direction can be obtained through parametric modeling by PYTHON, then crack growth length can be calculated by crack growth rate mode, the new crack tip is updated according to Formula (10) and (11). The code will automatically remodel and analyze the crack information until the structures arrived to its residual strength.

418

S. Su et al. Input the structure geometry information

Read crack position

Read load spectrum

FEA analysis

Records SIF, COD,T-stress

N

Crack will propagate

Y Calculate crack growth length, and update the crack tip

N

Record the cycle numbers

Unstable growth

Y End

Fig. 6. Diagram of crack growth analysis

6 Results and Discussion Two internal Non-collinear Cracks. The crack propagation path of two non-collinear cracks in 180 mm  90 mm rectangular plate was analyzed by developed code in Fig. 7. Because of the weak interference between two cracks in the initial state, the crack path is straight. With the increase of crack length, the interference effect is clear, the right tip of left crack deflects towards the upper part, as well the left tip of right crack deflects towards the down part, and the elliptical failure mode is formed in the central part. Afterwards, the propagation rate at outside tips of two cracks accelerates, the structure is finally failure. The life of the structure is 6954 cycles, which is in good agreement with the literature [Tu and Cai 1993].

Fatigue Crack Growth Prediction and Verification of Aircraft

Fig. 7. Crack growth path of plate with two internal collinear cracks

2500 KI at A 2000

KII at A KI at B

KI,KII

1500

KII at B

1000

500

0

-500 2000

3000

4000 5000 Number of cycles Fig. 8. Life-SIF curves

6000

7000

419

420

S. Su et al.

Figure 8 is the evolution of the SIFs at crack tip A and B. Clearly there is a pure mode I crack propagation problem in the beginning, the value of KII can be negligible. With the interference influence of interior crack tips, the value of KII at B increases as KI increases, when two values arrives the maximum, the interference effect between cracks is strongest. Yet the crack tips at the edges have little mixed-mode state. Plate with Seven Internal Collinear Cracks. Considering the symmetry of structures, the FEA model was built as Fig. 9, the crack growth rate parameters was chosen pffiffiffiffiffiffiffi C = 1:41254  1014 mm=ðMPa
mmÞ3:16 , n = 3.16 as in reference [Sun 2018]. We imagined the cracks propagating symmetrically along the midline of structure in the numerical simulation. Three specimens were tested to validate the analysis model. For simulating the prediction accurately, we chose Damin ¼ 0:5 mm in the crack growth analysis. Figure 10 shows the a-N curves of every crack tip. The experimental results of fatigue lives of plates are respectively 255385 cycles, 236536 cycles and 244704 cycles, and their failure modes are similar: crack propagates directly until cracks linking together, the structure is failure. The prediction result is 250395 cycles, presenting a good correspondence with the average life of test results (245522 cycles).

Fig. 9. FEA model of structures with collinear cracks

Fatigue Crack Growth Prediction and Verification of Aircraft

421

Fig. 10. Evolution of crack length vers number of cycles

Curved Panel with Two Collinear Cracks. According to the design of horizontal self-balanced test facility and the geometry of test specimen, FEM model was built and Fig. 11 shows the stress distribution of the structure without crack subjected to pressure. Uniform pressure can be obtained through this facility. After adding two cracks into FEA model of structure, the strain along hoop direction on top of skin region with cracks is approximately 828 microstrain, which is in good agreement with experimental results (843 microstrain). The crack growth rate parameters were chosen as

Fig. 11. Stress distribution of curved panel subjected to inner pressure

422

S. Su et al.

pffiffiffiffiffiffiffi c = 106:662 mm=ðMPa
mmÞ2:575 ,n = 2.575 [Guo 2008]. The prediction of fatigue life of structure is 97000 cycles, agreeing well with test results (95078 cycles).

7 Conclusion This paper presents a numerical algorithm of fatigue crack growth prediction on aircraft panels with MSD based on FEA in ABAQUS. Its function of automatic crack propagation and crack growth life calculation facilities engineers to evaluate the WFD problem in aging aircraft. In addition, a series of fatigue tests of aircraft fuselage panels have been conducted to validate the numerical simulation. Key findings of this research can be summarized as follows: • The effective SIF based on Tanaka formula is more adequate to present the mixedmode stress state at crack tip. • The designed numerical algorithm can efficiently and accurately predict the crack growth behavior of MSD problems. • The interference effects between multiple cracks play an important role in the crack growth path and life prediction of MSD problems. • The experimental results present that horizontal self-balanced test facility can be used to imposed the uniform inner pressure on aircraft fuselage panels.

References Guo, J.: Fatigue properties of 2X24 alloy sheets with high strength and high toughness. Master Thesis. Central South University (2008) Jones, R., Molent, L., Pitt, S.: Understanding crack growth in fuselage lap joints. Theoret. Appl. Fract. Mech. 49, 38–50 (2008) Labeas, G., Diamantakos, J.: Analytical prediction of crack coalesce in multiple site damaged structures. Int. J. Fract. 134, 161–174 (2005) Ma, S., Hu, H.: The mixed-mode propagation of fatigue crack in CTS specimen. Chin. J. Theoret. Appl. Mech. 38(5), 698–704 (2016) Qian, J., Fatemi, A.: Mixed mode fatigue crack growth: a literature survey. Eng. Fract. Mech. 55 (6), 969–990 (1996) Richard, H.A., Schramma, B., Schirmeisen, N.H.: Cracks on mixed mode loading – theories, experiments, simulations. Int. J. Fatigue 62, 93–103 (2014) Shaopu, S.U., Dengke, Dong, Zhang, Haiying.: Fatigue crack growth analysis based on abaqus/python. Sci. Technol. Eng. 15(27), 1671–1815 (2015) Sun, H.: Study on probability damage tolerance analysis method for multiple site damages in structures containing holes. Master Thesis. AVIC Aircraft Strength Research Institute (2018) Keisuke, T.: Fatigue crack propagation from a crack inclined to the cyclic tensile axis. Eng. Fract. Mech. 6, 493–507 (1974) Tu, S.T., Cai, R.Y.: A coupling of boundary elements and singular integral equation for the solution of the fatigue cracked body. Stress Anal. 239–247 (1993) Yan, X., Du, S., Zhang, Z.: Mixed mode fatigue crack growth prediction in biaxially stretched sheets. Eng. Fract. Mech. 43, 471–475 (1992)

Fatigue Life Prediction of CFRP Laminate Under Quasi-Random Loading Vitaly E. Strizhius(&) Fatigue Strength Department, JSC “AeroComposite”, 27 Polikarpova Str., Bld. 3, Moscow 125284, Russia [email protected]

Abstract. The results of fatigue test and fatigue life predictions of CFRP T300/5208 [45/0/-45/90]2s specimens with open holes under loading of quasirandom “TWIST” program with different levels of truncation of large and small loads were considered. It was noted that the fatigue life predictions made using the Palmgren-Miner rule showed an unacceptable accuracy of the results. In order to increase the accuracy of such predictions, two new prediction methods are proposed, formed by using two different non-liner fatigue damage accumulation models. Results of fatigue life predictions for considered specimens with use of new methods are presented. The results of the predictions are compared with the experimental data. Conclusions are made about the accuracy of predictions using the proposed methods. Keywords: Fatigue life predictions
CFRP laminate Specimens with open holes
Quasi-random loading




1 Introduction It is well known that the possibility of control the physical and mechanical characteristics of layered composites, including their fatigue resistance characteristics, is of paramount significance. However, reliable prediction methods for evaluating these characteristics are absent up to now, which is a serious obstacle to the implementation of layered composites in the manufacture of structural elements, in particular, in the manufacture of various elements of aircraft composite structures. A distinct problem is the lack of reliable calculation methods for fatigue life prediction of layered composites under complex fatigue loading spectra simulating a realistic flight loading of structural members of present-day aircraft. This type of programmed loading is represented, for example, by the well-known standardized quasi-random “TWIST” (Transport WIng Standard Test) program (De Jonge et al. 1973), which imitates the flight loading of the transport aircraft wing. The results of fatigue test of CFRP T300/5208 [45/0/-45/90]2s specimens with open holes under uniaxial loading of quasi-random “TWIST” program with different levels of truncation of large and small loads (Phillips 1981) are considered in this paper. The results of fatigue life predictions of the specimens under consideration performed with using the Palmgren-Miner rule are also considered. The results of the predictions are compared with the experimental data (Phillips 1981). It was noted that © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 423–431, 2020. https://doi.org/10.1007/978-3-030-21503-3_33

424

V. E. Strizhius

the fatigue life predictions made using the Palmgren-Miner rule showed an unacceptable accuracy of the results. In order to increase the accuracy of such predictions, two new prediction methods formed by using two different non-liner fatigue damage accumulation models are proposed. Results of fatigue life predictions for considered specimens with use of new methods are presented.

2 Results of Fatigue Tests As already noted above, in this paper the results of fatigue test of CFRP T300/5208 [45/0/-45/90]2s specimens with open holes under uniaxial loading of quasi-random “TWIST” program with different levels of truncation of large and small loads are considered. Frequency of occurrence of flight types and load cycles within each flight of “TWIST” program are given in Table 1 (De Jonge et al. 1973). Table 1. Frequency of occurrence of flight types and load cycles within each flight of “TWIST” program. Flight Frequency in one block Frequency of occurrence of flight cycles at the ten load levels type I II III IV V VI VII VIII IX X 1.60a 1.50 1.30 1.15 0.995 0.84 0.685 0.53 0.375 0.222 A 1 1 B 1 C 3 D 9 E 24 F 60 G 181 H 420 I 1090 J 2211 Total number of load cycles per 0 block Total number of load cycles per 0 block a

1 1

1 1 1

4 2 1 1

8 5 2 1 1

18 11 7 2 1 1

64 39 22 14 6 3 1

112 76 61 44 24 19 7 1

391 366 277 208 165 115 70 16 1

900 899 879 680 603 512 412 233 69 25 800 4170 34800 358665

0

5

18

52

152

0

5

23

75

227 1037 5197 39997 398662

Ratio of alternating load to the flight mean load.

All ground loads represented by a single load event equal to minus one-half the flight mean load. Ten flight types (severities) - each characterized by the number of load levels involved and the number of occurrences of each level. Load sequence repeats after 4000 flights; that is, the block length is 4000 flights. The average values of fatigue lives (Ntest median ) obtained in fatigue test of considered specimens by using baseline spectrum “TWIST” and truncated spectra are given in Table 2 (Phillips 1981). Flight mean gross-section stress rm ¼ 111 MPa for baseline spectrum and truncated spectra.

Fatigue Life Prediction of CFRP

425

Table 2. Results of the spectrum loading fatigue tests. Spectrum truncations None (baseline spectrum) Load level # X omitted Load levels # X–IX omitted Load levels # VII–X omitted Load levels # I–II omitted

Ntest median , flights 88655 75955 106555 111555 189962

3 Fatigue Life Prediction Methods For the purpose of the fatigue life prediction of the considered specimens, three fatigue life prediction methods were used. Fatigue Life Prediction Method #1. According to the data presented by Phillips (1981), we can conclude that for the purpose of the fatigue life prediction of the considered specimens he used the fatigue life prediction method #1. The main features of this method are presented in Table 3. Table 3. Main features of fatigue life prediction method #1. Main features Cycle counting Equation of basic S-N curve Constant life diagram Damage accumulation model

Using equations and models “Rainflow” counting method raðR ¼ 1Þ ¼ a þ b lg N See Fig. 1 P Dblock ¼ Nnii

Fatigue life prediction equation Npred ¼ ð1=Dblock Þ
NPB

Necessary notes to Table 3: 1. As the equation of the basic S-N curve the following equation was used: raðR¼1Þ ¼ a þ b lg N;

ð1Þ

where raðR¼1Þ is amplitude of cyclic stress (R = −1). 2. As the constant life diagram experimental data, which are presented in Fig. 1, were used (Phillips 1981). 3. As a model of the damage accumulation the well-known Palmgren-Miner rule was applied: Dblock ¼

X ni Ni

:

426

V. E. Strizhius

Fig. 1. Constant life diagram constructed from the results of the constant amplitude tests

4. Accumulated damage at the time of fatigue failure of the considered specimens: D ¼ Nblock  Dblock ¼ 1; where Nblock is fatigue life of considered specimens in the number of blocks of the “TWIST” program. Hence: Nblock ¼ 1=Dblock 5. Fatigue life of considered specimens in the number of flights is determined as Npred ¼ Nblock  NPB; where NPB = 4000 flights - the size of the “TWIST” program block. Fatigue Life Prediction Method #2. The main features of the fatigue life prediction method #2 are presented in Table 4. Necessary notes to Table 4: 1. The modification of equation of Beheshty - Harris - Adam (Beheshty et al. 1999; Harris et al. 1997; Harris 2003) is proposed as the equation of constant life diagram (CLD): raðR¼1Þ ¼

ra i
ruUTS
jrUCS jv ; ðrUTS  rm i Þu
ðjrUCS j þ rm i Þv

Fatigue Life Prediction of CFRP

427

Table 4. Main features of fatigue life prediction method #2. Main features Cycle counting Equation of basic S-N curve Constant life diagram

Using equations and models “Rainflow” counting method raðR¼1Þ ¼ a þ b lg N jv

r
ru
jr

Damage accumulation model

UCS rað1Þ ¼ ðrUTS ra mi i ÞUTS u
ðjrUCS j þ rm i Þv     2  k P A Nnii þ B Nnii Dblock ¼

i¼1

Fatigue life prediction equation Npred ¼ ð1=Dblock Þ
NPB

where raðR¼1Þ is amplitude of equivalent stress (R = −1); i is cycle type; ra i is amplitude of stress; rm i is mean stress; rUTS is ultimate tensile strength; rUCS is ultimate compressive strength; u = 2.18; m = 2.40 - fitting parameters (Beheshty et al. 1999). 2. As a model of the damage accumulation the Howe and Owen model (Howe and Owen 1972) was applied: Dblock ¼

k X i¼1

"    2 # ni ni A ; þB Ni Ni

ð2Þ

where A and B are parameters which values don’t depend on the loading level. Values of the parameters A and B of Eq. (2) should be determined using a linear regression analysis of the experimental data of Table 2. 3. The remaining provisions of the prediction method #2 are completely analogous to the corresponding provisions of the prediction method #1. Fatigue Life Prediction Method #3. The main features of the fatigue life prediction method #3 are presented in Table 5. Necessary notes to Table 5: Table 5. Main features of fatigue life prediction method #3. Main features Cycle counting Equation of basic S-N curve Constant life diagram Damage accumulation model

Using equations and models “Rainflow” counting method raðR¼1Þ ¼ a þ b lg N r
ru
jr

jv

UCS rað1Þ ¼ ðrUTS ra mi i ÞUTS u
ðjrUCS j þ rm i Þv  ci i k k h P P Di ¼ Ai  Nnii Dblock ¼

i¼1

i¼1

Fatigue life prediction equation Npred ¼ ð1=Dblock Þ
NPB

428

V. E. Strizhius

1. Special fatigue damage accumulation model (Strizhius 2016) is proposed as a model of fatigue damage accumulation: Dblock ¼

k X i¼1

Di ¼

 ci  k  X ni Ai  ; Ni i¼1

ð3Þ

where Di is the fatigue damage of CFRP laminate corresponding to the ith load level; ni is the number of applied loading cycles corresponding to the ith load level; Ni is the number of cycles to failure corresponding to the ith load level; Ai and ci are parameters which values depend on the loading level. The values of parameters Ai and ci should be determined using an iterative procedure in order to fit Eq. (3) to the experimental data of Table 2. 2. The remaining provisions of the prediction method #3 are completely analogous to the corresponding provisions of the prediction method #1.

4 Prediction Results Strength characteristics of the specimens, which are necessary to perform fatigue life estimates of CFRP laminate (Phillips 1981): • rUTS = 346.06 MPa; • rUCS = −324.72 MPa; • raðR¼1Þ = 320.56-27.4172 lgN - equation of type (1). Tables 6 and 7 shows simplified sequences of counted cycles extracted from loading of the quasi-random program “TWIST” using the “rainflow” counting method. It is assumed that such sequences can be used for engineering fatigue life predictions. In Tables 6 and 7 the following symbols are used: Table 6. Counted cycles of “TWIST” program (air stage). Loading level Loading amplitude n, cycles I 1:6  rm 0 II 1:5  rm 1 III 1:3  rm 2 IV 1:15  rm 9 V 0:99  rm 28 VI 0:84  rm 92 VII 0:68  rm 619 VIII 0:53  rm 3750 IX 0:37  rm 33710 X 0:22  rm 356454

m r – −0.34 −0.34 −0.34 −0.34 −0.34 −0.34 −0.34 −0.34 −0.34

a r – 0.51 0.442 0.391 0.3383 0.2856 0.2329 0.1802 0.1275 0.07548

min r – −0.85 −0.782 −0.731 −0.6783 −0.6256 −0.5729 −0.5202 −0.4675 −0.4156

Fatigue Life Prediction of CFRP

429

Table 7. Counted cycles of “TWIST” Program (“GAG” cycles).  min Type of flight Number of “GAG” cycles r A 1 −0.884 B 1 −0.85 C 3 −0.782 D 9 −0.731 E 24 −0.6783 F 60 −0.6256 G 181 −0.5729 H 420 −0.5202 I 1090 −0.4675 J 2211 −0.4156

max r 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17

Table 8. Values of the Parameters Ai and ci of Eq. (3). min Loading level r I −0.884 II −0.85 III −0.782 IV −0.731 V −0.6783 VI −0.6256 VII −0.5729 VIII −0.5202 IX −0.4675 X −0.4156

• • • •

Ai 1.05 1.05 1.05 1.05 0.9 0.9 0.9 0.9 0.9 0.9

ci 0.8 0.8 0.8 0.8 1.0 1.0 1.0 1.0 1.0 1.0

m ¼ rm =jrUCS j; r a ¼ ra =jrUCS j; r min ¼ rmin =jrUCS j; r max ¼ rmax =jrUCS j. r

Values of the parameters A and B of Eq. (2) of the Howe and Owen model (Howe and Owen 1972), obtained using a linear regression analysis of the experimental data of Table 2, are: A = 0.846; B = 914.949. Values of the parameters Ai and ci of Eq. (3), obtained an iterative procedure aiming to fit Eq. (3) to the experimental data of Table 2, are presented in Table 8. The results of fatigue lives prediction Npred are presented in Table 9. Figure 2 shows a comparison of the obtained values of fatigue lives predictions and experimental data. First of all Fig. 2 shows that fatigue life predictions, using Palmgren-Miner rule, showed the unacceptable accuracy of the estimations.

430

V. E. Strizhius Table 9. The results of fatigue lives prediction Npred .

Spectrum Truncations

Ntest median , flights

None (baseline spectrum) 88655 Load level # X omitted 75955 Load levels # X–IX omitted 106555 Load levels # VII–X omitted 111555 Load levels # I–II omitted 189962 (*) # of fatigue life prediction method

Npred , flights (1)* 1533730 1503910 1140140 903600 2374530

Npred , flights (2)* 88372 90189 93334 103708 209608

Npred , flights (3)* 87600 88720 90280 94760 204600

Fig. 2. A comparison of the obtained values of fatigue lives predictions and experimental data

5 Discussion Based on the analysis of the data presented in Table 9 and Fig. 2, a number of important conclusions can be drawn. 1. The experimental data presented in Table 9 and Fig. 2 show a significant increase in the fatigue life of the specimens under consideration when truncating, especially the high loading levels. This conclusion is extremely important in the development of quasi-random load programs for composite wings of modern transport aircraft. 2. Fatigue life predictions of the specimens under consideration, using PalmgrenMiner rule, showed the unacceptable accuracy of the estimations, with all prediction results showing an overestimated fatigue life of the specimens (in comparison with the experimental values). It should be specially noted that the predictions show the greatest error for cases of truncation of small loading levels.

Fatigue Life Prediction of CFRP

431

3. The accuracy of the fatigue life estimations of the specimens under consideration, obtained using methods # 2-3, can be considered acceptable.

6 Conclusion It was noted that the fatigue life predictions of CFRP T300/5208 [45/0/-45/90]2s specimens with open holes under uniaxial loading of quasi-random “TWIST” program with different levels of truncation of large and small loads made using the PalmgrenMiner rule showed an unacceptable accuracy of the results. In order to increase the accuracy of such predictions, two new prediction methods are proposed, formed by using two different non-liner fatigue damage accumulation models. Results of fatigue life predictions for considered specimens with use of new methods are presented. The results of the predictions are compared with the experimental data. Conclusions are made about the accuracy of predictions using the proposed methods.

References Beheshty, M.H., Harris, B., Adam, T.: An empirical fatigue-life model for high-performance fiber composites with and without impact damage. Compos. A: Appl. Sci. & Manuf. A30, 971–987 (1999) De Jonge, J.B., Schutz, D., Lowak, H., Schijve, J.: A standardized load sequence for flight simulation tests on transport aircraft wing structures. In: LBF Bericht FB-106 (NLR 73029U) (1973) Harris, B., Gathercole, N., Lee, J.A., Reiter, H., Adam, T.: Life prediction for constant-stress fatigue in carbon-fiber composites. Phil. Trans. Roy. Soc. (Lond) A355, 1259–1294 (1997) Harris, B.: A parametric constant-life model for prediction of the fatigue lives of fiber-reinforced plastics (Chap. 20). In: Harris, B. (ed.) Fatigue in Composites, pp. 546–567. Woodhead Publishing Ltd and CRC Press LLC, Cambridge (2003) Howe, R.J., Owen, M.J.: Accumulation of damage in glass-reinforced plastic under tensile and fatigue loading. In: Proceedings of the Eighth International Reinforced Plastics Congress, pp. 137–148. British Plastic Federation, London (1972) Phillips, E.P.: Effects of truncation of a predominantly compression load spectrum on the life of a notched graphite/epoxy laminate. In: Lauraitis, K.N. (ed.) Fatigue of Fibrous Composite Materials, pp. 197–212. ASTM STP 723, San Francisco (1981) Strizhius, V.: Fatigue damage accumulation under quasi-random loading of composite airframe elements. Mechanics of Composite Materials 52(4), 645–664 (2016)

Fatigue Life Simulation and Experiment of 2024 Aluminum Joints with Multi-Fasteners Interference-Fit Qingyun Zhao1(&), Yunliang Wang2, Hong Huang1, Sirui Cheng1, and Fenglei Liu1 1

AVIC Manufacturing Technology Institute, P.O. Box 340, Beijing, China zhaoqy_1997@163, hhong625@163, [email protected], [email protected] 2 Naval Aeronautical and Astronatical University, 188#, Erma Road, Yaitai, Shandong, China [email protected]

Abstract. A three-dimensional finite element model of 2024 aluminum hi-bolted joint was established. Hi-bolts installation and pre-tightening force applied were simulated, then distal alternating load on middle plate of the joint. Fatigue life was predicted with FE-SAFE soft, and verified by experiment. The simulation results show that middle plate is the weakest plate in the plates of the joint. Hoop residual compressive stress increases with raising interference, and decreases gradually from the hole to the outer edge. Fatigue life is longer with interference fit than clearance, and it is benefit for fatigue life improvement with interference from 0.08 to 0.14 mm. While fatigue test results show that failure origins from No. 2 hole of the middle plate. The fatigue life grow longer with interference from 0 to 0.11 mm, then shorten with 0.14 mm interference. The fatigue life is longest with 0.11 mm interference. The suggested interference for engineering is 0.08–0.11 mm for possible defects with 0.14 mm interference. Keywords: Hi-bolt
Interference-fit
Bolt joining
2024-T351 aluminum alloy
Pre-tightening force
Residual stress
Fatigue life

1 Introduction With international airline competition intensifying, requirement for aeroproducts is stringent. Structural joining is inevitable in the aircraft manufacturing due to the needs of design and maintenance. Mechanical joining technology, a technological methods of connecting the components into one piece, plays an important role in aerospace industry. Bolt joining is one of the most important joining methods in Mechanical joining technology, and is used widely with the features of simple structure, reliable connection and easy disassembly and assembly. Hi-bolt joining has obvious advantages compared with ordinary bolts, 39% weight reduction. High strength titanium hibolts are usually used in the assembling of bearing structure, such as aircraft wing and fuselage. On the one hand, structural weight can be reduced. On the other hand, interference fit joining can improve structural fatigue strength. © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 432–443, 2020. https://doi.org/10.1007/978-3-030-21503-3_34

Fatigue Life Simulation and Experiment

433

The relation of interference and fatigue life improvement can be gotten by test, but test data is dispersible with high cost due to assembly factor, hardly to get fatigue-life enhancement mechanism and influence rule. With the development of computer and finite-element technology, many researchers conduct bolt joining study by FEM. A finite element model of the anti-rust bolts with oil reservoir and helical groove was developed and its strength was analyzed with Patran/Marc by Guanlin Han [1]. A 3D model of composite laminates with bolted joint was created by using the software ABAQUS by Tong Liu [2]. Considering the tightening torque, the failure process and the linking strength of single-lap, single-bolt joints were calculated. Finite element simulation and experimental were conduct to find how the different preload affect the strain distribution in high strength bolt gusset plate by Zhiwen Lan [3]. The improved Hashin, the 3D failure criteria, was also introduced, and a new degradation model called Partition Degradation Model was put forward by Jianing Wang [4]. By using this method, the failure situation of single-lap laminated plates with the connection of raised head bolt and countersunk head bolt were calculated separately in the condition of existence of pre-tightening force. For titanium alloy M8 hi-lock bolts, squeezing installation test was conducted and the process of installation was simulated by Jiefeng Jiang [5, 6]. Several process parameters were analyzed by comparing squeezing force and protuberance sizes. The stress distribution around the hole was studied in single lap joint with different interference value by Zhen Yuan [7]. The strengthening mechanism of interference fit was researched and the theoretical basis of compound strengthening was put forward by Qingliang Zhang [8], which guides the application of related technology. T.N. Chakherlou, et al. [9, 10] have studied the effect of bolt interference fit on the fatigue life of lap joints, revealed that the fatigue life were improved by increasing the clamping force, and explained the trends which were observed in the experimentally obtained S–N curve behaviour. Generally theoretical and experimental research are scattered about bolts joining at present. System study is lack for stress distribution around hole under different interference. In this paper, FEM model for 4 bolts double lap joints was established. The effect of interference on stress around hole was simulated during installation, and distribution regularity of residual stress can be gotten. Then, stress was replace into FE-safe software and fatigue life was analyzed. Fatigue tests were carried out and fatigue life was obtained. Compared simulation and test results, results of fatigue life improvement was estimated. The best interference valve was suggested and theoretical guidance was provided for interference joining application.

2 Experimental Materials and Methods Hi-bolt is made of Ti-6Al-4V alloy with elasticity modulus E = 1.138  105 GPa and poisson ratio 0.3, whose surface is coated with aluminum coating. Material of joint plates is 2024-T351 aluminum alloy with elasticity modulus E = 7.924  104 GPa and poisson ratio 0.33, whose chemical composition is shown in Table 1. Tension properties is given in Table 2, and true stress-strain curve in Fig. 1. Hi-bolt shank diameter

434

Q. Zhao et al.

is 7.925 mm. Interference fit can be gotten by controlling hole’s diameter, and corresponding relation of hole’s dimension and interference is shown in Table 3. The samples are double lap joints with 4 bolts and 4 plates and symmetry structure, showed in Fig. 2. Table 1. Composition of 2024-T351 aluminium alloy (%, mass fraction). Cu Mn Mg Cr Zn Ti Fe Si 4.19 0.58 1.42 0.011 0.111 0.023 0.20 0.10

Table 2. Tension properties of 2024-T351 aluminium alloy. Rm/MPa Rp0.2/MPa A/% 471.1 364.0 18.3

600

Stress (MPa)

500 400 300 200 100 0 0.00

0.05

0.10

0.15

0.20

Strain (mm/mm)

Fig. 1. True stress-strain curves of 2024-T351 alloy.

Unit: mm 6 149 20

4 4

32

5

90 162 36

Fig. 2. Diagram of test sample with hi-bolts interference-fit.

Fatigue Life Simulation and Experiment

435

Table 3. Hole diameter with different interferences. No. 1 2 3 4 5

Hole diameter/mm 7.925 7.875 7.845 7.815 7.785

Interference/mm 0 0.05 0.08 0.11 0.14

3 Simulation Analysis of Installation Establishment of Finite Element Models. For symmetry of joints, boundary conditions and load, 1/4 joints were modeled and calculated as showed in Fig. 3. Due to complicated stress around hole, mesh was refined to guarantee simulation accuracy, at the same time reduce calculation and enhance computational efficiency. Three plates in the model were named upper plate, middle plate and lower plate. The hi-bolt neighbouring symmetry was No. 1 bolt and the hole was No. 1 hole. Away from the symmetry the hi-bolt was No. 2 and the hole was No. 2. Model was mesh generated by hexahedra element C3D8I and partitioned. Friction factor between bolt shank and hole was 0.15, and 0.40 between plates, bolt’s bearing surface of head and plates, and nut and plates.

Fig. 3. Finite element analysis model.

Simulation analysis of hi-bolt installation was carried out in two stages. First, hibolts displace downward until bearing surface of head touching upper plate, and touching was defined between bolt’s bearing surface of head and plates, and nut and plates. Then, thread joining was simplified into bind constraint between bolt and nut. At last, clamping force was imposed. Therefore, two finite element models and corresponding two stages analysis were finished. Data of stress and strain was transferred by predefined field between two stages.

436

Q. Zhao et al.

Stress Distribution When Installation. Cloud image can indicate stress distribution in model, and decide the dangerous position quickly. Equivalent stress distribution was given in Fig. 4 with 0.14 mm interference, after 4 KN clamping force was applied. Equivalent stress in the two holes of the middle plates was bigger than that in upper and lower plates, from which it can be concluded that the middle plate was the weakest position of fatigue.

Fig. 4. Equivalent stress diagram. (a) upper plate; (b) middle plate; (c) lower plate

After hi-bolt installation and 4 KN clamping force applied, hoop stress rh can be drawn as Fig. 5. From Fig. 5, it can be seen that hoop stress distribution in Path-U-1 of No. 1 hole and Path-U-4 of No. 2 hole of upper plate, was roughly the same. The same trend happened in Path-L-2 of No. 1 hole and Path-L-5 of hole of lower plate. Compared stress distribution in No.1 hole of upper plate and lower plate, upper plate was

Fatigue Life Simulation and Experiment

437

compressed due to interference fit at the beginning and stress increase gradually with raising interference. While stress was tensile stress around No. 1 hole of lower plate when interference was small, and tensile stress turn to compressive stress when interference became bigger. The stress distribution difference was due to the different constraints from entering to exiting.

150

(a)

0 -50

hole diameter7.925mm hole diameter7.875mm hole diameter7.845mm hole diameter7.815mm hole diameter7.785mm

-100 -150 -200 -250

100 50

hoop stress [MPa]

hoop stress [MPa]

150

(b)

100 50

hole diameter7.925mm hole diameter7.875mm hole diameter7.845mm hole diameter7.815mm hole diameter7.785mm

-100 -150 -200 -250 -300 -350

Path-U-1

-300 -350

0 -50

Path-U-4

-400

-400

0

0

1

2

3

4

5

6

7

8

1

2

distance from hole edge [mm]

(d) hoop stress [MPa]

hoop stress [MPa]

200

100 0 -100

hole diameter7.925mm hole diameter7.875mm hole diameter7.845mm hole diameter7.815mm hole diameter7.785mm

-200 -300

5

6

7

8

9

300

Path-L-2

200

4

distance from hole edge [mm]

300

(c)

3

9

Path-L-5

100 0 hole diameter7.925mm hole diameter7.875mm hole diameter7.845mm hole diameter7.815mm hole diameter7.785mm

-100 -200 -300 -400

-400 0

1

2

3

4

5

6

7

distance from hole edge [mm]

8

9

0

1

2

3

4

5

6

7

8

9

distance from hole edge [mm]

Fig. 5. Hoop stress after clamping force applied. (a) along Path-U-1; (b) along Path-U-4; (c) along Path-L-2; (d) along Path-L-5

4 Fatigue Life Fatigue Life Analysis Process. The basic step of fatigue analysis was given in Fig. 6. At first, stress and strain results of hi-bolt installation were obtained from postprocessor, then load and cyclic material properties were defined in postprocessor, following life was calculated for every load according fatigue criteria, at last determine whether the damage begins according to damage theory.

438

Q. Zhao et al. Creat component finite element model

Material characteristics

Load history

Stress-strain analysis Fatigue damage analysis

Fatigue life

Fig. 6. Fatigue life estimation process by FEM.

Fatigue life calculation process by fe-safe was given in Fig. 7. When calculating, many factors were considered, such as mean stress, stress concentration, notch sensitivity factor, initial stress, surface smooth finish. Many mean stress modifying methods in fa-safe was adopted, which can judge dangerous position in the joints.

Fatigue performance parameters of materials

Fatigue load spectrum

Fe-safe

Stress

Finite element analysis

Finite element modeling

Results of Fatigue analysis

Fig. 7. Analysis process by fa-safe.

Load Application. By implicit dynamics in abaqus, stress and strain distribution and change rule were calculated when distal alternating load shown in Table 4, applied to hi-bolt joints. Table 4. Load of fatigue test. Maximum load Maximum stress Stress ratio Load frequency f = 10 Hz Pmax = 44.4 kN rmax = 205.56 MPa R = 0.1

55% Static load, i.e. Pmax = 44.40 KN, was defined as the maximum fatigue load. Load application was divided into two steps. Firstly, in t ¼ 0:05 s, load-remote for

Fatigue Life Simulation and Experiment

439

middle plate was smoothly increased from 0 to rmax ¼ 205:56 MPa. Secondly, for one load cycle t ¼ 0:1 s, load for middle plate was changed periodically. 4.1

Fatige Life Simulation Results

Stress Distribution in Fatigue Test. Stress and strain can be gained after hi-bolt joints loading was calculated by implicit dynamics. For 5 interference shown in Table 3, longitudinal stress rx distribution of middle plate was given in Fig. 8 when load reached to rmax ¼ 205:56 MPa. It can be seen that the largest stress concentration was in No. 2 hole of the middle plates.

Fig. 8. Longitudinal stress distribution of middle plate with max. load. (a) interference 0; (b) interference 0.05; (c) interference 0.08; (d) interference 0.11; (e) interference 0.14

440

Q. Zhao et al.

Fatigue Damage Prediction. For hi-bolt interference fit joints with 5 interference value, cloud image of fatigue life predicted by fe-safe was shown in Fig. 9, in which fatigue life was logarithmic life, i.e. lgNf. It can be seen that No. 2 hole in the middle plates was the weakest positon and fatigue life from simulation was given in Table 5.

Fig. 9. Cloud image of fatigue life. (a) interference 0; (b) interference 0.05; (c) interference 0.08; (d) interference 0.11; (e) interference 0.14 Table 5. Fatigue life by FEM. Interference/mm 0 0.05 0.08 0.11 0.14 Fatigue life 3502.55 3751.78 4991.46 24075.04 20260.22

Fatigue Life Simulation and Experiment

441

Simulation results show fatigue life with hi-bolt interference fit was longer than with clearance fit, which demonstrates interference fit contribute to prolong fatigue life of joints, the best results with 0.11 mm interference. Fatigue life declined with 0.14 mm interference. Fatigue Test Results. The fatigue test equipment was 200 KN MTS810 hydraulic servo fatigue testing machine. Static tensile failure test was carried out at the load application rate of 2 KN/s and the average failure load was obtained Pb = 113.23 KN. Then fatigue tests were carried out with Pmax = 62.15 KN, Pm = 34.18 KN,△P = 27.97 KN,f = 10 Hz. Fatigue failure specimen were given in Fig. 10, and failure was from No. 2 hole in the middle plates for the joints in Fig. 2. That coincides with the highest stress concentration in No. 2 hole from simulation as showed in Fig. 8. Subjected to cyclic load, usually fine microcracks origin from the position of the stress concentration and then extend gradually and at last fracture occures.

Fig. 10. Fatigue failure specimen.

The curve of interference and fatigue life gained from fatigue tests, were shown in Fig. 11. It can be seen that fatigue life with hi-bolt interference fit at first increase and then decrease, higher than that with clearance fit when interference is 0.08. Fatigue life reaches to the maximum value with 0.11 mm interference and then decrease. With 0.11 mm interference, fatigue life has been raised 6 times compared with clearance fit. Fatigue test is different from simulation results in data, but the general trend of fatigue life changing with interference is the same. Fatigue life from simulation and tests all increase with raising interference and reach to the maximum when interference is 0.11 mm, while decrease with 0.14 mm interference. Large interference can cause severe plastic deformation around holes and initial defect is introduced, which reduce fatigue life. If no defects, specimen need to be polished for large extrusion, which causes cost increase. Considering engineering application and fatigue life enhancement effect, interference from 0.08 to 0.11 mm is suggested.

442

Q. Zhao et al. 140000 25000 result from simulation result from experiment

120000

100000

80000

15000

60000 10000

Fatigue life (times)

Fatigue life (times)

20000

40000 5000 20000

0 -0.02

0 0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

Interference (mm)

Fig. 11. Fatigue life comparison with simulation and test.

5 Conclusion In this research, finite element simulation analysis, fatigue life prediction and fatigue test were carried out for 2024 aluminum alloy joints, and the following conclusions can be drawn. (1) Hi-bolts installation simulation results show that equivalent stress in middle plates is greater than in upper plates and lower plates. (2) Fatigue simulation results show that stress concentration of No. 2 hole in the middle plate is the most serious. Fatigue life from interference joints is longer than that from clearance joints. Interference 0.08 to 0.14 mm is benefit for enhancement of fatigue life. (3) Fatigue failure position of No. 2 hole in the middle plates is verified by fatigue tests. Fatigue tests show that fatigue life first increase, then decrease with hi-bolt interference joining, and reach to the maximum value with 0.11 mm interference. It is suggested that interference 0.08 to 0.11 mm is used in engineering for possible defects with 0.14 mm interference. Acknowledgements. Project group members, senior engineer Yufeng WANG, Professor Huadong LIU,should be thanked, for their work in this project.

References 1. Han, G.L., Li, C., Sun, Y.: Finite element modeling and strength analysis of anti-rust bolts in helicopter. Adv. Aeronaut. Sci. Eng. 8(2), 130–134 (2017) 2. Liu, T., Xu, X.W., Lin, Z.Y.: Research on the thick composite laminate bolt joint strength. J. Mech. Strength 39(2), 353–359 (2017)

Fatigue Life Simulation and Experiment

443

3. Lan, Z.W., Lin, X.P., Lv, C., et al.: Influence of friction high-strength bolt connection strain distribution under varietal preload. J. Nanchang Univ. (Eng. & Technol.) 39(1), 43–49 (2017) 4. Wang, J.N., Zhao, M.Y., Zhou, Y.H.: Research on Influence of pre-tightening force to laminate bolt joint strength. Adv. Aeronaut. Sci. Eng. 3(2), 189–194 (2012) 5. Jiang, J.F., Dong, H.Y., Bi, Y.B.: Analysis of process parameters influencing protuberance during interference fit Installation of hi-lock bolts. Acta Aeronautica et Astronautica Sinica 34(4), 936–945 (2013) 6. Jiang, J.F., Dong, H.Y., Ke, Y.L.: Maximum interference fit size of hi-lock bolted joints. J. Mech. Eng. 49(3), 145–152 (2013) 7. Yuan, Z., Hu, W.P., Meng, Q.C.: An investigation about interference fit effect on stress field of lap joints. Aircr. Des. 36(1), 38–41 (2016) 8. Zhang, Q.L.: Fatigue enhancing theory and simulation study of interference fitting in lightweight aircraft structures. Xi’AN: Northwestern Polytechnical University (2014) 9. Chakherlou, T.N., Mirzajanzadeh, M., Vogwell, J.: Experimental and numerical investigations into the effect of an interference fit on the fatigue life of double shear lap joints. Eng. Fail. Anal. 16, 2066–2080 (2009) 10. Esmaeili, F., Chakherlou, T.N., Zehsaz, M.: Prediction of fatigue life in aircraft double lap bolted joints using several multiaxial fatigue criteria. Mater. Des. 59, 430–438 (2014)

Influence of Heat Treatment on Near-Threshold Fatigue Crack Growth Behavior of High Strength Aluminum Alloy 7010 M. S. Nandana1, Bhat K. Udaya1, and C. M. Manjunatha2(&) 1

2

Department of Metallurgical and Materials Engineering, National Institute of Technology Karnataka, Surathkal, P.O. Box 575025, Mangalore, India [email protected], [email protected] Structural Technologies Division, CSIR-National Aerospace Laboratories, P.O. Box 560017, Bangalore, India [email protected]

Abstract. In this study, aluminum alloy 7010 was subjected to three different ageing treatments i.e., peak ageing (T6), over ageing (T7451) and retrogression and re-ageing (RRA) to study the influence of precipitate microstructure on the fatigue crack growth rate (FCGR) behavior. The microstructural modifications were studied by using TEM to examine the change in size and morphology of the precipitates. The size of the precipitates in the matrix range from 16–20 nm in T7451, 5–6 nm in RRA and 2–3 nm in T6 alloys, respectively. The FCGR tests were performed on standard compact tension (CT) specimens as per ASTM E647 standard in a computer controlled servo-hydraulic test machine with applied stress ratio, R = 0.1 and loading frequency of 10 Hz. The crack growth was measured by adopting compliance technique using a CMOD gauge attached to the CT specimen. The fatigue crack growth rate was higher in T7451 and lowest in RRA treated alloy. The RRA treated alloy showed higher ΔKth compared to T7451 and T6 treated alloys. The measured ΔKth was 11.1, 10.3 and 5.7 MPam½ in RRA, T6 and T7451 alloys, respectively. In the nearthreshold regime, the RRA treated alloy exhibited nearly 2–3 times reduction in the crack growth rate compared to the T6 alloy. The growth rate in the RRA alloy was one order lower than that of the T7451 condition. The surface roughness of RRA treated alloy was more pronounced. The reduction in FCGR observed in RRA alloy was correlated to partial crack closure due to tortuous crack path and partially due to increased spacing between the matrix precipitates. The reduction in near-threshold FCGR and increase in ΔKth is expected to benefit the damage tolerant capability of the aircraft structural components under service loads. Keywords: Damage tolerance Microstructure
RRA


Fatigue
Heat treatment


© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 444–451, 2020. https://doi.org/10.1007/978-3-030-21503-3_35

Influence of Heat Treatment on Near-Threshold Fatigue Crack Growth Behavior

445

1 Introduction Al-Zn-Mg-Cu alloys have drawn much attention towards its application in aerospace industry due to its low weight and high specific strength. These alloys are generally age hardened by suitable heat treatment processes. The aircraft structural material is required to possess good fatigue strength along with better corrosion resistance (Williams and Starke 2003). Since, aircraft components get exposed to harsh environment during their service condition, the components made of these alloys are prone to grain boundary corrosion. These high strength alloys are prone to stress corrosion cracking (SCC) when they are heat treated to peak aged condition (Silva et al. 2012). Although the alloy treated to over aged condition attains good SCC resistance, its peak strength is lost due to the coarsening of precipitates in the matrix (Fooladfar et al. 2009; Rout et al. 2015). Cina (1974) patented a reversion heat treatment process called retrogression and re-ageing. This heat treatment imparts the alloy with mechanical characteristics similar to that of the T6 alloy and corrosion characteristics similar to that of the over aged alloy. It has been reported that the microstructure of the alloy has direct influence on the fatigue crack growth characteristics in the near-threshold regime (Borrego et al. 2004). Efforts are being made to enhance the fatigue crack growth resistance of these alloys by heat treatment and addition of the rear earth elements, such as Er (Bai et al. 2010; Desmukh et al. 2006). A limited number of studies have reported the influence of microstructure on FCGR behavior of the Al-Zn-Mg-Cu alloys (Bai et al. 2011; Chen et al. 2012; Wang et al. 2014). In this study we demonstrate the influence of various heat treatments on the microstructural modifications and its effect on the fatigue crack growth behavior.

2 Materials and Methodology 2.1

Alloy Heat Treatment

The alloy used for the present study was received in over aged condition (T7451) in a rolled plate form. The composition of the received alloy is shown in Table 1. The received alloy was re-solutionized at 490 °C for 6 h and subsequently aged at 120 °C for 24 h to attain peak ageing (T6) condition. The retrogression is attained by further heating the peak aged alloy at 200 °C for 20 min. This process was followed by reageing at 120 °C for 24 h to complete the RRA process. The entire heat treatment process was conducted in a molten salt bath furnace (error ± 5 °C) to apply a uniform heating across the thickness of the test samples. Table 1. Chemical composition of 7010 alloy (in wt%) Zn Mg Cu Zr Fe Si Al 6.3 2.21 1.65 0.124 0.21 0.073 Balance

446

M. S. Nandana et al.

2.2

Alloy Microstructure

The microstructure of the alloy subjected to different heat treatment conditions were studied by using transmission electron microscope (TEM: JEM 2100, JEOL make) operated at 200 kV. The TEM samples were prepared by thinning down the alloy sample to foil of thickness 100 lm. Further the sample foil was punched into 3 mm diameter disc followed by lapping and dimple grinding. The dimpled sample was ion milled using PIPS instrument (Gatan make) by applying beam energy of 5 kV, maintaining the beam angle at 5° top and 5° bottom. 2.3

Fatigue Crack Growth Tests

The specimen blanks of size 65  65  15 mm3 were initially cut from the alloy block to prepare the compact tension (CT) test samples. The blanks were machined to CT specimen of dimensions shown in Fig. 1. The CT specimens were then subjected to respective heat treatment process prior to conduct fatigue crack growth rate tests as per ASTM E647 (ASTM E647-13a1 2015) standards. The FCGR tests were conducted in a computer controlled 100 kN servo-hydraulic fatigue test equipment (Instron make) at room temperature and laboratory air atmosphere. The tests were conducted at loading frequency of 10 Hz with an applied stress ratio of 0.1 (R = rmin/rmax). The crack length was measured as per the compliance method using a CMOD gauge fitted to the specimen. The CT specimen was initially pre cracked from the notch root for about 2 mm under cyclic loads. FCGR test, with decreasing ΔK was performed until the growth rate of 1  10−6 mm/cycle (near-threshold) was reached. The same specimen was then subjected to constant ΔP (increasing ΔK) test until failure of the test specimen.

Fig. 1. Standard compact tension (CT) specimen used for FCGR tests (dimensions in mm)

Influence of Heat Treatment on Near-Threshold Fatigue Crack Growth Behavior

447

3 Results and Discussion 3.1

Microstructural Details

TEM micrographs of the alloy in T7451, T6 and RRA heat treated conditions are presented in Fig. 2(a–c). The precipitates in the matrix of the T6 treated alloy (Fig. 2b) was observed to be smaller in size, ranging in 2–3 nm. These precipitates were found to be densely occupied with their inter-particle spacing of about 3 nm. The RRA treated alloy (Fig. 2c) is observed to contain relatively larger sized precipitates along with fewer smaller sized precipitates in the matrix. The average diameter of the bigger sized precipitates are measured to be about 12 nm, these are present along with smaller precipitates of size 6 nm. These precipitates were distanced apart in RRA treated alloy by about 6–7 nm. In the over aged condition (Fig. 2a) the precipitates were coarsened to a greater extent such that the measured average size of the precipitates stood at

Fig. 2. TEM microstructure of 7010 alloy subjected to; (a) T7451, (b) T6, (c) RRA treatments

448

M. S. Nandana et al.

22 nm. In this condition the precipitates are spaced too apart by about 12 nm. The tensile strength of T7451, T6 and RRA treated alloys are 518 MPa, 581 MPa and 586 MPa respectively (Nandana et al. 2018). 3.2

Fatigue Crack Growth Behavior

The FCGR behavior of 7010 alloy under different heat treatment conditions are presented in Fig. 3. The results infer a speedy crack growth behavior in T7451 condition. Unlike T7451 the crack growth rate of T6 and RRA alloys are similar in Paris regime. In the near-threshold regime, the alloy in all three heat treatment conditions exhibit a significant difference in FCGR. For observation, the crack growth rate was determined from the FCGR data at a constant ΔK level of 12 MPa√m, is presented in the Table 2. The Paris constants C and m were also determined from the FCGR data, is presented in Table 2. It is evident from the FCGR data, that the reduction in crack growth rate is by about 2 times in RRA alloy when compared to T6 treated alloy. When compared to T7451 condition, the crack growth rate is one order lower in RRA alloy. The positive shift of threshold stress intensity factor (SIF) range ΔKth is also noted after RRA treatment. An increase in the ΔKth from 5.78 MPa√m in T7451 condition to 10.7 MPa√m in T6 and 11.2 MPa√m in RRA conditions is observed.

Table 2. FCGR data in 7010 alloy at different heat treatment conditions Alloy condition

C

m

T7451 T6 RRA

1.34  10−7 2.5  10−9 4.5  10−11

2.92 4.1 5.36

da/dN = C(DK)m mm/cycle DK = 12 (MPa√m) 1.8  10−4 5.9  10−5 2.5  10−5

Threshold SIF ΔKth (MPa√m) 5.78 10.7 11.2

Fig. 3. Fatigue crack growth rate behavior of 7010 alloy; (a) FCGR curve, (b) load-CMOD curve

Influence of Heat Treatment on Near-Threshold Fatigue Crack Growth Behavior

449

The Fig. 3(b) represents the load-CMOD data for different heat treated alloys at range of stress intensities ΔK from 12–25 MPa√m. It infers that there exists a crack closure effect on the FCGR behavior of T6 and RRA treated alloys. However, no crack closure is seen for T7451 condition. The trend implies that the crack closure load decreases with increase in ΔK from the near-threshold regime of 12 MPa√m to 25 MPa√m. The measured crack closure load (Pcl in Fig. 3b) is higher in RRA alloy compared to T6 alloy for any given ΔK. The surface roughness of fatigue fractured samples were measured by using a confocal microscope, is shown in Fig. 4(a–c). The measured roughness was higher in the RRA treated alloy compared to that in other heat treated conditions. The average roughness measured stood at 18.7 lm, 24.7 lm and 31.7 lm in T7451, T6 and RRA conditions, respectively, as shown in Table 3.

Fig. 4. Roughness profile of 7010 alloy subjected to; (a) T7451, (b) T6, (c) RRA treatments

Table 3. Surface roughness measured on the fatigue fractured specimens Trials Roughness T7451 Avg Sa 1 14.3 18.7 2 21.5 3 20.4

Roughness (peak) Sp (lm) (avg) Sa (lm) T6 RRA T7451 T6 RRA Sa Sa Avg Sp Avg Sp Avg 26.1 24.7 27.4 31.7 100.2 128 150.6 132 152.2 163 25.1 35.4 112.9 114.6 195.5 22.9 32.2 171 131.2 141.3

450

M. S. Nandana et al.

The fractured surface was sectioned to observe under the light microscope, is presented in Fig. 5. The nature of the crack propagation was observed in the nearthreshold regime, which implies a tortuous crack path in the RRA treated alloy when compared T7451 and T6 alloys. Though the alloy in T6 and RRA conditions possess similar tensile strengths (Nandana et al. 2018), the surface roughness and tortuous nature of crack propagation in the RRA alloy justifies the existence of the partial crack closure that is observed in the load-CMOD curves (Fig. 3b). In the near-threshold regime of FCG, the microstructure also plays a role on the fatigue crack growth rate (Borrego et al. 2004). The matrix of the RRA alloy consists of the precipitates that are slightly coarsened and more distanced apart. During cyclic loading the dislocations glide forward and backward promoting cyclic slip (Wang et al. 2014). Since the microstructure of RRA alloy reveals higher inter-precipitate distance, an enhancement in the reversible cyclic slip is promoted (Xia et al. 2016). Whereas in the T7451 condition, the alloy loses its ultimate tensile strength by about 13%, also there exists no roughness induced crack closure, as evidenced from the load-CMOD curves. Even though the precipitates in the matrix are distanced apart compared to the RRA alloy, the microstructural influence on FCGR is very negligible and hence higher fatigue crack growth rate in T7451 condition.

Fig. 5. Near-threshold FCG behavior of the alloy 7010 subjected to; a) T7451, b) T6, c) RRA treatments

Influence of Heat Treatment on Near-Threshold Fatigue Crack Growth Behavior

451

4 Conclusions • RRA treatment results with partial crack closure due to surface roughness by tortuous crack path. • The microstructure of RRA treated alloy lead to reduction in fatigue crack growth rate by 2–3 times compared to the T6 alloy. • An increase in threshold SIF range, ΔKth is observed in the RRA treated alloy. Acknowledgements. All the authors would like to thank the Director NITK and the Director CSIR-NAL for the support provided during the research work.

References ASTM E647-13a1.: Standard test method for measurement of fatigue crack growth rates. ASTM Intern., West Conshohocken, PA, 1–49 (2015) Bai, S., Liu, Z., Gu, Y., Zhou, X., Zeng, S.: Microstructures and fatigue fracture behavior of an Al-Cu-Mg-Ag alloy with a low Cu/Mg ratio. Mater. Sci. Eng., A 530, 473–480 (2011) Bai, S., Liu, Z., Li, Y., Hou, Y., Chen, X.: Microstructures and fatigue fracture behavior of an Al–Cu–Mg–Ag alloy with addition of rare earth Er. Mater. Sci. Eng., A 527(7–8), 1806–1814 (2010) Cina, B.: Reducing the susceptiibility of alloys, particularly aluminium alloys, to stress corrosion cracking. US Patent 3856584 (1974) Borrego, L.P., Costa, J.M., Silva, S., Ferreira, J.M.: Microstructure dependent fatigue crack growth in aged hardened aluminium alloys. Int. J. Fatigue 26(12), 1321–1331 (2004) Chen, X., Liu, Z., Lin, M., Ning, A., Zeng, S.: Enhanced fatigue crack propagation resistance in an Al-Zn-Mg-Cu alloy by retrogression and reaging treatment. J. Mater. Eng. Perform. 21 (11), 2345–2353 (2012) Desmukh, M.N., Pandey, R.K., Mukhopadhyay, A.K.: Effect of aging treatments on the kinetics of fatigue crack growth in 7010 aluminum alloy. Mater. Sci. Eng., A 435–436, 318–326 (2006) Fooladfar, H., Hashemi, B., Younesi, M.: The effect of the surface treating and high-temperature aging on the strength and SCC susceptibility of 7075 aluminum alloy. J. Mater. Eng. Perform. 19, 852–859 (2009) Nandana, M.S., Udaya Bhat, K., Manjunatha, C.M.: Effect of retrogression heat treatment time on microstructure and mechanical properties of AA7010. J. Mater. Eng. Perform. 27(4), 1628–1634 (2018) Rout, P.K., Ghosh, M.M., Ghosh, K.S.: Microstructural, mechanical and electrochemical behaviour of a 7017 Al-Zn-Mg alloy of different tempers. Mater. Charact. 104, 49–60 (2015) Silva, G., Rivolta, B., Gerosa, R., Derudi, U.: Study of the SCC behavior of 7075 aluminum alloy after one-step aging at 163 & #xB0; C. J. Mater. Eng. Perform. 22, 210–214 (2012) Wang, Y.L., Pan, Q.L., Wei, L.L., Li, B., Wang, Y.: Effect of retrogression and reaging treatment on the microstructure and fatigue crack growth behavior of 7050 aluminum alloy thick plate. Mater. Des. 55, 857–863 (2014) Williams, J.C., Starke, E.A.: Progress in structural materials for aerospace systems. Acta Mater. 51, 5775–5799 (2003) Xia, P., Liu, Z., Bai, S., Lu, L., Gao, L.: Enhanced fatigue crack propagation resistance in a superhigh strength Al–Zn–Mg–Cu alloy by modifying RRA treatment. Mater. Charact. 118, 438–445 (2016)

Multiaxial Fatigue Behavior of 30HGSA Steel Under Cyclic Tension-Compression and Reversed Torsion Daniel Dębski1(&), Krzysztof Gołoś1,2, Marek Dębski1,2, and Andrzej Misztela2 1

2

Warsaw University of Technology, ul.Narbutta 84, 02-524 Warsaw, Poland {daniel.debski,krzysztof.golos}@pw.edu.pl Institute of Mechanised Construction and Rock Mining, ul. Racjonalizacji 6/8, 02-673 Warsaw, Poland

Abstract. Many critical mechanical parts in aerospace, automobile and other industries are subjected to complex cyclic loading during their service life. The 30HGSA steel is one of the most commonly used for manufacturing of this highly loaded structures. The 30HGSA has gained a great interest since it exhibits very good strength properties, high hardness, abrasion resistance and contains a trace amount of nickel. Despite its wide applications and superb characteristics the data about material’s behavior under monotonic and combined cyclic in-phase tension-compression and torsion loading is not available in the literature. The paper aims to fill that void by providing a thorough experimental and numerical analysis of the 30HGSA steel. We will examine and compare Gough–Pollard (GP) and Dębski–Gołoś–Dębski failure criteria in the form of limit curves (DGD-LC) and evaluate high-cycle fatigue models. The obtained experimental high-cycle limit curves will be used to make comparison with the above mentioned failure criteria. The results has shown better agreement between experimental data and DGD-LC model than with GP approach. Keywords: Multiaxial fatigue
Complex cyclic loading Limit curves of complex stress state


Fatigue criteria


1 Introduction Mechanical components in aerospace, automobile industry etc. under complex cyclic loading during their service life. One of the most encountered problems is proposing a method of accounting for this fatigue (Taveres and de Castro 2017; Wanhill 2017). The paper aims to study the material’s behavior under cyclic in-phase tensioncompression and torsion for 30HGSA steel. The subgroup of steels: 30HGS, 30HGSA, 35HGS and 35HGSA is unique because it has great properties even though it does not contain Ni. Nickel in steels/Cr-Ni or CrMo/ is used to significantly increase the hardenability of steel, the brittleness threshold and to improve the resistance to twisting and impact. The XHGS subgroup was designed to replace alloys for thermal treatment without additional nickel and molybdenum additives. The material is cheaper than nickel based © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 452–460, 2020. https://doi.org/10.1007/978-3-030-21503-3_36

Multiaxial Fatigue Behavior of 30HGSA Steel

453

steels like 34HNM, 45HN or 20HN3A and exhibit very good strength properties, high hardness and abrasion resistance. The 30HGSA grade is a high hardenability and wear resistant steel, as evidenced by the addition of manganese in the chemical composition. Additionally, elongation and impact resistance is maintained at a high level. All of those reasons lead to common usage of 30HGSA steel for manufacturing parts in aviation or automobile, heavy machinery or military industries. Therefore it is desirable to examine behavior of this steel under different loading paths, especially under cyclic loading of various combined stresses (Kocańda and Szala 1997). Determination of accurate cyclic deformation properties typically requires significantly more time and cost, as compared to monotonic tension or hardness tests which are relatively simple and inexpensive. Therefore the properties obtained from tension tests are more commonly available than for torsion properties. For that reasons for some materials including 30HGSA steel the data in the literature are not widely available. In the paper uniaxial and multiaxial fatigue properties of 30HGSA will be discussed. The experimental data for monotonic and complex tension and torsion in high cycle fatigue will be presented. An extensive effort has been made to model multiaxial behavior of metals under complex cyclic loading. In recent decades many multiaxial fatigue criteria and fatigue damage models have been reported. Various conclusions have been reached by the different investigators concerning the applicability of the propose theories of failure to the results of fatigue tests under combined stresses. In the present paper Gough–Pollard (GP) (Gough and Pollard 1935; Gough 1950; Dębski 1987; Dębski et al. 2002; also Dębski et al. DGD-LC 2018) failure criteria will be examined. Both criteria avoided the difficulty by including as arbitrary constants the values of the fatigue strengths in bending and in torsion. In the present paper this models will be extended and used to describe fatigue strength in cyclic tension/compression and in reversed torsion for 30HGSA steel. The high-cycle limit curves for 30HGSA steel based on GP and extended DGD-LC failure criterion will be analyzed.

2 Specimen and Test Set-Up The investigated specimen is a pipe of 16 mm in diameter made of steel 30HGSA (GOST 4543-71) that is used in aviation. Table 1 shows the chemical composition of the 30HGSA used in this work by weight. Table 1. Composition for 30HGSA, weight percentage for GOST 4543-71. C Mn Si P S Cr Ni Fe 0.28–34 0.8–1.1 0.9–1.2 20 lm depth, sub-surface cracks, surface oxidation or contamination. To remove defects as described, rectifications/rework can be performed as part of a normal production cycle.

Is the Civil Aerospace Industry Ready to Implement Laser Shock Peening

653

4 Experimental Approach Can Low Energy LSP Systems Provide Required Fatigue Mechanical Performance? Low energy LSP system, (  200 mJ, pulse width of  25 nsec) and associated small laser spot size (< 1 mm diameter), can determine high compressive stress in the near surface of typical Aluminum alloys and compression depth beyond 1 mm. The higher attenuation rate associated to spherical shock wave (small spot size) is recovered by the increased peening coverage larger than for LSP associated to high energy (typically 2–3 layers treatment, Furfari et al. 2017). Investigated Materials, Specimen Geometry and Test Condition and Setup. 7175T7531 aluminum alloy was investigated. All coupons were extracted in the LT-ST orientation at the mid-thickness position from plate material. The specimens used for the residual stress characterization were 70 mm x 80 mm x 10 mm with LSP area placed in the center of the coupon of 20 mm x 20 mm (ref. to Fig. 4).

Fig. 4. Coupon orientation from plate material (left); Residual stress coupon with typical ZigZag LSP peening pattern (i.e. scanning and stepping direction)

The four point bending specimens used to demonstrate fatigue life enhancement after LSP treatment are described in EN6072 (2010). The surfaces subjected to the maximum stresses were covered with either shot peening or LSP extending the peened surface to half thickness from each side of the coupon. Shot peening parameters were intensity of Almen strip 0.20-0.24 mmA, steel balls peening media 600 lm diameter and coverage of minimum 100%. The LSP treatment was performed using a green laser (wavelength 532 nm) with peak energy per pulse of 70 mJ, pulse width of 8 nsec. Spot diameter at the focus plane has been varied to investigate the influence on the compressive residual stress profile. The LSP coverage is performed by scanning the surface to be treated with high percentage overlap (typically > 60% of the spot size) and advancing the line with similar overlap in the so called stepping direction. The residual stress characterization was carried out by means of X-ray diffraction (XRD) technique, to assess the residual stress on near surface (i.e. < 20 µm depth), and by means of incremental hole drilling (ICHD) method to obtain the depth profile up to 1 mm into the material. Figure 5 describes exemplarily the measurement location for both techniques inside the LSP area. The gauge area for the X-ray method was 2 mm diameter and two lines of measurements placed at 2 mm and 10 mm inside the peened

654

D. Furfari et al.

area. The residual stresses near the surface obtained by XRD were reported as arithmetic average measurements from 14 readings resulting from 2 lines and 7 spots measurements and in some cases up to 18 measurements (2 lines and 9 spots readings). The incremental hole drilling measurements were performed with 1.9 mm hole diameter at the position described in Fig. 5.

Fig. 5. Residual stress measurements locations inside the LSP area for X-ray diffraction method (left) and incremental hole drilling (right)

5 Experimental Results and Discussion Residual Stress Measurements in 7175-T7351 Material. The residual stresses profiles up to 1 mm depth have been measured by incremental hole drilling after LSP. The effect in terms of compressive residual stress induced by LSP using a pulsed laser of 532 nm wavelength at pulse width of 8 nsec and 45 pulses/mm2 coverage with increasing spot diameter from 0.6 mm up to 1.1 mm are shown in Fig. 6.

Fig. 6. Residual stress profiles after LSP in 7175-T7351 material using 532 nm laser at 0.8 nsec pulse width and 45 pulses/mm2 at increased spot size.

The residual stresses at near surface, measured by X-ray diffraction, are influenced by the peening strategy, resulting in higher compressive residual stress component in the stepping direction than in the scanning direction. The effect of increasing the spot size results in a decreased power density (in GW/cm2) which is the major parameter of LSP as discussed in Furfari (2014), Hombergsmeier (2015) and Furfari et al. (2017). Reducing the power density results in lower compressive stress at the near surface. The

Is the Civil Aerospace Industry Ready to Implement Laser Shock Peening

655

compressive residual stress at near surface reach a maximum value at power density level above 1.5 GW/cm2 as shown in Fig. 7. Similar threshold (i.e. 1.5 GW/cm2) is demonstrated for the residual stresses at 0.3 mm and 1 mm depth (Fig. 7 right). The results summarized in Figs. 6 and 7 demonstrate the capability to induce compressive residual stress in excess of 1 mm depth into 7175-T7351 material using low energy LSP system, providing a minimum power density of 1.5 GW/cm2 is applied.

Fig. 7. Residual stress in 7175-T7531 as function of power density near the surface (left), and at 0.3 mm and 1 mm depth (right).

Fatigue Life Results in 7175-T7351 Material. The fatigue life response of four point bending coupons made of 7175-T7351 material and subjected to constant amplitude loading condition with R ratio of 0.1 is reported in Fig. 8.

Fig. 8. Wöhler curves of coupons made of 7175-T7351 material in as machined condition (reference), after shot peening and LSP (70 mJ, 0.7 mm, 45 pulses/mm2)

Fatigue lives up to complete failure are reported in the Wöhler curves with three conditions investigated: as machined (reference), after conventional shot peening and after LSP with 70 mJ energy per pulse, spot diameter of 0.7 mm and 45 pulses/mm2. At increasing stress levels the benefit in terms of fatigue lives up to failure for surface

656

D. Furfari et al.

mechanical process as shot peening and LSP diminish. The fatigue lives after either conventional shot peening or LSP at all stress levels have shown a significant extension of fatigue lives up to final failure.

6 Conclusions The major constrains for in-service implementation of LSP at MRO environment have been reviewed. Heavy and complex laser shock peening systems are not compatible with surrounding operational environments such as a typical MRO. Easy transportability, modularity, quick assembly and disassembly capabilities, availability of spare parts and consumables are the key features to enable the development for MRO environment. Health and safety, together with the ATEX requirement represent one of the most challenging constraints to be fulfilled. Key features to enhance the development of LSP portable device capable to apply this surface technology at typical MRO environment were suggested. Engineering requirements that must be fulfilled beside any laser specific requirements (e.g. laser type, laser beam characteristic, laser beam quality etc.) were described in terms of residual stress profiles, fatigue life improvement and surface quality. It has been demonstrated that even using low energy LSP system (< 200 mJ) and small spot diameter (< 1 mm) it is possible to introduce high compressive residual stress in excess of 1 mm depth in 7175-T7351 aluminum alloy. The response in fatigue of this material treated with low energy LSP system has shown a dramatic extension of fatigue lives compared to not treated material (reference). These results open the door for development of portable low energy LSP system providing that ALL the other requirements and constrains as described in this paper are respected.

References Furfari, D.: Laser shock peening to repair, design and manufacture current and future aircraft structures by residual stress engineering. Adv. Mater. Res. 891–892 (2014), 992–1000, © (2014) Trans Tech Publications, Switzerland (2014) Hombergsmeier, E., Furfari, D., Ohrloff, N., Heckenberger, U.C., Holzinger, V.: Enhanced fatigue and damage tolerance of aircraft components by introduction of residual stresses – a comparison of different processes. In: 27th ICAF Symposium, Jerusalem, 5–7 June 2013 Liu, Q., Yang, C.H., Ding, K., Barter, S.A., Ye, L.: Fatigue Fract. Eng. Mater. Struct. 30(2007), 1110–1124 (2007) Frank, A.D., Smith, D.L., Kuster, R.L.: Investigation of laser shock processing – executive summary. In: Technical Report AFWAL-TR-80-3001, Vol. I, Executive Summary to Final Report – July 1978–October 1979 (1980) Furfari, D., Ohrloff, N., Hombergsmeier, E., Heckenberger, U.C., Holzinger, V.: Laser shock peening as surface technology to extend fatigue life in metallic airframe structures. In: 35th ICAF Conference, Nagoya, Japan, 7–9 June 2017 EASA Part-145: Maintenance Organization Approvals (2012)

Is the Civil Aerospace Industry Ready to Implement Laser Shock Peening

657

MIL-STD-810G: Department of Defense Test Method Standard. In: Environmental Engineering Considerations and Laboratory Tests (2014) MacGillivray, Lt.Ken, Dane, B., Osborne, M., Bair, R., Garcia, W.: F-22 laser shock peening depot transition and risk reduction. USAF ASIP Conf., 30/11–02/12, S. Antonio, TX (2010) Polin, L., Bunch, J., Caruso, P., McClure, J.: Full scale component tests to validate the effects of laser shock peening. In: ASIP 2011, Nov 29–Dec 1, San Antonio, TX (2011) Sano, Y., Akita, K., Masaki, K., Ochi, Y., Altenberger, I., Scholtes, B.: Laser Peening without Coating as a Surface Enhancement Technology. JLMN-J. Laser Micro/Nanoeng. 1(3) (2006) Uehara, T., Sano, Y., Chida, I., Yoda, M., Mukai, N., Kato, H.: Laser peening systems for preventive maintenance against stress corrosion cracking in nuclear power reactors. In: Proceedings of the 16th International Conference on Nuclear Engineering, ICONE16, Orlando, Florida, USA, 11–15 May 2008 Gujba, A.K., Medraj, M.: Laser Peening Process and Its Impact on Materials Properties in Comparison with Shot Peening and Ultrasonic Impact Peening. Materials 7, 7925–7974 (2014). https://doi.org/10.3390/ma7127925 Clauer, A.H., Fairand, B.P., Wilcox, B.A.: Pulsed laser induced deformation in an Fe-3wt Si alloy. Metall. Trans. A 1977(8), 119–125 (1977) Fairand, A.H., Clauer, B.P.: Interaction of laser-induced stress waves with metals. In: Proceedings of the ASM Conference Applications of Lasers in Materials Processing, Washington, DC, USA, 18–20 April 1979; ASM International: Materials Park, OH, USA (1979) IEC 60825-1, (International Standard): Safety of laser products – part 1: Equipment classification, requirements and user’s guide (2001) IEC 60079-10, (International Standard): Explosive atmospheres-Part10-1: Classification of areas - Explosive gas atmospheres (2015) IEC 60079-28, (International Standard): Explosive atmospheres – Part 28: Protection of equipment and transmission systems using optical radiation (2015) Slater, J.M., Brian E.: Characterization of high power lasers. In: Laser Technology for Defense and Security VI. Ed. Mark Dubinskii & Stephen G. Post. Orlando, Florida, USA: SPIE, 2010. 76860 W-12. ©2010 COPYRIGHT SPIE—The International Society for Optical Engineering (2010) Prime, M.B.: Cross-sectional mapping of residual stresses by measuring the surface contour after a cut. J. Eng. Mat. & Techn. 123, 162–168 (2001) Prime, M.B., Sebring, R.J., Edwards, J.M., Hughes, D.J., Webster, P.J.: Laser surface contouring and spline data-smoothing for residual-stress measurement. Exp. Mech. 44(2), 176–184 (2004) Dorman, M., Toparli, M.B., Smyth, N., Cini, A., Fitzpatrick, M.E., Irving, P.E.: Effect of laser shock peening on residual stress and fatigue life of clad 2024 aluminium sheet containing scribe defects. Mater. Sci. Eng., A 548(2012), 142–151 (2012) ASTM E837-13: Standard test method for determining residual stresses by the hole-drilling strain-gage method (2013) EN 6072: Aerospace series – metallic materials – test methods – constant amplitude fatigue testing (2010) AC25.571-1D Advisory Circular: Damage Tolerance and Fatigue Evaluation of Structure (2011) DIN EN ISO 4287: Geometrical product specifications (GPS) – surface texture: profile method – terms, definitions and surface texture parameters (2010)

Fatigue Life Prediction at Cold Expanded Fastener Holes with ForceMate Bushings Yan Bombardier(&), Gang Li, and Guillaume Renaud National Research Council Canada, Aerospace Research Centre, 1200 Montreal Road, Ottawa, ON K1A 0R6, Canada [email protected] Abstract. ForceMate high interference fit expanded bushings, made by Fatigue Technology Inc. (FTI), are used by aircraft designers and maintainers to improve the fatigue and wear resistance of holes. While the fatigue life improvement (LIF) resulting from the installation of ForceMate bushings has been demonstrated experimentally, no analytical methods have been officially approved yet to take full benefit from the beneficial state of interference and residual stress resulting from this technology. To address this gap, a methodology was developed to analytically determine the LIF resulting from the installation of ForceMate bushings by explicitly taking into account the residual stresses and the effect of high interference fit bushings. To achieve this, a three-dimensional residual stress field is obtained from finite element process modelling of the ForceMate installation; fatigue crack nucleation lives are calculated; crack propagation analyses are conducted to calculate the resulting crack shapes and stress intensity factors, and the crack growth predictions are performed. This methodology was demonstrated on a CF–188 bulkhead at the holes attaching the main landing gear uplock mechanism. Based on this analytical study, a LIF of 6.6 was predicted for crack nucleation and 4.9 for crack growth from a 0.254 mm quarter-circular crack to a through-the thickness crack. While there are several aspects of this analytical study that need to be validated experimentally, the calculated LIF correlates well with the LIF typically obtained with the ForceMate system. Keywords: Residual stress ForceMate


Fatigue
Cold working
Crack growth


1 Introduction Background. The ForceMate system is commercialized by Fatigue Technology Inc. (FTI) as an alternative to shrink and press fit bushing installation methods. As opposed to shrink fit bushings that are installed using a cryogenic shrink fitting process, the ForceMate bushing is a lubricated clearance fit bushing that is radially expanded into the hole using an oversized tapered mandrel, creating a high interference fit. Depending on the relative yield strength between the bushing and parent structure, the expansion process may also cold expand the parent material, leaving beneficial compressive residual stresses in the parent structure. © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 658–673, 2020. https://doi.org/10.1007/978-3-030-21503-3_53

Fatigue Life Prediction at Cold Expanded Fastener Holes

659

Aircraft designers and maintainers use the ForceMate high interference fit expanded bushings to improve the fatigue life and the wear resistance of holes. For cold expanded holes, the Joint Service Specification Guide (DoD 1998) states that the limits of the beneficial effects to be used in design should be no greater than the benefit derived by assuming a 0.127 mm radius corner flaw from one side of the non-expanded hole containing a neat fit fastener, instead of a 1.27 mm rogue flaw. While this assumption has been shown to be significantly conservative, the allowance for smaller initial flaws provide longer fatigue crack growth lives without having to substantiate the resulting life improvement factor (LIF) with extensive testing. This assumption is however not allowed for ForceMate bushings, as no analytical methods have been approved for taking any benefit of this technology despite that large LIF values have been demonstrated experimentally for several applications (Reid 2011). To address this gap, the National Research Council Canada (NRC) has been developing analytical methods and tools to quantify the residual stresses in cold worked structures for fatigue life predictions. Process simulation models developed by NRC were initially used to calculate the residual stresses induced by the split sleeve hole cold expansion process (Renaud et al. 2017). This methodology has been recently been expanded to simulate the installation of ForceMate bushings and was evaluated on the CF-188 case study briefly described in the next section. Case Study. Two bulkheads of the Royal Canadian Air Force (RCAF) CF-188 aircraft, referred to as Y488 and Y497, have been prone to fatigue cracking at fastener holes attaching the main landing gear (MLG) uplock mechanism. Over the years, several structural modifications and inspection procedures have been developed for these holes to prevent, delay, or repair fatigue damages. Most aircraft were modified by oversizing the holes and installing ForceMate bushings. These aircraft are now subjected to a repeated inspection beyond the expected safe life limit of the un-modified holes. Based on the evidence that the full benefit of the ForceMate bushings was not considered in the determination of the inspection schedule, the methodology developed by NRC for calculating the LIF of holes was used to analytically evaluate if the repeated inspections were actually required. For this study, the normalized spectrum was derived from the wing root bending moment from the IARPO3a loading spectrum, which was developed for the International Follow-On Structural Test Project (IFOSTP). The spectrum was arbitrarily normalized to a maximum applied stress of 137.9 MPa and the resulting LIF were calculated for demonstration purposes only. Objective. The main objective of this study was to develop a methodology for determining the LIF resulting from the installation of ForceMate bushings. The developed methodology consists of the following tasks: (1) calculate the residual stresses in a structure resulting from a ForceMate bushing installation, (2) calculate the crack nucleation and crack growth lives for a structure with a residual stress free open hole and a ForceMate bushing, and (3) calculate the crack nucleation and crack growth LIF for the ForceMate configuration based on the residual stress free open hole configuration.

660

Y. Bombardier et al.

2 Residual Stresses Methodology. To simplify the LIF calculation, mode-I cracks were assumed to nucleate and propagate along a plane that is normal to the loading axis. To calculate the residual stress distribution, a three-dimensional finite element model (FEM) was developed to simulate the installation of the ForceMate bushing. The residual stresses contributing to crack opening and closing were extracted along the assumed crack plane and were used for the life predictions. Calculation of the residual stress distribution is complex and requires advanced modelling techniques to solve this highly non-linear problem. A methodology was developed based on the finite element (FE) method to calculate the residual stress distribution. FE contact analyses were conducted to model the installation process, including the mandrel insertion and bushing reaming. Contact pairs were defined between the mandrel and the ForceMate bushing and between the ForceMate bushing and the plate. The steel mandrel was assumed to be rigid due to its high modulus and high strength. While MSC.Marc was used for this simulation, other solvers can be used as long as they meet the following requirements: geometric and material non-linear capabilities, contact modelling, and element deletion to simulate reaming. In addition, parametric modelling capabilities were included in the FEM to modify the geometric and analysis parameters without having to re-create a model every time a new analysis was required. Geometry. For the CF-188 case study, a simplified Y488 bulkhead representation was modelled using a 8.89 mm thick flat plate with a width of 76.20 mm. The total length of the plate was set to 152.40 mm. The hole was centrally located within the plate and its initial diameter was set to 9.47 mm. For simplification and comparison purposes, the open hole and ForceMate configurations were simulated using the same hole diameter. The symmetry of the problem was used to reduce the number of degrees of freedom of the FEM and to reduce the computational time. Consequently, only a quarter of the plate was modelled. The geometric model is illustrated in Fig. 1 with the mandrel and the ForceMate bushing. As shown in Fig. 1, the cracking plane, defined as the XZ plane, is normal to the loading axis, defined as the Y –axis. For a crack propagating within the XZ plane, the crack opening and closing residual stresses would correspond to the normal stresses along the Y -axis. For simplicity, only the component of the residual stresses responsible for crack opening or closing were extracted from the finite element analysis (FEA). The geometry of the mandrel defines the level of cold working induced into the bushing and the plate. As the geometry of the mandrel is proprietary to FTI, a geometry was developed for the CF-188 case study based on the target cold expansion ratio by setting the final bushing wall thickness and inner diameter after reaming to 1.65 mm and 6.35 mm respectively. Other dimensions were then obtained using a trial and error approach to achieve a 4% cold expansion level in the bushing.

Fatigue Life Prediction at Cold Expanded Fastener Holes

661

9.47 mm outer diameter 5.84 mm inner diameter 1.81 mm wall thickness

8.89 mm

Mandrel ForceMate bushing (initial)

Mandrel pull direction Fig. 1. Finite element model of the simplified Y488 bulkhead geometry.

Contact. Segment-to-segment contact entities were used for two zones: (1) between the mandrel and the bushing and (2) between the bushing and the plate. The segmentto-segment contact algorithm does not rely on the slave-master concept and provided better stress continuity at the contact interface. The bilinear Coulomb friction model was used for both contact zones with a coefficient of friction of 0.1 for the two contact zones. It was found that the coefficient of friction had no significant effect on the residual stresses in the plate and that using a lower friction coefficient improved the convergence of the contact algorithm. Meshing. The plate and bushing were meshed using eight-noded, isoparametric, threedimensional brick elements with trilinear interpolation functions, referred to as Element Type 7 in MSC.Marc. This type of element is recommended for contact problems. The FE meshes of the plate and bushing are shown in Fig. 2. As shown, biased mesh densities were used to provide smaller elements around the hole while providing smooth transitions to regions away from the point of interest. The FE mesh of the bushing, shown in Fig. 2b, was subdivided into two concentric volumes: the inner and outer volumes. The inner volume, 0.254 mm thick, contained the elements that were deactivated during the reaming step of the simulation. Material. Elastic-plastic material models were required to model the plate and bushing materials. For the case study, isotropic hardening behaviour was assumed for both materials. The bushing was assumed to be made of AISI 4130 steel and the plate material was made of 7050-T7451 aluminium. The material stress-strain relationship for the elastic-plastic behaviour was defined using tabular plastic strain data.

662

Y. Bombardier et al.

a) Plate

b) Bushing

Fig. 2. Overview of the finite element model.

Loads and Boundary Conditions. Symmetric boundary conditions were applied to the planes of symmetry and the displacements along the Z-direction were constrained at the exit surface of the bushing to simulate the nosecap of the puller. The load steps outlined in Table 1 were used to simulate the ForceMate installation (Steps 1–3) and remote loading (Steps 4–7). Table 1. Load Steps of the finite element analysis. Step 1:

Step 2: Step 3:

Step 4:

Step 5:

Step 6:

Step 7:

Mandrel Pulling: Pull the mandrel through the ForceMate bushing by enforcing mandrel displacements. Once this load step is completed, the mandrel exits the bushing, leaving residual stresses in the bushing and in the plate Release Mandrel Contact: Release the contact interface between the mandrel and the ForceMate bushing and the nosecap boundary conditions Reaming: Deactivate the elements located within the inner volume identified in Fig. 2b. Once this load step is completed, the residual stress distribution in the bushing and the plate are slightly modified as the internal forces are rebalanced after the deactivation of the inner volume of the bushing Tensile Remote Loading: Apply tensile uniform remote loading. This changes the stress around the hole by combining the effect of residual stress and remote loading Unloading: Unload the plate by applying zero remote loading. The residual stresses are modified if the combination of stress due to remote loading and residual stress exceeds the yield strength. Otherwise, the residual stresses should be identical to the results obtained after Step 3 Compressive Remote Loading: Apply compressive uniform remote loading. This changes the stress around the hole by combining the effect of residual stress and remote loading Unloading: Unload the plate by applying zero remote loading. The residual stresses are modified if the combination of stress due to remote loading and residual stress exceeds the yield strength. Otherwise, the residual stresses should be identical to the results obtained after Step 3

Fatigue Life Prediction at Cold Expanded Fastener Holes

663

Mandrel Pull Direction

Results. The crack opening and closing stress and strain distributions along the crack plane identified in Fig. 1 were extracted from the FEA for the fatigue and crack growth analyses. Stress contour plots of the plate after Load Steps 1 and 3 are provided in Fig. 3. As shown, there is a slight residual stress relaxation between Steps 1 and 3 due to the reaming process performed in Step 3. After the ForceMate bushing installation, the simulation results predicted a 1.2 mm thick layer of compressive residual stresses, which is balanced by tensile residual stresses away from the hole bore. The compressive residual stress values at the bore and the maximum tensile residual stresses found Fig. 3b are plotted in Fig. 4 as a function of position along the thickness. As shown in Fig. 4, the stresses at the bore oscillate near the exit surface of the mandrel (from Z = 0.0 mm to 1.1 mm) due to numerical artefacts from the contact analysis when the mandrel leaves the bushing. However, based on the trend of the residual stress variation as a function of thickness, the exit surface of the mandrel (Z = 0.0 mm) appears to provide the lowest residual stress, i.e. most compressive. Consequently, the observed oscillations may not have any impact on the calculated fatigue life as cracks were assumed to nucleate from the bore at the entrance surface of the plate. Additional results are presented in the following sections to support the crack nucleation and crack growth life predictions.

-207

-165

-124

-83

-41

0

41

83

124

165

207

(MPa)

Mandrel Pull Direction

a) Residual stresses along hoop direction after mandrel pulling (Step 1)

-207

-165

-124

-83

-41

0

41

83

124

165

207

(MPa)

b) After reaming (Step 3) Fig. 3. Stress contour plots for Load Steps 1 and 3.

3 Fatigue Life Methodology. Fatigue crack nucleation lives were calculated for the open hole and ForceMate configurations for all the nodes located along the bore of the hole. NRC developed a fatigue crack nucleation analysis tool for structures with residual stresses

664

Y. Bombardier et al. 200 150

Residual stress (MPa)

100 50 0

Lowest compressive residual stress, excluding the oscillations at the exit surface

Residual stress oscillations at exit surface

-50 -100 -150 -200 -250 -300

Stress at bore

-350 0

1

2

3

4

5

Maximum tensile stress 6

7

8

9

-position along the thickness (mm)

Mandrel exit

Mandrel entrance

Fig. 4. Variation of residual stresses induced by the ForceMate installation as a function of position along the thickness.

subjected to variable amplitude loading spectra. The fatigue life was calculated using the Smith-Topper-Watson (SWT) equation to include the mean stress effect (Smith et al. 1970). The SWT strain-life equation is: " rmax 2a ¼

r02 f ð2Nf Þ2b E

#

h i þ r0f 20f ð2Nf Þb þ c elastic

plastic

where rmax is the maximum stress at maximum strain, 2a is the strain amplitude, Nf is the number of cycles to a 0.254 mm crack, E is the elastic′ modulus, r0f is the fatigue strength coefficient, 20f is the fatigue ductility coefficient, b is the fatigue strength exponent, and c is the fatigue ductility exponent. The cycles to a 0.254 mm crack, Nf , were obtained for each cycle using a rootfinding numerical method. The maximum stress, rmax and the strain amplitude, 2a , were obtained by scaling the FEA strains, that include the effect of residual stresses, according to the peak and valley loads for each cycle. Bilinear scaling was used as the stress and strain responses were a function of loading direction (tensile versus compressive loads) as the hole was filled with the ForceMate bushing. Fatigue Model Parameters. For the case study, the SWT equation parameters were obtained by least-square fitting strain life data for Al7050-T74 aluminium. Fatigue Life Analyses. The fatigue damage indices per spectrum pass were calculated using Miner’s rule cumulative damage model for the open hole and ForceMate configurations. The damage indices were calculated for all the nodes along the bore and are graphically presented Fig. 5 for both configurations. The fatigue life calculation performed using the FEA results indicated that:

Fatigue Life Prediction at Cold Expanded Fastener Holes

665

– For the open hole configuration, the maximum damage index was obtained at the midplane of the plate. This suggests that a surface crack would more likely nucleate at the midplane (Z = 4.45 mm) than a corner crack. While this could be counterintuitive, it is important to highlight that this analysis does not consider many degrading conditions, such as initial fabrication defects, fretting, and corrosion; and – For the ForceMate configuration, the maximum damage index was obtained at Z = 8.08 mm. This is located 0.81 mm below then entrance face of the mandrel. This correlates with the residual stress distribution shown in Fig. 4, where the highest residual stress was obtained at the same location.

Damage index per spectrum pass

3.5

Open Hole

3

Nucleation site

2.5 2 1.5 1

Nucleation site

0.5

ForceMate

0 0

2

Mandrel exit

4

6

Z-position along thickness (mm)

8

Mandrel entrance

Fig. 5. Damage index as a function of the position along the bore.

As the damage index is inversely proportional to fatigue life, the predicted LIF for the ForceMate bushing was calculated by dividing highest damage index obtained from the open hole configuration (3.123 per spectrum pass) by the highest damage index obtained from the ForceMate configuration (0.474 per spectrum pass), yielding a LIF of 6.6. As the analyses were conducted using the fictitious maximum remote spectrum stress of 137.9 MPa, a sensitivity study was performed on the spectrum stress scaling factor. For the case study, increasing the maximum remote stress by 10% decreased the LIF by 15%, whereas decreasing the maximum remote stress by 10% increased the LIF by 26%. Consequently, it was conservative to select a high spectrum scaling factor for calculating a LIF.

4 Crack Growth Initial Crack Assumption. Based on the results presented in the previous section, a crack would more likely nucleate 0.81 mm below the entrance of the mandrel along the bore. To simplify the crack growth analyses, it was assumed that the crack would

666

Y. Bombardier et al.

nucleate at the intersection between the bore and the entrance surface of the mandrel. The assumed crack nucleation location, superimposed onto the residual stress distribution, is presented in Fig. 6. Based on this global FEM coordinate system, a quartercircular corner crack with a radius of 0.254 mm was assumed to nucleate at X = 0 mm and Z = 8.89 mm.

Fig. 6. Assumed crack nucleation site based on the residual stress field.

Crack Propagation Tool. The crack growth analyses required the following information: the stress intensity factors (SIF) as a function of crack size, and a crack growth model. The SIF can be estimated for a crack growing from an open hole in a residual stress free structure using closed-form solutions available in the literature (Tada et al. 2000). For a structure with residual stresses, the evolution of the resulting crack shape is often more complex and the effect of residual stress on the calculated SIF is more difficult to quantify. To overcome this challenge, FE based approaches were used to accurately predict the evolution of the crack shape within a residual stress field and to calculate its associated SIF values. At the time of writing, crack propagation using MSC.Marc was still under investigation. In the interim, a numerical tool developed by Engineering Software Research & Development (ESRD), CPAT, was used to simplify the process for incrementally propagating a three-dimensional crack in a solid. CPAT was developed to simulate three-dimensional planar crack growth at an open fastener hole located in structures with predefined simple geometries. CPAT has the capability to include the effect of residual stresses when propagating the crack and provides an easy-to-use interface that eliminates manual operations for the preparation of crack

Fatigue Life Prediction at Cold Expanded Fastener Holes

667

growth simulations using the FE method. CPAT builds the FEM, performs the FEA using the StressCheck solver, post-processes the results, and propagates the crack automatically. Crack Propagation Analyses. The CPAT program was used to calculate the SIF solutions as a function of crack size from a quarter-circular crack propagating from an open hole. The simulations were completed with and without residual stresses to calculate the resulting LIF. The CPAT simulations captured the evolution of crack shape from the initial crack geometry. The simulations were interrupted when the partthrough crack transitioned to a through-the-thickness crack. For the CF-188 case study, the Y488 bulkhead was modelled as a flat rectangular plate with a width of 71.1 mm and a thickness 8.89 mm. The 9.40 mm diameter hole was centrally located within the plate. The plate was 304.80 mm long. The material properties for 7050-T73651 aluminium were obtained from the AFMAT material database (tabular lookup da=dN data - 7050-T73651 L-T Lab air Plate) (LexTech 2018). The residual stress profile used in CPAT was fitted using a Legendre polynomial surface. Due to the residual stress oscillations observed near the exit surface of the mandrel, the residual stress profile was manually modified to smoothen the residual stresses in this area to reduce the error in the region of interest, i.e. entrance surface of the mandrel. The fitting parameters were selected by trial and error to minimize the absolute residual stress error in the neighbourhood of the crack nucleation region. The evolution of crack shapes from the initial flaw at an open hole are provided in Fig. 7 for both simulated cases (with and without residual stresses). As shown, the crack propagation analysis without residual stresses (Fig. 7a) resulted in an elliptic crack shape, where the crack tip growing along the bore of the hole grew faster than the crack tip along the surface of the plate. As shown in Fig. 7b, the shape of the crack growing within the residual stress field resulted in an almost circular crack shape due to a reduction of the crack growth rate along the bore. The reduction of crack growth rate was caused by the compressive residual stresses along the bore of the hole that reduced the effective SIF range, DK.

a) Open hole without residual stresses

b) Open hole with residual stresses

Fig. 7. Crack front prediction using CPAT for an open hole condition.

668

Y. Bombardier et al.

Effect of High Interference Bushing. Although the effect of residual stresses were included in the SIF calculated using CPAT, the presence of the high interference ForceMate bushing was missing. This section describes the methodology developed to include the effect of the high interference bushing in the hole. First, the loading direction affects the local load path around the hole, which modifies the crack opening stresses and strains. This effect is illustrated in Fig. 8 with FEA results obtained from a zero clearance bushing inside a hole. Under tensile remote loads (Fig. 8a), the load path goes around the bushing and the bushing prevents transverse deformations. For this case, the maximum tangential stress is lower than the one obtained for an open hole scenario and is located, as expected, at 90° from the loading direction. Under compressive remote loads (Fig. 8b), the load is partially transferred by contact through the bushing, which significantly reduces the stresses around the hole in the structure. The magnitude of load transferred by contact through the bushing increases as the relative stiffness of the bushing increases (e.g. steel bushing versus aluminium bushing). Consequently, the magnitude of the tangential stress is significantly lower under compressive remote loads compared to tensile loads for the same remote load magnitude.

a) Tensile load

b) Compressive load

Fig. 8. Local load path effect due to filled hole.

If the hole is filled with a high interference bushing instead of a zero clearance bushing, the gap that is visible between the bushing and the plate in Fig. 8a, could be non-existent when remote loading is applied. Furthermore, this gap only appears beyond a certain load magnitude. The simulation of high interference bushings increases the complexity of the analysis as the bushing is already partly taken into account by the residual stress distribution due to the installation. Consequently, the effect of the high interference bushing due to remote loads had to be isolated from the residual stresses induced during its installation. To isolate these effects, the FEA results obtained with MSC.Marc were used to determine the effect of the ForceMate bushing on the tangential stress distribution under remote loading.

Fatigue Life Prediction at Cold Expanded Fastener Holes

669

The residual stresses at the midplane are shown in Fig. 9 with the stress distributions under a tensile and compressive stress of 137.9 MPa. The effect of the ForceMate on the stress distribution due to remote loads was then obtained by subtracting the residual stresses from the stress distributions under remote loads, such that linear superposition would apply. The calculated tensile and compressive stress distributions due to remote loading only with a ForceMate bushing are compared in Fig. 9 against the stress distribution at an open hole. As shown, the presence of the ForceMate bushing reduced the tangential stress at the bore due to remote loading by approximately 28%, compared to the stresses at an open hole. 500

28% Crack opening stress (MPa)

300 100 -100 ForceMate (no remote load) ForceMate under tensile load ForceMate under compressive load Calc. tensile load stress distribution Calc. compressive load stress distribution Open hole under tensile load Open hole under compressive load

-300 -500 -700 0

2

4

6

Radial distance from the edge of the hole (mm)

8

Fig. 9. Tangential stress distribution at a hole with a ForceMate bushing.

In order to conduct the crack growth analysis, a geometric correction factor, bFM , was developed to correct the open hole SIF solution obtained using CPAT to simulate the effect of the ForceMate high interference bushing under remote loading. This was done using the Gaussian integration method by scaling the SIF based on the stress distributions shown in Fig. 9. The process was repeated for tensile and compressive loads and the resulting bFM correction factors are illustrated in Fig. 10. For a 2.54 mm crack, the ForceMate reduced the SIF by approximately 21% based on the open hole configuration. The total SIF, K, for the ForceMate configuration is calculated as pffiffiffiffiffi K ¼ KRS þ bO=RS bFM r pc where KRS is the residual SIF calculated with CPAT, bO=RS is the geometric correction factor calculated with CPAT due to remote loading, r is the remote stress, and c is the surface crack length. Note that the bO=RS factor includes the effect of remote loading

Y. Bombardier et al. 7.0

2.0

6.0 5.0

1.6

4.0 3.0

1.4

2.0

(

-correction factors

1.8

)

670

1.2

1.0 0.0

1.0

-1.0

0.8 0.6

-2.0 0

2

4

6

Crack size (mm)

8

10

-3.0

Fig. 10. b-correction factors and K RS for stress intensity factor calculation.

without residual stresses, but considering the effect of the crack shape resulting from the residual stress field. The bO=RS and KRS values are presented in Fig. 10. For the open hole configuration without residual stresses, the total SIF is calculated as pffiffiffiffiffi K ¼ bO r pc, where bO is the geometric correction factor calculated with CPAT due to remote loading without residual stresses. Crack Growth Analyses. AFGROW version 5.03.03.22 was used to conduct the crack growth simulations. The total SIF was calculated by superimposing the SIF due to residual stresses, KRS , onto the SIF due to external loads. The effect of residual stresses was included in AFGROW using the Residual K feature from the residual SIF calculated using CPAT. The SIF due to remote loading was defined with geometric correction factors, bO or bO=RS , using the User-Input Beta feature. The bFM correction developed for modelling the effect of the ForceMate bushing under remote loading was compounded to bO using the K-Solution Filters feature in AFGROW, which scales the calculated SIF according to the loading direction (tension versus compression) and crack size. The crack growth simulations were conducted using the Harter-T crack growth model that was used for the CPAT analyses. The generalized Willenborg crack growth retardation model was enabled using the default option for aluminium (SOLR = 3.0). Despite that the fact that retardation model reduces the crack growth rate and provides longer crack growth lives, it was found to have insignificant impact on the calculated LIF for the tested case study as the open hole and ForceMate configurations were both affected by the retardation model. The crack growth results normalized to the residual stress free open hole configuration are presented in Fig. 11. As shown, the prediction obtained with the ForceMate correction, bFM , and the residual stress effect, KRS , provides significantly longer crack

Fatigue Life Prediction at Cold Expanded Fastener Holes

671

growth life compared to the open hole configuration with and without the residual stress effect. This demonstrates the impact of the tangential stress reduction presented in Fig. 9 due to the high interference bushing. These results suggest that the beneficial effect of residual stresses alone in an open hole may have less impact on crack growth than the effect of the high interference bushing. The crack growth LIF are directly obtained from Fig. 11 as the crack growth curves were normalized to the open hole configuration without residual stresses. As the beneficial effect of the ForceMate is localized in the region surrounding the bushing, the LIF decreases as a function of crack size. In addition, the tensile residual stresses away from the bore reduces the beneficial effect of the ForceMate as the crack grows. This is clearly illustrated in Fig. 11 with the ForceMate configuration, where the LIF decreases from 26 to approximately 4.9 as the final crack size increases from 0.279 mm to 9.81 mm (through-the-thickness crack) from the initial crack size of 0.254 mm.

Normalized cycles to crack size

30 25

ForceMate with residual stress

20 15 10 5

Crack growth LIF

Open hole with residual stress

7

(from 0.254 mm to 2.5 mm)

Open hole (LIF = 1)

0 0.25

2.5

Crack size (mm)

Fig. 11. Effect of high interference bushing on crack growth curves.

5 Concluding Remarks A methodology was developed for calculating the LIF resulting from installation of a ForceMate bushing. The methodology covered the crack nucleation phase and the crack growth phase from a 0.254 mm quarter-circular corner crack. One of the key inputs for the determination of the LIF was the analytical calculation of the residual stresses resulting from the installation of a ForceMate bushing. This was achieved using the FE method by simulating the installation process (mandrel pull and bushing reaming). The analyses required using contact algorithms and include the effects of geometric and material non-linearities. As demonstrated, the resulting LIF is

672

Y. Bombardier et al.

sensitive to the accuracy of the calculated residual stresses. While this is challenging, it is necessary to experimentally validate the calculated residual stresses to gain confidence in the resulting LIF. Current experimental residual stress measurement efforts for quantifying the residual stresses due to cold working (Carlson et al. 2017) are being leveraged by the authors in order to substantiate the FE-based approach for calculating residual stresses. The current methodology for calculating the crack growth life within a residual stress field with a high interference bushing relies on simplifying modelling assumptions, which may reduce the accuracy of the crack growth predictions. Specifically, the effect of residual stresses and the presence of a high interference bushing are currently treated independently. Ideally, the crack propagation analysis should be conducted within the three-dimensional calculated residual stress field, allowing residual stress redistribution as the crack grows into the structure. While the proposed methodology provides a workaround solution, the level of confidence could be greatly increased if the number of steps and simplifying assumptions are reduced. Based on the crack growth analyses, the modelling of the ForceMate bushing highlighted the impact of filling a hole as opposed to assuming an open hole. As most damage tolerance analyses are performed by assuming an open hole, longer crack growth life could be obtained by including the effect of the filled hole, provided that the filling material and tolerances are correctly taken into account. In addition, other effects, such as fretting, clamp-up stress, and assembly stresses, should be considered. The calculated LIF for the CF-188 MLG uplock case study are summarized in Table 2 for the crack nucleation and crack growth phases. While there are several aspects of this analytical study that need to be validated experimentally, the calculated LIF seems to provide a conservative estimation based on experimental results obtained for many applications (Reid 2011).

Table 2. Summary of the calculated life improvement factor. Fatigue phase Fatigue crack nucleation to a 0.254 mm quarter–circular corner crack Fatigue crack growth from a 0.254 mm quarter–circular corner crack to a through-the-thickness crack

Life improvement factor 6.6 4.9

References Carlson, S., Andrew, D., Pilarczyk, R.: Making significant strides towards engineered residual stress implementation (ERSI). In: 2017 Aircraft Structural Integrity Program Conference, Jacksonville, FL, USA, 27–30 November 2017 DoD (Department of Defense): Joint service specification guide: aircraft structures, Rep. No. JSSG-2006, p. 497 (1998) LexTech: AFMAT crack growth database (2018). https://www.afgrow.net

Fatigue Life Prediction at Cold Expanded Fastener Holes

673

Reid, L.: Applying the damage tolerance approach to expanded bushing and rivetless nut plate installations. In: ICAF 2011 Structural Integrity: Influence of Efficiency and Green Imperatives, pp. 797–810. Springer, Dordrecht, Netherlands (2011) Renaud, G., Liao, M., Li, G.: Verification and validation of analytical methods to determine life improvement factor induced by engineered residual stresses. In: 29th ICAF Proceeding, pp. 2224–2233 (2017) Smith, K.N., Watson, P., Topper, T.H.: A stress-strain function for the fatigue of metals. J. Mater. 5, 767–778 (1970) Tada, H., Paris, P.C., Irwin, G.R.: The Stress Analysis of Cracks Handbook, 3rd edn. ASME Press, NewYork, NY, USA (2000)

Why Should We Encourage Usage of Interference-Fit Fasteners at Airframe Structural Joints? Carmel Matias(&) and Ekaterina Katsav Fatigue and Damage Tolerance Department, Engineering and Development Group, Israel Aerospace Industries (IAI), Ben-Gurion International Airport, 70100 Lod, Israel [email protected]

Abstract. It is well established that using interference-fit fasteners will obtain longer fatigue lives to airframe structures, compare to using transition-fit fasteners (or close-tolerance) and certainly clearance-fit fasteners. But, common practical manufacturing considerations, drive to less usage of the interference-fit fasteners (due to various installation difficulties of these fasteners being applied into the corresponding holes in the structure layers). In addition, concerns may be raised whether the fatigue advantage is actually being kept for any practical interference-fit installation method (or even fatigue disadvantage may occur due to installation procedures). It seems that there is lack of information regarding the influence on fatigue lives for the different practical installation methods of the interference-fit fasteners. This study presents testing results supported by analyses, for the influence on fatigue life, of the following two main parameters: (I) The fastener-to-hole fit level. (II) Two different common manufacturing practice for fastener installation methods of: hand plastic hammering and pneumatic steel hammering. The study shows that the fatigue advantage of interference-fit fasteners, is being kept even for the more aggressive installation method. The study results show that whenever fatigue life improvements are needed for structural joints, usage of interference-fit fasteners for these joints, is a good option to achieve it. Keywords: Interference-fit Fatigue


Fasteners
Structural-joints
Manufacturing


1 Introduction It is well established that, structural joints using interference-fit fasteners will obtain longer fatigue lives compare to using transition-fit, close-tolerance or clearance-fit fasteners (Crews 1975; Mann et al. 1983; Finney 1993; Chakherlou et al. 2010). However, common practical manufacturing considerations, drive to less usage of the interference-fit fasteners. This is mainly due to interference-fit fasteners installation difficulties (such as corresponding holes alignment for fastener insertion, etc.). These difficulties are amplified for principal structural joints containing three member layers, as is for double shear splice joints (attachments of two separate spar sections or double © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 674–691, 2020. https://doi.org/10.1007/978-3-030-21503-3_54

Why Should We Encourage Usage of Interference-Fit Fasteners

675

shear skin splices, etc.). This means that were we the most might need interference-fit fasteners, usually we will not use them. For such joints to have interference-fit fasteners, the fasteners will be installed using pneumatic steel hammering, in somewhat “aggressive” practical manner. The IAI Process Specification (IAI Process Specification 2008) that is being used for installation of interference-fit Hi-Lok fasteners, calls for fastener installation either by tapping the head with a plastic mallet or by driving the pin with a light rivet gun. But, according to the practical manufacturing experience for production of structural joints as mentioned above, driving the interference-fit pins into the holes is usually achieved by using pneumatic rivet guns that could not be considered as “light” methods. Concerns were raised that practical “aggressive” installation procedures, might damage the holes such that the fatigue advantage will be insignificant, or even present fatigue disadvantages. There is lack of information for the influence on fatigue lives of the different practical installation methods for these interference-fit fasteners. Due to these concerns, lack of relevant information and manufacturing requirements, there is tendency not to benefit of interference-fit fasteners in design of such joints. The present study addresses this concern by performing fatigue tests accompanied with STRESSCHECK finite element (StressCheck 2014) analyses, and strain-life fatigue analyses using “FATLAN” computer program (FATLAN 1992). There are quite many variables and parameters that would influence the fatigue life of these structural members, such as: the interference-fit level, different levels of load transfer (between the layers by the fastener), fastener clamp-up level, fretting and friction between faying surfaces, and variable-amplitude loading. The present study investigates the influence upon fatigue life of different fastener installation methods accompanied by controlling the dominant parameter of the fastener-to-hole fit level (interference-fit, transition-fit). This study avoided additional parameters involvement. Investigation of different fastener installation methods done under one controlled dominant parameter enables better understanding of the phenomena. As so, the study avoided including fasteners load transfer by uniformly induce the cyclic tensile loading into both the two layers of the tested specimen. The interference-fit studied in this article scope, deals with fastener-to-hole plasticity induced phenomena due to fastener hammering installation methods, and not the phenomena related to “shrink-fit” fastener installation methods. The tested specimen features are representative of typical structural members. The study presents experimental investigation supported by analyses, for the influence on fatigue life, of the following two parameters: (1) Fastener-to-hole fit level, per % of diametric interference (positive value means that the fastener diameter is larger than the hole diameter, vice versa is clearance). (2) Two different fastener installation methods of: Hand plastic hammering and pneumatic steel hammering as is practically used at the manufacturing line. The study shows that the fatigue advantage of obtaining longer fatigue lives is being kept for common manufacturing practice of fastener-to-hole interference-fit installation using pneumatic steel hammering method. Significant improves in the

676

C. Matias and E. Katsav

fatigue life are being presented for interference-fit fasteners installed via pneumatic steel hammering compare to transition-fit fasteners installed via plastic hand hammering (and the reference open-hole specimens). The study suggests positive relation between interference level to fatigue life, i.e., higher interference (per specification) contribute to longer fatigue lives, even for using “aggressive” pneumatic steel hammering. The high analytical fatigue life results shown for interference-fit configuration (relative to open-hole configuration), suggests that any “aggressive” practical installation method, will not be a factor that is having significant influence on the fatigue life. The study suggests that any damage that might be induced to the hole by an “aggressive” installation procedure (as flaws in the hole bore) will have negligible growth. The conclusion from that study is that we should encourage usage of interference-fit fasteners, at structural joints, whenever fatigue life improvements are needed.

2 The Experimental Investigation The purpose of the test program was to enquire the effect of interference-fit fastener holes and practical manner of the fasteners installation on the fatigue life, for specimens representing of typical aircraft structure features and fasteners installation procedures. Test Specimens. The test program included specimens that are composed of two layers of AL7050-T7451, 6.35 mm thick plates, attached by 4 Titanium Hi-Lok fasteners of 7.94 mm diameter (HLT336AP10-8). The specimens are representing of typical aircraft structure features and installation procedures. In addition, “open-hole” specimens were also tested to be used as a reference base-line. Each “open-hole” specimen was also composed by the same two layers of AL7050-T7451 6.35 mm thick plates, but not attached via fasteners, and just aligned the two layers next to each other at the tensile machine (so that holes at each layer will be concentric). The specimens were manufactured to be grouped into four distinct specimen feature type, separated by hole-tofastener interference level and fastener installation method. Total of 13 specimens (26 plates) were tested. The total of 13 specimens were composed of 4 specimens for the “open-hole” group (group A), and 3 specimens for each of the three other groups, as presented at Table 1. Table 1, presents the different specimen features and dimensions, and the number of specimen tested per each type group. The data of the holes diameter in the specimens and the fastener diameter are in accordance to specifications (IAI Document No. 25G045/020929, Holes Specification Report) and (“hi-Lok” “hi-tigue” ® pin, Drawing No. HLT336) respectively. The fasteners installation was performed according to IAI’s common manufacturing practice for the different fastener-to-hole interference levels and according to the IAI’s Process Specification (IAI Process Specification PS 070500, Installation of Hi-Lok Fasteners). The two different fastener installation methods were applied via regular IAI manufacturing procedures, tools and personal. Figure 1, presents the specimen dimensions. All specimens were made with the grain direction parallel to the specimen longitudinal axis (L-T direction per ASTM E399 nomenclature), as presented at Fig. 1.

Why Should We Encourage Usage of Interference-Fit Fasteners

677

Figure 2, presents the corresponding tools used for the two different fastener installation methods of Hand Plastic and Pneumatic Steel Hammering. The Figure presents also specimens tested via the tensile machine at the IAI Ground Testing Center.

Fig. 1. Specimen dimensions.

Table 1. Tested specimen features and dimensions.

Specimen Type

A

Open hole 5/16" Dia.

Fastener Application N/A

Hole size "d" diameter [inch] Min.

0.3085

Max

0.3125

Fasteners diameter [inch] N/A

B Transition-fit

Hand Plastic Hammering

Min.

0.3110

Min.

0.3115

Max

0.3125

Max

0.3120

C

Interferencefit

Hand Plastic Hammering

Min.

0.3085

Min.

0.3115

Max

0.3100

Max

0.3120

D

Interferencefit

Pneumatic Steel Hammering

Min.

0.3085

Min.

0.3115

Max

0.3100

Max

0.3120

"e" [inch]

e/d

0.67

2.1

3

1.05

3.4

1

0.67

2.1

3

0.67

2.1

3

0.67

2.1

3

Quant ity

678

C. Matias and E. Katsav

Table 1 – Notes (diameter data is presented exactly as appear in the specifications): (1) All holes were drilled & reamed in a set of the two plates layers held together. (2) Each hole drilled and reamed was inspected for its quality and specific dimensions. Each hole was given a specific identification (ID No.), and was followed accordingly throughout the test. (3) For the specimen group types of C and D that includes 6 specimens (sets of two plates layers per specimen) drilled with 4 holes at each specimen (total 48 holes), sorting procedure was done to these 6 specimens (keeping as a set the drilled two plates layers) to decide if a specimen belongs to group type C or type D. This sorting procedure was performed before installation of the fasteners. The specimens were sorted by actual reamed hole diameter so that, 3 of the specimens with larger hole diameter were chosen to the hand plastic hammering. For each specimen all its 8 holes were measured, and for each specimen the smallest hole diameter (of the 8 holes) was the one to determine the group. It was not allowed for a specimen to be with combined fastener application methods. (4) Hole diameters that are specified at the table are per spec. (“hi-Lok” “hi-tigue” ® pin, Drawing No. HLT336). The actual final reamed holes diameter that was achieved for each specimen group ranged as follows: Type A – 0.3090” to 0.3114”, Type B – 0.3106” to 0.3125”, Type C – 0.3086” to 0.3100”, Type D – 0.3086” to 0.3090”. (5) There was no need for sorting procedure of the fasteners, as all were having the same exact diameter measured (according to calibrated Caliper device used).

Fig. 2. Specimen fasteners installation tools and testing tensile machine.

Why Should We Encourage Usage of Interference-Fit Fasteners

679

Test Loading Spectrum. The specimens were fatigue tested using constant amplitude cyclic tensile loading of ratio R = 0.05 (R = Min._Load /Max._Load), and peak stress of 111.07 MPa. In addition, each 500 cycles, of the cyclic loading specified above, was followed by 5 cycles of about 20% higher loading (with the same R = 0.05), which served as a “simple” set of load markers to enable future performance of detailed Fractographic analysis (to failed specimens that had developed cracks at fastener holes). Thus, the fatigue testing cyclic loading was composed of loading spectrum blocks as presented at Table 2. The specimens were cyclic loaded as they were held by clamping jaws, so that each layer, was applied with the same loading. No load transfer was induced by the fasteners. Each specimen, before it was cyclic tested, had undergo a calibration procedure. Each specimen had 4 strain gages, bonded at the center line on both faces and both sides (thickness dimension). Figure 3 presents the strain gages arrangement. Calibration procedure included stepped loading application up to a static load of 50% maximum spectrum loading. The strain data was recorded. Accordingly, the specimens were aligned (as needed) until difference between two opposite (face to face or side to side) strain gage readings were lower than 50 micro strains. Table 2. Tested Loading Spectrum Block.

*Note: The loads specified at Table 2 were relating to the nominal specimen dimensions (of specimen thickness and width). Each specimen tested, was measured for its actual thickness and width dimensions, and specimen specific cyclic loading were adjusted in order to achieve the stresses specified at Table 2.

Fig. 3. Specimen strain gages arrangement.

680

C. Matias and E. Katsav

3 The Experimental Investigation In order to properly evaluate the test results, each hole of each specimen that had undergone fastener installation, was evaluated for its actual interference level, and was followed up during the test. For this purpose, measurements were taken for the actual fasteners and holes diameter. It was found out that there were no measured variations in the fasteners diameter. As so, the fastener diameter was taken into account at the “mid.”-specification diameter. Regarding the specimen holes, sorting procedure was done as specified in the Specimen Section Table 1. Accordingly, the interference level produced for each specimen was known. Strong correlation between cracking findings to the interference level, was noticed. It was found out that cracks appeared at holes of which were having lowest interference level (among the 8 holes per specimen). Table 3, presents the specimens geometrical dimension data and the number of cycles to failure for each specimen. In addition, for each specimen having fasteners installed, it is specified for all its 8 holes (per specimen), the average, the minimum and the maximum interference level (the “–” sign indicates of clearance level). The Table presents also for each specimen type, the “average” number of cycles-to-failure according to Weibull distribution 63% probability data per the test results. In addition, the Table information is being addressed by specific notes to present exceptions, elaborations and further information about the data. Since strong correlation was found between the cracking findings at the holes, to the lowest interference levels (among the 8 holes per specimen), Table 3 highlights the following two columns: “Cycles to specimen failure” & Minimum % of interference. Test Results for the Open-Hole Specimens. 4 open-hole specimens were tested. These specimens presented cycles to failure results ranged from 122,000 cycles up to 585,000 cycles. The result of 585,000 cycles seems to be an exceptional result. But, since no test anomaly was evident for this “high-life” result specimen, this result was taken into account for the results evaluation. Applying the Weibull distribution to these test results (not including the Spare-Specimen result) gave a very low shape parameter of 1.5. As we would expect for aluminum alloys fatigue life Weibull shape parameter of 4, this clearly indicates that we got too much variance level for these results. The 63% Weibull probability result per these test data (not including the Spare-Specimen result) gave an “average” of 325,000 cycles to failure (if the “high-life” result specimen was not accounted an “average” of 150,000 up-to 200,000 cycles to failure would have been expected). Figure 4 presents typical type A specimen cracking at hole. The SpareSpecimen result is presented as an additional data-point to look at (par a different test). Test Results for the Transition-Fit-Hole & Plastic Hand Hammering Specimens. The 3 specimens of type B presented cycles to failure results that ranged from 325,000 cycles up to 357,000 cycles. Applying the Weibull distribution to these test results gave shape parameter of 4.0 (as expected for aluminum alloy fatigue life). The 63% Weibull probability result per these test data gave an “average” result of 344,500 cycles to failure. Figure 5 presents all the Type B specimens cracking and failure results. It is seen for specimens B1 and B3, that they had developed primary and secondary cracking at hole (at the two specimen layers) up to specimen failure. When comparing the cracking pattern of specimens B1 and B3 to the cracking pattern of all the specimens of type A (the open hole specimens), the following differences can be seen:

Why Should We Encourage Usage of Interference-Fit Fasteners

681

• Type B specimens cracking pattern – Symmetry (similarity) is noticed between the 2 layers for cracking pattern. At both layers there is primary cracking from hole to the edge and secondary cracking at other side of the hole, almost at the same cracking rate for the 2 layers. This can be explained by the presence of fastener at the hole that for each increase increment of crack length at one layer the fastener immediately shares corresponding loading increments with the other layer. • Type A specimens cracking pattern – There is no symmetry (similarity) between the 2 layers for cracking pattern. Hole to edge primary cracking is only in one layer. Secondary cracking is at other side of that hole, and in the other plate layer. This can be explained by lack of fastener at the holes of these specimens. For specimen B2 it is seen that the specimen had failed not due to cracking at holes section, but due to cracking at section next to the tensile machine clamping jaws. In addition, one hole has been cracked all-the-way to the edge. Since that at the time of specimen failure there was a cracked fastener hole for the entire edge ligament, that crack had been taken into account for the specimen type B statistics. The corresponding interference level for the fastener holes that had primary cracking for specimens B1, B2 and B3 were found out to be −0.14%, −0.14% and −0.24% respectively (“−” sign means that these interference levels actually are clearance levels). Test Results for the Interference-Fit-Hole & Plastic Hand Hammering Specimens. The 3 specimens of type C gave each one, a different type of result, so no statistics could be made out of these test results. C1 specimen, as presented at Fig. 6, had developed a crack up to its failure at the tensile machine clamping jaws, and no cracking was detected at any of the fastener holes. The minimal interference level (per 8 fastener holes) was 0.89% (maximal interference was 1.02%). This test result can be explained as stress concentration at the clamping jaws is higher than at the fastener holes having 0.89% to 1.02% interference. C3 specimen, as presented at Fig. 6, had developed cracking that initiated at fastener holes, simultaneously at the two plate layers that share the same fastener (due to fastener load transfer). Also here, crack rates are similar for the two plates layers. The specimen had failed at that fastener hole section. An “anomaly” (not expected) cracking scenario is presented, that the primary cracking is for a crack growing from one fastener hole to the other adjacent fastener hole (and not from the fastener hole to the plate near edge). The minimal interference among the 8 fastener holes was 0.56%, and the maximal was 1.02%. The crack initiated at the fastener hole having the minimal interference level. C2 specimen was not tested. The specimen was defected getting a permanent bow deflection due to erroneous machine compression order of beyond control limiter. Test Results for the Interference-Fit-Hole & Steel Pneumatic Hammering Specimens. Figure 7 presents a typical specimen of type D, tested up to specimen failure. For all type D specimens, primary cracking up to section failure had been occurred at the tensile machine clamping jaws. For all type D specimens, no cracking had been developed at any of the fastener holes. For two of these specimens, secondary cracking had been developed at a section positioned at some distance below the fasteners line section (apparently caused by fastener head acts as “hard-point” under out-of-plan bending at one layer due to cracking near the tensile machine clamping jaws at the

682

C. Matias and E. Katsav

other layer that compose the specimen). The 3 specimens of type D presented cycles to failure results that ranged from 640,000 cycles up to 840,000 cycles. Applying the Weibull distribution to these test results gave shape parameter of 4.0 (expected for aluminum alloy fatigue life). The 63% Weibull probability result per these test data gave an “average” result of 766,000 cycles to failure. All type D specimen fastener holes (total 24 holes) interference ranged from 0.89% to 1.02%. Fatigue Life Results vs. Fastener-to-Hole Interference & Fastener Installation Method. Figure 8, presents the specimen fatigue lives vs. interference data & method of fastener installation. The interference data presented for each specimen, are per the lowest interference level of its 8 holes (due to cracks related to lower interference). Each different specimen type test result, is marked, in Fig. 8, by a different distinct filled marker shape. The Weibull probability “average” results per each different specimen type are marked by the corresponding distinct marker shape, with a hollow marker shape. For all specimen results that no cracking had been developed at any of their holes, an arrow is added to their test result marker (expected that due to cracking in holes, life would had been longer). The test results presented at Fig. 8 show that: • The average life result of the Transition-fit Hand Plastic Hammering specimens (Type B specimens), is slightly higher than the average life result of the Open-hole specimens (Type A specimens), but not by a much high factor. If we consider that the cracked fastener holes at Type B specimens, were at these holes having negative interference-fit levels (i.e. clearance-fit levels), these results can be interpreted that such holes will behave not much different then open holes. Accordingly, we can say that fatigue behavior of fastener holes having clearance-fit levels will be not much different, then fatigue behavior of open holes. • The average life result of the Interference-fit Pneumatic Steel Hammering specimens (Type D specimens), is significantly higher than the average life result of the Open-hole specimens (Type A specimens). In addition, we should keep in mind that the test results of the Type D specimens did not develop any cracking at any of their fastener holes. As so, these specimens are expected to present much higher fatigue life results for failures to occur at these specimens Type D interference levels (but unfortunately these tests did not capture this). More evaluations for this issue will be presented at the following section. • Regarding type C specimens (the ones of Interference-fit Plastic Hand Hammering specimens), we can say nothing about an average life result due to large variation type in these three specimens, of which one was damages so that it could not be tested, one had failed due to cracking near the Tension Machine Clamping Jaws, and one had failed due to “abnormal” cracking at the fasteners holes (i.e. not expected and not explained cracking scenario, of which primary crack developed from one hole to the adjacent hole, while expected to be developed from hole to near edge). Accordingly, the test results discussion will not include the results of type C specimens. Fatigue Life Test Results Discussion. A Relative Life Ratio (RLR) can be a term used to present the ratio of an average fatigue life of a specific fastened holes configuration over the average fatigue life of the “open-hole” specimens configuration (of which are regard as a reference base-line).

Why Should We Encourage Usage of Interference-Fit Fasteners

683

Accordingly, we can denote a term of RLRTestTypeD=OH to specify the ratio of the average fatigue life test result of the Type D fastened holes specimens (the configuration of * 1% interference & pneumatic steel hammering) over the average fatigue life test result of the Type A specimens (the configuration of the open-hole specimens). As so, according to the test results, it can be said that: RLRTEST Type D=OH [ 766; 000=325; 000 ¼ 2:36 It can be seen at Fig. 8 that as fastener-to-hole interference level is higher, the fatigue life also tends to be higher. Accordingly, we can say that a clear relation of interference level to fatigue life, is identified, of which, higher interference levels Table 3. Test Results – Dimensions /Cycles to failure /Interference Levels.

684

C. Matias and E. Katsav

contribute to higher fatigue lives, and this is even with using pneumatic steel hammering (“aggressive” method), for the fasteners installation. (a) If no note is specified, then, the specimen had failed at holes section, Otherwise, cracking was as specified by notes: (1) to (5), below. (b) Strain measurements taken at load of 50% of the maximum cyclic load. (c) The spare specimen material is AL7075-T7351 (AL7050-T7451 per all other). (1) Specimen failed at section next to the Tension Machine Clamping Jaws, and, at time of failure, one holes had cracked entirely through its edge ligament. (2) Specimen failed at section next to the Tension Machine Clamping Jaws. At time of failure no crack was detected at any hole.

Fig. 4. Typical type A specimen tested and cracked up to failure

Fig. 5. The three type B specimens tested and cracked up to failure

Why Should We Encourage Usage of Interference-Fit Fasteners

685

(3) Defected specimen of permanent bow deflection (over limiter compression). (4) No crack detected at any hole. 1st plate cracked & failed at Clamping Jaws section. 2nd plate cracked & failed at section near the holes below fastener head. (5) Abnormal cracking of primary growth from hole to hole (expected: hole to edge).

Fig. 6. C1 and C3 specimens tested and cracked up to failure

Fig. 7. Typical type D specimen tested and cracked up to failure

686

C. Matias and E. Katsav

Fig. 8. Cycles to failure per interference and fastener installation method

4 Numerical Analyses and Evaluations The analysis used 2D STRESSCHECK FEM (“StressCheck” FEM Computer Program, Version 10.1, June 2014, by ESRD) to simulate the cyclic loaded tested specimen having all their geometrical features, representing the four 7.94 mm diameter holes, hole-to-edge distance of 17 mm (e/d = 2.1), for the following 3 different configurations: – Open holes specimen. – Interference-fit fastener hole specimen for a level of 1% interference. – Interference-fit fastener hole specimen for a level of 0.5% interference. All analyses used p-version elements up to level of 8. Linear analysis was conducted for the open hole specimens, while non-linear analysis was performed for the interference-fit specimens. Half specimen was modeled via symmetry conditions in center-line. The analyses used automatically constructed tetragonal mesh of 436 elements for the interference-fit and 124 elements for the open hole analyses. Material was represented by elasto-plastic AL7050 Ramberg-Osgood properties, E = 0,300 ksi, m = 0.33, ry = 64 ksi and n = 15.2. The interference-fit fastener was modeled by Titanium material properties of E = 16,000 ksi and m = 0.31. A special element represented the fastener inside the hole. This element can transmit the interference displacements.

Why Should We Encourage Usage of Interference-Fit Fasteners

687

For the interference-fit analyses, non-linear runs were conducted using incremental steps. First stage was that the special element representing the interference-fit fastener was attached to the hole with application of a displacements (corresponding to 1% and 0.5% interference levels), using three incremental steps. Second stage was the application of the remote loading, as traction applied to the upper edge of the plate in three incremental steps. The remote loading was applied to simulate the test cyclic loading of Max. cyclic level of 16.1 ksi and Min. cyclic level of 0.805 ksi (R = 0.05). The FEM had undergone validation procedures, via comparison to strains measured. FEM & Fatigue Analyses Results, and Discussion. FEM result for open hole specimen give a net section geometrical stress concentration factor at the holes of Ktn = 2.88 (gross section of Ktg = 3.58). For this configuration the Fatigue Notch Factor KN, is evaluated (according to IAI Fatigue Manual, Feb. 2003) as KN = 2.11. Fatigue life result based on strain-life analysis done (using “FATLAN” – “FATigue Life Analysis” Computer Program August 1992) give 135,000 cycles. This analyzed fatigue life result for the open hole specimen correlates relatively well with the tests results. In order to obtain analyzed fatigue life results for the Interference-fit specimens, it is needed to know the geometrical stress concentration factor at the hole. Figures 9 and 10 present FEM analyses stress gradient results, along a line that passes through the hole center, starts at the hole edge and ends at the specimen center line, for 1% and 0.5% Interference-Fit levels, respectively. The stresses per distance from the hole center normalized to hole diameter, are presented at Figs. 9 and 10, for loading cases of: – Residual stress induced due to fastener installation, with no loading applied. – The Residual stress + 111.07 MPa remote loading per Max. spectrum loading cycle. – The Residual stress + 5.55 MPa remote loading per Min. spectrum loading cycle. Figure 11 presents hoop stress around the hole, for the residual stresses induced due to fastener installation of 1% interference, with no remote loading application. The Figure also presents Max. principal stress results, for residual stresses per 1% fastener interference, combined with 111.07 MPa remote loading. The following can be seen: – The residual stresses induced combined with remote loading presents significant stress concentration reduction at hole edge (relative to open hole configuration). – The residual stresses combined with remote cyclic loading presents reduction in local cyclic stresses R-ratio at the hole edge (relative to applied cyclic loading). These effects drive to fatigue advantages of interference-fit fasten joints for obtaining longer fatigue lives relative to transition-fit or clearance-fit fasteners. The reduction in the stress concentration and in the local spectrum cyclic stresses R-ratio at hole edge, is more dramatic for the 1% compare to the 0.5% interference-fits.

688

C. Matias and E. Katsav

Fig. 9. Stress gradient results for 1% Interference FEM

Fig. 10. Stress gradient results for 0.5% Interference FEM

Why Should We Encourage Usage of Interference-Fit Fasteners

689

Fig. 11. 1% Interference-Fit FEM analysis stress results plot

Table 4 presents stresses for hole edge & maximum gradient along the line passing through hole center (hole edge-specimen center) for loaded interference configurations. The Table presents also, for the hole edge only, the corresponding stress concentration factors (per Max. spectrum load) and the corresponding local R-ratio, for the two interference-fit levels. Note that, even though along the stress gradient we get higher stress than at the hole edge, we address only the hole edge stress to calculate the stress concentration and the local R-ratio, for fatigue analyses. This is due to the fact that usually hole edge periphery is the one prone to cracking. Table 4. Stresses, Kt’s and Local R-ratio for loaded interference configurations

690

C. Matias and E. Katsav

The above Kt’s & local R-ratio correspond well with data per study done by NASA (Crews 1975). Using that data for 0.5% & 1% interference levels, to perform strain-life fatigue analyses using “FATLAN” computer program, we get fatigue life results in the range of 38,500,000 cycles to 1*1010 cycles. Relative Life Ratio (RLR). RLR specifies ratio of fatigue life for any given structural configuration (interference, etc.), to a reference fatigue life of open-hole configuration. The term expresses amount of fatigue life advantage exists at a structural configuration. According to the analyses and the test results presented, the following can be said: • The theoretical Interference-fit RLR according to analyses is: [Any method & 0.5%-1% Interference fatigue life] / [Open-hole fatigue life] = [38,500,000 cycles to 1*10^10 cycles] /135,000 cycles = 280 to 74,000 • The practical Interference-fit RLR according to this study test is: [Average 0.89%–1.02% Interference specimen life] / [Average Open-hole life] > 766,200 /325,000 cycles = 2.36 The sign of “ > ” is due to the fact that all Interference-fit specimen failed by cracking at the tensile machine clamping jaws, and no crack was initiated at any of their holes. • Practical Interference-fit RLR according to literature (T. N. Chakherlou et al., ELSEVIER 2010, for example): [% Interference* & loading levels tested life] /[Open-hole tested life] = 2.8 – 5.3 * Not including different common practical fastener installation methods.

5 Conclusion The study shows that “aggressive” installation practice of pneumatic steel hammering, keep obtaining longer fatigue life for interference-fit fasteners usage. Analytical fatigue life advantage significantly greater then what we see for practical applications, suggests that fastener installation method is not a significant influence factor (some extent of flaws in the hole that might be induced is insignificant for fatigue life). For maximizing the fatigue advantage, usage of pneumatic steel hammering for 1% interference is preferred over usage of light hand plastic hammering for less interference (up to 0.5%).

References ASTM E399 - 09 Standard Test Method for Linear-Elastic Plane-Strain Fracture Toughness KIc of Metallic Materials “FATLAN” – “FATigue Life Analysis” Computer Program, version B.1.1 August (1992) “hi-Lok” “hi-tigue” ® pin, protruding tension head, titanium, 1/12” grip variation, Drawing No. HLT336, Rev.: 10, Date: D.P.S. 12-23-87 IAI Process Specification PS 070500, Installation of Hi-Lok Fasteners, Revision: D2, 13 February 2008

Why Should We Encourage Usage of Interference-Fit Fasteners

691

IAI Document No. 25G045/020929, Holes Specification Report, Rev.: A, 15 February 2004 IAI Fatigue Manual, TR SAT/650/020554, Issue B, February 2003 Finney, J.M.: “Cold Expansion and Interference for Extending the Fatigue Life of Multi-Layer Metal, Joints”, Department of Defense, Defense Science and Technology Organization, Aeronautical Research Laboratories, Mewourne. Victoria, Research Report 17 October 1993 Mann, J.Y., Machin, A.S., Lupson, W.F., Pell, R.A.: “The Use of Interference-Fit Bolts or Bushes and Hole Cold Expansion for Increasing the Fatigue Life of Thick-Section Aluminum Alloy Bolted Joints”, Department of Defense, Defense Science and Technology Organization, Aeronautical Research Laboratories, Mewourne. Victoria, Structures Note-409, August 1983 John H. Crews, Jr.: “Analytical and Experimental Investigation of Fatigue in a Sheet Specimen With an Interference-Fit Bolt”, NASA Technical Note – TN D-7926, Langley Reseurcb Center, Humpton, Va. 23665, July 1975 “StressCheck” FEM Computer Program, Version 10.1, June 2014, by ESRD Chakherlou, T.N., Mirzajanzadeh, M., Abazadeh, B., Saeedi, K.: An investigation about interference fit effect on improving fatigue life of a holed single plate in joints. European Journal of Mechanics A/Solids, ELSEVIER 29, 675–682 (2010)

Full Scale Fatigue Testing of Aircraft and Aircraft Components

Analysis Prediction and Correlation of Fiber Metal Laminate Crack Growth in Semi-Wing Full-Scale Test Willy R. P. Mendonça(&) and Danielle F. N. R. da Silva Embraer S.A., São Paulo, Brazil [email protected]

Abstract. This paper aims to demonstrate the correlation between simulation and experimental results obtained for artificially inserted crack propagation in the Full-Scale fatigue test of a semi-wing, developed by Embraer. The cracks were inserted on the lower wing skin, which was manufactured in Fiber Metal Laminates (FML), in order to validate the analysis methodology in this material. The application of cracks in a Full-scale test allowed the evaluation of several damage scenarios that could hardly be reproduced with high fidelity in panel tests, by incorporating: design details without simplifications and load redistribution among structural components. The analysis and test results will be substantiated and discussed. Keywords: FML Delamination


Full-Scale
Semi-Wing
Crack
Propagation


1 Introduction FML are a new class of material that exhibit excellent damage tolerance (crack growth performance and residual strength), while preserving similarity with aluminum structures. As the average aircraft economic life is over 30 years and there are new perspectives of increasing this time, the inspection and maintenance costs are important design drivers. FML are formed of aluminum sheet (aluminum alloy 2000 series are the most common) and a composite material (usually epoxy polymer as matrix and a glass fiber type S2 as reinforcement) bonded together under pressure and heat, forming a compact laminate. The FML has very low fatigue crack growth rate when compared to the monolithic metal. This rate depends on the laminate lay-up and on the fiber/metal volume. Several studies about crack propagation in FML have been published in the last decade, mostly related to simple coupons or subcomponents. A full-scale fatigue test provides an effective way to account mechanical behavior of an airframe structure, in which real structures with a realistic spectrum simulation allow the proper interaction of all structural components. In the context of a full-scale fatigue test, the Embraer R&D department developed a metallic advanced semi-wing concept under the dimensions and functionalities of a © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 695–707, 2020. https://doi.org/10.1007/978-3-030-21503-3_55

696

W. R. P. Mendonça and D. F. N. R. da Silva

typical regional aircraft wing. This advanced semi-wing was fabricated using an existent wing, by replacing some primary structures with parts manufactured with new technologies. The new primary structures were elected with the objective to prove the concept as well as to understand the behavior of the components during the real flight conditions test simulation. Mainly, the new technologies included in this semi-wing were: Friction Stir Welding (FSW), Structural Bonding and FML. The FML and Metal Bonding were applied in the lower wing skin as shown in Fig. 1.

Fig. 1. Detail of the semi-wing.

In more details, the lower wing panel was manufactured with: • FML skin - AA2524-T3 aluminum sheets alternately bonded to uni-directional fiber glass plies embedded in an adhesive system layers; • Extruded stringers - AA2024-T351 bonded to the skin. Figure 2 illustrates the semi-wing used in the full-scale test and highlights the lower wing panel manufacture in FML with bonded reinforces. In parallel with the test, it was developed and performed the analysis prediction for crack propagation in the FML lower wing skin. The aim of this paper is to evaluate the correlation between the analysis prediction and these experimental results obtained for artificially inserted crack propagation in this full-scale fatigue test. In the future developments, the methodology for this new material application will be used in optimized FML structures concepts. In the following chapters the full-scale fatigue test description, the methodology, results, discussions and finally the comments and conclusions will be presented.

Analysis Prediction and Correlation of Fiber Metal Laminate

697

Fig. 2. Lower wing panel assembled in semi-wing article.

2 Full-Scale Test The full-scale test of advanced semi-wing covered durability and after the propagation tests. The durability test was performed with a complex wing spectrum loading condition. This phase lasted 120.000 Flight Cycles (FC), which corresponds to two nominal lifetimes of a reference regional aircraft operation. The semi-wing was bolted at wing root to a rigid fixed structure. A computer controlled loading system was used in the test and it comprises 6 electro-hydraulic actuators mounted on the loading frames. The loading frames are parallel or almost parallel to the spars and are connected to crossbeams that attach the loading pads. The loading pads are the interface of the loading system to the semi-wing. The loads from the actuators were calculated to generate a representative fatigue load spectrum that was based on the reference regional aircraft. The load application system was designed using whiffletree. Figure 3 provides a view of the test setup. During the durability test, no cracks initiated naturally in FML after the 120.000 FC. In the sequence of durability test, crack propagation phase initiated, for which a constant amplitude loading was applied for 110.000 load cycles. The simplification in the spectrum (constant amplitude loading), was justified to contribute for methodology validation (Alderliesten 2017a). Some artificial damages were introduced in this phase, depicted in Figs. 4 and 5. The cracks propagations were monitored during the test progress, such that many damage scenarios could be evaluated in depth.

698

W. R. P. Mendonça and D. F. N. R. da Silva

Fig. 3. Full-Scale setup.

Fig. 4. Artificial damages location in the semi-wing.

The results presented in this paper, as well the analysis prediction were obtained from the damage #6, that simulates a failure in the inspection window frame, made of FML.

Analysis Prediction and Correlation of Fiber Metal Laminate

699

Fig. 5. Detail of artificial damages introduced in the semi-wing.

3 Methodology The methodology for crack growth analysis in FML has been developed to enhance the crack growth prediction for complex structures and complex loads. This methodology was applied to predicted crack growth from artificial damages inserted in the Semiwing full-scale test. The development happened through several steps, with successive improvements in the aircraft representativeness (see Fig. 6). The increments sought to incorporate geometric details and load complexity. The mathematical models applied in the crack growth simulation in coupons, followed the analysis methodology developed in the University of TUDelft (Alderliesten 2005; Gupta et al. 2013; Khan et al. 2011; Rodi et al. 2009). This methodology yielding an accurate crack length and delamination predictions. Figure 7 depicted a delamination in a coupon tested.

700

W. R. P. Mendonça and D. F. N. R. da Silva

Fig. 6. Development flow chart.

Fig. 7. Delamination in a coupon tested.

For crack growth simulation in complex structures, complementary developments were implemented in the original methodology, establishing an analysis process that uses finite element analysis (FEA) to provide functions necessary to correctly represent mathematically a complex structure. The simulation flowchart is shown in Fig. 8. Following the flowchart presents in Fig. 8, the analysis begins with the failure scenario evaluation, for which the FEA shall be created. For this evaluation, it is important to define the crack growth path based on the aircraft structure knowledge. The FEA will provide: • The far field stress for simulation (rff); • The normalized stress intensity factor (SIF) solution function b = f(a), which can be obtained by one of the methods: Energy Release Method, Virtual Crack Closure (VCCT) or J-integral (Irwin 1957; Broek 1988); • The secondary bending function Mx = f(a), which depends on the layup, structure offset and damage mode.

Analysis Prediction and Correlation of Fiber Metal Laminate

701

Fig. 8. Analysis process - flow chart.

The functions obtained from the FEA and the other parameters (Loads, Materials, Layup and SN data) should be input in in-house developed crack growth program, which uses the widely disseminated formulations published by TUDelft (Wilson 2013; Spronk 2013). The algorithm is able to predict crack growth in different balanced or unbalanced FML layups; and thus it is applicable for practical configurations in structures and reinforcements. Figure 9 shows a flowchart of the crack growth algorithm, and it is similar to presented by (Wilson 2013) in his thesis. The in-house program applies the geometry factors and reference loads obtained from FEA as input data, which are used for the SIF solution computation and calculation of loading in all plies, through the classical laminate theory (CLT) (Alderliesten 2017b). The software will calculate the crack growth and the delamination simultaneously, generating the following results: curves da/dN vs. a, a vs. N and delamination geometry.

702

W. R. P. Mendonça and D. F. N. R. da Silva

Fig. 9. Crack growth algorithm flow chart, based in (Wilson 2013).

4 Results and Discussion Different damage scenarios were inserted into the FML lower wing panel. Crack growth experimental data made it possible to calibrate the analysis methodology developed and previously tested in coupons and subcomponents. For this study, the artificial crack inserted in inspection window frame, damage #6, it was selected to analysis, as shown in Figs. 5 and 10.

Fig. 10. Artificial crack in inspection window frame.

Analysis Prediction and Correlation of Fiber Metal Laminate

703

The expected crack growth path passes through different FML thicknesses and under a bonded stringer onto the skin (see Fig. 11A). Based on the expected path to the crack growth, it was created the FEA to obtain the functions, as shown in Fig. 11B.

Fig. 11. Crack growth path and FEM.

The functions obtained with this FEA considered that the stringer remained intact and bonded to the FML skin. It was observed after test that there were no cracks in the stringer, and the integrity of bonding between stringer and skin shall be verified with non-destructive and destructive inspections (C-scan and metallography) during the teardown. In another similar damage scenario, in a different inspection window (damage #1, see Fig. 5), the adjacent stringer was cut as part of the damage. Underneath the sectioned stringer, relevant secondary bending was observed, causing debonding in stringer/FML interface during the test. For the first prediction made, there was doubt about stringer influence in differential crack growth in each ply. The most conservative scenario idealized was that all FML metallic plies were cracked simultaneously, with same rate. The FEA construction was based on this premise. Figure 12 shows the calculated b(a); whereas the Mx(a) function is negligible in this simulation, since the intact stringer restricts the secondary bending. However the most likely scenario is that the metallic plies closest to the stringer should have a lower crack growth rate. The ratio between crack growth rates can be evaluated by fractographic examination of cracked surfaces in teardown inspection, after tests. Figure 13 shows the final crack length at external and internal plies. At internal ply the crack was under of the stringer and was not possible inspection the length after 45.000 load cycles. It is observed in the calculated b, that values dropped when the crack goes under the stringer, this will imply in a relevant delay in the crack growth.

704

W. R. P. Mendonça and D. F. N. R. da Silva

Fig. 12. Normalized SIF solution for damage #6

Fig. 13. Final crack length

Analysis Prediction and Correlation of Fiber Metal Laminate

705

Fig. 14. Simulation vs test propagation results

Figure 14 shows the simulation using this normalized SIF function. By plotting the simulated results with the experimental results, it was verified that the retard occur as predicted, which made the premise of simultaneous cracks growth in the same rate a conservative premise. The structures surrounding the damages were monitored through strain gages. The aim was to calibrate the real far field stress and also to verify the FEA quality in representing the stress field in structure. Based on the observed data, the simulation far field stress was adjusted. The experimental far field stress was slightly lower than that calculated in the FEA. It is observed that up to 60.000 load cycles the results has excellent correlation. The correlation was not conservative in the stretch that propagates under the stringer, even though it was very satisfactory.

706

W. R. P. Mendonça and D. F. N. R. da Silva

Results compared show that the damage mechanism has a complex propagation mode, which must consider the stringer debond and asymmetric crack growth rate in the b(a) solution. The simulation result adhered to the experimental results. The delamination correlations have not yet been performed, as they await teardown inspection results. The teardown inspection will also clarify the issues related to debond and asymmetric crack growth. With these results a new update in the FEA can be carried out to verify the influence on the propagation simulated results under the stringer.

5 Conclusion The study demonstrated the importance of evaluating the experimental results from Full-Scale tests, where all the structure complexity is incorporated. These results enabled to verify the details of complex damage modes; involving cracking, delamination, debonding and loading redistribution simultaneously. Another important result obtained with the Full-scale test was to elucidate the influence of the bonded stringer on the laminate, therefore establishing an analysis criterion that satisfies the simulation of the damage propagation under the stringer. All six correlations performed with results from the Full-scale test required adjustments in the pre-established method and criteria. The change of load flow within the wing structure was the main driver for adjustments. The final results represent a great enhancement to the knowledge of the FML technology. The experimental results evidenced the importance of coupling local and global effects in the mathematical modeling to reach prediction results accuracy. In this development, the global structural representativeness was achieved with the accurate estimation of the b = f(a) This is only attained with the experience obtained from the results correlation. Finally, the test results demonstrated that the analysis methodology is efficient in predicting FML crack propagation in complex structures.

References Alderliesten, R.: Fatigue crack propagation. Chapter 9 in Alderliesten, R. (ed.) Fatigue and Fracture of Fibre Metal Laminates, pp. 175–220. Springer, Delft, The Netherlands (2017a) Alderliesten, R.: Stress and strain. Chapter 4 in Alderliesten, R. (ed.) Fatigue and Fracture of Fibre Metal Laminates, pp. 59–76. Springer, Delft, The Netherlands (2017b) Alderliesten, R.: Fatigue crack propagation and delamination growth in Glare. PhD Thesis, Delft University of Technology (2005) Broek, D.: The Practical Use of Fracture Mechanics. Fracture Search Inc., Galena, OH, USA. Kluwer Academic Publishers (1988) Gupta, M., Alderliesten, R.C., Benedictus, R.: Crack paths in fibre metal laminates: role of fibre bridging. Eng. Fract. Mech. 108, 183–194 (2013) Irwin, G.R.: Analysis of stresses and strains near the end of a crack traversing a plate. J. Appl. Mech. 24, 361–364 (1957)

Analysis Prediction and Correlation of Fiber Metal Laminate

707

Khan, S.U., Alderliesten, R.C., Benedictus, R.: Delamination in Fiber Metal Laminates (GLARE) during fatigue crack growth under variable amplitude loading. Int. J. Fatigue 33, 1292–1303 (2011) Rodi, R., Alderliesten, R., Lazzeri, L.: Analytical modeling of Fiber Metal Laminate stiffened by bonded straps: the effect of external stiffening elements on the fatigue crack growth in Fiber Metal Laminates. VDM Verlag Dr. Müller (2009) Spronk, S.W.F.: Predicting fatigue crack initiation and propagation in Glare reinforced frames. PhD Thesis, Delft University of Technology (2013) Wilson, G.S.: Fatigue crack growth prediction for generalized fiber metal laminates and hybrid materials. PhD Thesis, Delft University of Technology (2013)

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning to ¼ Life C. André Beltempo1(&), Alexandre Beaudoin2, and Robert Pothier2 1

2

Aerospace Research Centre, National Research Council Canada, 1200 Montreal Road, Ottawa, ON K1A 0R6, Canada [email protected] Bombardier Experimental Ground Test Department, Bombardier Aerospace, 1800 Marcel-Laurin, Montreal, QC, Canada {alexandre.beaudoin, robert.pothier}@aero.bombardier.com

Abstract. This paper describes the commissioning campaign on the Bombardier Global 7500 durability and damage tolerance test (DADTT), as conducted by the experimental department of Bombardier Aerospace (BAEX). A rapid test rig commissioning time and smooth increase of cycle rate was a key objective in order to meet the ¼ life milestone for certification. Due to schedule changes, the time available for commissioning the loading apparatus and achieving the initial target ¼ life milestone was reduced from the original plan, with approximately three months’ time allocated to this goal. In anticipation of this aggressive testing milestone, the BAEX test team employed several new approaches and techniques to the DADTT that had not been applied on previous BAEX fatigue tests. A technical liaison from the National Research Council Canada was also on-site prior to and during commissioning activities. This paper opens with a summary of the objectives of the G7500 DADTT test conducted at BAEX, and focuses in particular on the techniques and approaches used by the BAEX team to achieve a rapid, efficient and productive test commissioning phase, such that the target cycling rate was achieved in a faster timeframe than anticipated. These techniques included: judicious data-informed hydraulic and pneumatic hardware selections; informed design choices to minimize mass and actuator count; hydraulic and load controller training and procedure generation on a dedicated independent test platform; extensive hardware-in-the-loop tuning to maximize performance; and using the global finite element model (GFEM) of the test article, coupled with a simple pneumatic model, to better estimate initial test load transition times. As a result of the techniques employed, the ¼ life milestone was successfully achieved, contributing to the final certification package for the world’s largest purpose built business aircraft, which received Type Certification from Transport Canada on September 28, 2018. Keywords: Fatigue testing


Full scale testing
Bombardier

© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 708–722, 2020. https://doi.org/10.1007/978-3-030-21503-3_56

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning

709

1 Introduction 1.1

Background

The Global 7500 was launched by Bombardier Aerospace in October of 2010 as an ultra-long range, four-zone cabin, high transonic speed, business jet. The aircraft design has several new features, including an all-new high speed transonic wing, next generation GE Passport engines, and the first four zone cabin in a dedicated business jet. The aircraft is certificated to Canadian Aviation Regulations (CARs), Part V, which is based on the FAA Part 25 airworthiness standards. As is required for any new aircraft program, the Global 7500 was subjected to the required structural test programs to show compliance with the appropriate sub-sections of the aforementioned regulations. In addition to a full scale complete aircraft static test (CAST) and a variety of component tests, a full scale durability and damage tolerance test (DADTT) was planned, with 2 lifetimes of fatigue cycling of durability testing to be followed by 1 lifetime of damage tolerance fatigue cycling. Following these, a residual strength test was to be performed to demonstrate the residual strength of the airframe, and finally, a detailed tear-down inspection would be conducted to verify the structure was free of widespread fatigue damage (WFD). At the time of writing, the G7500 fatigue test had completed 2/3 of the fatigue cycling program, and is on track to complete the remaining DADTT milestones on schedule. The test is being conducted at Bombardier Experimental (BAEX) Ground Test Department, in Montréal, Québec, Canada. 1.2

Schedule

In order to achieve Type Certification, one of the critical milestones in the Global 7500 development schedule was the completion of ¼ life of DADTT fatigue cycling and the related inspections of the structure. In order to mitigate schedule slippage risks during the crucial commissioning and ¼ life cycling phase, the BAEX structures testing team began a comprehensive investigation of methods, techniques and approaches to ensure a smooth commissioning and achieve a faster cycling rate earlier. The investigation covered both technical/design considerations, as well as resource/personnel considerations. In brief, these focused on 4 key areas: 1. During conceptual design, the test rig loading approach was evaluated in order to maximize cycle rate and minimize inspection efforts; 2. A dedicated fatigue test training platform was designed, commissioned and used for personnel training as well as to develop a robust hydraulic commissioning strategy. 3. Hydraulic loading and pneumatic pressurization system design options were explored in depth, including physical testing of configurations and hardware in the test lab prior to DADTT build. 4. Estimating initial load transition times based on hydraulic and pneumatic system performance estimation. The BAEX team leveraged some of the accumulated expertise of the National Research Council Canada (NRC) in this area to establish

710

C. André Beltempo et al.

appropriate initial start times to minimize test shutdowns and achieve a better ‘starting’ point for the cycling rate. This combination of initiatives was successfully employed to reduce the commissioning time from the original estimate by a factor of 2. 1.3

Global 7500 DADTT

The Global 7500 DADTT test consists of a full-scale production aircraft, built to production specifications, with all primary structural elements, excepting flight control surfaces. The test is conducted in accordance with the test plan and requirements, including 2 lifetimes of durability and 1 lifetime of damage tolerance testing. The test article is loaded with more than 130 computer controlled servo-hydraulic actuators, acting through a variety of load introduction systems including collars, straps, fittings, whiffle-trees and mounts. The aircraft fuselage is sealed and pressurized to the aircraft operating conditions during the fatigue cycles. The aircraft was restrained in all degrees of freedom via a system of straps and hard mounts through the fuselage and the landing gear. Figure 1 shows an overview of the loading arrangement.

Fig. 1. G7500 DADTT test loading overview

2 Rig Design for Test Acceleration For the G7500 fatigue test, rig design was considered from a perspective of test schedule reduction at the start. As a result, design changes that could still apply the required test loading and pressure, yet reduce cycle rate or inspection time were incorporated early. This led to several changes in practice from previous Bombardier

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning

711

full scale tests. In particular, the wing and h-stab loading system was substantially lightened by using a bonded tension/compression pad sets, allowing the wing loading to be counterbalanced hydraulically. BAEX design also ensured that the pressurization orifices were enlarged to allow very large flow rates during pressurization or depressurization. The wing-loading collar systems were substantially lightened by employing a bonded tension/compression loading pad system, loading from the upper or lower wing surface, depending on loading vector. A typical loading system is shown in Fig. 2.

Fig. 2. Typical outboard wing loading system

This system leveraged previous Bombardier loading collar designs, but resulted in a solution which effectively halved the moving mass on the highest deflection locations of the article. Previous NRC research had demonstrated that minimizing moving mass, particularly mass between the load cell and the test article, results in superior test cycling performance [1]. This design change provided the additional benefit that it reduced the wing mass to a level below which it was not necessary to passively counterbalance the wings. This meant that the wings were only hydraulically counterbalanced, resulting in even less moving mass during the cycling operations. Finally, by loading the wings from a single side, it provided better access for wing inspections, thereby reducing the inspection labour hours required. A critical design change involved using very large pressurization orifices on the 6 allotted windows, which ensured the pressure control system would not be limited by the performance of the connected pressurization hardware more than orifice size.

712

C. André Beltempo et al.

3 Dedicated Test Platform In recognition of the complexity of the test, BAEX undertook some efforts to maximize the value of the test commissioning time by designing a dedicated test platform to develop appropriate commissioning methodologies and train operators in real-time, hands on closed-loop hydraulic system signals analysis and tuning. BAEX objectives for this platform were to verify the predicted performance of single versus double ended actuators, establish best practices for various features of the load control system, including filters, cross-coupling compensation and profile segment optimization, develop best practices for tuning and commissioning and validate various hydraulic configurations including accumulator placement and supply pressure changes. This resulted in a very robust and thorough commissioning plan, increased initial operator competence during commissioning, and allowed operators to focus on unexpected events during commissioning rather than learning the systems on-the-job. Additionally, having an available test platform to trial more advanced control tuning methods and algorithms as they arose on the DADTT without risking a priceless test article allowed for a faster iteration of solutions rather than attempting more low-risk trials on the test article. The dedicated test platform consisted of a cantilevered aluminum beam with several bolt-hole locations available for load introduction. Up to 6 actuators could be used to apply loads, with a 2  6 arrangement allowing for both torque and bending to be applied (see Fig. 3). This configuration was designed to induce similar phase coupling that would be expected in an aircraft wing under force feedback load control. In order to maximize the benefit of the platform, external training and collaboration was scheduled with the test load control vendor providing expert assistance, and NRC participating to explore specific items of interest unrelated to the G7500 DADTT. Lessons learned from the test platform that were incorporated into the DADTT commissioning plan included: • individual actuator excitation for frequency response analysis and output filtering; • repeatable methods to analyze real-time feedback and error signals to maximize system response; • a consistent methodology for analyzing and incorporating phase lag compensators, including integral gain, filters and cross-coupling compensation; • a procedure for generating appropriate cross-coupling compensation terms and identifying when this is most appropriate; • performance data for various hydraulic system changes, including changing system supply pressure, accumulator placement in the supply and return lines; • methods and best practices for creating custom phase compensators, and when these may be necessary; and • methods to identify various transient effects on the real-time feedback signals, such as mechanical pin-slop or backlash.

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning

713

Fig. 3. Dedicated test platform rendering

4 Data Informed Hydraulic/Pneumatic Design Selections Following major rig design efforts, the hydraulic load introduction and pneumatic systems were designed, with hydraulic hardware selections informed from data derived from the dedicated test platform and pneumatic hardware selections informed from several experimental trials conducted on the G7500 Complete Aircraft Static Test (CAST) test rig. For the hydraulic selections, the major equipment selections included manifold routing and cylinder selection. Cylinders were largely selected from BAEX’s existing stock; however, for the wing tips, new double-ended low friction actuators were employed, given the apparent performance increase demonstrated on the dedicated test platform. An additional complex consideration was the selection of hydraulic servo-valves. Servo-valves are, to some extent, the ‘heart’ of the hydraulic loading system, since they drive the flow rate to either end of the loading actuator, resulting in the desired loads. In particular, although a large flow rate is desired to achieve a rapid displacement, the requirement to achieve precise load application is a countervailing objective, as a smaller servo-valve allows a finer control of load and consequently higher software control gains. As a result, appropriate servo-valve selection is a crucial consideration in a hydraulically driven load controlled fatigue test. Unfortunately, servo-valve response is also very difficult to analytically model. Valves have a non-linear response, and in addition are in various states of physical wear, which will affect their frequency response. In order to select hydraulic valves, BAEX used some of the accumulated

714

C. André Beltempo et al.

experience and expertise from NRC to generate a coupled load transition time hydraulic flow model of the test. The method to develop this model is described in further detail in the following ‘Transition Time and Optimization’ section. The output of this model provided an estimate of the transition time for each of the spectrum load lines, based on flow parameters, an error estimate, the effective calculated stiffness of the attachment location, an estimate of the control effort, and the valve and actuator sizes. By using this model, the ‘limiting actuator’ could be determined, as actuators that were slowing the test down could be identified and valves in the model changed to increase the transition time. An example plot of the output of the model (see Fig. 4), indicates the estimated valve usage over the course of a specific flight for the load spectrum for each of the over one hundred hydraulic positions. As can be seen in the figure, several positions show maximum valve effort, i.e. they peak at a normalized usage value of 1, while others are under-utilized. In the interest of achieving optimal control and cycle rate, valves could be up-sized at the limiting (high peak) locations, while under-utilized locations could be downsized, leading to finer control and better tracking.

1

1.0

0.9 0.9

0.8 0.7

Percentage of Maximum Velocity

Normalized Valve Usage

0.8

0.7 0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0.0

WV WV1 WV3 WV5 W 7 WVV9 WV11 WV13 WV15 WV17 WV19 WV21 WV23 WV25 WV27 WV29 3 SF 1 V SF 1 V SA 3 V SA 1 SW V3 V SS 1 V SS 1 V3 SS V SS 5 V WL7 WF1 WA1 WV1 WV2 WV4 WV6 WV 8 WV10 WV12 WV14 WV16 WV18 WV20 WV22 WV24 WV26 WV28 WV30 3 SF 2 V SF 2 V SA 4 V SA 2 SW V4 V SS 2 V2 SS V SS 4 V SS 6 V WL8 WF2 WF WA2 WRSPD 2 SP 1 D1 FV 1 FV 2 FV 3 FV 4 FV 5 FV 6 FV 7 FV FV 8 D FV 1 D2 FV D FV 3 D FV 4 D FV 5 D FV 6 D FV 7 U CK FVD7 PIT 8 U FL P 1 FL L 1 FL R 2 FL L 2 FL R 3 FL L 3 FL R 4 FL L 4 FL R 5 FL L 5R FL 6 FL L 6 FL R 7 FL L 7R F MG L8 M G L1 D MG MGD1 LH 3 RE MG MG L2 D MG MG 2 RH D4 R NG E D NG 1 NG L1 R EN E T EN 1 L EN 1 L EN 3 EN L5 V EN 1 V EN 3 T EN 2 L EN 2 L EN 4 EN L6 V EN 2 V HS 4 V HS 1 V HS 3 V HS 5 V HS 7 HS V9 V HS 11 D HS 1 V HS 2 V HS 4 V HS 6 HS V8 V HS 1 0 V H S 12 D VS 2 L VS 1 L VS 2 L VS 3 VS L4 D RU 1 RUL1 L EL 2 ML E V1 GU LV2 PL K

0

Actuator ID

Fig. 4. Normalized valve utilization for a single flight type, initial valve selections

Following 17 iterations of the valve selections, the valve usage estimate resulted in much more evenly distributed valve usage (see Fig. 5), which, presuming a sufficiently accurate model, would indicate a better test initial performance. Similarly, one would expect that ‘right-sized’ valve selections would lead to ‘limiting actuators’ from a variety of locations in the test, and not a single location. This would be another indication of appropriate valve selection. A normalized percentage comparison of the number of times a specific actuator location was the ‘limiting actuator’ for the initial and final valve configuration is shown below in Fig. 6. The initial selections resulted in many instances of a few actuators causing the majority of the test slowdowns, while the final selection distributed this amongst an increased number of actuators.

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning

715

1

1.0 0.9 0.8 0.8

0.7 0.7

Percentage of Maximum Velocity

Normalized Valve Usage

0.9

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0.0

WV WV 1 WV3 WV5 W 7 WVV9 WV11 WV13 WV15 WV17 WV19 WV21 WV23 WV25 WV27 WV29 3 SF 1 V SF 1 V SA 3 V SA 1 SW V3 V SS 1 V SS 1 V SS 3 V SS 5 V WL7 WF1 WA1 WV1 WV 2 WV 4 W 6 WVV8 WV10 WV12 WV14 WV16 WV18 WV20 WV22 WV24 WV26 WV28 WV30 3 SF 2 V SF 2 V SA 4 V SA 2 SW V4 V SS 2 V2 SS V SS 4 V SS 6 V WL8 WF2 WF WA2 WRSPD 2 SP 1 D1 FV 1 FV 2 FV 3 FV 4 FV 5 FV 6 FV 7 F FV V8 D FV 1 D2 FV D FV 3 D FV 4 D FV 5 D FV 6 D FV 7 U C K FVD 7 PIT 8 U FL P 1 FL L 1 FL R 2 FL L 2 FL R 3 FL L 3 FL R 4 FL L 4 FL R 5 FL L 5 FL R 6 FL L 6 FL R 7 FL L 7R F MG L8 MG L 1 D MG MGD1 LH 3 R MG E MG L 2 D MG MG 2 RH D4 R NG E D NG 1 NG L1 R EN E T EN 1 L EN 1 L EN 3 E N L5 V EN 1 V EN 3 T EN 2 L EN 2 L EN 4 EN L6 V EN 2 V HS 4 V HS 1 V HS 3 V HS 5 V HS 7 HS V 9 V H S 11 D HS 1 V HS 2 V HS 4 V HS 6 HS V8 V HS 10 V HS 12 D VS 2 L VS 1 L VS 2 L VS 3 VS L4 D RU 1 RUL1 L EL 2 ML E V1 GU LV2 PL K

0

Actuator ID

Fig. 5. Normalized valve utilization for a single flight type, final valve selections BD700 Flight 1 Limiting Actuator Counts

MLGUPLK ELV1 HSV8 ENT1 NGD1 MGD2 FVD4 FV6 WRSPD1 SFV4 SFV3 WV13 MGLHRE WL2 WV30

BD700 Flight 1 Limiting Actuator Counts HSV6 SSV7 SFV3

SSV8

SSV2 WA1

FVD7

SSV1

FVD8 WA1

NGD1

ENV2 MGD3 SFV4

MGRHRE VSL4

FVD2 MGD4

SAV4

MLGUPLK

SAV3

WL1

MGD1 FVD8

WV14

Fig. 6. Limiting actuators for a single flight type, initial valve selections (L), final valve selections (R)

As a result of these efforts, 60% of the original BAEX valve selections were altered to provide better initial performance. During the commissioning period to the original ¼ life milestone, only 1.4% of the valves were swapped for flow or control reasons, indicating that the final selections were adequate for the performance requirements and that the model had sufficient predictive capability to provide valuable insight into hydraulic system performance. It is also noteworthy that predicted total cycle times over the 17 valve iterations also generally decreased, as shown in Fig. 7. Note that iteration 11, although showing better performance than the final configuration (iteration 17), was achieved using too many 5 gpm servo-valves, which exceeded the available stock at BAEX. As a result, a constraint was applied to the solution to ensure that the maximum stock levels of the various flow servo-valves was taken into account.

C. André Beltempo et al.

Valve Selecon Iteraons

716

17A 17 16 15 14 13 12 11 9 2 1

1.86 1.79 1.79 1.76 1.72 1.12 1.85 1.94 1.58 1.47 1.00

0

0.5 1 1.5 Normalized Cycle Rate (Inial Configuraon = 1)

2

Fig. 7. Normalized predicted cycling rate for selected servo valve configurations

The G7500 DADTT was also required to simulate cabin pressurization, in order to accurately reproduce flight loading conditions. As a result, a pneumatic system was also incorporated into the test. In order to provide an overall estimate of the test cycle rate, and to validate pneumatic design choices, a simple model of the pneumatic system was created. This also led to some internal design decisions as to the optimal cabin filling strategy. Referring to Fig. 8, allowing the cabin to fill and empty over the course of several load transitions would impose a less severe requirement on the pressurization system, as a ramp from zero pressure to peak pressure and back, however, this was not optimal from a structural analysis perspective, where a more accurate simulation during load transitions would involve a change in pressure corresponding to the appropriate simulated load condition, shown as ‘Ideal’ in the figure. However, the requirement to fill the aircraft cabin in a single load step would likely lead to an overall increased cycle time, as the fill rate is fixed based on the existing hardware. In order to de-risk the pressurization/depressurization cycles as a limiting step, a ‘stair-step’ pressurization/ depressurization profile similar to Fig. 8 was used. Further, BAEX opted to use large air accumulator reservoirs with a nominal pressure of 100 psi for pressurization to force a high flow rate for filling, and maximum allowable orifice sizes for depressurization to compensate for the reduced pressure difference when emptying the cabin. Six windows were used for pressurization and depressurization, two for pressurization with smaller orifices, and four for depressurization with larger orifices. A comparison between the fill and dump orifice sizes is shown in Fig. 9. The fill side windows were sized based on the much higher forcing pressure from the accumulator system, and all fill lines were constructed with hard-lines due to the high system fill pressure. The pressurization model was a relatively complex combination of the pneumatic system, including the response time and maximum flow assumptions of the volume booster (low flow), automax (high flow). The model simulated the response of each of these systems and then estimated the time to achieve the target pressure in a given

Normalized Pressure

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning

717

1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 Load Steps Ideal

Ramp

Stair-Step

Fig. 8. Idealized individual flight spectrum with 3 pressure cycle options

Fig. 9. Pressurization orifices (L – 2 windows), depressurization orifices (R - 4 windows)

transition time. In order to simplify the cycling rate analysis, the model was checked against empirical performance and found to be broadly linear in terms of pressurization, with only depressurization requiring a more complex treatment. As a result, the pressurization sequence was estimated to be linear for 70% of the transition time, with only the first and last 15% of the transition estimated to decay as ¼ sin curves. Functionally this translates to the following equation for pressurization time: tp ¼ ð0:3Þ


 Dp Dp ð1:58Þ þ ð0:7Þ Rp Rp

Where, Dp is the desired pressure change Rp is the maximum pressure change rate

718

C. André Beltempo et al.

Note that 1.58 is the constant relationship difference between the time it would take for a ramp of similar slope to the sin curve at its inflection point to achieve the target versus a sine. The maximum pressure change rate was based on an empirical scale calculation obtained from the CAST test. Depressurization did not respond linearly, since although the response of the valves was similar to those used for pressurization, the flow rate was much more sensitive to the lower pressure differential between the cabin and the exterior environment. The non-linear decay effect of the depressurization was accounted for using the equation: pffiffiffiffiffiffiffiffi DPI ffi f ðDPI Þ ¼ pffiffiffiffiffiffiffi P MAX

PR

Pressure Drop From Inial Pressure (Δ psi)

Where: PI is the initial starting pressure; PMAX is the maximum initial pressure; and PR is the pressure response The effect of the initial pressure on the depressurization rate was significant, and could not be ignored, as shown in the chart in Fig. 10, which compares the time required to depressurize from 11 psi, or from 5 psi. As shown in the figure, depressurizing from 5 to 0 psi takes almost 6 s (a DP of 5), yet depressurizing from 11 to 6 psi (also a DP of 5) takes only *2.1 s, with all other values held constant.

-10.9 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0

2

4

6

8

Time (s) 5 psi Start

11 psi Start

Fig. 10. Theoretical time to depressurize from 11 or 5 psi.

10

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning

719

In order to simplify the depressurization calculation, the decay equation was multiplied by the same sin response function used for pressurization for all pressures from the maximum G7500 test pressure to zero: P ¼ ð0:3Þf ðDPI Þð1:58Þ þ ð0:7Þf ðDPI Þ Where P is the pressure drop rate in psi/s; f ðDPI Þ is calculated from the depressurization equation above A cubic curve was fit to the resulting table provided an estimate for the psi/s depressurization response at a given pressure, which took into account the decay rate and penalization function above. P ¼ 0:00131P3 þ 0:031488P2  0:3579P  0:28584 Where P is the pressure drop rate in psi/s; P is the pressure value Finally, in order to avoid performing an iterative calculation for each estimate, the pressure value P was taken to be 1/3 of the value between two given pressure drop points. The calculation for pressure for each load condition was then a function of the change in pressure DP from the previous load line. If there was a positive change in pressure, then the DP was multiplied by the empirically estimated pressure increase from the CAST test. If there were a negative change in pressure, then PEST was calculated using the equation above, and then used in the cubic equation to calculate a psi/s. The DP was then divided by this to get a depressurization time estimate. The estimates above were checked against the complex model, and found to be in good agreement, sufficient to generate relative estimates.

5 Initial Transition Time Estimation and Optimization A major element in load controlled fatigue testing is the amount of time taken to transition to each end level. Assuming all of the relevant operating control loop tuning has been carried out, there is still a requirement to define the length of time expected for each load transition. Transition times are crucial to a smooth running fatigue test. If they are too long, the test may take substantially longer to complete than desired, and if they are too short, then they may lead to system instability and test shutdowns as error limits are exceeded. Although there are available tools within the load control system to calculate transition times, primarily based on an assumed linear loading rate per second, these have no a priori knowledge of the test article, and as such, do not include an estimate of the effective stiffness or other characteristics of the test article in a transition time estimation model. NRC Canada is the joint patent holder with MTS Systems for Cross-Coupling Compensation (CCC), a powerful phase compensation technique that is offered as part of the MTS Systems AeroPro load control software suite. As a result, NRC has substantial background and experience in control system simulation and actuator performance, which are used at NRC for initial transition time estimation.

720

C. André Beltempo et al.

To support the BAEX team, for the G7500, NRC used data from the Finite Element Model of the G7500 aircraft to extract key stiffness and frequency response parameters to incorporate those into NRC’s algorithm for estimating transition time. The model incorporated hydraulic parameters, such as valve and actuator size, test controller effort, allowable error, article response, and CCC effects and estimated transition times for each actuator based on the provided G7500 fatigue spectrum. As discussed in the valve selection discussion above, the algorithm also calculated the ‘limiting’ actuator for each transition and output the time penalty associated with it. Due to the complexity of the G7500 FEM, several areas of the aircraft had system response characteristics that were judged to not be representative of the test article, and were altered using engineering judgment to achieve stable results. In particular, the landing gear stiffness was substantially different, since the G7500 DADTT used dummy gear rather than actual aircraft parts. In addition, several assumptions were made with regards to estimating control effort and the effect of CCC, based upon empirical work previously performed at NRC. Incorporating the valve selections and pneumatic system performances discussed in the above section, allowed NRC to provide BAEX with an estimate of the ultimate ‘ideal’ performance of the G7500 DADTT, as well as initial transition times that were based on the model. These transition times provided a much more realistic starting estimate than would otherwise have been the case, and reduced the amount of errorinduced hydraulic and pneumatic shutdowns during the commissioning phase. The outcome of this work was that the time required to achieve target cycling rates was reduced substantially from the estimated baseline.

6 Results A critical milestone for the test was the ¼ life design service goal deadline. This would ensure sufficient time to analyze the resultant test and inspection data in order to meet the certification schedule target. The scheduled plan for the G7500 fatigue test provided 5 months from test rig and article build completion (i.e. hydraulics and pneumatic first ‘on’) to achieve the ¼ life DSG. The employment of all of the methods and techniques described in the paper above resulted in a faster initial cycling rate than predicted, allowing a shorter duration of time to achieve ¼ life DSG. This was despite the fact that the actual commissioning time exceeded the original estimate by 16%. In the event, the actual test progression from hydraulics ‘on’ to ¼ life was achieved faster than the estimate, being reached within 81 days, as opposed to the estimated 88. This is illustrated by plotting the original estimate start on the actual hydraulics ‘on’ date, as shown in Fig. 11, below. Recently, the NRC predicted theoretical transition time and cycling rates were compared against BAEX achieved cycling rates. At the time of writing (end of 2nd life DSG), the G7500 DADTT is cycling 6% slower than the NRC initial prediction, which is an excellent agreement between the original estimate and the achieved cycle rate. A normalized comparison of the various estimates and measured data for the G7500 DADTT as compared to other Bombardier programs is shown in Fig. 12. Note that the cycle rates are normalized to the original G7500 DADTT business case estimate.

Bombardier Global 7500 Fatigue Test Cycle Rate Commissioning

721

% of Design Service Goal (Life)

0.3 0.25 0.2 0.15 0.1 0.05

Commissioning Period

0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75 78 81 84 87

Actual Test Progression

EsƟmated Test Progression

0.25 DSG for CerƟficaƟon - Day 82

Fig. 11. Original estimated test schedule overlaid on actual test start DATE and actual test progression to ¼ life DSG

G7500 NRC Esmate

2.68

G7500 @ 2nd Life

2.54

G7500 @ 1st Life

1.90

G7500 Stretch Goal Esmate

1.50

G7500 Achieved @ 1/4 Life

1.34

G7500 Business Case Esmate

1.00

CRJ700

1.43

Challenger 300

1.11

CRJ200

1.00

Global Express

0.60

0

0.5

1

1.5

2

2.5

3

Fig. 12. Normalized cycle rate (larger is faster) G7500 estimates and other Bombardier tests comparison

722

C. André Beltempo et al.

In conclusion, the combination of designing for mass minimization, data informed hydraulic/pneumatic hardware selections, technique development on a dedicated test platform and employment of NRC derived algorithms for hydraulic and pneumatic transition time estimation resulted in a successful commissioning effort for the G7500 DADTT, and it has achieved the fastest cycling rate of any Bombardier product DADTT to date.

Reference 1. Hewitt, R.L.: Effects of fixturing mass on dynamic strain errors in structural fatigue testing. J. Test. Eval., JTEVA 28(6), 462–472 (2000)

Changing the Philosophy of Full-Scale-FatigueTests Derived from 50 Years of IABG Experience Towards a Virtual Environment Gerhard Hilfer(&), Olaf Tusch(&), Don Wu(&), and Michael Stodt(&) IABG mbH, Einsteinstrasse 20, 85521 Ottobrunn, Germany {hilfer,tusch,wu,stodt}@iabg.de

Abstract. IABG has continuously developed new techniques in order to improve time, cost and quality for full-scale fatigue tests. With the advancement of virtualisation of the aircraft development and certification process, questions have to be raised as to how full-scale fatigue tests can be incorporated into these processes and how virtualisation can benefit from physical testing. Virtualisation will speed up aircraft development to its next level and will change time, cost and planning expectations. Will full-scale fatigue testing (FSFT) still fit into this context? Authorities are demanding physical tests for new aircraft types for good reasons. Testing experience in fact confirms that some significant issues were first detected during the full-scale fatigue tests. The purpose of full-scale fatigue tests may, however, be changing. The amount of available but unused information gathered during fatigue runs is still sizable, and only by changing the approach to fatigue tests the return on investment could be increased. It will be required to introduce agile methods and processes to the preparation and performance of the full-scale fatigue test to align the test with the dynamic evolution of the aircraft’s requirements and design. In the end this could result in a new philosophy of full-scale fatigue tests, moving from flight-by-flight testing towards purely artificially triggered load sequences in order to demonstrate the correctness and reliability of the virtual qualification tools which will be of crucial importance for any virtually based certification. This paper, however, is aiming as well at describing some of the more near-term benefits the FSFT can contribute in view of the long-term goal. Keywords: Virtual testing
Machine learning Certification
Agile processes


Full-scale fatigue test


1 Introduction The present paper discusses, based on 50 years of IABG experience in testing and simulation within and beyond the aerospace industries, the possible future options and challenges of full-scale fatigue tests (FSFT) and their benefits to the aircraft industry.

© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 723–735, 2020. https://doi.org/10.1007/978-3-030-21503-3_57

724

G. Hilfer et al.

Having run and supported a variety of FSFTs, a prevailing outlook at the end of the more recent campaigns was that in some future FSFT would no longer be necessary, as virtual models would become increasingly capable and powerful and would fully replace the fatigue tests. As the improvements in virtual modelling are in fact impressive, then why would FSFT remain a must for aircraft development and certification?

2 Evolution of Full-Scale-Fatigue-Tests Over the past decades, Full-Scale-Fatigue-Tests (FSFT) have evolved in various ways. Rocketing computing power has increased the measurement and control performance notably, leading to faster testing and higher sampling rates of strain gauges and other sensors. Twenty years ago a measurement campaign of 3,000+ strain-gauge channels took weeks and was realisable only under more or less static conditions. The test data easily fit on a 1,44 MB-floppy disk. Today 3.000+ channels can be recorded with >100 Hz sampling rate with nonstop cycling. The full set of data is available for each load point, and data can easily fill Terabytes of memory space. Similar developments were made at the control and monitoring systems. Even if the pure testing duration of three design service-lives, i.e. without any downtime, of a major FSFT still takes around one year, the accuracy has clearly improved and the number and complexity of loading scenarios has significantly increased. Similarly, the inspection has improved notably. For example, a much better visualisation of the processed signals of the ultrasonic and eddy current inspection devices has led to faster inspection and better reporting means. Despite all technological gains, however, even though the positions of damages and instrumentations are stored within the virtual model, this interface is still managed manually until today. All these technological advancements aside, one issue has remained unchanged over the last 50 years of FSFT: unplanned down time occurs overwhelmingly due to issues not foreseen, and due to failures first detected during FSFT (Wong 2017). Many of these failures will likely be attributable to human errors or misjudgements. Also, the time needed for analysis and subsequent remedial actions has not been reduced significantly. This alone bears potential for future optimisation, let alone other aspects. In recent years, the validation of virtual models, so far mainly for derivates of the same aircraft type, has increasingly gained importance. Load cases or loading scenarios of potential derivates were simulated within the FSFT in order to validate the virtual models beyond their use for the specific aircraft type under test. These models have successfully been used for the certification of derivatives without performing a dedicated FSFT.

Changing the Philosophy of Full-Scale-Fatigue-Tests Derived

725

3 Evolution of Virtual Qualification Capabilities Numerical analysis has developed dramatically over the last decades, not so much in terms of the underlying principles, but the more so in terms of performance and accuracy, with some of the analyses running on simple workstations and, depending on the complexity of the models even in real time (or faster). Not surprisingly, simulation techniques have improved the preparation and performance of FSFT similarly significantly. FSFT are controlled by a closed-loop control system more or less optimised by tuning the feedback signals and adjusting control algorithms. In addition, advanced control concepts of today have virtual models implemented to calibrate the control system much more efficiently. Ultimately, this enables the tests to run at a higher speed and a much better accuracy. IABG uses its own multi-physics model combining: • structural dynamics model of the aircraft specimen and load introduction; • a nonlinear model of the hydraulic actuators and piping; • thermodynamic model for the air inflation and pneumatic installation. The multi-physics model, as shown in Fig. 1, allows for dynamic simulations of the test rig behaviour under real test conditions. It also contains a complex control algorithm, which in turn enables new dimensions of test speed and accuracy.

Fig. 1. Virtual test rig supported by a multi-physics model

So far, the main application of this multi-physics model is to speed up the test rig development and the test performance (Schorr et al. 2017). However, there is a potential to extend its use as a complete digital representation of the test. In what follows, we explore the potential of the FSFT as means to support this virtualisation process.

726

G. Hilfer et al.

4 Exploiting the Potential of the FSFT So, the next logical step towards an increased usage of virtual testing appears to be to strengthen the cross-link between virtual models used in the aircraft development and the FSFT and its related virtual models. In what follows, we attempt to outline how this could significantly improve the value of and the benefits derived from the FSFT. 4.1

Hardware-in-the-Loop Testing

Hardware-in-the-loop (HIL) testing, combining real hardware with a simulated operational input and feedback loop, is a good approach to enhance the benefits of physical testing. Today, it is often used to test control units of mechatronic systems. The use of real hardware ensures genuine functionality and provides realistic feedback to the control unit under test. In this way, the nominal working environment of the control unit can be simulated. During the test, other effects such as increased friction, icing or sensor malfunctions can be introduced to test the control unit’s behaviour under extraordinary conditions. But HIL testing can also be thought of in another way, i.e. the control system of a test set-up for a hardware unit (e.g. a landing gear) can be interfaced with a virtual representation of attached systems/components (e.g. the aircraft body) not readily available, possibly because they are still under development. The test of the hardware can be controlled using (more) realistic model output data while receiving and processing feedback from sensors installed on the hardware (Fig. 2). A closed-loop control can be realised which offers significant advantages compared to a pre-determined testing program without feedback. The test results can be used not only to evaluate the hardware but also to provide useful data for the improvement of the systems represented in the test only by their virtual models. 4.2

Model Validation

Today, at the beginning of an FSFT a static test campaign is being performed which is used to validate the FE models. Usually, this is done using all or a selection of the constituting load cases of the fatigue program. It could be worth considering applying additional specifically composed load cases to limited areas of the structure. By doing so, not only modelling deficiencies, but also errors resulting from sensor misalignments, faulty labelling or other sources could be identified and resolved more efficiently before they may detrimentally affect the test execution or the interpretation of test results. All structural areas of interest (or even the entire structure) can be checked with this process (Fig. 3).

Changing the Philosophy of Full-Scale-Fatigue-Tests Derived

727

Fig. 2. IABG HIL landing gear test set-up (Keil et al. 2017)

Fig. 3. Iterative model validation approach

After validation of the FE models in this initial static test campaign, individual reassessments of fatigue-critical locations are made based on the measured structural response to the fatigue load cases. By doing so, a final check of the test spectrum against the design spectrum is made. It should be noted that this comparison is made on an individual measurement of the load cases and does not cover subsequent changes of the structural response or the loading program. It would therefore be a significant improvement if a continuous analysis of the FSFT data stream could be performed in real time and in an automated way. The comparison of the factually applied load spectrum with the design spectrum could thus be permanently updated while the test is running.

728

G. Hilfer et al.

Further improvements will become possible when manufacturing and assembly features can be captured in a more mature way. Then, it would even become possible to predict emerging damages and thus anticipate required repairs. The repair solution as well as the repair procedure and the manpower planning could be established while the test is still ongoing, thus contributing towards the reduction of downtimes. The scope of planned inspections could potentially be reduced based on a comprehensive virtual monitoring of the test structure. In the same way, an anomaly emerging during the test could be discovered early and encountered by dedicated inspections or additional static test campaigns to reveal its cause. Obviously, this can be understood as a preparatory step towards developing a digital twin (see below). With increasing predictive capabilities of virtual models, scatter factors, load enhancement factors or environmental knock-down-factors, which all aim at producing aircraft safety in areas of uncertainty, might be revised in terms of potential reductions. Next to a preferable elimination of uncertainties, an appropriate handling of remaining uncertainties offers a significant potential for design optimisation and will hence play a significant role in the future (Harris 2017). 4.3

Data Science and Machine Learning

Today, probably less than 1% of the data gained/gainable from an FSFT is actively used. Whilst the test data can nowadays be made available to the analysts quite easily, the required capability to perform such real-time analysis (including interpretation and extrapolation) on huge amounts of data has yet to come forth. IABG is currently developing tools to exploit the potential of the FSFT data in order to streamline test execution, improve test-related processes and support the test evaluation for its customers. Data science methods, including machine learning, will be employed to enable a retrospective as well as a predictive analysis of the test data (Fig. 4). In conjunction with more detailed specimen data provided by the client, this will significantly improve the availability of processed test data containing evaluated and augmented information on the test progress and the status of the structure. This will enable recommendations for immediate actions as well as for the continuation of the test. As an immediate benefit, it will enhance test performance efficiency. Additionally, it will support the test evaluation work of the customer and will therefore increase the overall benefit of the test. The current work of IABG is also aiming at assessing the needs with respect to the test instrumentation as well as the required metadata (labels) for an application of data science and machine learning for FSFT methods. Test instrumentation requirements for classic structural investigations may be significantly different from those suited to satisfy a big data approach. The provision of appropriate labels which are essential for machine learning algorithms is a major challenge as the evaluation of the data is governed by them and their associated quantifiers instead of the ‘classic’ engineering judgement, which would invoke additional information and evaluation criteria in a selective and context-oriented way. Further intensive research work is required to fully exploit the potential of machine learning for the FSFT.

Changing the Philosophy of Full-Scale-Fatigue-Tests Derived

729

Fig. 4. Data science approach for structural testing

4.4

Structural Health Monitoring and Predictive Maintenance

The FSFT could also be employed to help develop and to test structural health monitoring (SHM) systems for next-generation aircraft. Whilst the nature of an FSFT performance (e.g., non-representative accelerations, sheltered environmental conditions) may require some adaptations regarding the handling, evaluation and interpretation of sensor data, the FSFT nevertheless offers the closest representation of inservice operations and should therefore be considered a valuable platform for SHM systems test and validation. Addressing the growing significance of big data for analysing complex interacting systems and deriving operational conclusions, the FSFT can also contribute as it provides a database which describes in a rather straight-forward manner the relation between certain loading conditions and the structural response. This database can be used for example in machine learning (ML) algorithms for the above-mentioned realtime analysis of the test data. Anomalies can be identified very quickly once the MLalgorithms have been trained on a baseline structure. During subsequent stages of ML applications, algorithms could be trained as well on data showing deviations born from typical damages. Thus, an automatic identification of damages can be performed by correlation with known damages. Based on this knowledge, FSFT data could also be exploited to develop tools for the interpretation of in-service data, particularly in case of new aircraft types lacking operational experience. If in the future a changing structural response could be detected already in-flight, proactive actions could be taken. In the same way, operational data could be analysed in correlation with FSFT data in order to improve predictive maintenance approaches.

730

4.5

G. Hilfer et al.

Digital Twin

Future virtual product development capabilities will have an impact way beyond the development and qualification process itself. Within the industry, there exists a great push towards establishing a so-called digital twin to supporting in-service operations. This digital twin should be a perfect virtual representation of each physical copy of the product incorporating in particular individual manufacturing, assembly and configuration data. All data gained during operation of the real counterpart should be fed into the digital twin in order to have comprehensive access to all relevant information concerning current configuration (including repairs), health status, upcoming maintenance activities, remaining lifetime etc. Such a model would be a huge improvement to current operations and logistics management. Understandably, the establishment of such a powerful digital twin will require significant efforts during the development and validation phase. As an FSFT is typically being performed on a close-to-serialproduction specimen, it can serve as a demonstration object for the structure-related models of the digital twin. Additionally, the FSFT environment could be augmented by model-based representations of other systems usually not required for the FSFT itself (such as the drive system of control surfaces, the landing gears or the engines). In this way, the digital twin could be fed with artificially generated but context-relevant data from these additional systems thus providing a more comprehensive test environment for the digital twin. 4.6

Future Benefits from FSFT

These days, major changes in future aircraft development and operation becomes apparent. It is fuelled by high expectations from the collection and use of data and from the creation of a consistent virtual development chain. As the FSFT provides a close-toreality testbed, it can definitely promote this envisaged future of aircraft development by offering the required playground for advanced technologies. As such, the FSFT has a potential for further exploitation well beyond today’s value. As highlighted above, it can propel developments in fields not less than: • • • • • • • •

Model validation Real-time data analysis Training of ML algorithms Life consumption tracking Damage prediction SHM testing Provision of database and algorithms for MRO Digital Twin test & verification

FSFT should therefore be conceived not only as an essential pre-requisite for aircraft certification but also as a crucial asset along the way towards virtualisation.

Changing the Philosophy of Full-Scale-Fatigue-Tests Derived

731

5 How to Develop the FSFT As outlined throughout the previous sections, FSFT has the potential of assisting the product development process in many ways. However, the nature of FSFT being currently scheduled towards the end of the process hampers the exploitation of this potential. With this in mind, two questions need addressing: (a) how can FSFT provide meaningful feedback to the design process at all, and (b) how can FSFT provide meaningful feedback at an earlier stage of the design process? To both questions, agility appears to be a viable answer. 5.1

Introducing Agility to FSFT

At first sight, introducing agile processes to the FSFT development and performance may seem unconventional at the least, as this implies a dramatic change in the way FSFT has been handled in the past. Many parameters are traditionally defined very early along and best left unchanged throughout the test planning process, such as load paths, loading programs, instrumentation and inspection. Any modifications to these may have significant impact on the design of the test rig and the preparation of the test. To think that these parameters should be subject to last-minute changes within an agile process appears unfathomable. However, agility aims at achieving the best results at each particular point in time. So, why should we exclude the FSFT from the concept of agility? In view of its potential and its future purpose, it appears prudent to revise the concept of the FSFT instead, to ensure that some of the expectable changes will no longer affect those planning aspects associated to long lead-time for their realisation. To start with, such a FSFT concept must provide flexibility in the choices and changes of loading configurations. Also, such concepts must make most use of virtual models to bridge the residual gap between the operational envelope of the FSFT and the ultimately desirable test configuration. With these aspects in mind, the FSFT can and in fact must be integrated into the agile product development process of the aircraft. By this, a conceptual framework can be defined to which the aircraft designers as well as the test engineers can adhere. During later execution of the test, this framework would provide further benefit for quick identification of options for potential on-the-fly changes. So, the key for an enhanced future utilisation of the FSFT in an agile development environment will be a dedicated frontloading of test, qualification, certification and operations aspects in order to define the above mentioned framework and enable FSFT development to become a fully integrated part of the aircraft development. In this way, the FSFT development can employ the database consistently used for the aircraft development and the related virtual models. As data continuity becomes a primary goal of future development programs, the FSFT development data and the test results should and must be an integral part of this concept. All data and models developed for the FSFT could be updated accordingly as and when new datasets become available, in the spirit of continuous integration. Deviations from the pre-defined framework could be

732

G. Hilfer et al.

identified and resolved early, thus ensuring that the FSFT stays fully aligned with the overall development progress. Virtual and physical testing could be conceptually synchronised and thus exploited to the maximum extent. An intensified virtualisation of the development process will affect the FSFT in many ways (Fig. 5). With agile processes in place, the development of the FSFT can be continuously adopted to the progress made in other areas. However, the concept of virtual product and agile processes will raise other new questions, such as:

Fig. 5. Interaction of advanced product development aspects with the FSFT

• how to synchronise and update requirements and data used by a huge number of suppliers and engineers involved in the development? • how to achieve a buy-in of the customer to the agile development concept, will he trust in the advantages and will he be able to provide the required support? • how to ensure a consistent certification approach when the most consistent element of the development process was “change”?

Changing the Philosophy of Full-Scale-Fatigue-Tests Derived

733

As pointed out earlier, data continuity and data management will be a key issue for future aircraft programs relying on agile processes. Data storage management, data access management, revision and release processes, standardised interfaces as well as a bi-unique nomenclature and meaning of all data are just some of the aspects to be addressed. While this may be resolved with due diligence, a monitoring and control of the extremely dynamic development process could prove itself a major challenge as agile processes impose an intrinsic danger of compromising the initial idea/specification of the new product. A high-level program management focusing on the adherence to the core elements of the development goals and on conflicting tradeoff situations will be vital. A key element of success in agile development is, however, to closely involve the customer and continuously request his feedback, his acceptance, and most of all his decisions and resolution of impediments as they arise. With respect to the certification process, an agile product development process needs to ensure an integrated development and verification/qualification toolchain and a strong data continuity backbone. Despite all intermediate changes, these two elements can provide the required consistency needed in front of the authorities. As agile processes intrinsically provide a continuous and early feedback and validation of results, this can even be a significant advantage for the avoidance and detection of errors and for the acceleration of the qualification and certification phase.

6 Certification – No Challenges? Authorities currently require a new FSFT whenever there is a lack of experience with similar designs. Essentially, this means that new aircraft types have to be qualified by means of a test as long as authorities are not fully convinced that the numerical analysis can provide the same level of confidence. This is because an FSFT demonstrates physical evidence for the structural integrity of the complex aggregation of design, manufacturing and assembly features, while simulation is limited to a prognosis based on the numerical representations, which do not feature the factual scatter but instead reproduce design, manufacturing and assembly features in an identical manner all over the model. Nevertheless, there is already a relevant number of certifications granted without having performed an FSFT because a previous certification process of a similar design was deemed sufficient to rely on an evolutionary numerical verification. In order to achieve a virtually based certification, the following gaps will have to be bridged: • The understanding and the simulation quality of physical phenomena need to be improved; fatigue mechanisms as well as manufacturing and assembly features should be of major interest in this respect. • Sufficient evidence has to be provided to show that uncertainties in design data, modelling features or computational limitations are covered in the design so that they do not adversely affect the safety of the aircraft; standardized comparative calculations may be appropriate to assess and qualify the level of trustworthiness and acceptability.

734

G. Hilfer et al.

• A continuous verification chain needs to be established and documented; models, tools and methods which produce certification-relevant data need to be validated based on previously proven models, tools, methods or tests if they do not have their own heritage of proven applications. • All data used must have a back-traceable, validated origin; all certification-relevant results must be produced with a consistent, up-to-date dataset which has been used throughout the entire certification-relevant verification chain. • A coherent approach must be presented and demonstrated that the results of the FSFT are suited to provide final confirmation and evidence of the correctness and reliability of the virtual qualification chain.

7 Conclusion For new aircraft types, FSFT will remain a mandatory task, but there is a significant potential to improve the return on investment. Testing has to be integrated much stronger in the product development process and its interaction with virtual modelling should be intensified. The interaction should be focused on making the virtual models more robust, on reducing uncertainty, and on preparing a gradual transition to an intensified use of virtual testing. Currently only a fraction of the data that could theoretically be gathered during an FSFT is being analysed, thus forfeiting valuable additional information. A systematic data analysis could detect failures in virtual models, in the test setup and in the structure itself much earlier, which would reduce down times during FSFT and accelerate the process of verification. Establishing real-time monitoring and evaluation using the huge amounts of data generated by an FSFT is an imminent task in this context. The current flight-by-flight test philosophy could be revisited in order to allow for early verification of the virtual models and subsequent real-time assessment and analysis of test data. With increasing predictive capabilities of virtual models, the flight-by-flight loading program could even become obsolete as new approaches to the FSFT could result in shortened testing times while at the same time providing much more valuable information for fatigue demonstration, future design and modelling improvements, SHM systems development, certification, MRO and the development of the digital twin. IABG is actively pursuing the development of methods and tools to increase the return on investment of full-scale fatigue testing and to anticipate future aircraft development needs in various regards. Next generation FSFT will push test efficiency to its next level by analysing the airframe 24/7 and providing processed data and operational recommendations in real-time. All test data collected will have augmented information, be ready for use in any data science environment. By successfully implementing the suggested advancements of this paper, FSFT will remain a substantial asset on the long way to virtual certification.

Changing the Philosophy of Full-Scale-Fatigue-Tests Derived

735

References Harris, L.: The challenges in airbus to replace full scale aircraft fatigue testing by predictive virtual testing, ICAF June 2017, ISBN 978-1-510-85610-3, AIRBUS (2017) Keil, A., Karl, M., Anderl, T.: Automatisierung von Systemprüfständen für Flugzeugfahrwerke auf der Basis von VeriStand, Virtuelle Instrumente in der Praxis VIP 2017, ISBN 978-38007-4441-1, IABG (2017) Schorr, F., Tusch, O., Wu, D., Mösenbacher, A., Reimann, M., Urban, A., Stodt, M.: Fatigue testing of new generation wide body aircraft at benchmark level, ICAF June 2017, ISBN 9781-510-85610-3, IABG (2017) Wong, A.: Blueprint TITANS: a roadmap towards the virtual fatigue test through a collaborative international effort, ICAF June 2017, ISBN 978-1-510-85610-3, Defence Science & Technology Group (2017)

Combined Static and Fatigue Tests of the Full-Scale Structure of a Transport Aircraft K. S. Shcherban1(&), A. A. Surnachev1, M. V. Limonin1, A. G. Kalish2, and O. V. Chuvilin2 1

2

Central Aerohydrodynamic Institute, Zhukovsky 1, Zhukovsky, Russia [email protected] Ilyushin Aviation Complex, Leningrad Avenue, 45г, Moscow, Russia [email protected]

Abstract. The article discusses a non-traditional approach in strength tests of the full-scale structure of a transport aircraft, which consists of combining static and fatigue tests on one object. The test object included a full span wing with installed pylons, the middle part of the fuselage and the main landing gear. The tests were carried out on the bench, which allowed reproducing both static cases of loading and variable loads of flight cycles. To confirm the static strength, the structure of wing half span was loaded with limit loads and simultaneous strain measurement. The data of strain gauges verified the finite element model (FEM), on the basis of which the prediction of the stress-strain state of the structure was made with the ultimate loads. To confirm the strength capacity of the upper wing panels under the stability conditions, tests of full-size wing panels were carried out. After loading with the limit load, fatigue tests were carried out which are required to confirm the service life. Keywords: Full-scale structure
Static test
Fatigue test
Residual strength
Stress-strain analyses
Buckling

1 Introduction In connection with the designing of a new modification of the Il-76 transport aircraft, it became necessary to experimentally confirm the static strength of the wing structure, as well as fatigue life of the wing, engine suspension and main landing gear of the modified Il-76MD-90A aircraft. To carry out the necessary tests, non-conventional approach was applied, which consisted in combining static and fatigue tests one full scale structure of airplane.

2 Object of Testing The test object (Fig. 1) included: (a) a full span wing, on which pylons for PS-90A-76 engine with dummy power engines are installed, flaps rails with dummy carriages; © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 736–746, 2020. https://doi.org/10.1007/978-3-030-21503-3_58

Combined Static and Fatigue Tests of the Full-Scale Structure

737

Fig. 1. Object of testing

(b) the middle part of the fuselage with installed fairing along the left side; (c) the main landing gear with dummy wheels. The wing of the aircraft is swept, wing box type, trapezoidal shape with a change form along the trailing edge. On the axis of symmetry of the aircraft, the wing is conventionally divided into right and left half span of wing. The wing has two technological joints at a distance of 2,4 m from the axis of the aircraft, which divide the wing on the center section and the two out parts of the wing. The wing connected with the fuselage on the power frames. The basis of the structure of the wing box prefabricated-monolithic structures. The wing box is divided by ribs into 12 fuel tanks, two drain tanks and two dry sections. Tanks are completely sealed. The joint of the out part of wing with the center section is by a set of fittings connecting the upper and lower panels. On the lower surface of the wing, there are attachment points for engine pylons. On the pylons are mounted dummy engine for the application loads. For the application the forces to the rails of the flaps, dummy carriages are mounted on them. The test object includes the middle section of the fuselage, which is closed from the ends by sealing plates. Sealing plates are designed for sealing the section of fuselage and for the application of loads. The middle section are formed by a transverse set frames and longitudinal - stringers and skin. This part of the fuselage is sealed and designed for an operating overpressure up to 0,05 ± 0,002 MPa. The fuselage section is a circle with a diameter of 4,8 m. There is a cargo cabin in the fuselage. The center section is fastened to the upper part of the fuselage along the power frames by the front, middle and rear spars, respectively. At the bottom of the fuselage is installed a fairing, which closes the attachment points of the main landing gears and their wheels in the retracted position. Landing gears are attached to the lower parts of the frames. The landing gears of the test object consists the four main supports, which are identical in structure. Each left support is a reflection of the corresponding right support. Fastening the landing gear to the fuselage is by the traverse. The traverse is structurally made in the form of a horizontal part (two shoulders with cylindrical pins) and a vertical cylindrical part with two bronze bushings. The main landing gears is fixed in the extended position, the shock absorbers are filled with oil and the compression is in the parking position. The wheels of the landing gear are replaced by their dummies, to which the load is applied.

3 Test Procedure The tests were carried out in two stages. At the first stage, the wing was loaded with the limit load of loading case “A” with simultaneous strain measurement, and at the second stage the cyclic loading with flight cycles “At the height” and “At the ground” was

738

K. S. Shcherban et al.

performed. At the first stage, the right wing half span was loaded up to limit load, left half span to maximum loads which simulated during fatigue tests. At the second stage fatigue test were carried out. When testing for fatigue, the wing, the rails of the flaps, the engine suspension, the middle part of the fuselage, and the main landing gear were cyclically loaded. Simultaneously with the cyclic loading, an overpressure was simulated in the cabin. The loading was carried out by blocks of variable loads consisting of 7 flight cycles “At the height” and one flight cycle “At the ground”. The flight cycle “At the height” is divided into two types: B1—in which, at the “Pre-preparation” stage, two cycles of the engine thrust are reproduced alternately on the external and internal engine, and B2 - in which there is no engine thrust. In one block of loads from the 7 flight cycles “At the height”, variant B1 is repeated 1 time and variant B2-6 times. In flight cycles “At the height” (B1 and B2), the “Reverse engine” mode was reproduced only on external power plants. In flight “At the ground” - the “Engine reverse” mode was reproduced both on external and internal engines. In the flight cycle “At the ground” at the air stage there is no simulation of overpressure of the fuselage.

4 The Complex for Testing The complex for testing (Fig. 2) provides for performing both static strength tests and fatigue tests. The complex includes a set-up of multichannel loading, a compressor station, an oil pump station, an information-measuring system, a non-destructive testing instrument and a programmer.

Fig. 2. The complex for testing

Combined Static and Fatigue Tests of the Full-Scale Structure

739

When testing for static strength, the loading system provides loading up to limit loads. When testing for fatigue, the loading system provides simultaneous cyclic loading of the wing, rails of flap, engine suspension, the middle part of the fuselage, the main landing gears. To balance the active loads, which simulate the aerodynamic and inertial loads that occur in flight, special loading channels prevent the aircraft from moving as a whole. There is pressurization system which provides overpressure in cabin. The set-up for multi-channel loading (Fig. 3) includes digital servo cylinders and load tree systems that load the wing, flap rails, engine suspension, and main landing gears. For simultaneous loading of the airframe 93 servo cylinders are applied.

Fig. 3. The set-up for multi-channel loading.

To load the wing, the hydraulic cylinders are hinged mounted on the force floor and load the lower surface of the wing by transferring both pulling and pushing forces with a lever system and nodes glued to the bottom surface. On the end sections in the wing sections, the hydraulic cylinders are inclined in such a way that the force is perpendicular to the bottom surface when maximum deflection.

740

K. S. Shcherban et al.

For loading the flaps rails, the hydraulic cylinders are hinged on the force floor and load the rails by transferring the pushing force to the carriage installed in the position corresponding to the flap extended position. The cylinders are inclined in such a way that the force is perpendicular to the rail. To load the fuselage, the hydraulic cylinders are hinged on the force floor and load the left and right sides by transferring pulling forces by the load tree system and nodes attached to the skin of the fuselage at the floor level in the fuselage. For loading the middle section of the fuselage, bending moments and shearing forces that arise from the missing front and tail parts of the fuselage apply vertical forces to the sealing plates by means of hydraulic cylinders that are fixed to the force floor and nodes on the sealing plates. For loading the suspension of the external engine, eight hydraulic cylinders are applied, which load the engine with aerodynamic and inertial loads in the longitudinal and vertical directions. Aerodynamic loads are applied at the center of the engine pressure, and inertial loads at the center of its gravity. For loading in the longitudinal and lateral directions, the hydraulic cylinders are mounted on supports and the forces are transmitted to the dummy of engine. The engine thrust and lateral forces were reproduced by a pair of hydraulic cylinders acting at an angle, which loading in such a way that the resultant force maintained a horizontal direction as the dummy of engine moved during bending of the wing. For loading in the vertical direction, the hydraulic cylinders are fixed on the force floor, and the force is transmitted to the dummy of engine. The landing gears are loaded in vertical, lateral and longitudinal directions. For loading in the vertical direction, the aircraft is suspended on the power portals by the dummy of wheels. Suspension provides free deformation of the landing gears. For loading the main landing gears in the longitudinal and lateral directions, the hydraulic cylinders are fixed to the supports, and the forces are transmitted to the dummy wheel at the level of its axis. For balancing the aircraft on the ground and flight modes of the flight cycle, the aircraft is fixed in the longitudinal direction with the help of fasteners that are attached to the force floor on one side and attached to the front and rear sealing plates on the other.

5 Researching the Stresses of the Wing Box Structure As a result of the finite element calculation of the stress-strain state of the wing loaded with the loading case “A”, the distributions of normal stresses along the stringers in the upper panels (Fig. 4) and lower panels (Fig. 5) was obtained.

Combined Static and Fatigue Tests of the Full-Scale Structure

741

Fig. 4. The distributions of normal stresses in the upper panels, MPa

Fig. 5. The distributions of normal stresses in the lower panels, MPa

For experimental verification of the calculation results, 9 wing sections were selected (Fig. 6). In these sections, strain gauges were mounted on the inner and outer sides of the lower and upper panels and on the stringers, as well as on the spars, on the right wing half span.

742

K. S. Shcherban et al.

Fig. 6. The sections where, strain gauges were mounted.

As an example, Fig. 7 shows the installation of strain gauges in the most loaded Sect. 6a.

Fig. 7. The scheme of mounting strain gages in Sect. 6a.

The strain measurements were performed under step loading with a load distribution that corresponds to the loading case “A”. The right half span of wing was loaded up to the limit load. The left half span of wing of the wing was loaded up to the maximum load of the flight cycle “At the ground.” Balancing relative to the longitudinal axis was carried out by loading the fuselage.

Combined Static and Fatigue Tests of the Full-Scale Structure

743

The stresses obtained by strain measurement were compared with the stresses obtained by calculating the FEM. The results of the comparison for the limit load for the most loaded Sect. 6a are shown in Fig. 8.

Fig. 8. Normal stresses in the wing panels under loading with the limit load of the loading case “A”

From consideration of Fig. 8 it can be noted that there is a good agreement between the stresses obtained by measuring and the stresses obtained by calculating the FEM. So the compressive stress in the upper panels, obtained by calculating the FEM, reaches −288,5 MPa, and slightly exceeds the stress obtained by measure, −275,7 MPa. The deviation does not exceed 4,5%. The greatest tensile stress in the lower panels obtained by calculating the FEM reaches 233,4 MPa. The stresses obtained by strain measurement reaches 247,5 MPa and exceeds the stresses obtained by calculating the FEM by 6%. A similar picture was observed for all considered sections. This allowed concluding that the FEM calculation results can be used to predict the stress state of the wing structure with a correction of the calculation results.

6 Tests of Full Scale Panels for Stability To determine the strength capacity of compressed upper wing panels, three span panels were tested for stability. Testing of three-span panels made it possible to most fully realize the working conditions of the panel in the structure, since such tests take into account the initial deflection of the panel, the mutual influence of spans, moments at the junction of the panel with ribs, ribs stiffness, fastener strength. For testing, panels with a length of 3 distance between the ribs and the width of 5 stringers were selected. Two panels were connected to the box to provide real conditions of support on the ribs. The box mounted on the testing machine is shown in Fig. 9.

744

K. S. Shcherban et al.

Fig. 9. The box mounted on the testing machine

The panels included in the box, were tested in turn. The panel was mounted on the support plates with end planes in such a way that the center of gravity of the panel section coincided with the central power line of the testing machine. Monitoring over the possible eccentricity, in case of uneven loading across the width was carried out using strain gauges mounted on the panel. The tests were carried out on a Riehle-300 testing machine. In the process of loading, according to the indications of strain gauges mounted on the panel, loads and stresses were determined, corresponding to the local loss of stability of the panel elements and the overall loss of stability of the panel as a whole. In the test, measurements were made by laser displacement sensors. Tests of two panels were carried out before the general loss of stability. The loss of the carrying capacity of the panels resulted from the general loss of stability of the three-span panel and its fracture in the middle span (Fig. 10). The fracture of the panels was accompanied by the fracture of the stringers and the rupture of the rivets.

Combined Static and Fatigue Tests of the Full-Scale Structure

745

Fig. 10. The panel after general loss of stability

As a result was defined the critical stresses for the total loss of stability and local buckling. Critical stresses for monolithic panels are also obtained by calculation. The form of the total loss of stability obtained by calculation is shown in Fig. 11. Comparison of the critical stresses obtained by calculation and testing has shown their good agreement.

746

K. S. Shcherban et al.

Fig. 11. Buckling of for monolithic panels obtained by calculation

7 Conclusion Applying non-traditional approach in strength tests of the full-scale structure of a transport aircraft, which consists of combining static and fatigue tests on one object allowed to significantly reduce the duration of the test and the cost of their conduct. This became possible due to the fact that one object was used for testing and the tests were carried out on a single universal set-up. The reliability of the tests was achieved by the accompanying stress researches by experimental and computational methods and testing of full-scale panels.

Conception of Modular Test Stand for Fatigue Testing of Aeronautical Structures Andrzej Leski, Wojciech Wronicz(&), Piotr Kowalczyk, and Michał Szmidt Institute of Aviation, Krakowska Av. 110/114, 05-256 Warsaw, Poland [email protected]

Abstract. Fatigue tests of specimens and components are a necessary part of structural development in aerospace but they are expensive. It can be a problem especially in the case of low cost projects or students researches. Most of them are conducted on testing machines with simple specimens, usually with loads limited to tension mode. The paper presents the Modular Test Stand which was design and developed to decrease the cost of fatigue tests and to test specimens with more complex load condition. The stand consists of three identical sections which are structures similar to the airframe, namely the wing box. Sections are connected, and during a test are loaded in the same manner by bending or twisting moment. The whole section structural node, a particular joint or a skin can be an object of testing. Based on FE calculations, the design of the stand was developed. The desired requirement were uniform stress distribution in skin panels and axial stress level during bending equal to 100–120 MPa. Two stands were constructed - one for bending and one for torsion. Displacements and shearing strains were measured in the central part of the middle skin panel during torsion with the use of Digital Image Correlation method. The measurement correlated very well with FE calculations and confirmed uniform strain distribution in the panel. The stand can be used to examine joining methods, materials but also structures with damages or repairs as well as various types of SHM sensors. The main advantage is a possibility of testing up to six specimens at the same time (double side of three sections) which reduces the cost of a single test. Additionally, a more complex load state can be achieved compare to simple specimens. Keywords: Fatigue


Test
Airframe
FE
DIC

1 Introduction Fatigue tests of specimens and components are a necessary part of an aircraft development (Ansell 2015; Schorr et al. 2015) as well as an introduction of new improvements and solutions into aircraft structures (Leski 2017; Leski et al. 2015). Despite the enormous development of methods of analysis, experimental testing remains the primary method of proving fatigue strength (Schijve 2009). This is caused by the risk of a crash in the case of failure, as well as a limited ability to properly identified fatigue behavior through analysis (Swift 1999). At the same time fatigue tests © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 747–761, 2020. https://doi.org/10.1007/978-3-030-21503-3_59

748

A. Leski et al.

are definitely more expensive than static tests since they are more complicated and last longer. The majority of fatigue tests are conducted on testing machines with simple specimens, usually with loads limited to tension (Okada et al. 2015; Skorupa et al. 2010). Introducing a more complex state of loads (compression, torsion and bending) requires a preparation of a more complicated specimen (e.g. sub–component) and a dedicated testing stand (Aoki et al. 2011; Molent et al. 2009; Tsukigase et al. 2015). The high cost is a problem especially in the case of low cost projects or students researches where industry is not involved.

2 Conception of the Modular Test Stand The Modular Test Stand (MTD) was proposed by the authors to decrease the cost of fatigue tests and enable simple specimens to test specimens with a more complex load condition. The conception assumes that during one experiment, several specimens can be tested and it is possible to test a simple specimen which is mounted in a more complex structure. The MTD consists of three identical sections, which are connected in such a way that each section is loaded in the same manner. The section is a structure similar to airframe, namely wing box, and consists of two spars, ribs and flat skins. The scheme of the MTD is shown in Fig. 1. The whole section, a structural node, particular joint or a skin can be an object of testing. The single section has dimensions of 600  600  150 mm. Each section can be quite easily disconnected and replaced. It is also possible to use one or two sections only.

Fig. 1. Scheme of three sections with fixings.

The sections mounted on the stand can be loaded by a bending or twisting moment (not simultaneously). One end is fixed and the other is loaded with a moment induced by two hydraulic actuators. During a fatigue test, wing are usually loaded in several points distributed along wingspan (Brzęczek et al. 2014). As a result, bending moment

Conception of Modular Test Stand for Fatigue Testing

749

as well as shearing force are not constant. Since all panels should be loaded in the same way, the moment in the MTD is constant along the wingspan and shearing force is not present. During tests of panels, structural joints or singular section such condition is acceptable and in the case of some types of examination, e.g. composite patch repairs, it is even more beneficial. The MTD was designed for bending as a primary loading mode, and torsion was assumed as a possible additional loading mode. Because there was definitely higher interest in testing with torsion, it was decided to build two stands, one for bending and another one for torsion. If the loading sequence consists of bending and twisting cycles, the wing box will be moved from one stand to the other with fixing which makes such an operation quite simple. Configurations of the stands are presented in Figs. 2 and 3. In both cases, the actuators are vertical which eliminates the necessity of balancing them.

Fig. 2. Configuration of stand for bending.

Fig. 3. Configuration of stand for torsion.

The proposed stands can be used to determine fatigue behavior of the whole sections (e.g. made in various technologies), selected structural nodes, joining methods (e.g. riveting, bonding, Friction Stir Welding, etc.) or materials (metal, composite, FML). The MTD can also be used for fatigue tests of structures with damages or repairs

750

A. Leski et al.

(e.g. composite patch repairs) as well as with various types of sensors (e.g. used for Structural Health Monitoring). The main advantage is a lower cost and a more complex load state compared to simple specimens.

3 Design The modular Test Stand has been designed with the use of FE analyses in order to obtain quite uniform stress distribution in the tensioned skin under bending, with the assumed level of 100–120 MPa, which is typical in aircraft design (Müller 1995; Müller and Hart-Smith 1997). Additional limitations were the maximum value of force and displacement ranges of the selected actuators (25 kN/500 mm). After several iterations, the final FE model, presented in Fig. 3, has been developed. The model consists of three sections and fixing connected with each other. One section is assembled from two ribs, two spars and skins. Stringers (two per each skin) were introduced to reduce buckling. Sections are connected by spars, and additionally, thicker skins without stringers. Fixings at both ends are identical. Loading (bending or twisting moment) is introduced by a pair of forces applied to beams (H–sections). At the other end, fixing was constrained in two areas (see Fig. 4.). Linear four-node FE shell elements were used for the wing box and fixings. Beams were modelled by linear two-node elements. Elastic material models of the 2024-T3 alloy (E = 72,000 MPa, m = 0.33) and steel (E = 210,000 MPa, m = 0.3) were assumed. Rivets and screws, that join elements, were not modelled (FE nodes were merged). Analyses were performed with the nonlinear implicit algorithm (MSC Nastran Sol 400) for bending and twisting separately.

Fig. 4. FE model of stand.

Conception of Modular Test Stand for Fatigue Testing

751

Quite uniform axial stress (x) distribution has been achieved in tensioned skins of sections during loading. Figure 5 presents a map of x-stresses in tensioned panels, Fig. 6 shows a graph of x-stress distribution in their axis. The assumed stress level of 100–120 MPa was obtained.

Fig. 5. Axial stress distribution in tensioned skins of sections during bending.

Fig. 6. Axial stress distribution in axis of tensioned skins during bending.

752

A. Leski et al.

Based on the FE calculations, the detailed CAD model were prepared. It was assumed that the designed wing box will be used for tests of skins. To facilitate mounting and exchange skin panels after a test, screws were used to join them with the rest of the riveted structure. Each section could be replaced, e.g. with a fully riveted or composite section. The model is shown in Fig. 7.

Fig. 7. Three sections with fixing, upper skins of one section and one connection were hidden.

4 Installation and Verification After manufacturing, the wing box was installed at first in the stand for torsional loading. Additional support with a bearing was introduced in the axis of rotation to ensure clear torsion, without bending. It is shown in Fig. 8. The dummy, steel specimen was installed in the stand for bending loading in order to configure and check the loading system. The stand is presented in Fig. 9. The stand for torsion was equipped with two 10 kN actuators (SAVAD 10–200), the stand for bending with two 25 kN actuators (MTS 201.17–500). Actuators were controlled by the MTSFlexTest60 controller.

Conception of Modular Test Stand for Fatigue Testing

Fig. 8. Test stand for torsion with wing box.

Fig. 9. Test stand for bending with dummy specimen.

753

754

A. Leski et al.

After installation, the wing box was statically loaded with a torsional moment (9.1 kNm) and one of the skin panel was measured with the use of full-field method (Digital Image Correlation) (Sutton et al. 2009). The two camera system (3D DIC) was used to investigate out-of plane displacements and shearing strains of the panel. The aim of the measurement was to validate if the proposed experimental set up gives symmetrical displacements and produces constant shearing strains during torsion loading. In the experiment, two Grasshopper 5 Mpx (Pointgrey) cameras equipped with 25 mm focal length lenses were used. Cameras were placed under the experimental setup and the field of view was set at its middle section. Frames during the measurement where taken every 0.2 s. The measurement registered one full cycle of torsion loading which took 135.6 s. During this cycle, torsional moment increased to 9.1 kNm, then decreased to zero and reached the same value at the opposite direction and then again fell to zero. At both extremes (+9.1 and −9.1 kNm) the moment was constant for a few seconds. It can be seen in the graph of vertical displacements of points as a function of time (upper plots in Figs. 11 and 12). The DIC displacement and strain analysis were performed with GOM Correlate Professional 2018 software (GOM GmbH). The subset of 27 pixels and 17 pixels step was used in a correlation algorithm. The sections of the panel with rivets were cut out from the analysis to prevent high noise of the measurement which occurred near the edges of the rivets. The DIC measurement was taken for middle panel. Configuration of the experiment is shown in Fig. 10.

Fig. 10. Configuration of DIC measurement, side (upper) and top (bottom) view.

Conception of Modular Test Stand for Fatigue Testing

755

Six points of the panel was chosen for plotting out-of plane displacements (Z direction) and shearing strains exy. Additionally, displacements and strains were plotted along two cross-sections of the panel for two extrema of loading: at 36th second and 102nd second (Figs. 11–14). The displacements in points show a highly symmetrical movement in both cycles of loading. The points on the horizontal centre line of the panel, which is above the axis of torsion, stays close to zero showing the expected behaviour. On the cross-sections, the distortion of initially flat panel can be observed which is caused by local buckling of the skin. On the cross section in the axis of torsion, displacement in the centre of the panel caused by this phenomenon was present (Figs. 11 and 12).

Fig. 11. Displacements Z for first extrema of loading (36th second), upper plot – displacements vs time, lower plot – displacements on cross-sections (Section 1 – vertical cross-section, Section 2 – horizontal cross-section, directions correspond to directions of axes) in 36th second.

Shearing strains of the surface of the panel reached the maximum value of about exy = 0.0006 for both extrema of loading. The uniform strains can be observed in the middle part of the panel. Closer to the edges of the panel the strains have different values. It can be especially seen in the first extrema of loading (36th second). Second cycle of loading with extrema at 102nd second shows less distortion of strains (Figs. 13 and 14).

756

A. Leski et al.

Fig. 12. Displacements Z for second extrema of loading (102nd second), upper plot – displacements vs time, lower plot – displacements on cross-sections (Section 1 – vertical crosssection, Section 2 – horizontal cross-section, directions correspond to directions of axes) in 102nd second.

Fig. 13. Strains exy for first extrema of loading (36th second), upper plot – strains vs time, lower plot – strains on cross-sections (Section 1 – vertical cross-section, Section 2 – horizontal crosssection, directions correspond to directions of axes) in 36th second.

Conception of Modular Test Stand for Fatigue Testing

757

Fig. 14. Strains exy for second extrema of loading (102nd second), upper plot – strains vs time, lower plot – strains on cross-sections (Section 1 – vertical cross-section, Section 2 – horizontal cross-section, directions correspond to directions of axes) in 102nd second.

FE model described at the beginning of this paper was used to analyse the experiment with DIC measurement. Figures 15–16 present comparison of vertical displacements (z) and shearing strains along Sect. 1 defined above, obtained experimentally and numerically. A good agreement between experimental and numerical vertical displacements was obtained, however, FE model is more rigid, which is caused mainly by neglecting joints (rivets and screws). The very good agreement was obtained for shearing strains in the central part of the panel for both extremes, and in the case of the second extremum (102 s), also outside this area. During the first extremum (36 s), measured strains outside the central part have significantly lower values than measured during the second extremum (102 s) and obtained numerically. In the numerical model, the region of influence of stringers (decrease of strains in this region) is considerably wider than while measured experimentally.

758

A. Leski et al.

Fig. 15. Displacements Z on cross-Sections 1 a) first extrema of loading (36th second), b) second extrema of loading (102nd second).

Conception of Modular Test Stand for Fatigue Testing

759

Fig. 16. Strains exy on cross-Sections 1 a) first extrema of loading (36th second), b) second extrema of loading (102nd second).

5 Conclusions The conception of the Modular Test Stand has been proposed by the authors to decrease the cost of fatigue tests and enable to examine quite simple specimens in complex loading conditions. Based on the FE calculations, the design of the stand was developed, which meets assumed requirements; uniform stress distribution in skin panels and axial stress level during bending equal to 100–120 MPa. Two stands were constructed - one for bending and one for torsion. Displacements and shearing strains

760

A. Leski et al.

were measured in the central part of the middle skin panel during torsion, with the use of Digital Image Correlation method. The measurement correlates very well with the FE calculations and confirmed uniform strain distribution in the panel. This indicates that the stand was appropriately designed and manufactured and is usable for various test types. The next step of the project will be an installation of the wing box in the stand for bending and verification of strain distribution with the use of DIC and strain gauges. Acknowledgements. Authors would like to express their gratitude to the Laboratory of Structures team for their help at design stage and during the test execution as well as for drawings and pictures of the stands. We would also like to thank the management of the Materials and Structures Research Center and the Center for Composite Technologies of the Institute of Aviation for enabling this research. The researches were financed from the subsidy granted by the Polish Ministry of Science and Higher Education for statutory activities of the Institute of Aviation.

References Ansell, H.: Structural integrity assessment of gripen NG aircraft. In: Proceedings 28th ICAF Symposium–Helsinki, pp. 610–624. VTT, Helsinki (2015) Aoki, Y., Hirano, Y., Sugimoto, S., Iwahori, Y., Nagao, Y., Ohnuki, T.: Durability and damage tolerance evaluation of VaRTM composite wing structure. In: ICAF 2011 Structural Integrity: Influence of Efficiency and Green Imperatives, pp. 561–572. Springer (2011) Brzęczek, J., Gruszecki, H., Pieróg, L., Pietruszka, J.: Selected aspects related to preparation of a fatigue test plan of a metallic airframe. Fatigue Aircr. Struct. 2014, 88–94 (2014) Schorr, F., Stodt, M., Plate, T.: A350 XWB EW test - combined static and fatigue testing of composite aircraft structures within one test set up. In: Proceedings 28th ICAF Symposium– Helsinki, pp. 855–860. VTT, Helsinki (2015) Leski, A.: Localization of sound sources during full scale fatigue test of the vertical stabilizer with the acoustic holography technique. Fatigue Aircr. Struct. 2017, 17–25 (2017) Leski, A., Kurdelski, M., Reymer, P., Dragan, K., Sałaciński, M.: Fatigue life assessment of PZL-130 Orlik structure – final analysis and results. In: Proceedings 28th ICAF Symposium– Helsinki, pp. 294–303. VTT, Helsinki (2015) Molent, L., Barter, S.A., White, P., Dixon, B.: Damage tolerance demonstration testing for the Australian F/A-18. Int. J. Fatigue 31, 1031–1038 (2009) Müller, R.P.G.: An experimental and analytical investigation on the fatigue behaviour of fuselage riveted lap joints. Delft University of Technology (1995) Müller, R.P.G., Hart-Smith, L.J.: Making fuselage riveted lap splices with 200-year crack-freelives. Fatigue in New and Aging Aircraft, pp. 18–20. Edinburgh, Scotland, EMAS (1997) Okada, T., Liao, M., Machida, S., Li, G., Renaud, G.: WFD evaluation of riveted lap joint. In: Proceedings 28th ICAF Symposium–Helsinki, pp. 250–263. VTT, Helsinki (2015) Schijve, J.: Fatigue damage in aircraft structures, not wanted, but tolerated? Int. J. Fatigue 31, 998–1011 (2009) Skorupa, M., Skorupa, A., Machniewicz, T., Korbel, A.: Effect of production variables on the fatigue behaviour of riveted lap joints. Int. J. Fatigue 32, 996–1003 (2010)

Conception of Modular Test Stand for Fatigue Testing

761

Sutton, M.A., Orteu, J.J., Schreier, H.: Image Correlation for Shape, Motion and Deformation Measurements: Basic Concepts, Theory and Applications. Springer Science & Business Media (2009) Swift, S.: Gnats and camels: 30 years of regulating structural fatigue in light aircraft. In Structural Integrity for the next Millennium. Citeseer, Bellevue, Washington, USA (1999) Tsukigase, K., Fukuoka, T., Kumagai, K., Nakamura, T., Taba, S.: Curved panel fatigue test for MRJ-200 pressurized cabin structure. In: Proceedings 28th ICAF Symposium–Helsinki, pp. 276–286. VTT, Helsinki (2015)

Full Scale Fatigue Testing for Mitsubishi Regional Jet Koji Setta(&), Toshiyasu Fukuoka, Kasumi Nagao, and Keisuke Kumagai Mitsubishi Aircraft Corporation, Nagoya Airport, Toyoyama-cho, Nishikasugai-gun, Aichi 480-0287, Japan [email protected]

Abstract. Mitsubishi Aircraft Corporation is performing the full-scale fatigue testing (FSFT) for Mitsubishi Reginal Jet (MRJ) type certification. Main objective of this test is to show freedom from wide spread fatigue damage (WFD) during the life of aircraft and establish Limit of Validity (LOV). Prior to the test, WFD susceptible structures are defined based on the stress distributions and structure configurations, and they are fully covered by this test. Test duration of FSFT is 240,000 flights (3  DSG of 80,000 flights). The flight-byflight loading spectrum is newly designed for MRJ and its loads occurrence data was verified by flight test data. Furthermore, to reduce the test duration, low loads omission was applied based on the results of some spectrum verification tests. During fatigue test, scheduled inspection consistent with MRJ maintenance program is planned. Since the main objective of FSFT is no WFD substantiation, no artificial crack is introduced during FSFT. Therefore, damage tolerance evaluation (i.e. crack growth analysis validation) is separately conducted by sub-component level testing. For example, damage tolerance substantiation for fuselage structure was performed by using curved panel test facility. This facility can simulate the pressurization load with axial load and their loading sequence can be customized for each test. Major detail design points of fuselage structure such as the lap/butt-joint, cut-out structure and repaired structure are individually evaluated by this type of tests. Based on the test results, potential fatigue critical locations and crack growth behaviors are efficiently investigated, and they significantly contribute to the crack growth analysis validation necessary for CFR/CS 25.571 compliance. Keywords: Full scale fatigue test Damage tolerance


Wide spread fatigue damage


1 Introduction Mitsubishi Aircraft Corporation and Mitsubishi Heavy Industries, Ltd. are developing and producing the next generation regional jet airplane as Part 25 airplane with 70–90 seat capacity. Material selection for the MRJ structure is shown in Fig. 1. Fuselage and wingbox are made of conventional aluminum alloys, such as 2000 series and 7000 © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 762–770, 2020. https://doi.org/10.1007/978-3-030-21503-3_60

Full Scale Fatigue Testing for Mitsubishi Regional Jet

763

series. Therefore, fatigue and damage tolerance characteristics will be the structural key drivers to achieve appropriate durability and operatability.

Fig. 1. MRJ90 structure material usage

2 Evaluation Approach Building block approach shown in Fig. 2 is applied to F&DT substantiation of MRJ. Fatigue evaluation including WFD is conducted by full scale airplane of component tests, and DT evaluation (mainly for the analysis validation) is basically conducted by coupon, element, subcomponent and component level tests. Based on the results of these activities, maintenance program necessary for continues airworthiness is established.

3 Full Scale Fatigue Test Overview Full scale fatigue test began later 2018 at Nagoya, Japan (Fig. 3). Test airplane is fully structural representative production MRJ except for dummy components which assessed by separated component tests. This test will simulate 240,000 flights (3  DSG of 80,000 flights) and finally support the establishment of LOV (Limit of Validity) of MRJ in future. Fatigue tests of composite components (e.g. control surfaces and stabilizers) were separately conducted since LEF (Load Enhancement Factor) is applied.

4 Test Loading Full scale fatigue test airframe is subjected to flight-by-flight spectrum loading during full scale fatigue testing. Flight-by-flight spectrum is defined as the typical and representative flight profile of regional jet. Both ground and flight operating sequence is applied in order as shown in Fig. 4. To simulate realistic load distributions on the airframe, 124 actuators and pressurization supplying systems are installed as show in

764

K. Setta et al.

Fig. 2. Fatigue and damage tolerance substantiation approach

Fig. 3. MRJ90 full scale fatigue test setup

Full Scale Fatigue Testing for Mitsubishi Regional Jet

765

Fig. 4. Flight profile considered in spectrum loading

Fig. 5. Flight types are classified into 13 types considering the difference of load severity, and they are randomly mixed in the block consist of 8,000 flights. FSFT will repeat 30 blocks to achieve 3DSG. To investigate the response of airframe, approximately 3,000 channels of strain are continuously monitored during the test.

5 Test Spectrum Verification To reduce test duration of FSFT, low load omission is necessary. However, change of flight-by-flight spectrum configuration may affect fatigue behavior of airframe. Therefore, Mitsubishi aircraft conducted the S-N curve investigation of major structural joints of PSE (Principal Structural Elements) as shown in Fig. 6, and also some spectrum verification tests to verify the appropriateness of test spectrum. Through such activities, test specific loading spectrum which is equivalent to the original full spectrum was established.

6 Prediction Analysis for WFD Prior to start FSFT, WFD susceptible structures are defined by analysis. They can be selected based on the structural configuration and/or the stress distribution. In addition, detailed FEA is also effective and applicable to estimate fatigue life and identify fatigue critical portions. Figure 7 shows the example of detailed FEA at fuselage butt joints

766

K. Setta et al.

Fig. 5. Layout of loading actuators and test jigs for FSFT

(one of the WFD susceptible structure) for the fatigue criticality evaluation. Stress concentrations for each fastener holes at splice plate were calculated and compared to estimate WFD behavior (life and location). In addition, this may also supports establishment of inspection program for FSFT including teardown inspection.

7 Thermal Stress at Hybrid Joint Thermal stress generated at the hybrid structural joint of composite and metallic structure due to the difference of CTE (Coefficient of Thermal Expansion) may affect fatigue characteristics of airframe. It is not feasible to demonstrate it during FSFT, therefore separated component testing was conducted to validate thermal stress analysis methodology. Full scale composite horizontal stabilizer was installed in the environmental chamber and strain behavior during thermal cycle was investigated. Through this process, thermal stress spectrum generation process was established as shown in Fig. 8.

Full Scale Fatigue Testing for Mitsubishi Regional Jet

767

Fig. 6. S-N curve investigations and test spectrum verifications

(a) FEA Model

(b) FEA Results

Fig. 7. Detailed FEA for the fuselage butt joint (WFD susceptible structure)

8 Full Scale Level Damage Tolerance Tests Since the main objective of FSFT is no WFD substantiation, no artificial crack is introduced during FSFT. Therefore, damage tolerance evaluation (i.e. crack growth analysis validation) is separately conducted by sub-component level testing. For example, damage tolerance substantiation for fuselage structure was performed by using curved panel test facility [1]. This facility can simulate the pressurization load with axial load and their loading sequence can be customized for each test. Major detail

768

K. Setta et al.

Fig. 8. Thermal stress evaluation process for hybrid structure

Fig. 9. Combined loading fatigue test fixture for fuselage curved panel

design points of fuselage structure such as the lap/butt-joint, cut-out structure and repaired structure are individually evaluated by this type of tests [2]. Figure 10 shows an example of comparison of crack growth curve between test and analysis for the fuselage lap splice joint. Analysis well simulates the actual crack growth behavior with an appropriate conservatism (Fig. 9).

Full Scale Fatigue Testing for Mitsubishi Regional Jet

769

Fig. 10. Example of crack growth analysis validation (fuselage lap splice)

9 Conclusion Mitsubishi Aircraft Corporation is successfully started FSFT and conducted various supporting activities based on the building block approach. To simulate realistic operational loads in FSFT, the typical and representative mission profile for regional jet was defined, and flight-by-flight loading spectrum was developed based on the spectrum verification activities. Since FSFT cannot cover all operational loads such as thermal stress induced by the difference of CTE at structural joint portions, some separated evaluations have to be done. Thermal stress at composite-metallic hybrid joints was evaluated by component level testing, and successfully validated their analysis methodology. Fatigue and DT assessment to establish maintenance program will be done based on these analysis methods. After the completion of 3DSG fatigue testing of FSFT, teardown inspection will be conducted to investigate existence of WFD, and finally LOV of MRJ will be established. For the damage tolerance substantiation including repair configurations, subcomponent level tests are being conducted. Fuselage curved panel tests are one of the major DT testing, and they are very effective to justify the fatigue and crack growth analysis methodology.

770

K. Setta et al.

References 1. Tsukigase, K., Fukuoka, T., Kumagai, K., Nakamura, T., Taba, S.: In Curved Panel Fatigue Test for MRJ-200 Pressurized Cabin Structure, Proceedings of the 28th ICAF Symposium, Helsinki, 3–5 June, pp. 276–286 (2015) 2. Setta, K., Fukuoka, T., Kumagai, K., Nakamura, T., Taba, S.: In Structural Damage and Repair Assessment for MRJ Aircraft, Proceedings of the 29th ICAF Symposium, Nagoya, 7– 9 June, pp. 1286–1292 (2017)

Full-Scale Fatigue and Residual Strength Tests of the Composite Wing Box of a Passenger Aircraft K. S. Scherban1(&), A. Yu. Zakharenkova1, V. V. Konovalov1, S. V. Kulikov2, and V. E. Strizhius2 1

Central Aerohydrodynamic Institute, Zhukovsky 1, Zhukovsky, Russia [email protected] 2 Aerocomposite Company, Moscow, Russia [email protected]

Abstract. Fatigue and residual strength tests were performed of the wing box, which was manufactured using a polymer composite material according to a new infusion technology. There are investigated influence of the technological effects (porosity, delamination, etc.) and operational (impact) defects on the durability and strength of the composite structure. Keywords: Composite Residual strength


Impact
Delamination
Infusion
Fatigue


1 Introduction The experience in the design, testing and operation of aircraft structures demonstrates that the advantages of polymer composite materials (PCM) it cannot be fully realized due to the restriction the level of allowable stress associated with the brittle nature of fracture, sensitivity to damage in production and operation, heterogeneity of PCM, instability of technological processes, and environmental influences. Over the past decades, the variety forms of delamination of composite materials became a critical topic of research, and the results of these studies are the subject of many reviews [1–3]. However, these studies do not provide an answer to the problems that has to be solved when designing a structure made from polymeric materials, and which includes: – the strength of the joints of metal and composite parts depends on a large number of parameters that are associated with the use of a composite material (different sensitivity of the metal and composite to different stress levels, fretting in the contact zone, low plastic deformability and, as a result, uneven distribution of forces over fasteners); – barely visible damage of composite parts can grow during compression cyclic loading and lead to a significant decrease in compressive strength; – cyclic loading in typical operation can lead to a decrease in the strength capacity of the composite structure.

© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 771–787, 2020. https://doi.org/10.1007/978-3-030-21503-3_61

772

K. S. Scherban et al.

This has required to fatigue and residual strength test of the full scale wing box for the purpose of experimental proof of basic manufacturing techniques, assembly and quality control of the composite part of the MS-21 aircraft.

2 Object of Testing The test objects were two identical wing box of a passenger aircraft (Fig. 1). The test object included the loading box, the wing box and the dummy of the center section. In the wing box, the upper and lower panels, the front spar and the end part of the rear spar, as well as two ribs were made from PCM and were monolithic integral structures, each of which was made in one technological process by infusion technology.

Fig. 1. Test object

The composite structural elements were made from PCM based on epoxy resin reinforced with carbon fibers. The test object was design by company “Aerocomposite” with company “Irkut” in the framework of a common project of the composite wing of the MS-21 aircraft.

3 Test Procedure The test procedure provided testing of two wing boxes, each of which was tested in the following sequence: – – – – – –

visual and instrumental inspection of the structure in the state of delivery: application of impact damage with energy from 90 to 240 joules; fatigue tests up to 60,000 flight cycles; loading up to 120% of the limit load of the load case “A”; fatigue tests up to 120,000 flight cycles; loading up to 120% of the limit load of the load case “A”.

Full-Scale Fatigue and Residual Strength Tests

773

In fatigue tests, the structure was cyclic loaded by a flight cycle, which included 3 pulsating cycles simulating air flight modes and 1 ground-air-ground cycle (GAG). Minimum cycle GAG simulated the lowest load during the ground phase, and the maximum - the highest load during the flight phase.

4 Test Set Up Fatigue tests were carried out on a special set up (Fig. 2), which simulated the variable loading of typical operation of a passenger aircraft. For testing, the wing box was jointed with the dummy of the center section, which was attached to the reinforced colonnade.

Fig. 2. Test set up

The loading was carried out by a multichannel system of electrohydraulic loading, which consisted of 14 digital servo-cylinders controlled by the CanBAS fieldbus. The loading box was loaded with four servo-cylinders. This made it possible to reproduce the bending and tortion moments arising in the cantilever part of the wing. The reproduction of the distributions of shear forces, bending and torsion moments arising on the ground and flight modes was provided by 10 servo-cylinders. The servocylinders were attached to the load tree systems which applied forces to the wing box.

774

K. S. Scherban et al.

The loads were applied to the lower surfaces at 36 points, which were located at the points of intersection of the ribs with the front and rear spars. To reproduce the specified loads, an control system was used, which is a program complex that provides synchronization of servo cylinders by the CanBAS digital bus, check the accuracy of reproducing the specified loads, produce and receive positional signals.

5 Researching the Deformations of the Wing Box Structure The deformations of wing box structure were measured by strain gauge sensors. The deformations were measured both the maximum and minimum GAG cycle. The measured deformations e% arising on the upper panels is shown in Fig. 3, and in the lower panel in Fig. 4. From the figures, it can be seen that the greatest deformations on the top panel occurred between the ribs 9–10 and reached −0.22% in the flight mode and +0.031% in the ground mode. The greatest deformations on the lower panel appeared near the manhole and reached +0, 29% in the flight mode and −0.04% in the ground mode.

Fig. 3. Deformations on the upper panel (flight modes/ground modes)

Full-Scale Fatigue and Residual Strength Tests

775

Fig. 4. Deformations on the lower panel (flight modes/ground modes)

6 Inspection of the Wing Box Before Testing Inspection included visual external and internal structure inspection and ultrasonic inspection of the upper panel. As a result of inspection of the wing box № 1, the following technological defects were detected: – on the outer surface of the upper panel the defects in the form of poor-quality gluing of fiberglass, and on the bottom panel - scratches; – on the inner surface of the panels fiberglass rupture and non-gluing to the panel; – areas with increased damping of ultrasonic signals along the front and rear edges of the upper panel (increased porosity and/or insufficient impregnation of individual layers is possible). According to the results of ultrasonic inspection of the wing box № 2, the following defects were detected: – defects in the base of the T-stringers on the upper panel; – non-gluing of the surface layer on the front spar wall. Further tests were carried out with the detected technological defects.

776

K. S. Scherban et al.

7 Impact Damaging the Structure Before the tests, three categories of impact damage were performed on the upper and lower panels from the outside: visually undetectable (BVID-impact energy 90 J), hard to detect (small VID- impact energy 140 J) and visually detectable (large VID- impact energy 240 J). For impact damaging was used pile driver with vertical falling load Dynatup 9250HV INSTRON Company with impact energy of 2.7- 1600 J. Impacts were cylindrical steel striker with hemispherical tip diameter of 25 mm. The damaged zone was inspected by the ultrasonic echo method (ultrasonic testing) to determine the size and shape of the cracking zones. SiteScan D 20 ultrasonic detector with THM 2-10 Z2 MHz converter was used for testing. As a result of the impact, a dent was formed on the surface, and it was detected the cracking of the matrix (Figs. 5 and 6).

Fig. 5. Scheme of impact damage on the upper surface

From a consideration of the figure it may be noted that in the result on the surface has appeared the dents (0,2–0,6 mm), while in the panel the cracking, the size of which reached 170  110 mm. These results showed the very limited possibility of visual detection in operation.

Full-Scale Fatigue and Residual Strength Tests

777

Fig. 6. Scheme of impact damage on the lower surface

It was formed histogram of the depth of the dent versus the impact energy (Fig. 7), and a histogram of the area of the cracking zone versus the impact energy (Fig. 8). From the consideration of histograms, it can be noted that both the depth of the dent and the area of the cracking zone depend not only on the value of the impact energy, but also on the local stiffness in the impact zone. The greatest depths of dents and areas of cracking were observed between the stringers.

Fig. 7. Histogram of the depth of the dent versus the impact energy

778

K. S. Scherban et al.

Fig. 8. Histogram of the cracking area versus the impact energy

8 Tests of Wing Box №1 At the first stage of fatigue tests after 36,000 flight cycles, damage to both the lower and upper panels were detected. Four types of damage were detected in the upper panel (Fig. 9):

Fig. 9. The damaging of the upper panel

Full-Scale Fatigue and Residual Strength Tests

779

(a) delamination of the stringers in the regular zone; (b) detachment of the stringer with simultaneous it delamination in the regular zone; (c) detachment of the stringer with simultaneous it delamination in the end of the stringer in the impact damaged zone; (d) detachment of the stringer with simultaneous it delamination in the zone of the stringer ending without impact damage. On the lower panel the reinforcing of the manholes detached and the stringer delaminated (Fig. 10).

Fig. 10. The damaging of the lower panel

After repairing, in order to confirm the residual strength, the wing box was loaded up to 120% of the limit load. After loading, no additional damage was detected. However, in the process of loading with loads exceeding the limit loads, according to the indications of acoustic emission monitoring sensors, an increase in acoustic activity was observed, which could be due to cracking of the matrix. Further cyclic loading was carried out with reduced 20% loads on the “flight mode” until the 120,000 flight cycles. The scheme of damages observed at this stage is shown in Fig. 11.

Fig. 11. The scheme of damages after 120,000 flight cycles

780

K. S. Scherban et al.

From the figure it is clear that the size of the damage that was identified before the tests, as well as the impact damage, remained virtually unchanged up to 120,000 flight cycles. It is important to note that the damage growth was also not observed in the repairing zones. However, during tests, numerous damages were additionally detected in the base of the stringers. The greatest damage of the stringer was observed for stringer № 2, which had increased up to length equal to the distance between the ribs 7–8. The damaged structure was statically loaded. After holding the structure for 40 s at 120% of the limit load, a structure fractured in the section between the ribs 9–10. Diagram of all damage the wing box after testing is shown in Fig. 12.

Fig. 12. The main structural damage of wing box structure

Inspection detected the following main structural damage: – – – –

the fracture the upper panel between the ribs 9 and 10; the fracture the front and rear spars; detachment and fracture the stringers in the zone from rib 8 to rib 11; the fracture and deformation of the bolts in the zone of fracture of the upper panel along the front and rear spars.

In the process of loading, vertical displacements of the upper panel were measured by the photogrammetric method and acoustic emission monitoring was carried out. The vertical displacements of the upper panel between the ribs 9–11 within four milliseconds before fracture is shown in Fig. 13.

Full-Scale Fatigue and Residual Strength Tests

781

Fig. 13. The vertical displacements of the upper panel between the ribs 9–11 before fracture

It can be seen from the figure those 4 ms before the fracture without an increase in the load, the panel buckled. The deflection of upper panel between the ribs 10–11 reached 18 mm. It can be assumed that the panel buckling caused the fracture of the bolts on the rear spar and, subsequently, the local fracture of the panel in section at the rib 9. This is also confirmed by an increase in the intensity of acoustic emission signals from the zone of fracture.

9 Tests of the Wing Box №2 The tests of the wing box №2 were carried out according to the same method, as the tests of the wing box №1. Strains measuring the upper panel (Fig. 14) and the lower panel (Fig. 15) showed that on the upper panel the greatest compressive deformations were measured between the 9–12 ribs and reached 0.15%, and on the lower panel tensile stresses reached 0.19% on the manhole reinforcing. It was measured that along the edge of the upper panel near the rib 9, compressive local strains reached 0.21%. After fatigue testing until 30000 flight cycles, the limit load was applied. Periodic measurements of deformations in the process of fatigue tests showed that during test, the deformations remained practically unchanged, with the exception of the edge of the upper panel in the zone of rib 9 (Fig. 16).

782

K. S. Scherban et al.

Fig. 14. Strains in the upper panel

Fig. 15. Strains in the lower panel

It can be seen from the figure that after application the limit load along the edge of the upper panel in the zone of rib 9, the compression deformation suddenly decreased and with further cyclic loading it decreased so that after 50,000 flight cycles decreased by 10 times compared to the initial one. It can be assumed that the sudden reduction of the compressive strain under the application of an limit load is due to local buckling of the upper layer of the panel, which was caused by the delamination of the edge of the

Full-Scale Fatigue and Residual Strength Tests

783

Fig. 16. The deformation on the edge of the upper panel in the zone of rib 9.

panel. Further cyclic loading led to an increase in the delamination zone and, as a consequence, an increase in the buckling of the upper layer of the panel and a change in the compressive deformations on the surface. Periodic ultrasonic testing of the top panel during the test up to 53,577 flight cycles detected damages, which are shown in Fig. 17. In the process of cyclic loading up to 30,000 flight cycles, damage was growing in the base of stringers, as well as local detaching of the stringer at a length no more than 50 mm and the delaminating of stringers at a length of up to 80 mm. It can be noted that the detachment in the base of stringer occurred both in the zones of initial damage and outside these zones. After loading the limit load there was an insignificant growth the delamination of the stringers, the detachment of stringers in the new zones, and the delamination of the panel near the rib 9 along the edge of the panel. The delamination of the upper panel appeared on an area of 120  65 mm. In the zone of delamination, the upper layer of the composite failed at a length of 35 mm. The crack growth in the zones of impact damage was not observed. Further cyclic loading up to 52,576 flight cycles led to an increase in damage in the stringers, and the maximum damage length reached 1080 mm. Also, zones of stringers detachment have grown and additional zones of detachment and delamination the stringers have appeared; it length reached 50 mm. The impact damages and delamination of the panel edge at the rib 9 have not grew. After 52,577 flight cycle, the upper panel fractured in a section at the rib 9 from the zone of its delamination at a length of 670 mm. Simultaneously, the panel delaminated on an area of 83  200 mm on both sides of the fracture and the stringer detached (Fig. 17).

784

K. S. Scherban et al.

Fig. 17. Damages of wing box №2 after 52,577 flight cycles

10 Analysis of Test Results According to the analysis of data presented in [4, 5], the fatigue curve for the case of regular loading with symmetric cycles can be represented as follows: ea 1 ¼ c þ m1g N, Whereea 1 - amplitude of symmetric cycles of normal deformation c, m –constant of material. When loading with asymmetrical deformation cycles with amplitude ea and average em, the amplitude of the equivalent symmetric cycle ea -1 can be determined by the formula: ea 1 ¼

ea e2UTS e2UCS ðeUTS  em Þ2 ðeUCS þ em Þ2

;

where eUTS, eUCS- tensile and compressive break down deformations. Using the above ratios, the results of fatigue tests of wing box № 1 and № 2 are presented in the form of a fatigue curve (Fig. 18). The curve can be approximated by the relation: ea 1 ¼ 0; 47  0; 0521g N; % For comparison, the graph shows the curve of fatigue of a specimen with a free hole made of panel material: PCM based on Cycom 977-2 matrix and Tenax ® IMS carbon fiber. Laying (54/36/10), panel thickness t = 8 mm. From the graph it can be noted that the durability of the wing box is *15 times less than the durability of the specimens. It could be the additional influence of structural and technological factors that are not taken into account when testing the specimens. It should also be noted more than twice the difference in durability of the 1-st and 2-nd wing box, which indicates an increased scattering of fatigue characteristics.

Full-Scale Fatigue and Residual Strength Tests

785

Fig. 18. The fatigue curves

From the comparison of fatigue damage of the upper panels of wing box №1 and №2, it can be noted that the damages was of the same types, but they differed in size, damage zones and durability before their detection. So if, during testing of wing box №1, detaching of the stringers’ was observed in areas of increased compression deformations of *−0.2%, then during testing wing box №2, detaching was observed even in the root part of wing, in which compression deformations did not exceed −0.1%. In addition, in contrast to the wing box №1 on the wing box №2, a new type of damage appeared in the form of delamination along the edge of the upper panel at the rib 9. Monitoring of the damage growth by the ultrasound method showed an insignificant increase in damage over 15,144 flight cycles, and then sudden growth, which led to the loss of the carrying capacity of the upper panel (Fig. 19).

Fig. 19. Damage growth in wing box №2

786

K. S. Scherban et al.

This allows us to conclude that, firstly, there is an increased scattering of fatigue characteristics, and secondly, there is practically no observed growth of damage from the initially detectable size by the ultrasonic method. It should be noted that the fracture of the panel was preceded by the rupture of the bolts of the longitudinal junction of the panel with the spar, which could have caused the local loss of stability of the panel, which led to its fracture. The same fracture dynamics was observed when wing box №1 was tested for residual strength. At the first loading up to 1,2 limit load after 36,000 flight cycles, no damage was detected, both fasteners and panels. At the second loading up to 1,2 limit load after 120,000 flight cycles with simultaneous rupture the bolts the upper panel was fractured. It can be assumed that as a result of cyclic loading, local panel cracking occurred in the area of the bolts, which led to the bending of the bolts in the hole and, as a result, fatigue damage to part of their section. Under loading up to 1,2 limit load, part of bolts ruptured and part of bolt heads pulled through the panel, which caused a local loss of stability of the panel and its fracture. It should be noted that if the fracture of the wing box №1 occurred at a compression deformation of −0.46%, then the fracture of the wing box № 2 occurred under two time less compression deformation of −0.21%, which indicates an increased dispersion of the characteristics of the residual strength of the structure damaged by cyclic loading.

11 Conclusion 1. It was performed fatigue and residual strength tests of a high-loaded composite structure with operational damages, which was a technological prototype of the wing box of the MS-21 aircraft. 2. The test results demonstrated a number of significant features of the composite structures of this class, including: – impact damage to both the upper panels and the lower panels practically does not growth under cyclic loading, and also does not cause fracture when tested for residual strength; – cyclic loading in the compression zones on the upper panel causes delamination of the stringers’ with simultaneous their detachment, delamination of the edge of the panel and damage in the base of the stringers, and in the tension zones on the lower panel delamination the reinforcing of manholes; – damage to the structure by cyclic loading leads to a significant reduction in the strength of the wing box, as a result of reduced available compressive stresses compared to available compressive stresses of non-damaged structure; – both fatigue characteristics and residual strength characteristics have increased scattering. 3. The identified features defined areas for further work to improve the manufacturing technology and in-depth study of the strength of the composite wing in the framework of the MS-21 project.

Full-Scale Fatigue and Residual Strength Tests

787

References 1. O’Brien, T.K.: Towards a damage tolerance philosophy for composite materials and structures, NASA-TM-100548 (1988) 2. Tay, T.E.: Characterization and analysis of delamination fracture in composites: an overview of developments from 1990 to 2001. Appl. Mech. Rev. 56(1) (2003) 3. Pagano, N.J., Schoeppner, G.A.: Delamination of polymer matrix composites: problems and assessment. In: Comprehensive Composite Materials, pp. 1–96, Elsevier Science, Oxford (2000) 4. Polymer matrix composites: Materials Usage Design and Analysis. In: Composite materials handbook (CMH-17), vol. 3, SAE International (2012) 5. Edited by Bryan Harris: Fatigue in Composites. Science and Technology of the Fatigue Response of Fibre-Reinforced Plastics. Woodhead Publishing Ltd, Cambridge, England (2003)

Full-Scale Fatigue Testing from a Structural Analysis Perspective Derk Daverschot1(&), Paul Mattheij1, Mathias Renner2, Yudi Ardianto1, Manuel De Araujo3, and Kyle Graham4 1

2

Airbus Operations GmbH, Hamburg, Germany [email protected] Airbus Defence and Space, Bremen, Germany 3 Airbus Operations SAS, Toulouse, France 4 Airbus Operations Ltd., Filton, UK

Abstract. Generally speaking, full-scale fatigue tests are used to demonstrate ‘Means of Compliance’ (MoC) for Type Certification. Aircraft are designed in accordance with fatigue and damage tolerance requirements; the main purpose of the fatigue test being to provide the physical evidence necessary to validate design assumptions. Located at the top of the test pyramid, these tests come with significant investment. The test objective is therefore not only limited to compliance with regulations, but also aims to obtain highly important experience of the airframe providing significant benefit to future applications. This paper presents the structural analysis view of full-scale fatigue testing, which drives large parts of the test definition, execution, exploitation and use of test outcomes. Evolution of the general fatigue test approach, alignment of fatigue requirements with test execution and exploitation of the test results for Airbus aircraft is explained. Finally, the paper also aims to capture details regarding future development of full-scale fatigue testing. Keywords: Full-scale fatigue test Test damage management


Airbus
Structural analysis


1 Introduction The primary objective of Full-Scale Fatigue Testing (FSFT) of aircraft structures is to validate fatigue & damage tolerance analyses with test evidence. These tests are thus implemented to provide ‘Means of Compliance’ (MoC) with Type Certification (TC) requirements of new programs by providing evidence of structural behaviour in response to fatigue loading for the life of the aircraft and potentially beyond. FSFT demonstrates both durability and damage tolerance characteristics, including Widespread Fatigue Damage (WFD) [1]. As well as delivering MoC, FSFT also provides an important opportunity for the testing of new technologies and design principles in the fully-assembled condition; one successful example being the application of the FibreMetal Laminate (FML) (GLARE®) in the mega-liner barrel test [2] and A380 FSFT [3]. Further to this, [4] provides an example of the introduction of a new design © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 788–800, 2020. https://doi.org/10.1007/978-3-030-21503-3_62

Full-Scale Fatigue Testing from a Structural Analysis Perspective

789

principle; the A400M FSFT being part of the MoC for the A400M FML-reinforced main frame. Airbus FSFT aim to reflect as closely as possible the structure and loading expected of in-service serial aircraft. Therefore, to achieve a high level of representativeness and ensure maximum benefit, dedicated attention is given to the definition of the test specimen by structural analysis specialists: • Numerous varieties of artificial damage are implemented; these include the installation at test start of standard repairs and possible manufacturing non-conformities, as well as artificial damage items designed to demonstrate damage tolerance behaviour and large damage capability. More detail is given in the next section. • The definition of instrumentation is driven by structural analysis requirements; gauges and sensors are installed as appropriate to correlate stress predictions with and/or without damage, to monitor damage evolution, measure load redistribution, deformation, rotation, deflection and etc. To enable the correct interpretation of test results, selection of the right instrumentation and its application to the most relevant locations is key. • Definition of the loading program gives careful consideration to the external loading of the structure (and resulting internal loads), as well as the boundary conditions. More detail is given in the section ‘Loading Program Definition’. This paper concentrates on the structural analysis aspects of Airbus FSFT of metallic structures. Airbus has accumulated extensive experience with large full-scale testing and structural testing in general, from which has been derived a vast amount of engineering data as used for all Airbus aircraft types. FSFT was carried out for each new program, starting with the Airbus A300 up to the current A350 [5]. Airbus general philosophy is to test more than the minimum requirement from regulations of 2 times design or extended service goal. Airbus has carried out a multitude of large component (e.g. engine pylon, Horizontal Tail-Plane (HTP), flap, etc.) fatigue tests. Furthermore, full-scale or large component fatigue tests for composites are usually carried out separately due to differences in fatigue load spectra. One exception to this is the HTP which incorporates a ‘hybrid’ metallic/composite structure; in this case, two separate test load spectra are utilised (one composite and one metallic). These are subsequently applied to a single test specimen; the composite test phase is executed first before metallic parts are renewed and the metallic fatigue test phase is carried out. Some tests were separated into different aircraft sections in order to enable the use of both less complex loading programs and load application systems. This separation gives the added benefit of reducing the total time for execution of the complete test campaign.

2 Test Preparation and Execution The complete procedure of Airbus FSFT comprises the following phases (see Fig. 1), that are accompanied or defined by structural analysis:

790

1. 2. 3. 4. 5. 6.

D. Daverschot et al.

Test Preparation Calibration Fatigue Test Phase Damage Tolerance (DT) Test Phase Residual Strength Test Campaign Tear Down & Storage

Fig. 1. Description of the complete procedure of a full-scale fatigue test

Test Preparation The preparation phase is key to the success of the test, both in achieving the primary objective of Type Certification and extracting the maximum possible return on investment. Top-level program requirements and the experience of the test lab provide the overriding drivers of test definition; this in-turn influences structural analysis objectives with necessary consideration being given to definition of the testing period, loading, data acquisition, and the test rig itself. Once practical constraints are frozen, development of the loading program is initiated based on the theoretical fatigue spectrum to be used for certification. In conjunction, the test-rig is designed with close support from structural analysis specialists to ensure that rig structure and load introduction systems are appropriate given the duration of the test and the load cases to be applied. The test specimen itself is designed to be as close as possible to the anticipated serial standard to ensure representativeness for certification, as well as justification work of later derivations or modification; this could include: • Secondary structural parts, e.g. large system brackets & fairings holding the potential to influence internal load distribution and fatigue behaviour of principal structural elements. • Artificial damage installation, both that used to simulate anticipated natural damage occurrence and that applied to verify assumptions made in analyses for certification (more detail on this is given in the section ‘Test Damage Management’). • Standard repairs and allowable damage to be included in the Structural Repair Manual (SRM). • Manufacturing non-conformities such as scratches, dents, oversizing of holes, etc. Instrumentation of the intact structure is defined with respect to global loading but aims to focus on ‘hot-spots’ identified during fatigue and damage tolerance analyses. Instrumentation for pre-planned artificial damage is installed initially at the given

Full-Scale Fatigue Testing from a Structural Analysis Perspective

791

location on the pristine structure to enable comparison of pre and post-damaged behaviour. In some areas, additional specialised sensors or gauges are applied to measure specific parameters such as deflection, deformation and load distribution. Finally, test load introduction is monitored by dedicated load cells. To conclude the preparation phase, an extensive inspection program is defined to ensure all relevant damage is logged and followed-up. This can include several levels of scrutiny depending on the nature of the damage, e.g. general visual, detailed visual and special detailed (utilising Non-Destructive Test (NDT) methods). Calibration Phase The purpose of calibration is to verify accuracy and test performance through corroboration of strain and load measurement as well as to check stress predictions at specific locations. Additionally, this phase serves to ensure that all systems are operating as anticipated, including rig performance and data capture. Fatigue Phase The fatigue test phase begins only after the approval of the test-lab and structural analysis specialists. During this phase, fatigue loading is applied as specified between stoppages for inspections while a dedicated team ensures efficient reactive support to damage findings. (Further details can be found in the section ‘Test Damage Management’). Data is collected from measurements and all relevant damage is documented accordingly. Damage Tolerance Phase In FSFT, this phase generally runs concurrently with that of fatigue; artificial damage items being installed at stages in load cycling defined by the target inspection intervals. These damage items are specified to ensure that they cannot interfere either with eachother or naturally occurring damage and to avoid load introduction or nonrepresentative areas. In this way, the behaviour of the structure in response to damage is explored; crack growth analysis methodology and load re-distribution is verified. Residual Strength On completion of the damage tolerance test phase, load-cycling is stopped. Subsequently, residual strength of the structure could be investigated, to be decided case by case; this can utilise both naturally occurring damage resulting from the fatigue and damage tolerance phases as well as previously installed artificial damage. In some cases, these damage items are artificially extended or entirely new artificial damage is installed specifically to address residual strength requirements. In all cases, appropriate static load cases are applied in order to verify residual strength assumptions through demonstration of load-carrying capacity in the presence of large damage, including the fail-safe behaviour of structural features employing a multiple load path design concept. Tear-Down In this phase, the test specimen is disassembled and inspections are carried out in locations that could not be accessed during test execution, e.g. areas covered by other components such as doublers or test equipment. Extensive and detailed non-destructive inspections are carried out to identify any damage that contributes to the correlation of

792

D. Daverschot et al.

analysis predictions. Where relevant, extracted structure is documented and retained for potential further analysis in the future.

3 Loading Program Definition Spectra for FSFT represent a variety of load cases within the normal operation of the aircraft. In general, there are four basic load conditions which contribute to the definition of fatigue loads. These are: 1. Steady-state ‘1G’ load cases representing various ground and flight stages. 2. Incremental loads or ‘disturbances’ to steady conditions due to, e.g. gusts, manoeuvers, bumps, etc. 3. Cabin pressurization. 4. Mechanically simulated thermal loading for discrete load introductions. A ‘flight-type’ concept is adopted for simulation of different flight conditions and severities. The flight severity distribution follows the same principle as TWIST spectrum [6, 7]. Fatigue test spectra are derived from the analytical spectra used for certification. The test spectra differ from the analytical in that they must be simplified to contain significantly fewer loading points in order to achieve a realistic test schedule; i.e. the fewer the loading points, the less time taken to apply the spectrum (and vice versa). The simplification process involves eliminating low amplitude/high occurrence cycles (see Fig. 2) while maintaining fatigue representativeness (i.e. reproducing as closely as possible the fatigue damage of the analytical spectra). In the event that underloading results from simplification of the spectra, load levels can be increased to compensate.

Fig. 2. Example of stress-time history simplification

Full-Scale Fatigue Testing from a Structural Analysis Perspective

793

Validation of the simplification is achieved by comparison of fatigue and damage tolerance predictions obtained using the test spectra to those obtained using the analytical spectra. It is not necessary to calculate predictions at all locations of the aircraft structure but rather at specific fatigue-representative locations or ‘pilot points’ distributed throughout the airframe [8]. The validation process takes into account each simplification step in test spectra development. Figure 3 gives an example of the location of pilot points in the fuselage of the A380.

Fig. 3. Example of fatigue-representative locations in A380 fuselage structure

Despite simplification, test spectra contain all the realistic conditions that the aircraft structure is normally subjected to. In some FSFT, this reaches a high level of complexity; e.g. for A400M, specific events were included during load cycling such as operation of the aircraft’s loading ramp and cargo door.

4 Test Damage Management Management of damage is an important part of FSFT; a well thought out approach will ensure that maximum benefit is taken from the test and that any analysis can be conducted in an efficient manner. Examples of the Airbus approach to manage damage findings are given in the following case studies. Natural Damage A380 Window Frame In this example a fatigue crack occurred in one of the passenger window frames of the A380 FSFT. The crack developed in line with expectations at a location previously identified as exhibiting high stress levels. The area was therefore heavily instrumented with strain gauges from the start of the test; these being monitored on a regular basis. The location can be seen in Fig. 4.

794

D. Daverschot et al.

Fig. 4. Natural damage at an A380 window frame equipped with strain gauges (external view (left) and internal view (right))

ε [micro strains]

These regular strain gauge measurements taken during typical flights and at pure pressure cycles allowed the history of damage development to be traced back to crack initiation as can be seen in Fig. 5.

Fig. 5. A380 window frame and fuselage skin strain gauge data indicating crack initiation

Full-Scale Fatigue Testing from a Structural Analysis Perspective

795

This example shows how the regular evaluation of strain gauge data during testing can indicate damage initiation and thus provide an important trigger for inspection. In order to take full advantage of this approach, an in-depth knowledge of the structure and instrumentation is required, and in particular an understanding of potential fatigue ‘hot-spots’; to provide this knowledge, Smarter Testing & Simulation (ST&S) can be used in a supportive capacity to FSFT. For more details, see the section ‘Future Testing’. Fatigue crack growth was monitored for a certain number of cycles and the damaged window frame was subsequently extracted and replaced. Residual strength testing was then performed via simulation of the fully failed window frame, successfully demonstrating load transfer from the window frame to the FML-skin and thus proving the multiple load path capability of the design. In order to reduce stress at the window-frame location, a serial modification was subsequently implemented in the form of a skin doubler underneath it. To ensure continuous airworthiness for pre-modification aircraft, a specific inspection task for the area was defined in the maintenance document Airworthiness Limitation Section (ALS) for the affected structure. A400M Fuselage Frame This example concerns a fatigue damage finding comprising of a crack in a fuselage frame initiating at a ‘mouse-hole’ observed in the A400M FSFT. This can be seen in Fig. 6.

Fig. 6. Natural damage example: A400M FSFT fuselage frame

Crack initiation was detected after 17,500 simulated flights on the right-hand frame; an immediate inspection on the opposite side of the aircraft revealed similar damage. Following risk assessment, it was decided to monitor both cracks through regular inspections and the installation of ‘crack-wires’; a mesh of these devices enabled

796

D. Daverschot et al.

automatic measurement of crack propagation, as well as providing early warning to stop the test should the crack reach a length where instability could occur. After monitoring crack propagation for around 6,000 simulated flights, both cracks were repaired by stop-drills and the installation of doublers. As with all natural damage occurring during FSFT’s, these cracks were thoroughly investigated and evaluated for potential consequences to test continuation as well as impact to the fleet. In this case, the study comprised a loading investigation, root cause analysis, tear down inspection and fracture surface examination which led to confirmation that the stress concentration at the ‘mouse-hole’ feature provided the driving factor in initiation. Tear-down inspection of the part was carried out in the disassembled condition and revealed no re-initiation, either on the repair doubler holes or at the stop drill. This proved to be an excellent opportunity to verify the use this type of repair for in-service aircraft as a temporary solution. Artificial Damage For all Airbus FSFT, a significant part of the test is focused on investigating artificially introduced damage. The nature of these types of damage displayed significant variation, including for example: • Artificial saw-cuts to monitor and correlate crack propagation. • Removal of junction parts to explore the consequences to residual strength. • Introduction of in-service damage such as dents and scratches. A400M Fuselage Skin For artificial damage definition and monitoring an example is given hereafter that was applied on the A400M FSFT. The time of introduction of these damage items can vary greatly; one example being the implementation of a longitudinal saw-cut in the centre fuselage skin first introduced after 15,000 simulated flights, see Fig. 7.

Fig. 7. Artificial damage example: A400M FSFT fuselage skin

Full-Scale Fatigue Testing from a Structural Analysis Perspective

797

The structure showed good damage tolerance behaviour with insignificant initial growth rate. Consequently, this damage was extended twice until 22,500 simulated flights when natural growth began. In this configuration, the test was continued until the end of the test at 27,500 simulated flights and fully matched predictions. With a final crack-length of almost 0.5 m, this damage was subsequently utilised to evaluate residual strength. Artificial damage representing the type typically found in-service (such as dents and scratches) is usually introduced before the start of the test. For the A400M FSFT, dent locations were carefully selected based on their potential criticality whereas scratches were not specifically positioned but rather made use of accidental introduction during manufacturing and assembly of the test specimen. As with natural damage, all artificial damage is assessed by structural analysis specialists for its potential consequences to the flying fleet. Thanks to FSFT, inspection

Fig. 8. Lap joint repair inspected during tear-down (upper picture) and subsequent crack findings (lower picture)

798

D. Daverschot et al.

programs can often be alleviated and baseline justifications adopted for certification and maintenance can be revised. Tear-Down Tear down of the test specimen is the very last phase of FSFT but is essential to ensuring that any wide-spread fatigue damage and hidden damage in the structure are detected and documented. Specific areas of the structure are selected for partial dismantling whereby fasteners are removed to allow access for NDT inspection of holes and overlapping surfaces. In areas with high secondary bending or application of fatigue improvement methods at holes, cracks may initiate some distance away from the hole itself. In such cases, it is standard practice to completely disassemble the structure to ensure crack detection independent of the crack path. Figure 8 provides an example of a complete disassembly of a fuselage lap joint repair showing fastener rows selected for inspection. Crack initiations were subsequently detected at fastener holes in the skin at the run out of the repair.

5 Smarter Testing & Simulation Traditionally, simulation of fatigue comes with a large computational cost. However, with the ever increasing availability of large computing power (e.g. CPU’s, RAM and large data storage), this exercise can now be performed at a Global Detailed Finite Element level, thus opening the opportunity to exploit a range of features leading to what can be described as large-scale Smarter Testing & Simulation (ST&S). It is anticipated that these enhanced digital capabilities will influence the traditional test pyramid concept (supporting fatigue, damage tolerance and residual strength substantiation) and lead to improved predictive simulations at much greater resolution, while improving fatigue ‘hot-spot’ identification during aircraft development. The capability to model fatigue in large scale components at highly refined levels means that a much better understanding of fatigue behaviour can be obtained early on in the development phase and much earlier than the completion of physical tests. The concept of creating a ‘fatigue digital twin’ also brings opportunities to visualise and map fatigue damage severity not previously possible with traditional methods. An example of this can be seen in Fig. 9. In summary, physical testing will be increasingly supported by the careful application of ST&S allowing the following benefits: • Tests can be optimised to focus on key locations where ‘hot-spots’ have been identified, e.g. tailoring of instrumentation and load introduction systems. • A better understanding of the behaviour of the structure in response to fatigue can be obtained earlier on in the development phase allowing for expedited modifications to design. • Development time and costs can be radically reduced. • Repair solutions can be assessed prior to installation.

Full-Scale Fatigue Testing from a Structural Analysis Perspective

799

Fig. 9. Damage severity example: Fatigue equivalent stress (left) and spectrum severity (right)

6 Conclusion Beginning with the A300 and through to the A350, Airbus has amassed world-class knowledge and experience of airframe FSFT that provides the basis for the highest levels of safety. Test evidence obtained by FSFT underpins the definition and evolution of the structural maintenance program, demonstrates airframe structural capability and provides Means of Compliance to airworthiness regulations. Through careful definition of FSFT and subsequent results analysis, Airbus continues to refine fatigue and damage tolerance methods while continuously optimising design. In recent years, FSFT has also been used to validate the implementation of Extended Service Goals (ESG) to Airbus aircraft, thus generating added value for customers. In summary, FSFT consistently provides the foundation for an intimate understanding of how fatigue influences airframe structures and continues to maintain its high importance in aircraft development. Looking to the future, recent developments in computing offer opportunities to study fatigue in ways not previously possible. Through the simulation of large-scale components and even whole aircraft, ‘fatigue digital twins’ hold the potential to overhaul the traditional test-pyramid concept. ST&S will play an increasing role in structural analysis and physical testing and provide the means to radically reduce development times and costs. Ultimately, as ST&S for fatigue becomes more widely understood and associated analytical techniques more refined, it will develop into an indispensable asset for airframe structural analysis.

800

D. Daverschot et al.

References 1. Buchweitz, I., Santgerma, A., Turrel, N.: Airbus on the way to show compliance with WFD regulation. In: Brot, A. (ed.) Proceedings of the 27th ICAF Symposium, vol. I. Jerusalem, Israel (2013) 2. Borgonje, B., Concepción Escobedo Medina, M.: Lessons learnt from the full-scale fatigue test “Megaliner barrel” – F&DT analysis of the GLARE® structure. In: Proceedings of 24th ICAF Symposium, Naples (2007) 3. Bosch, P., Nielsen, T., Radiant, Y.: Test program for the A380 major fatigue test. In: Proceedings of the 23rd ICAF. Hamburg, Germany (2005) 4. Plokker, M., Daverschot, D., Beumler, T.: Hybrid structure solution for the A400M wing attachment frames. In: Proceedings of the 25th ICAF – Rotterdam (NL) (2009) 5. Bösch, P., Eyre-Jackson, D.: A new experience of fatigue testing with the A350 XWB. In: Proceedings of the 29th ICAF Symposium. Nagoya (2017) 6. de Jonge, J.B., Schültz, D., Lowak, H., Schijve, J.: A Standardized Load Sequence for Flight Simulation Tests on Transport Aircraft Wing Structures. National Aerospace Lab. NLR TR73029; LBF-Bericht FB-106 (1973) 7. Lowak, H., de Jonge, J.B., Franz, J., Schültz, D.: MiniTWIST, a Shortened Version of TWIST. National Aerospace Lab. NLR, MP-79018; LBF-Report TB-146 (1979) 8. Graham, K., Artim, M., Daverschot, D.: Aircraft fatigue analysis in the digital age. In: Proceedings of the 29th ICAF Symposium. Nagoya (2017) 9. Fraunhofer-Institut für Betriebsfestigkeit und Systemzuverlässigkeit LBF report TE4-256/74 “Beziehung zwischen Truncation Level und Inspektionsintervall”, Birrenbach Koshorst

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests Risto Laakso1(&), Jussi Kettunen2, and Juha Lähteenmäki2 1

2

VTT Technical Research Centre of Finland Ltd., P.O. Box 1000, 02044 VTT Espoo, Finland [email protected] Patria Aviation Ltd., Lentokonetehtaantie 3, 35600 Halli, Finland {Jussi.Kettunen,Juha.Lahteenmaki}@patria.fi

Abstract. Until 2017, there was no certainty about the fatigue life of Hawk tailplanes in FINAF’s flight conditions. Then full-scale fatigue tests were performed to determine if the FINAF is required to procure more tailplanes, and to extract evidence, which could be used to increase the structural inspection interval times. The tests were executed with two 4000 FH flown tailplanes and the goal was to achieve additional 2000 FH with a scatter factor of 5. Test loads were applied with actuators feeding both buffeting and maneuvering symmetrically at the same time. Test’s spectrum was based on the FINAF OLM strains and on the usage spectrum of the FINAF flights 2014–2015. Limited NDIs were done after every 200–340 EFH and full inspections after every 1000 EFH. Several damages, such as broken rivets and cracks in spars and angles, arose. Following the testing, the tailplanes were subjected to RSTs with the load corresponding the ultimate design load. The tailplanes passed the RSTs without noticeable additional damages. Centre sections were torn down for more detailed inspections. Some fault indications were obtained from the buttstraps, but all the defects were very small. Seven cracks were found on the skins and one location could be determined as the critical location. The centre joint survived the test period. The residual strength was sufficient with a 20 mm crack at the skin rivet hole, which was estimated to be the most loaded. The tests gave solid basis for increasing the TP’s acceptable usage life by 1000 FH. It was possible to determine the crack propagation rate to verify the structural inspection period to be applied. Considerable cost savings will be achieved, because the inspections can now be optimized. In addition, now it is known that the current number of TPs is sufficient with the additional 1000 FH for the targeted HW life cycle, and no additional procurement is required. Keywords: Hawk


Tailplane
FSFT
NDI
Lifetime

1 Introduction As there was no certainty about the Hawk Mk 51/51A/66 tailplane’s (TP) fatigue life in Finnish Air Force’s (FINAF) flight conditions, the FINAF initiated full-scale fatigue tests (FSFT) to be performed on the Hawk (HW) Mk 51/51A/66 (Fig. 1) tailplanes. The FINAF’s main objective was to have conclusive results to determine if the FINAF © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 801–815, 2020. https://doi.org/10.1007/978-3-030-21503-3_63

802

R. Laakso et al.

is required to procure more TPs to keep its fleet operational for all its intended life cycle. The secondary objective was to obtain data about the crack growth rate, which could be used to increase the related structural inspection interval times. This was mainly achieved by means of the periodic non-destructive inspections (NDI).

Fig. 1. The Hawk Mk 51/51A/66. (V. Tuokko, FINAF)

The tests were performed by Patria Aviation Ltd. (prime contractor) with Elomatic Ltd. as a subcontractor for the load distribution plate design and manufacturing work, and VTT Ltd. with VTT Expert Services Ltd. (at present Eurofins Expert Services Ltd.) (main subcontractors) during year 2017. The design work and manufacturing, as well as load spectrum basics, were mainly done in 2016. The structural tests were conducted with two tailplane units, which both had flown approx. 4000 flight hours (FH). The goal was to test the TP’s up to 10,000 equivalent FH (EFH) to achieve additional 2000 FH with a scatter factor (SF) of 5. The first unit had undergone structural repairs during its normal operational cycle at 3330 FH; the latter had repaired just before the test at 4000 FH. The tailplane’s fatigue life is determined by its critical primary structural components: the upper and lower centre buttstrap plates, and centre spar (e.g. middle area in Fig. 4). The stop criteria of the test was either (1) final cracking of the centre line buttstrap joint; (2) damage in a location other than the centre joint so that there is no point of repairing the structure and extending the test; (3) finishing the fatigue test with 10,000 EFH. The calendar duration of the fatigue tests, including all NDI and related activities, was from May to July (TP no. 1), and from August to October (TP no. 2) (2017). Both the tests lasted 10,064 EFH (Laakso 2017) of which the first TP’s qualified test duration was 7000 EFH and the second TP’s 8585 EFH.

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests

803

2 Methods Load Distribution: Aerodynamic Loads The aerodynamic resultant load location was determined with the Hawk’s computational fluid dynamics (CFD) and finite element (FE) models in four different flight conditions (Table 1) (Hoffren 2016). The cases were selected based on strain gauge data from the operational loads measurements (OLM), where tailplane encountered high stress levels (tension, compression) as well as flutter phenomenon and asymmetric flight maneuver (Fig. 2). Table 1. Tailplane’s CFD-case based flight conditions. (J. Hoffren, Patria Ltd.) Case Ma AoA [o] nz-arvio h [m] P [o/s] Model type

1 0,85 0,992 4,5 1100 0 half

2 0,55 9,00 6 1000 0 half

3 0,75 −4,50 −2,5 2000 0 half

4 0,70 0 1,2 2000 180 full

The Hawk’s existing CFD model was updated and the aerodynamic resultant load was evaluated with comparison to the original equipment manufacturer’s (OEM) load case data. From the CFD results, it can be seen that the detached airflow from the wing (Fig. 3) hits the tailplane approximately at the same flight parameters as in the measured data to create buffet loading. This gave good confidence in using CFD loading to estimate stress distribution at tailplane’s FE model (Hoffren 2016). Load Distribution: Loading Aerodynamic loads were transferred to the FE model, and based on the stress distributions (max/min principal stress, von Mises), the appropriate location for one resultant load on each side could be determined. The location of the resultant load was selected by the criteria of obtaining similar stress distribution in both models (CFD loading and single load vector). This corresponded well also to the stress distribution of the OEM’s load case (Figs. 4 and 5) (Lähteenmäki et al. 2017b). Due to the tailplane’s fixed wing structure (i.e. no separate control on left-hand (L/H) and right-hand (R/H) sides, the tailplane moves as one plate), the need for asymmetrical loading was small and thus, for the test system’s simplicity, it could be omitted. In addition, because the tailplanes had already flown 4000 FH before the tests, and with SF of 5, this was considered satisfactory. Load Distribution: Plate Design Based on the location of the resultant and the anticipated stress distribution in the centre joint area, FE model of the FSFT’s load distribution was made (Fig. 6). The design of the load distribution plate was based on the three-dimensional (3D) laser scanned geometry (Fig. 7), and the size of the plates was determined so that they would not fatigue in the tests.

804

R. Laakso et al.

Fig. 2. Example of tailplane’s flight parameter data. (J. Hoffren, Patria Ltd.)

Fig. 3. Detached airflow from wing collides with TP. (J. Hoffren, Patria Ltd.)

As the load was brought into the TP from only one point per side, the load needed to be distributed on a larger surface area with strong plates. Because the inner structure can be damaged in real usage, the stress levels were analyzed, and it was clear that the plate needed to extend over nearby spars and ribs. It was accepted that the tests could not be completed without causing damage to the skin plates and inner structure, so preparatory plans for quick ‘n dirty repairs were developed. The plates were bolted through the TP and tightened to a predefined torque to avoid free play. This introduced initial compression to help the inner structure to withhold the generated tension and shear loadings without buckling.

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests

805

Fig. 4. Centre skin plates’ max principal stress distribution (magnitude and direction) with the OEM’s load case. (J. Lähteenmäki, Patria Ltd.)

Fig. 5. Centre skin plates’ max principal stress distribution (magnitude and direction) with the CFD defined resultant. (J. Lähteenmäki, Patria Ltd.)

Supporting Structures and Load Implementations On the grounds of measures, and TP’s attachment and load distribution points, a specific test rig was designed and manufactured. The tests were installed under a portal loading frame, where the rig was bolted on the laboratory strong floor (Fig. 8). Besides, for symmetrical loading, the rig was designed for a lifetime of at least two fatigue and residual strength tests (RST). Required lateral movement leeway was included in the mounting brackets so there was no need for a separate sliding structure for the rig. Because the open structure caused some undesired lateral displacements during the

806

R. Laakso et al.

Fig. 6. FE model of the load distribution. (J. Lähteenmäki, Patria Ltd.)

Fig. 7. Validation of the FE model by means of laser scanned 3D measures from a tailplane. (J. Lähteenmäki, Patria Ltd.)

Fig. 8. The test set-up of the tailplane. (J. Juntunen, Eurofins Ltd.)

tests, the rig lugs were supported by a horizontal plate, which was detached during NDIs. The upper attachment rod was adjustable, which provided desired position angles of the tailplane (Fig. 9).

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests

807

Fig. 9. Bottom lug’s attachment including horizontal adjustment within the flanges (lug’s sleeve bearing with clearance not shown), first tailplane in the alignment process, and final version of the vertical beam’s top end. (VTT Ltd.)

Test loads were applied with two servo-hydraulic actuators. The actuators were rated at 50 kN and had an end-to-end stroke of 250 mm. They were attached from their upper ends to the adjustable crossbar of the load frame (Fig. 8). Above-mentioned bolted load distribution plates were mounted on the surfaces of the tailplane, where the actuators were attached forming laterally a 10-degree incline relative to the vertical. The purpose was to ensure not only the perpendicular installation to the tailplane’s surface, but also the retention of the load-sharing plates and the minimization of the lateral forces applied to the bolts of these plates. Because of the lateral tilt, the fasteners in top ends of the actuators had to be ball joints (Laakso et al. 2017b).

808

R. Laakso et al.

FWD

Instrumentations, Data Collection, and Calibrations The instrumentation was identical between the tested tailplanes. Strain gauge (SG) channels S01 and S02 (Fig. 10) were mounted to the left and right hand sides on the top skin of the tailplane. S15 was the primary gauge, which was located on the top skin at the tailplane’s centre box (Fig. 10). The channels S01, S02, and S15 were instrumented in accordance with the OLM documentations (e.g. Liukkonen et al. 2004); all the other SG instrumentation was new and made with new drawings. The new strain gauge channels were SHP1 (identical to S15, but in front of it), SHP2 (identical to S15, but on lower surface of the tailplane), and SHP3 (one-grid gauge for verification purposes) (Fig. 10).

R/H

Fig. 10. Strain gauges S02, S15, SHP1, and SHP3. S01 (not in view) is symmetrical to S02; SHP2 (not in view) is below S15 on lower surface of the tailplane. Red circle: prime test area of interest (R/H). (VTT Ltd.)

As the figures shows (Figs. 8 and 9), load cells were attached between the load distribution plates and the actuators, but also the adjustable upper attachment rod was fitted with a load cell due to the need to check the torsional loads of the test arrangement. The tests were running in a displacement control mode, which forced the movement of the tailplane to predetermined limits in case of e.g. system failure or damage of the structure. Despite the use of this, force limits were also kept on. Support reactions, or, for example, the test rig’s internal transitions, were not measured (Laakso et al. 2017d).

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests

809

The tests were monitored remotely on-line via the measurement computer to ensure (when necessary) proper operation, and to capture the data for further analyses. In addition, the purpose of the data acquisition was to continuously record all the signals during the tests, primarily from S15 and the load cells. The system thus recorded data from the channels SHP1 and SHP2 in case of any damage of S15, but also for calibration needs of the tailplane’s FE model. Data from S01 and S02 was recorded mainly for comparison purposes between (1) the OLM system, (2) fatigue test arrangement, and (3) FE analyses. SHP3 or the support bar’s load cell data had no function in the fatigue test itself, but these signals were also collected all the time. The system was intended to be run under load control, but displacement control was chosen because the load distribution plates weighted too much. Calibrations, pretests, and control parameter adjustments were made stepwise (actuators both individually and simultaneously). Before any part of the spectrum was introduced even at reduced levels, specific test signals were used as inputs to make sure that the system is responding in a correct way. After the whole spectrum was pushed through, the resulted fatigue damage value was calculated from the measured test signal S15. In all, the first tailplane was calibrated twice (initialization, calibration after structural repairs) and the second tailplane was calibrated four times (initialization, calibration after repairs, readjusting of the actuator forces, and calibration after repairs). Calibration results showed sufficient linearity and low enough hysteresis. This enabled final generation of the load spectrum so, that the centre boxes at the measurement point S15 were subjected to the same stresses as the flying OLM system with good accuracy. Loading Spectrum The test spectrum was based on (1) the strain gauge signal from the top skin of the tailplane near centre box (S15, Fig. 10) of the two FINAF OLM Hawks: HW-368 (Mk 66) and HW-319 (Mk 51A), and (2) on today’s real, FH and syllabus based (current and future) usage spectrum of the FINAF flights. The developed spectrum was not any generic basic operational spectrum (BOS), because it was edited further to match the needs of the FSFT’s command system. The spectrum was verified with OLM stress life analysis (e.g. Bäckström et al. 2009; Liukkonen et al. 2004; Viitanen et al. 2007). Both buffeting and maneuvering loading components were fed to the test (sideways symmetry) at the same time (Fig. 11, Laakso et al. 2017a). All Mk 66 OLM flights by then and all the last Mk 51 OLM flights were taken into account in compiling the spectrum. The primary aim was to generate a representative test spectrum of approximately 62.5 flying hours using mainly the Mk 66 flights and following the FINAF’s accurate mission based average usage spectrum (FH based distribution; flights 2014–2015). New fatigue life calculations were made for all the selected flights, and fatigue index (FI) values with these calculation results were used to fine-tune the spectrum. According to associated flight reports, the total duration of these selected flights was 68.05 h.

810

R. Laakso et al.

Fig. 11. Shape of the loading spectrum (equivalent to 68 EFH). Computational fatigue (background curves) cumulates effectively all along the line. (VTT Ltd.)

All irrelevant turning points, flights, ground runs, and other non-damaging parts of the selected measurement data were filtered, and effects of the removals were checked by means of fatigue life calculations. As can be seen from Fig. 11, calculated damage increases practically throughout the spectrum. Because of the force-stress conversion factor, the spectrum represented actuator forces [kN] within the desired S15 range. The measured spring constants (per side and per load direction) of the tailplane were also taken into account. Spectrum, after editing, corresponded only as a whole, and in the sense of the tailplane’s fatigue, to a total of 68 h of correctly distributed flying. As mentioned, the result was a specific spectrum of the actuator forces for these Mk 66 tailplane fatigue tests only. More precisely, this was a set of Instron’s control (time series) files for the particular, displacement controlled fatigue test. The repeatedly run test spectrum was provided only with the entire test system used, including the loading frame, moving masses, and all dynamic properties. S15 was monitored to check if the applied load levels were as demanded and equal to the Mk 66 OLM S15 in-flight levels, and, as can be deduced, the use of once formed spectrum for the other, or structurally damaged, worn, or repaired, tailplane was not successful. This was because of the relatively significant stiffness differences per side and per load direction, i.e. due to changes of the spring constants, and due to varying force-stress dependences. Direct load adjustments via the controller’s computer (i.e. no longer through the mechanical calibrations) were eventually inevitable because of the remarkably rapid changing of the state of the structure even within single spectrum only. With the spectrum(s) obtained, the tailplanes were tested as desired. The S15 ratios measured by the OLM system versus the tested tailplanes corresponded each other with good accuracy. Fatigue Tests and NDI The tests were performed full-time 24/7 with normal monitoring resource allocations. As a rule, fatigue testing was timed out of the working hours in order to provide NDI inspectors with the best possible working time, minimize daytime testing noise, and make the progress generally as fast as possible. Thus, and as usual, the night-time monitoring of the tests was based on the carefully set system safety limits for position and load: remote access to control processes was not possible (hardware-level control). The operation stopped automatically and in a controlled way, when the next EFH target

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests

811

was reached, or when exceeding the pre-determined limits of force or displacement. Even during the daytime fatigue testing there was only occasional supervising for the test arrangement, but this was in line with normal practice, and did not result in any measures to the test system or to performance. The tailplanes were constantly fastened to the rig and to the actuators. However, the above-mentioned horizontal plate between the vertical supports was removed during the NDIs allowing inspectors to work underneath the tailplane. As no defects had ever been found in the centre box in the FINAF’s use before, it was uncertain whether the fatigue test would result in damages at all, and then how much time NDIs would eventually take, if faults were found. However, it was to be expected, that in final stages of the tests, NDI would require much more working time compared to checking the structure at early stages without any defects. The actuators were used, if necessary, to cause static loads (crack openings) during NDI checks to ensure better detection. NDI’s of the presumed critical areas were done after every 200–340 EFH in a limited extent, and full inspections after approximately every 1000 EFH. The NDI methods were ultrasonic and endoscope inspections as well as visual examinations. The difference in resolutions of the two endoscopes used caused additional difficulties in following the observations inside the structure. Anyway, in most cases it was difficult or even impossible to estimate the progress of the damages because the sightings could not be obtained exactly in the positions of previous endoscope inspections. As the test progressed, several damages, such as broken rivets and cracks in the spars and angles, were found. Figure 12 (left) shows a broken rivet located on the R/H front spar, close to the frame three near the end of the test (9724 EFH). The broken rivets were replaced, cracked items strengthened, and cracks end-drilled, when possible (Fig. 12, right: end-drillings at 2992, 5576, 6936, and 8636 EFH). Besides spare parts for the tailplane structures, also some spare parts for the test rig were needed. Those were installed, and repairs were made, during the test breaks. The test arrangement lasted these fatigue and residual strength tests as planned.

Fig. 12. Broken rivet at 9724 EFH; R/H front spar, close to frame three (left). Continuous enddrillings to stop the crack growth in left front spar (right). (Endoscope views J. Lahti, Patria)

812

R. Laakso et al.

The effect of growing cracks and loosening structure on the loads and accumulated fatigue were calculated after e.g. 340 EFH periods. Load adjustments were made for the second tailplane to cover the changed structure stiffness due to the damages and repairs, and to ensure that the load level of the centre joint remained adequate (Laakso et al. 2017c). Following the fatigue testing, the tailplanes were subjected to a residual strength test with the load corresponding the ultimate design load defined by the manufacturer. Residual strength tests were quite similar comparing to the mechanical calibrations: the loading was applied stepwise by pushing both sides of the tailplane simultaneously down at specific load levels. The second tailplane was loaded in both directions; i.e., both by pushing and then pulling both sides on the predetermined steps. Residual strength test data was recorded as usual, and the tests were videoed. For both tailplanes, a breaking load of 32 kN was exceeded, and on the grounds of visual examination, the structures did not experience any additional damage. Teardown Inspections The tailplanes’ centre sections were torn down for inspections that were more detailed (Lähteenmäki et al. 2017a). Used NDI methods were eddy current, ultrasonic, and penetration liquid tests. The NDI findings during the tests and at the teardown inspections (11/2017–04/2018) were not identical in either of the tested tailplanes. There were numerous indications from scratches or surface cracks supposedly from assembly work and from small chafing which were found but could not be confirmed as fatigue cracks with penetration liquid tests. The most serious and confirmed crack was found on the same location in both tailplanes in the upper R/H skin plate (Fig. 13, red circle). No confirmed cracks were found in the upper or lower centre buttstrap plates, nor in the centre spar. Therefore, the crack’s location in the skin plate was determined to be the critical location.

3 Results The first crack indications from the centre joint were acquired in the NDI examination at 5236 EFH for the first tailplane, and at 6052 EFH for the second tailplane. At this point cracks could not be seen visually. As the test progressed, the largest cracks developed visible, and their growth rate could be monitored. Twenty-two crack indications in two tailplane centre joints were received in the final stages of the tests. Both tailplanes passed the residual strength test without noticeable additional damage. When the tailplanes were torn down, some new fault indicators were obtained from the centre buttstraps. However, all buttstrap defects could be confirmed very small, less than 0.5 mm, which could have been scratches originated during assembly or teardown, not real cracks. On the other hand, seven cracks were found on the skins, the longest 20 mm. The initiation of the crack could be traced back with using NDI reports during fatigue tests and the crack growth rate could be estimated. Therefore, fractography of the cracked surface was not needed.

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests

813

Fig. 13. Upper R/H skin plate; second tailplane. (J. Lähteenmäki, Patria Ltd.)

The most important result of the fatigue tests was that the tailplane centre joint lasted throughout the planned fatigue test period. The second important result was that the tailplane’s residual strength was sufficient with a 20 mm crack at the rivet hole, which was estimated to be the most loaded location in advance. Because the damages did not extend during the worst-case scenario, ultimate loading would not cause an immediate danger to the flight safety. Third, from the fatigue test it was possible to determine the crack propagation rate to verify the structural inspection periods to be applied for the FINAF HW fleet.

814

R. Laakso et al.

4 Conclusions Because there was no certainty about the Hawk Mk 51/51A/66 tailplane’s fatigue life in the FINAF flight conditions, specific FSFTs were justified. The main objectives were to determine if the FINAF is required to procure more tailplanes, and to obtain data about the crack growth time by means of the periodic NDI during the tests. The tests gave strong basis for increase tailplane’s acceptable usage life by 1000 FH. This is based on: (1) (2) (3) (4) (5)

Two test specimens where crack was initiated after 5000 EFH The load spectrum was based on measured operational (current and future) usage Test load levels were valid until 5000 EFH Usage of conservative scatter factor of 5 The damages did not extend during the residual strength tests i.e. the ultimate loading would not cause an immediate danger to the flight safety.

The significance of the research is that considerable cost savings will be achieved, when the inspections are optimized for the FINAF Hawk fleet. In addition, now it is known that the current number of tailplanes is sufficient with the additional 1000 FH for the targeted HW life cycle, and no additional tailplane procurement is required.

References Bäckström, M., Liukkonen, S., Laakso, R., Viitanen, T., Koski, K., Teittinen, T.: OLM/HOLM analysis environment, version 3. VTT Research Report No VTT-R-06669-08 (in Finnish). VTT Technical Research Centre of Finland Ltd, Espoo (2009) Hoffren, J.: Renewal of Hawk’s CFD grid and applying it in load calculations of the tailplane. Design Report No HW-S-0030 (in Finnish). Patria Aviation Ltd, Tampere (2016) Laakso, R.: HW TP FSFT. VTT Customer Report No VTT-CR-00791-17 (in Finnish). VTT Technical Research Centre of Finland Ltd, Espoo (2017) Laakso, R., Arasto, E., Varis, P., Juntunen, J.: HW TP FSFT: Loading Spectrum. VTT Customer Report No VTT-CR-00834-17 (in Finnish). VTT Technical Research Centre of Finland Ltd, Espoo (2017a) Laakso, R., Juntunen, J., Aitoniemi, J., Mäkinen, J., Salonen, L.: HW TP FSFT: Supporting Structures and Load Implementations. VTT Customer Report No VTT-CR-00832-17 (in Finnish). VTT Technical Research Centre of Finland Ltd, Espoo (2017b) Laakso, R., Varis, P., Arasto, E., Juntunen, J., Aitoniemi, J., Tuhti, A., Koskinen, T., Siljander, A., Lahti, J.: HW TP FSFT: Fatigue Tests and NDT. VTT Customer Report No VTT-CR00835-17 (in Finnish). VTT Technical Research Centre of Finland Ltd, Espoo (2017c) Laakso, R., Varis, P., Merinen, S., Eskola, S., Teittinen, T., Juntunen, J.: HW TP FSFT: Instrumentations, Data Collection, and Calibrations. VTT Customer Report No VTT-CR00833-17 (in Finnish). VTT Technical Research Centre of Finland Ltd, Espoo (2017d) Liukkonen, S., Viitanen, T., Laakso, R., Siljander, A., Teittinen, T., Bäckström, M., Savolainen, M., Ovaska, T.: FiAF Hawk Mk.51/51A OLM - Final Report. VTT Research Report No TUO33-032359. VTT Technical Research Centre of Finland Ltd, Espoo (2004)

Hawk Mk 51/51A/66 Tailplane Full-Scale Fatigue Tests

815

Lähteenmäki, J., Lahti, J.: Teardown Inspections of HW Tailplanes LYK11 and LYK45. Design Report No HW-L-0138 (in Finnish). Patria Aviation Ltd, Tampere (2017a) Lähteenmäki, J., Lahti, J., Miettinen, A., Mattila, M.: Centre Buttstrap’s Fatigue Test of the HW Tailplane. Design Report No HW-L-0137 (in Finnish). Patria Aviation Ltd, Tampere (2017b) Viitanen, T., Laakso, R., Bäckström, M., Janhunen, H., Merinen, S., Ovaska, T.: FiAF Hawk Mk.51/51A OLM Follow-up 2006, Final Report. VTT Research Report No VTT-R-06646-07 (in Finnish). VTT Technical Research Centre of Finland Ltd, Espoo (2007)

Progress on the Pathway to a Virtual Fatigue Test Ben Dixon1(&), Madeleine Burchill1, Ben Main1, Thierry Stehlin2, and Raphaël Rigoli2 1

Aerospace Division, Defence Science and Technology Group, Melbourne, Australia [email protected] 2 RUAG Aviation, Lucerne, Switzerland

Abstract. The Advancing Structural Simulation to drive Innovative Sustainment Technologies (ASSIST) collaborative program was initiated by Defence Science and Technology (DST) Group to foster improvements in fatigue life prediction technologies, with the long-term goal of a virtual fatigue test. It is based upon a growing series of airframe challenges, where participants are invited to test stateof-the-art fatigue prediction technologies on problems based on real aircraft structures and loads. Since each challenge is underpinned by a set of demonstrated test results, the predictive ability of all technologies can be accurately assessed. Furthermore, the collaborative forensic review of the challenge results is considered a key output, which will allow the limitations of predictive techniques to be better understood and addressed in future research. DST has established a collaborative online space for the ASSIST community at https://www.govteams.gov.au/. It is being used to share essential data associated with each challenge, provide a collaborative space for discussions, and post results and evaluations. The first ASSIST challenge has been completed and two further challenges have been released to the ASSIST community. The first ASSIST challenge, which was based on a fighter wing-root shear tie post, is described here. It demonstrated the critical importance of having accurate fatigue crack growth data, in addition to understanding the fatigue crack shape and local stress field when the crack depth is still very small. It is considered that future ASSIST challenges will provide many further insights, which will drive improvements to current fatigue life prediction technologies. It is envisaged that the growing database of ASSIST airframe challenges will provide an understanding of the accuracy that can be achieved with current fatigue life prediction methodologies. This is another important product of the program, because it enables the adoption of such methodologies on real aircraft structures. Keywords: Fatigue prediction


Fatigue testing
Finite element modelling

1 Background The underlying aim of a virtual fatigue test; to accurately predict the safe fatigue life of an airframe under service representative loading, is not new. In fact, the analytical prediction of fatigue lives has been a cornerstone of airframe design and certification © Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 816–830, 2020. https://doi.org/10.1007/978-3-030-21503-3_64

Progress on the Pathway to a Virtual Fatigue Test

817

from the time when fatigue degradation was first acknowledged to threaten the safe operation of aircraft. However, the fact that a certification full-scale fatigue test (FSFT) is still required by most civil and military airworthiness standards (e.g. [1–4]) for validation of analysis methods is evidence that fatigue predictions are, on the whole, not accurate or consistent enough to be relied upon alone. A certification FSFT serves two main purposes. Firstly, to generate sufficient fatigue damage across all parts of the airframe1 so that all locations that could develop potentially dangerous fatigue cracking in the service lifetime are identified. That is, to identify all airframe fatigue hot spots. Secondly, aircraft fleet managers need to be able to interpret the fatigue damage that occurred on the FSFT in order to accurately predict the development of fatigue damage in fleet aircraft. The occurrence of fatigue cracking in locations not previously identified as hot spots or earlier than expected during recent certification FSFTs (e.g. [5]) highlights that analytical predictions still cannot adequately replicate the demonstrated capability of a FSFT in performing these roles. The prediction of all hot spots is made difficult by the scale of an airframe, the complexity of its structural loading response and differences between the blueprint design and the as-built aircraft. Finite element models (FEMs) have proven to be useful for simulating the load carrying ability of an airframe, and by extension, predicting hot spots. However, in addition to current limitations of some current modeling techniques (e.g. load transfer at fasteners), their overall fidelity across the airframe tends to be limited by both human and computational capabilities that would be needed to accurately represent and calculate solutions for all locations. Traditionally, a coarse or loads FEM is used to evaluate the stress/strain response of the overall airframe design and provides boundary conditions for more detailed finegrid or sub-FEMs of the hot spots. This approach necessitates choosing hot spots without a detailed FEM, based on flight loads measurement, previous airframe experience, engineering judgement, etc. While this can be effective for airframes and loads resembling past cases, it can fall short for new design features and distinct loading environments. Predicting the stress/strain response of an airframe to flight loading can be complicated by factors such as: aeroelastic structural response to high frequency loading, indeterminacy, build and structural variations, and residual stresses. The sensitivity of fatigue life to small variations in local stress and the inaccuracy of state-of-the-art methods for predicting fatigue lives under a known, but complex variable amplitude (VA) loading spectrum are also key factors that impede the reliable identification of all hot spots via analysis. Furthermore, by extension, the same factors impact the accuracy in predicting the fatigue lives at hot spots under service conditions.

2 The ASSIST Program Research and development are ongoing to address the issues raised above and airframe designers, fleet managers and academia are three of the main areas of effort. Approximately two years ago, Dr. Albert Wong from Defence Science and Technology

1

Typically 2–5 times more damage than is expected in the average service lifetime.

818

B. Dixon et al.

Group (DST) initially proposed a pathway to a virtual fatigue test (referred to as TITANS) [6], a collaboration that would complement the many varied research streams within the global aerospace industry. To better capture the broader context, this DSTled program has been renamed the Advancing Structural Simulation to drive Innovative Sustainment Technologies (ASSIST) program. But ASSIST remains true to the original TITANS blueprint for a collaborative program, where interested researchers, engineers and designers can develop robust fatigue prediction methodologies. Some of the key features of the ASSIST program are: • ASSIST community members perform blind fatigue life predictions for airframe prediction challenges, which are based on realistic aircraft loads and structures. • The merits of each prediction methodology will be discussed within the community based on the demonstrated predictive ability versus actual test results. • Publication of the collaborative forensic review of the results and the current shortcomings of each methodology. • Development of a growing database of predictions for realistic aircraft loads and structures that can be used to establish error bands defining the expected accuracy and consistency for each methodology. This paper describes the philosophy behind the ASSIST program and early progress. Some of the key features of the program are described and these are demonstrated in action via the results of the first airframe prediction challenge.

3 A Pathway to a Virtual Fatigue Test It is considered that the inability to reliably characterise the expected error in fatigue life predictions for real aircraft structures and loading is one significant barrier to the adoption of a virtual fatigue test. When considering the pathway to a virtual fatigue test, consideration has to be given to airworthiness standards. These often have provision for analysis-only certification for limited airframe locations where representative testing cannot be achieved. Such standards prescribe penalty scatter factors (SFs, e.g. 2.0 by [3] and [7]), to compensate for inaccuracy in analytical predictions. However, the prescription of standard penalty SFs is questionable, since it assumes identical predictive accuracy for what may be very different prediction methodologies applied to structural configurations and loading conditions with varying levels of complexity. Depending on the problem under consideration, these penalty SFs may not be conservative enough to ensure an acceptable level of safety throughout the service life, or too conservative, and necessitate unnecessary and expensive fatigue management that can adversely impact availability. In comparison, the approach prescribed by FAR 25 [1], which requires tailoring SF increases based on individual circumstances, can be difficult to implement effectively without reliable error bands for the analytical methods employed. Furthermore, as penalty SFs must safeguard against the worst case predictions, they can be conservative for most analyses and this imposes penalties in terms of reduced component services lives, plus earlier and more frequent inspections. Thus, the error bands associated with the worst case analytical predictions need to be narrowed.

Progress on the Pathway to a Virtual Fatigue Test

819

In summary, it is considered two significant current barriers preventing the adoption of certification via a virtual fatigue test are the lack of reliable error bands when applying analytical techniques on real aircraft structures and the level of inaccuracy associated with the worst case predictions. It is considered that the framework of ASSIST, based around a growing database of fatigue predictions for realistic airframe problems can greatly assist the establishment and improvement of such critical error bands.

4 The ASSIST Framework and Current Progress The ASSIST Community. The ASSIST program is designed to be a multi-nation multidisciplinary collaborative effort and at its heart is the collaborative online space that has been established at the Australian Government GOVTEAMS portal: https:// www.govteams.gov.au. It is being used by DST to share the essential data associated with each airframe challenge, provide a collaborative space for discussions, and post results and evaluations. The site provides an opportunity to have both open and closed communities. Collaborative efforts are encouraged at any level and any arrangement, both within this portal or elsewhere – with the aim of publishing the key findings to drive the technology advances in the industry as a whole. To join the ASSIST community on the GOVTEAMS portal, interested parties are encouraged seek membership by contacting [email protected]. The Challenges. The ASSIST project is built around benchmark testing state-of-the-art fatigue life prediction techniques on real-world aircraft problems, known as airframe challenges. The effectiveness of the prediction methods are assessed against representative tests at coupon, component or full scale level. Furthermore, to facilitate the assessment process, additional data such as crack growth measurements, strain measurements and thermoelastic stress analysis (TSA) results are provided to participants when the results are released. The first ASSIST challenge has been completed and is described in the next Section. As listed in Table 1, there are two further DST-designed challenges that have been released via the ASSIST collaborative portal2. Each of the first three ASSIST challenges has features that are of direct relevance to Royal Australian Air Force (RAAF) aircraft, and thus, they should have applicability to other members of the Aerospace community. It can be seen from Table 1 that the challenges do not cover all aspects of the fatigue life prediction process, but rather, they focus on one or two aspects, so that specific parts of the prediction methodologies can be evaluated and improved upon. As outlined in the Table, for each challenge there is a timeline. First, a description of the challenge and all associated relevant data such as material, surface condition and loading conditions are supplied to the ASSIST community. Challenge participants are then given approximately six months to submit blind predictions, and at the end of this period, the results and relevant associated data are released to participants so that they can evaluate their own methods. The community will then forensically examine the 2

Challenge data packs can also be requested directly via email: [email protected].

820

B. Dixon et al. Table 1. Summary of the three ASSIST airframe structure challenges initiated to date.

Challenge name & details Challenge #1 - wing root shear tie (WRST) Material - AA7050-T7451, Spectrum – fighter aircraft with manoeuvre + buffet Test component – representative geometry coupons Challenge #2 – helicopter truncated spectra (Helo) Material - AA7050-T7451 Spectrum – high frequency flight loads Test component - hourglass coupons Challenge #3 – military transport aircraft (MTA) Material - AA7075-T7351 Spectrum - military transport aircraft Test component – wide flat panels, pre-cracked holes

Estimated dates Data posted Predictions due & on portal results released Challenge now closed.

Draft report for review

30 Mar 2019

01 Aug 2019

01 Dec 2019

30 Mar 2019

01 Aug 2019

01 Dec 2019

results of each challenge and at the end of the review period the community is encouraged to jointly publish key findings3. All members of the ASSIST community are encouraged to contribute their own airframe challenges through the ASSIST collaboration portal. DST will also attempt to incorporate the results of round robin fatigue prediction challenges in the literature (e.g. [8]) into the ASSIST database in order to leverage off past experience. Forensic Review. Detailed review of the results of the challenges is a key component of the ASSIST program that is necessary to understand the accuracy associated with particular methodology/problem combinations. It is also essential to better understand the limitations of the methodologies and focus efforts on ways to overcome these. By identifying where and how the prediction process falls short, we hope that the methodologies can be improved or at least efforts in research can be more effectively directed to overcome the most important shortcomings. For its own part, DST is trialing a detailed review process based on the fishbone method of root cause determination, as well as sensitivity analyses to evaluate the effects of specific parts of the prediction methodologies. However, DST does not claim that it alone will best evaluate these results and recognises that there are many centers of expertise in fatigue prediction

3

DST will seek permission for the use of all participants’ results in this publication.

Progress on the Pathway to a Virtual Fatigue Test

821

incorporating designers, operators and academia who could offer unique insights into the evaluations of the methodologies and the direction of further work needed for sustained continual improvement. It is intended that the evaluation be a collaborative process where all participants are encouraged to perform their own reviews on the submitted predictions and then there will be an active discussion of the conclusions that can be drawn from the challenge. While DST intends to publish a summary of findings with collaborators, it has learned from the review of the first airframe challenge that it can be difficult to explore the intricacies of multiple prediction techniques in a standard 10–12 page journal article. Also, the insights for the challenge have evolved over time, especially as new participants reviewed the results. Thus, DST plans to establish a wiki page that fully explores the results of each airframe challenge at the collaborative online space detailed above. The collection of the supplementary data described above (e.g. strain gauge measurements) is considered essential to the evaluation of the methodologies, because it allows specific components of the predictions to be evaluated. This is clearly illustrated via the assessment of the first ASSIST challenge in the next Section. DST encourages a robust discussion within the ASSIST community regarding what types of supplementary data should be collected during the testing phase of the airframe challenges in order to maximise the insights that can be gained during the evaluation process. Furthermore, DST is investigating the use of machine learning to improve its fatigue prediction capability and considers that maximising the relevant data characterising the conditions and observed fatigue behaviour of the test data added to the learning database will give such processes the maximum probability of success. It hopes that such discussions will lead to better datasets being used in the machine learning studies.

5 Airframe Challenge 1: Fighter Wing Root Shear-Tie Post Problem Description. The first challenge involved the fatigue life prediction of a test component representing a fighter aircraft wing root shear-tie post, shown in Fig. 1. Five components with geometry and material replicating those on an aircraft were manufactured and tested until failure under representative spectrum loading. The shear-tie post itself is a solid, cylindrical feature that is approximately 70 mm long with a 48 mm diameter. As illustrated in Fig. 1, each component was symmetrical and produced two fatigue test results. The components were machined from AA7050-T7451 plate and the surface was subsequently etched according to McDonnel Douglas Process Specification PS 13143. As illustrated in Fig. 1, for one of the test specimens, uniaxial and strip strain gauges were used to collect strain data to validate FEMs [9] and these data were also made available to the challenge participants. The test loading spectrum was based on inflight strain measurements and included the influences of wing shear loads caused by manoeuvres and aerodynamic buffet of the flaps. A test spectrum block comprised four repeats of a 66,920 turning point flight loads spectrum, followed by a set of marker loads (1,460 turning points). The marker loads made regular identifiable markings on the fatigue crack surfaces, which allowed the measurement of crack growth curves post-test using quantitative fractography. The results presented below are plotted against spectrum blocks.

822

B. Dixon et al.

Fig. 1. The component test specimen for the first ASSIST challenge.

Data Provided to Participants. The following information were available to the challenge participants: • test component geometry and loading conditions; • full definition of the loading spectrum; • standard material properties including Young’s Modulus, Poisson’s ratio, yield strength and fracture toughness; and • input deck for validated Stress Check ® FEMs of the cracked and uncracked component. Blind Predictions Submitted. Participants were requested to provide both fatigue life and crack growth rate data from an initial crack size, ai = 0.01 mm, along with any assumptions used for the predictions. Presented in this paper are fatigue life solutions from three independent participants, which are referred to as: DST FASTRAN, DST EASIGRO and RUAG AFGROW. Figure 2 plots all the predictions, noting that multiple solutions were submitted by each of the participants. Also shown in Fig. 2 are the crack growth measurements from the five shear-tie post component tests. Note that there were two results from each test due to the symmetric coupon design. The fatigue lives of the components (i.e. when the first shear-tie post in each failed) ranged from 10.9 to 16.9 spectrum blocks and the geometric mean life was 12.9 blocks. In comparison, the predictions ranged from 8.9 to 13.6 blocks4. This compares well to the requirement of predictions within 0.5 to 2 times the true fatigue life, as implied by the penalty SF for analysis of 2.0 prescribed by some standards (see above). It should also be noted that while the participants predicted the crack growth from a 0.01 mm starting crack depth, the average size of the discontinuities that started each lead crack was approximately 0.03 mm.

4

Noting that there may be differences between the failure criteria chosen by each analyst.

Progress on the Pathway to a Virtual Fatigue Test

823

Fig. 2. A comparison of crack growth from predictions and coupon test data.

An alternate comparison, where the growth rate (per block of test loading) is plotted versus crack depth, is shown in Fig. 3. It can be seen that the rate of growth predicted by the participants was (almost always) at or above the average rate measured from the coupons. This implies that had a larger initial crack size been assumed (i.e. 0.03 mm), most of the results would have been conservative versus the average component test result. This is also evident in Fig. 2, where the gradient of most of the crack growth prediction curves appears steeper on average than the measured crack growth curves, particularly for crack depths greater than 0.1 mm. While all participants had access to the same data and all used linear elastic fracture mechanics (LEFM) based methodologies, there was considerable variation in assumptions and software codes used. Tables 2, 3 and 4 summarise the key inputs and assumptions for the three main prediction methodologies, and in particular, the geometry/beta solution (b), material model (i.e. da/dN v DK) and crack growth algorithm (e.g. AFGROW) used. Additional information regarding each methodology can be found in the references given in Tables 2, 3 and 4. Review of Predictions Versus Test Results. In reviewing Figs. 2 and 3, several general statements can be made. • DST FASTRAN Case 1 and Case 2 predictions were conservative and the Case 2 solution was very conservative with significantly higher crack growth rates versus the experimental results. • The four RUAG AFGROW solutions predicted very similar crack growth rates, indicating the differences between the geometry factor solutions had little effect.

824

B. Dixon et al.

Fig. 3. A comparison of predicted and measured crack growth rates. Table 2. Summary of prediction methodology for participant: RUAG AFGROW. RUAG AFGROW All

Prediction model details •Material model – AFGROW tabular lookup approach [10] using DST16 crack growth data [11] with spectrum range-pair counted and no retardation •417.6 MPa per max. principal stress predicted by RUAG NX NASTRAN FEM of uncracked shear-tie post Stress gradient corrected •Beta solution - AFGROW classical surface crack in a rod rod (SG ROD) geometry factor solution, with stress gradient correction as predicted along the assumed crack path using NX NASTRAN Energy Release Rate •Beta solution - Crack path perpendicular to the major principal (ERR) stress field was defined through the FEM and crack assumed to be semi-elliptical with a/ba equal to 0.5. Crack was grown in 38 small increments and stress intensity factor was calculated via the change in elastic stored energy between crack depths using NX NASTRAN Aster AUTO •Beta solution - The software Code_Aster [12] was used to perform a three-dimensional simulation of crack growth through the component. Growth modelled in 39 steps, using Code_Aster’s capability of automatic crack growth and remeshing. Crack initially with a/b equal to 1.0 Aster PLAN •Beta solution - Code_Aster used as above, but the automatic crack growth was confined to a predetermined plane a a/b defines crack depth divided by semi-width for a semi-elliptical crack. When a circular rod geometry was assumed, the crack shape was the intersection between a semi-ellipse and the circular cross-section.

Progress on the Pathway to a Virtual Fatigue Test

825

Table 3. Summary of prediction methodology for participant: DST easigro. DST easigro All

WhiteChan 16

Prediction model details •Beta solution – Murakami and Aoki’s handbook solution for a semi–elliptical crack in a bar in bending (Serbb) [13]. Fixed ratio a/b = 1.0 •Crack growth – DST cycle-by-cycle crack growth code easigro [10]. Spectrum manipulation: peak load moved to front and the sequence was rainflow counted •413.2 MPa reference stress per maximum measured strain on component Material model – the equation below defined by White and Mongru [14, 15], with parameters fitted to the material dataset described at [11] h i
 3 2 da 1 dN ¼ exp aw log ðDKe Þ  bw log ðDKe Þ þ cw logðDKe Þ  dw  exp ðK1C ð1RÞDK Þew DK where: DKe ¼ ð1RÞ fw

WhiteBarter 14 HartmanSchjive

Material model – As above, but parameters fitted to material dataset described at [16] Material model - Hartman-Schjive equation as defined below, using the parameter values given in [17, 18]:  a thr Þ da= ¼ D pðDKDK ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi dN ð1Kmax Þ=A

NASGRO Material model - NASTRAN equation [10] using the parameters from [19] Forman Material model - Forman equation [20] using parameter values given in [21]

Table 4. Summary of prediction methodology for participant: DST FASTRAN. DST FASTRAN Case 1

Prediction model details

•Beta solution - AFGROW weight function approach (WFA) function [10]. Fixed (a/b = 1.0) semi-elliptical crack. •Material model - crack growth rate properties defined in [18, 22] •Crack growth - FASTRAN [23] cycle-by-cycle as implemented within the DST developed software CGAP [24]. Each load cycle defined by a min-max paira. The FASTRAN closure option was used, along with a constraint factor (a) of 2.0 •439.9 MPa reference stress, per maximum principle stress predicted by Stress Check ® FEM Case 2 As above, except for fixed ratio a/b = 0.6 for a < 1.27 mm a Where the spectrum has been reduced to a series of successive peak and valley loads.

• DST easigro solutions varied significantly. Since the same geometry factor solution was used in all cases, these variations can be attributed to the crack growth prediction methodologies that were employed. The two solutions that used da/dN data that were based on long crack measurements (i.e. the NASGRO and Forman equations) were the poorest at predicting the rate of crack growth of the experimental data. The DST easigro White-Barter 14 also poorly predicted the rate of crack growth.

826

B. Dixon et al.

• The RUAG AFGROW, DST easigro White-Chan 16 methodology and the DST FASTRAN Case 2 were considered to predict the experimental crack growth rates well. Two prediction methodology inputs were considered key to understanding the good or poor predictions of the test results, namely, the beta solutions and material crack growth rate data. Figure 4 provides a comparison of beta factors used by analysts in this challenge and plots crack depth on a logarithmic scale. Due to the approximately exponential rate of crack growth that occurred, with this scale, the horizontal axis approximately reflects the proportion of the fatigue life influenced by given geometry factor values. Hence, since the cracks were less than 1 mm deep for approximately 2/3 of the fatigue life, the geometry factors for this range of crack depths were very important for the total fatigue life predictions. Figure 4 also includes a geometry factor curve that was derived from the measured crack growth data for the components (labelled back calculated) as described in [9]. This can be considered the best estimate of the local stress affecting the lead crack for this geometry and applied loading, and is a useful reference against which the predicted geometry factors can be assessed. The differences between the FASTRAN Case 1 and Case 2 beta solutions illustrate the importance of accurately predicting the shape of the crack when it was small. These beta solutions were exactly the same, except for the aspect ratio of the semi-elliptical crack (a/b) that was assumed when it was less than 1 mm deep: 1.0 for Case 1 and 0.6 for Case 2. As a result, Case 2 was very conservative for these crack depths, while Case 1 predicted the back calculated beta solution far more accurately. The RUAG AFGROW SG ROD solution and DST easigro Serrb (Murakami) solution both assumed a crack in a circular rod, with a/b equal to approximately 1.0 for a < 1 mm. These two solutions provided similar results for crack depths less than 1 mm, which were even more accurate than the FASTRAN Case 1 solution. Another thing these solutions had in common was that the assumed loading reflected the applied bending on the shear-tie post. The DST easigro Serrb (Murakami) solution assumed generic bending of a rod, while the RUAG AFGROW SG ROD solution used the FEMpredicted stress decay and this more accurate representation led to a more accurate solution for crack depths above 1 mm. The other RUAG AFGROW solutions with beta calculated using the ERR approach or the Code_Aster software (Aster PLAN and Aster AUTO) also provided similar results for crack depths less than 1 mm as well as more accurate results for crack depths above 1 mm. All of these solutions were based on three dimensional FEM. There was less variability in the material crack growth rate data that was used for the predictions. Those predictions based on long crack growth rate data (i.e. DST easigro NASTRAN and Forman) resulted in less accurate predictions. All other predictions used data that were based on recent DST testing and measurements of small fatigue cracks. In particular, the best predictions were reliant on the crack growth rate data model reported at [11]. Finally, it should be acknowledged that the type of cycle counting used (i.e. cycle-by-cycle, rainflow or range-pair counting) would have affected the fatigue crack growth predictions, since each converted the VA spectrum into equivalent CA cycles differently. Generally, rainflow and range-pair counting extract similar cycles and result in more conservative fatigue predictions than the

Progress on the Pathway to a Virtual Fatigue Test

827

Fig. 4. A comparison of the geometry factors (b) used by challenge participants.

cycle-by-cycle counting that was used for the DST FASTRAN solutions. However, the precise effect in the present case has not been fully investigated and this will be an area for follow-up investigation.

6 Driving Innovative Aircraft Sustainment Improved Outcomes from Certification FSFTs. For many reasons, certification FSFTs cannot faithfully represent the spectrum loads that will occur at all structural locations, on all service aircraft. Thus, accurate test interpretation and Individual Aircraft Tracking (IAT) are needed to predict the fatigue life outcomes of service aircraft from the results of the FSFT. In line with the blueprint proposed by the digital twin concept [25], accurate IAT is especially aligned with maximising the fatigue sustainment outcomes of a FSFT. An important part of both test interpretation and IAT are fatigue prediction algorithms that weigh the loads experienced by fleet aircraft against those experienced by the FSFT up to failure5. The algorithms used by IAT systems are often simple, based on cumulative damage theories and without consideration of load order effects, while test interpretation methodologies tend to reflect the more sophisticated methods employed during airframe design. DST considers the insights gained from the ASSIST airframe challenges can be used to improve the accuracy of test interpretation and IAT fatigue algorithms alike to allow more cost efficient fatigue management outcomes from certification FSFTs.

5

The definition of failure will vary depending of the design/certification philosophy used.

828

B. Dixon et al.

More Efficient FSFTs. While certification FSFTs have some unique features that are difficult to replicate via analysis (e.g. faithful representation of a service aircraft’s build quality), they tend to be costly and take many years. In particular, downtimes for inspections and repairs can extend the program duration by more than 50%. If the ASSIST program provides improved hot spot identification, this can be used to reduce the impost of such downtimes. Specifically, better hot spot identification will allow more efficient monitoring and inspection of the FSFT, including monitoring during fatigue cycling using additional strain gauges and broad-field stress monitoring techniques (e.g. TSA) at potential hot spot regions. Furthermore, fatigue cracks can be detected earlier, allowing less complicated and more fleet-representative repairs. Finally, repairs could be pre-emptively designed for potential hot spots to allow faster implementation if cracking does occur. Streamlined Certification. While FSFTs are currently required to support certification, fatigue clearance via analysis alone is often necessary for some hot spots. However, the regulator’s task of compliance finding against relevant airworthiness regulations can be complicated, because of the uncertainty associated with the accuracy of fatigue prediction methodologies when applied to real aircraft structures. It is intended that ASSIST will provide a database of challenges that demonstrate the fidelity of current fatigue prediction methodologies for various airframe structures and loading conditions. These could be used to guide compliance finding activities and provide important insights to the designer and regulator alike as to which parts of the respective methodologies are most prone to error or have the most impact on prediction fidelity.

7 Summary The Advancing Structural Simulation to drive Innovative Sustainment Technologies (ASSIST) program has been initiated by the Defence Science and Technology Group (DST) to advance improvements in fatigue prediction technologies, with the ambitious long-term goal of a virtual fatigue test. This collaborative program is based around a growing series of airframe challenges, where participants apply state-of-the-art fatigue prediction technologies on problems that are representative of real aircraft structures and loads. Since the true fatigue performance associated with each challenge is established via testing, the predictive ability of all methodologies can be accurately assessed. Furthermore, the collaborative forensic review of the predictions, using additional data such as crack growth measurements, allows the limitations of the prediction technologies to be better understood and addressed in future research. It is also expected that the growing database of fatigue predictions for the airframe challenges will be useful for establishing reliable error bands for the different prediction methodologies. Such error bands are essential for full design and certification via analysis, because they allow adequate safeguards to be put in place to avoid unexpected service cracks.

Progress on the Pathway to a Virtual Fatigue Test

829

Some of the main features of the ASSIST program were illustrated via the completed first airframe challenge, which was based on a fighter aircraft’s wing shear-tie post. Three participants used different crack growth and local stress prediction techniques to make blind fatigue life predictions for a series of components that were tested with realistic aircraft loads. These predictions were then assessed with the aid of additional finite element analyses, strain measurements and crack growth measurements. In the present case, the fatigue cracks that failed the components were smaller than 1 mm for approximately two-thirds of the fatigue life, and therefore, high quality crack growth data based on measurements of small cracks were essential for accurate predictions. Additionally, a good approximation of the fatigue crack’s shape and the stress gradient affecting the crack as it grew through the post, especially when it was small, were also important. A collaborative online space has been established for the ASSIST program and this is where the details relating to each challenge are provided, members of the ASSIST community can discuss methodology strengths/weaknesses etc. and the reviews of the challenges are posted. Two further airframe challenges have been released and can be accessed via the ASSIST online space. Members of the Aerospace community are encouraged to join the ASSIST program and participate in the airframe challenges, as we have confidence that there are many people with the skills and experience necessary to make valuable contributions to advance airframe fatigue life prediction technologies through the ASSIST program. Acknowledgements. The authors would like to thank Dr. Albert Wong for providing the motivation for the ASSIST program and the authors of [9] for allowing the reproduction of their fatigue life predictions. They would also like to thank the Royal Canadian Air Force for providing loading spectrum data for the first ASSIST challenge. Additionally, they would like to thank Dr. Simon Barter for initiating the test program that was central to the first ASSIST challenge and the other DST staff that contributed to this test program.

References 1. Damage tolerance and fatigue evaluation of structure, AC 25.571-1D [Advisory Circular] Federal Aviation Administration (2011) 2. Department of Defense Joint Service Specification Guide: Aircraft Structures. JSSG-2006, U.S. Department of Defense (1998) 3. DEF STAN 00-970 Design and Airworthiness for Service Aircraft, Issue 5. UK. Ministry of Defence (2007) 4. Department of Defense Standard Practice - Aircraft Structural Integrity Program (ASIP) MIL-STD-1530D (USAF), Wright Patterson AFB Ohio, Department of Defense (2016) 5. Ball, D.L., Gross, P.C., Burt, R.J.: F-35 full scale durability modeling and test. Adv. Mater. Res. 891–892, 693–701 (2014) 6. Wong, A.K.: Blueprint TITANS: a roadmap towards the virtual fatigue test through a collaborative international effort. In: 29th ICAF Symposium, Nagoya, Japan: 7–9 June 2017, VTT Information Service (2017)

830

B. Dixon et al.

7. Fatigue, Fail-Safe, and Damage Tolerance Evaluation of Metallic Structure for Normal, Utility, Acrobatic, and Commuter Category Airplanes. AC 23-13A, [Advisory Circular] Federal Aviation Administration (2005) 8. Irving, P.E., Lin, J., Bristow, J.W.: Damage tolerance in helicopters. Report on the Round Robin Challenge. In: American Helicopter Society 59th Annual Forum, Phoenix (Arizona), USA (2003) 9. Main, B., Evans, R., Walker, K., Yu, X., Molent, L.: Lessons from a fatigue prediction challenge for an aircraft wing shear post. Int. J. Fatigue 123, 53–65 (2019) 10. Harter, J.A.: AFGROW Users Guide and Technical Manual, Version 5.02.01.18. Centreville OH, USA, LexTech (2014) 11. Burchill, M., Barter, S., Chan, L.H.: Improving fatigue life predictions with a crack growth rate material model based on small crack growth & legacy data. In: 17th Australian International Aerospace Congress (AIAC17), Melbourne, Australia (2017) 12. Abbas, M.: Introduction to code_aster. [Manual] Updated 28/9/2018 [Accessed 13/2/2019]; Revision 992db8bc5b15. Available from (2018). https://www.code-aster.org/V2/doc/default/ en/man_u/u1/u1.02.00.pdf 13. Murakami, Y., Aoki, S.: Stress Intensity Handbook. Pergamon, Oxford (1987) 14. White, P., Mongru, D., Molent, L.: A crack growth based individual aircraft monitoring method utilizing a damage metric. Struct. Health Monit. 17(5), 1178–1191 (2018) 15. White, P.: A guide to the program easigro for generating optimised fatigue crack growth models. DST-Group-TR-3583 [Technical report] Defence Science and Technology Group (2019) 16. Burchill, M., Barter, S., Amsterdam, E.: Improved predictions for combat aicraft fatigue life from a novel testing program. In: 15th Australian International Aerospace Congress (AIAC15), Melbourne, Australia (2014) 17. Jones, R., Molent, L., Barter, S.A.: Calculating crack growth from small discontinuities in 7050-T7451 under combat aircraft spectra. Int. J. Fatigue 55, 178–182 (2013) 18. Walker, K.F., Barter, S.A.: The critical importance of correctly characterising fatigue crack growth rates in the threshold regime. In: 26th Symposium of the International Committee on Aeronautical Fatigue, Montreal, Canada: 1–3 June 2011 (2011) 19. Forman, R.G., Shivakumar, V., Cardinal, J.W., Williams, L.C., McKeighan, P.C.: Fatigue Crack Database For Damage Tolerance Analysis. DOT/FAA/AR-05/15 [Technical report] Office of Aviation Research, Washington, D.C., Federal Aviation Authority (2005) 20. Forman, R., Hearney, V., Engle, R.: Numerical analysis of crack propagation in cyclicloaded structures. J. Basic Eng. Trans. of ASME (1967) 21. Schwarmann, L.: Material Data of High Strength Alumium Alloys for Durability Evaluation of Structures. Aluminium-Verlag, Dusseldorf, Germany (1985) 22. Walker, K.F., Wang, C.H., Newman, J.C.: Closure measurement and analysis for small cracks from natural discontinuities in an aluminium alloy. Int. J. Fatigue 82, 256–262 (2016) 23. Newman, J.C.J.: FASTRAN A Fatigue Crack Growth LIfe -Prediction Code Based on the Crack-Closure Concept, Version 5.4. Fatigue & Fracture Associates LLC (2013) 24. Hu, W., Wallbrink, C.: CGAP capabilities and application in aircraft structural life. In: 15th Australian International Aerospace Conference (AIAC15), Melbourne, Australia (2013) 25. Tuegel, E.J., Ingraffea, A.R., Eason, G.E., Spottswood, M.S.: Reengineering aircraft structural life prediction using a digital twin. Int. J. Aerospace Eng. 2011, 14 (2011)

Testing Approach for Over Wing Doors Using Curved Fuselage Panel Testing Technology Mirko Sachse1(&), Matthias Götze1, Silvio Nebel1, Sven Berssin2, and Christian Göpel2 1

IMA Materialforschung Und Anwendungstechnik GmbH, Dresden, Germany [email protected] 2 Airbus Operations GmbH, Hamburg, Germany

Abstract. The A321neo ACF (Airbus Cabin Flex configuration) contains newly designed Overwing Doors (OWD) that provide an automatic opening function for the case of evacuation. Due to the significant structural differences between the previous “hatch” design used on A319 and A320 aircraft and the new OWD design it has been decided to test the OWD’s and the surrounding fuselage structure to demonstrate the Fatigue and Damage Tolerance capabilities of the structure. This demonstration should be done by means of a fatigue and damage tolerance test. An appropriate test set-up had to be selected. Two generally different approached were investigated. This selection process led to the decision to follow the curved fuselage panel test method. Curved fuselage panel testing was developed in order to test undisturbed, regular panels. During the last years, this method was improved to allow testing of panels with major non-regularities, for instance door cut outs, floor beams or similar features. Keywords: Test load generation


Curved fuselage panel test

1 Introduction The A321neo ACF (Airbus Cabin Flex configuration) is a new derivative of the successful Airbus Single Aisle Family. Main differences compared to the A321neo aircraft are the two additional Overwing Doors (OWD), the re-positioning of the door 3 (moved four frame bays backward) and the deletion of the Pax door 2. These changes provide the airlines an improved flexibility in terms of seat arrangement and increase the efficiency of the aircraft. The latest EASA certification requirements ask for an automatic opening function for the two additional Overwing Doors for the case of evacuation. As such function is not provided by the hatches previously used on A319 and A320 aircraft, the doors and the surrounding structure had to be re-designed to accommodate the additional requirements.

© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 831–837, 2020. https://doi.org/10.1007/978-3-030-21503-3_65

832

M. Sachse et al.

2 Test Scenario Selection The definition of the test arrangement was significantly challenged by the fact, that the area of interest is located directly above the Centre Wing Box. In this area, the load flows are influenced by the stiff Centre Wing Box and the wing loads introduced through the front and the rear spar into the fuselage shell. These loads lead to deflections of the door surrounding structure that need to be represented through a simplified test set-up (Fig. 1) [3].

Fig. 1. A321 ACF OWD test area

In order to find the best possible test solution under Time, Cost and Quality aspects several potential test set-ups have been investigated by means of Detailed Finite Element (DFEM) Analysis. Two main scenario have been analyzed: • A so-called ¾ Barrel Test set-up, where a complete upper half of the fuselage is fixed on the lower edge to a stiff test-rig that is supposed to represent the Centre Wing Box (CWB) • A Panel test using enhanced facilities developed in frame of the MAAXIMUS research project [2] (Fig. 2).

Fig. 2. A321 investigated test concepts

Testing Approach for Over Wing Doors

833

The stress distributions for the two proposed test set-ups have been analyzed using Finite Element Models and compared to the stress distributions determined for the aircraft itself. This analysis revealed that the enhanced panel test set-up allows a better representation of the aircraft stress distribution as the ¾ Barrel concept does (Fig. 3).

¾ Barrel

Aircraft Model

Enhanced Curved Panel Test

Fig. 3. DFEM studies to support selection of test scenario

Consequently, the enhanced curved panel test set-up has been chosen to test the structure.

3 Development of Rig Adaptation The available test set-up had to be modified in order to incorporate the required panel. Modifications were necessary in terms of panel size and panel loading. The starting point was the test rig configuration developed in the MAAXIMUS project [1] and [2]. The test rig consists of • A fixed load frame, • A fixed bulkhead for panel installation at one side, • Moveable bulkhead for panel loading (3 DOFs displacement, 3 DOFs rotation) powered by six hydraulic actuators offering movement and loading capabilities of a hexapod, • A pneumatic system for pressure load application, • An internal hydraulic loading system, to ensure correct loading and boundary conditions in the panel (e.g. frame and skin spreaders). Adaptations to the frame load mechanism were necessary to account for the specific location of the panel inside the fuselage. Frame loading in such test concept is done by: • Restricting the rotation of the frame end or • Applying an opposite acting moment of the same value to both ends of the frame • Applying different moments to either side of the frame. While for the door panel tested previously only the door frames were loaded with different moments on either side, this panel required this loading mode for every frame in the area of interest (Fig. 4).

834

M. Sachse et al.

Fig. 4. Curved fuselage panel test rig

The development of the loading system, especially defining the required load levels for each loading station, was supported by a load determination process developed in the MAAXIMUS project [1].

4 Determination of Test Loads The determination of test loads is based on • Stress values for certain points (pilot points) of the structure, based on an aircraft FE model • Stress values for pilot points for each individual load component of the test rig (unitary load case) • Least square algorithm to determine the loads for each load component of the test rig (test load vector). For the area of interest pilot points are defined. These are locations, where the stresses or strains are extracted from the panel aircraft FE model as target values for the load determination process for a particular load case. Furthermore, a test rig FE model is built, which contains as an optimum the same model as the panel aircraft FE model in the area of interest. By loading the model with each individual load component that the test rig can apply, a set of stress or strain values for the same pilot points is derived. These are the unitary load cases.

Testing Approach for Over Wing Doors

835

For each load case an approximation algorithm is performed. It takes into account the loading capability of each load component. Therefore, local overloading is prevented in this stage already (Fig. 5).

Fig. 5. Test load determination process

This least square algorithm delivers a set of loads, the test load vector, for the load case in question. This load case can be used to predict strain values for the test by applying this load vector to the test rig FE model. Furthermore, these loads, when final, are used as command values for the actual test itself (Fig. 6).

Aircraft Panel FE Model

Test Rig FE model

Fig. 6. Comparison between target and approximation

The figure above shows the comparison of stresses between the aircraft panel model and the test rig model with loads the corresponding test load vector applied. The results shows a good match in the area of interest, based on an adequate pilot point selection. The deviations in the outer area are larger and mostly influenced by the boundary

836

M. Sachse et al.

conditions of the test rig, especially in the corner region. Such results are used to reinforce the panel locally. This leads to an iterative process for load determination.

5 Results Comparing Calculation and Test The test panel was equipped with about 400 strain gauge channels. These are used for two reasons: 1. Assess the quality of the achieved stress field in the area of interest of the panel for several significant load cases. 2. Make sure, that the stresses close to the boundaries do not cause failure of the panel during fatigue testing. The load cases for the comparison are taken from every flight stage. So, there are load cases without internal pressurization of the pressure box, as well as load cases with different pressure levels applied. As in opposition to the in plane mechanical loads the pressurization needs to meet more complex boundary conditions, this approach was chosen. So, a possible influence of pressurization on the quality of the test could be determined. The following diagram shows a comparison of the strain values of a single door corner strain gauge. Each data point represents one load case. The correlation between the values of prediction and measurement is quite good, given the complexity of the rig and the limited number of available control channels (Fig. 7).

Fig. 7. Comparison between target and measurement for a single strain gauge

Testing Approach for Over Wing Doors

837

6 Summary The test method developed by IMA Dresden for complex curved fuselage panel tests was successfully adapted to a new panel geometry and testing task. Specific boundary conditions like the location of the panel close to the center wing box were considered by the load determination process. Furthermore, the used method to determine test loads was successfully extended to a complete fatigue spectrum. One of the next steps will include integration of this test method into a virtual testing strategy for future fuselage designs and test rig concepts in general.

References 1. Werner, S., Goetze, M., Sachse, M., Stankovic, Z., Howes, L.: Comparison of numerical and experimental results for the door surround structure of a pressurized Fuselage. In: 29th ICAF Proceedings, Nagoya, Japan, June (2017) 2. Sachse, M., Nebel, S., Werner, S., Semsch, M.: State of the art curved fuselage panel testing. In: 29th ICAF Proceedings, Nagoya, Japan, June (2017) 3. Berssin, S.: Panel Test for new developed Airbus A321 ACF Overwing Door & Surrounding structure, XIX ICMFM, Porto, Portugal

Very High-Cycle Fatigue Characteristics of Cross-Ply CFRP Laminates in Transverse Crack Initiation Atsushi Hosoi1,2,3(&), Takuro Suzuki4, Kensuke Kosugi4, Takeru Atsumi4, Yoshinobu Shimamura5, Terumasa Tsuda6, and Hiroyuki Kawada1,2,3 1

Department of Applied Mechanics and Aerospace Engineering, Waseda University, 3-4-1, Okubo, Shinjuku-Ku, 169-8555 Tokyo, Japan [email protected] 2 Department of Materials Science, Waseda University, 3-4-1, Okubo, Shinjuku-Ku, 169-8555 Tokyo, Japan 3 Kagami Memorial Research Institute for Materials Science and Technology, Waseda University, 2-8-26, Nishiwaseda, Shinjuku-Ku, 169-0051 Tokyo, Japan 4 Department of Applied Mechanics, Waseda University, 3-4-1, Okubo, Shinjuku-Ku, 169-8555 Tokyo, Japan 5 Department of Mechanical Engineering, Shizuoka University, 3-5-1, Johoku, Naka-Ku, Hamamatsu-Shi, 432-8561 Sizuoka, Japan 6 Composite Materials Research Laboratories, Toray Industries, Inc, 1515, Tsutsui, Masaki-Cho, Iyo-Gun, Ehime 791-3193, Japan

Abstract. Fan blades are subjected to very high-cycle loadings during the design life, so it is essential to evaluate the giga-cycle fatigue characteristics of carbon fiber reinforced plastic (CFRP) laminates. In this study, the transverse crack initiation of the cross-ply CFRP laminates in very high-cycle fatigue region was evaluated using an ultrasonic fatigue testing machine. The fatigue tests were conducted at the frequency of f = 20 kHz and the stress ratio of R = −1. In order to suppress temperature rise of the specimen, the intermittent operation with the loading time of 200 ms and the dwelling time of 2000 ms was adopted. The fatigue life data to transverse crack initiation in very highcycle fatigue region was compared with the data of the fatigue test which was conducted at the frequency of f = 5 Hz and the stress ratios of R = 0.1 and −1 using a hydraulic control fatigue test machine. It was evaluated considering the influences of the stress ratio and the thermal residual stress by using the modified Walker model. The fatigue life to the transverse crack initiation of the cross-ply CFRP laminates in the very high-cycle region exceeding 108 cycles was on the extension of the test data in the low cycle region. Keywords: Giga-cycle fatigue


CFRP
Transverse crack
Ultrasonic

© Springer Nature Switzerland AG 2020 A. Niepokolczycki and J. Komorowski (Eds.): ICAF 2019, LNME, pp. 838–846, 2020. https://doi.org/10.1007/978-3-030-21503-3_66

Very High-Cycle Fatigue Characteristics

839

1 Introduction Carbon fiber reinforced plastic (CFRP) laminates are adopted not only for structural members of aircraft but also for fan blades of turbofan engines. The required level for these materials is high, and those are subjected to fatigue loadings of the giga-cycle order. However, when a fatigue test exceeding 109 cycles is conducted at the conventional frequency of f = 5 Hz, unrealistic time is required. The time required to reach the 109 cycles is 2315 days at the frequency of 5 Hz and 116 days at the frequency of 100 Hz. On the other hand, when the fatigue test is conducted at the frequency of f = 20 kHz, only 14 h are required to reach the 109 cycles. To obtain giga-cycle fatigue characteristics of CFRP laminates, it is required to conduct the fatigue test under high frequency. The giga-cycle fatigue characteristics of CFRP laminates can be obtained in short time by using an ultrasonic fatigue test machine. Authors (Hosoi et al. 2007, 2010) conducted fatigue tests at frequency of 100 Hz and evaluated transverse crack and delamination growth of quasi-isotropic CFRP laminates on the order of 108 cycles. We showed that the delamination growth behavior in the high-cycle fatigue region was different from the conventional one. Backe et al. (2015) developed ultrasonic fatigue testing facility for cyclic 3-point bending, and evaluated the fatigue life to delamination of a twill fabric CFRP laminates. Flore et al. (2017) developed the ultrasonic fatigue test method in the axial direction using quasiunidirectional GFRP laminates. The fatigue test was conducted at stress ratio of R = 0.1 until the specimen fractured. They showed that the fatigue life obtained by the ultrasonic fatigue test was on the extension of that obtained by the conventional servohydraulic fatigue test. Shimamura et al. (2018) succeeded in performing an axial tensile-compressive fatigue test of dumbbell shaped quasi-isotropic CFRP laminates by using an ultrasonic fatigue test machine. Since the authors have evaluated the high cycle fatigue characteristics at high frequencies, the experimental results evaluating the giga-cycle fatigue characteristics of CFRP laminates are very few. Also, as far as the authors know, there are no studies that evaluate giga-cycle fatigue characteristics for the transverse crack initiation, which is the initial damage of the CFRP laminates. Therefore, in this study, the transverse crack initiation of cross-ply CFRP laminates in the giga-cycle fatigue was evaluated using ultrasonic fatigue facility.

2 Experiments Specimens. CFRP laminates were formed with an interlaminar toughened prepreg (T800S/3900-2B) using an autoclave. The fiber volume fraction of the prepreg is Vf = 56%, the matrix resin is epoxy, and the prepreg thickness is 0.188 mm. The stacking sequence of the laminates was [0/904]S and the curing temperature was 180 °C. In order to perform the ultrasonic fatigue test, the specimen was designed to resonate at around the frequency of f = 20 kHz. The specimen including the tabs was designed to resonate in the longitudinal primary mode by using the finite element analysis (FEA) of the commercial code COMSOL Multiphysics. The analytical result is shown in Fig. 1. From the results of FEA, the specimen dimensions were determined as 70 mm long

840

A. Hosoi et al.

10 mm wide 1.88 mm thick. The metal tab was made of aluminum alloy A2017. Figure 2 shows the upper metal tab dimensions. The lower metal tab has a shape without the external thread. The CFRP specimen and tabs were bonded with an epoxy adhesive. For comparison, a fatigue test was carried out with stress ratios of R = 0.1 and −1 using a servo-hydraulic fatigue test machine. Figure 3 shows the specimen geometries. The gauge length of the specimen of R = −1 was set to 10 mm for the purpose of preventing bucking. The edge surfaces of the specimen were polished with emery papers and finished by buffing using a diamond powder having a particle diameter of 1 lm.

Fig. 1. Finite element analysis of cross-ply CFRP laminates resonating in the longitudinal primary mode.

Fatigue Test Conditions. Ultrasonic fatigue test was conducted to evaluate the fatigue life to the transverse crack initiation in giga-cycle region. By applying sinusoidal vibration of frequency of 20 kHz from the upper end of the specimen, the fatigue tests were carried out at strain ratio R = −1. The ultrasonic fatigue testing machine does not have a load cell. For this reason, the displacement amplitude of the metal tab attached to the bottom of the specimen was measured using an eddy current type gap sensor, and the stress and strain applied to the CFRP laminate was calculated by FEA as shown in Fig. 1. In addition to air cooling, intermittent operation was introduced in order to reduce the heat generation of the CFRP laminate by high frequency vibration. In this study, the loading and dwelling time were 200 ms and 2000 ms, respectively, and the effective frequency was about 1818 Hz. The temperature of the surface of the CFRP laminate under the fatigue test was measured by infrared thermography. In the fatigue test using the servo-hydraulic fatigue test machine, it was conducted under the conditions of the frequency of f = 5 Hz and stress ratio of R = 0.1 or −1. The loading stress level was set rmax /rti = 0.6–1.4 at the stress ratio of R = 0.1. Here, rmax and rti show the maximum stress applied in the specimen and the stress where the transverse crack is initiated under the static tensile loading. At the stress ratio of R = −1, the fatigue test was conducted within the range where the stress applied in the 90° layers is positive considering the residual thermal stress.

Very High-Cycle Fatigue Characteristics

841

Fig. 2. Schematic image of metal tab (upper side).

CFRP

35

GFRP

15

CFRP

40

10

GFRP

10

1.88

1.5

1.88

1.5

120 230

80

(b) R=0.1

(a) R= –1

Fig. 3. Specimen geometries for servo-hydraulic fatigue test.

3 Evaluation Model The authors (Hosoi and Kawada 2018) have proposed a model that can equivalently evaluate the fatigue life to transverse crack initiation in CFRP laminates with different laminate configuration and subjected to cyclic loadings at different stress ratios. The evaluation formula is shown in Eq. (1). ð90Þ1c

Ni ¼ g

rmax

ð90Þ

rti

ð90Þc

ra

!k ;

ð1Þ

where, the superscript (90) indicates the stress applied in the 90° layers in laminates, and the subscript, a, indicates the stress amplitude. η, k and c are material constants. This is based on the model proposed by Walker (1970). For brittle materials, the

842

A. Hosoi et al.

fracture is dominated by the maximum stress, so the value of c approaches 0. For ductile materials, the fracture is dominated by the stress amplitude, so the value of c approaches 1. In our previous study, c = 0.5 is appropriate for the CFRP laminates with epoxy resin as a matrix material.

4 Results and Discussions Displacement Amplitude and Temperature Measurement of Specimens. Figure 4 shows the measurement results of the displacement amplitude of the specimen at 2.4  106 cycles. It was confirmed that the CFRP laminate resonates at a frequency of 20 kHz by the test method constructed in this study. In addition, it was confirmed that there was no difference in the displacement amplitude at the start of the test and at the 107 cycles. It was thought that the fatigue test could be carried out under the condition where the mechanical stress was constant.

Fig. 4. Displacement amplitude of the bottom of the specimen during the ultrasonic fatigue test.

Figure 5 shows the surface temperature change at the center of the specimen during fatigue tests. It was confirmed that the surface temperature of the CFRP laminate increased due to amplitude of the specimen and became a steady state. Further, as the maximum initial strain at the central part of the specimen, e0, was larger, the specimen temperature was higher in the steady state. However, it was sufficiently lower than the glass transition temperature of the specimen. Considering the strain distribution, it is expected that the temperature rise will be the maximum in the central part of the specimen. However, as shown in Fig. 6, the temperature rise took a maximum value at the lower part than the center part of the specimen. This is due to the forced air cooling from the outside.

Very High-Cycle Fatigue Characteristics

843

Fig. 5. Surface temperature change of the center of the CFRP laminates during the ultrasonic fatigue test.

Specimen

70ºC

50 mm

20ºC

Fig. 6. Distribution of surface temperature of CFRP laminates during the ultrasonic fatigue test (emax = 0.204%).

Damage Observation. Figure 7 shows the photographs of the transverse crack occurred in the specimen subjected to the ultrasonic fatigue test observed with optical microscopy and soft X-ray photography. In this study, the transverse crack as observed in the conventional servo-hydraulic fatigue test was observed as in Fig. 7 (a). It was confirmed that the transverse crack occurred at the edge surface of the specimen passed through to the width direction from Fig. 7 (b). As can be seen from Fig. 7 (b), the transverse crack occurred in the CFRP laminate was located slightly lower than the center of the specimen. This is thought to be caused by the influences of the initial defect, position aberration of the metal tab and the mixture of the bending modes.

844

A. Hosoi et al.

Transverse crack

Loading direction

5 mm

(b)

500 µm

(a) Fig. 7. Observation of transverse crack occurred in the specimen subjected to the ultrasonic fatigue test under the test condition of emax = 0.152%, R = − 1 and N = 4.3  108 cycles: (a) optical microscopy; (b) soft X-ray photography.

Evaluation of Fatigue Life to Transverse Crack Initiation in the Giga-Cycle Region. Figure 8 shows the fatigue life to transverse crack initiation of the specimen evaluated using the ultrasonic fatigue test machine and the servo hydraulic fatigue test machine. The vertical axis is shown using initial maximum strain, emax. In the ultrasonic fatigue test, the strain applied in the specimen is distributed in the longitudinal direction because the specimen resonates in the first-order longitudinal vibration mode. In Fig. 8, the vertical axis emax represents the strain at the center of the specimen at the start of the fatigue test. Since the servo hydraulic fatigue tests were conducted by load control, the strain of the specimen was calculated with a mixture law. As a result of the fatigue tests, it was confirmed that the fatigue life to transverse crack initiation became longer as the initial maximum strain decreased in the giga-cycle region. In addition, the fatigue strength at R = −1 was decreased compared with that at R = 0.1. The fatigue characteristics affected by residual thermal stress and stress or strain ratio. Therefore, the fatigue life to the transverse crack initiation was evaluated taking these effects into

Fig. 8. Fatigue life to transverse crack initiation under fatigue loading.

Very High-Cycle Fatigue Characteristics

845

Fig. 9. Relationship between the normalized stress and the fatigue life to transverse crack initiation considering the effects of residual thermal stress and stress ratio.

account using the Eq. (1). The result is shown in Fig. 9. The solid line shows the fitting curve obtained from the Eq. (1). Here, the material constants were η = 3.22  10−3, k = −20.2, and c = 0.5, respectively. It was found that the experimental results by the servo hydraulic fatigue test and the ultrasonic fatigue test exist on the same theoretical curve. Moreover, in the strain range of this study, the clear fatigue limit with respect to transverse crack initiation was not observed.

5 Conclusions The giga-cycle fatigue characteristics in transverse crack initiation of interlaminar toughened cross-ply CFRP laminates were evaluated in this study. The experimental plots of the transverse crack initiation in the giga-cycle fatigue region obtained by ultrasonic fatigue test at frequency of 20 kHz were on the same theoretical curve as those in low cycle fatigue region obtained by servo hydraulic fatigue test at frequency of 5 Hz. In addition, no clear fatigue limit in transverse crack initiation was observed within the test conditions of this study.

References Backe, D., Balle, F., Eifler, D.: Fatigue testing of CFRP in the Very High Cycle Fatigue (VHCF) regime at ultrasonic frequencies. Compos. Sci. Technol. 106(16), 93–99 (2015) Flore, D., Wegener, K., Mayer, H., Karr, U., Oetting, C.C.: Investigation of the high and very high cycle fatigue behaviour of continuous fibre reinforced plastics by conventional and ultrasonic fatigue testing. Compos. Sci Technol. 141(22), 130–136 (2017)

846

A. Hosoi et al.

Hosoi, A., Arao, Y., Karasawa, H., Kawada, H.: High-cycle fatigue characteristics of quasiisotropic CFRP laminates. Adv. Compos. Mater. 16(2), 151–166 (2007) Hosoi, A., Sato, N., Kusumoto, Y., Fujiwara, K., Kawada, H.: High-cycle fatigue characteristics of quasi-isotropic CFRP laminates (Initiation and propagation of delamination considering the interaction with transverse cracks). Int. J. Fatigue 32(1), 29–36 (2010) Hosoi, A., Kawada, H.: Fatigue life prediction for transverse crack initiation of CFRP cross-ply and quasi-isotropic laminates. Materials 11(7), 1182-1-16 (2018) Shimamura, Y., Hayashi, T., Tohgo, K., Fujii, T.: Very high cycle axial fatigue testing of CFRP laminates by using ultrasonic fatigue testing machine. In: Proc., the 7th Int. Conf. on the Fatigue of Composites, Vicenza Italy (2018) Walker, K.: The effect of stress ratio during crack propagation and fatigue for 2024-T3 and 7075T6 aluminum. In: International, A.S.T.M. (ed.) Effects of Environment and Complex Load History on Fatigue Life; ASTM STP 462, pp. 1–14. West Conshohocken, PA, USA (1970)

Application of Optical Fiber-Based Strain Sensing for the Full-Scale Static and Fatigue Tests of Aircraft Structure U. Ben-Simon1, S. Shoham1, R. Davidi2, N. Goldstein1, I. Kressel1(&), and M. Tur2 1

IAI Engineering Division, Ben-Gurion International Airport, Tel-Aviv, Israel [email protected] 2 School of Electrical Engineering, Tel-Aviv University, Tel-Aviv, Israel

Abstract. An optical fiber-based Rayleigh backscattering distributed strain sensing system was adopted as the main structural integrity monitoring tool for airframe Full-Scale fatigue and ultimate tests. The strain signature along all major structural elements, as measured by the optical fibers, at each loading step was recorded and analyzed in real time. A specially developed human interface enabled easy tracking of emerging damage-related non-linear phenomena. This sensing concept reduces the need for adding hundreds of electrical strain gauges and eliminates intermediate conventional structural inspections during test, all leading to reducing test duration and cost. Keywords: Distributed sensing


Full-Scale tests
Optical fiber

1 Introduction This paper presents the application of Rayleigh backscattering [1–6] distributed fiber optic strain sensing to both fatigue and ultimate tests of aircraft structure. Such a system, having a spatial resolution of