Download as pdf or txt
Download as pdf or txt
You are on page 1of 70

Resilience of civil infrastructure

systems developments in testing


sensing monitoring and control First
Edition. Edition (Structural Engineer)
Zhishen Wu (Editor)
Visit to download the full and correct content document:
https://ebookmeta.com/product/resilience-of-civil-infrastructure-systems-development
s-in-testing-sensing-monitoring-and-control-first-edition-edition-structural-engineer-zhi
shen-wu-editor/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

29 Single and Nigerian INCOMPLETE First Edition


Naijasinglegirl

https://ebookmeta.com/product/29-single-and-nigerian-incomplete-
first-edition-naijasinglegirl/

Internet of Things and Access Control Sensing


Monitoring and Controlling Access in IoT Enabled
Healthcare Systems 1st Edition Shantanu Pal

https://ebookmeta.com/product/internet-of-things-and-access-
control-sensing-monitoring-and-controlling-access-in-iot-enabled-
healthcare-systems-1st-edition-shantanu-pal/

Infrastructure Sustainability Through New Developments


in Material Design Construction Maintenance and Testing
of Pavements Proceedings of the 6th GeoChina
International Conference on Civil Transportation
Infrastructures 1st Edition Anand Tapase
https://ebookmeta.com/product/infrastructure-sustainability-
through-new-developments-in-material-design-construction-
maintenance-and-testing-of-pavements-proceedings-of-the-6th-
geochina-international-conference-on-civil-transporta/

Reliability-Based Analysis and Design of Structures and


Infrastructure (Resilience and Sustainability in Civil,
Mechanical, Aerospace and Manufacturing Engineering
Systems) 1st Edition Ehsan Noroozinejad Farsangi
(Editor)
https://ebookmeta.com/product/reliability-based-analysis-and-
design-of-structures-and-infrastructure-resilience-and-
sustainability-in-civil-mechanical-aerospace-and-manufacturing-
Security of Cyber Physical Systems State Estimation and
Control 1st Edition Chengwei Wu

https://ebookmeta.com/product/security-of-cyber-physical-systems-
state-estimation-and-control-1st-edition-chengwei-wu/

Security and Resilience in Cyber Physical Systems


Detection Estimation and Control Masoud Abbaszadeh

https://ebookmeta.com/product/security-and-resilience-in-cyber-
physical-systems-detection-estimation-and-control-masoud-
abbaszadeh/

Analysis of Legal Argumentation Documents A


Computational Argumentation Approach Translational
Systems Sciences 29 Hayato Hirata

https://ebookmeta.com/product/analysis-of-legal-argumentation-
documents-a-computational-argumentation-approach-translational-
systems-sciences-29-hayato-hirata/

Fortschritte der Physik Progress of Physics Band 29


Heft 8

https://ebookmeta.com/product/fortschritte-der-physik-progress-
of-physics-band-29-heft-8/

Fortschritte der Physik Progress of Physics Band 29


Heft 4

https://ebookmeta.com/product/fortschritte-der-physik-progress-
of-physics-band-29-heft-4/
Experimental Vibration Analysis
for Civil Structures
Testing, Sensing, Monitoring, and Control
Taylor and Francis Series in Resilience and
Sustainability in Civil, Mechanical, Aerospace and
Manufacturing Engineering Systems
Series Editor:
Mohammad Noori
Cal Poly San Luis Obispo

PUBLISHED TITLES
Resilience of Critical Infrastructure Systems
Emerging Developments and Future Challenges
Zhishen Wu, Xilin Lu, and Mohammad Noori
Experimental Vibration Analysis for Civil Structures
Testing, Sensing, Monitoring, and Control
Jian Zhang, Zhishen Wu, Mohammad Noori, and Yong Li

For more information about this series, please visit: www.routledge.com


Experimental Vibration Analysis
for Civil Structures
Testing, Sensing, Monitoring, and Control

Edited by
Jian Zhang, Zhishen Wu, Mohammad Noori, and Yong Li
MATLAB ® is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks does not warrant the
accuracy of the text or exercises in this book. This book’s use or discussion of MATLAB ® software or related products
does not constitute endorsement or sponsorship by The MathWorks of a particular pedagogical approach or particular
use of the MATLAB ® software

First edition published 2021


by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742

and by CRC Press


2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

© 2021 Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, LLC

Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers
have attempted to trace the copyright holders of all material reproduced in this publication and apologize to
copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been
acknowledged please write and let us know so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or
utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including
photocopying, microfilming, and recording, or in any information storage or retrieval system, without written
permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.com or contact the
Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that
are not available on CCC please contact mpkbookspermissions@tandf.co.uk

Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.

ISBN: 978-0-367-54746-2 (hbk)


ISBN: 978-1-003-09056-4 (ebk)

Typeset in Times
by codeMantra
Contents
Preface............................................................................................................................................... xi
Editors ............................................................................................................................................ xiii
Contributors ..................................................................................................................................... xv

Chapter 1 Wavelet-Based Damage-Sensitive Features Extraction................................................ 1


A. I. Silik, Mohammad Noori, and W. A. Altabey

Chapter 2 Deep Learning for Automated Damage Detection: A Novel Algorithm in


CNN Family for Faster and Accurate Damage Identification.................................... 13
Wael A. Altabey and Mohammad Noori

Chapter 3 Seismic Protection of Cultural Relics Using Three-Dimensional


Base-Isolation System ................................................................................................ 27
Bai Wen and Dai Junwu

Chapter 4 Combined Actuator-Shake Table Test with Optimized Input Energy ....................... 33
Farhad Behnamfar and Mohammadreza Najar

Chapter 5 Design Spectra for Structures Subjected to Passing Underground Trains ................. 43
F. Behnamfar, A. A. Yazdani, and A. Bakhtiar

Chapter 6 Frequency-Domain Fast Maximum Likelihood Estimation of Complex Modes....... 53


Binbin Li and Yan-Long Xie

Chapter 7 A Full Version of Vision-Based Structural Identification .......................................... 63


Chuan-Zhi Dong and F. Necati Catbas

Chapter 8 Damage Recognition of Wood Beam Based on Curvature Modal Technology ........ 71
Chunyu Qian, Baozhuang Zhang, Junlong Zhu, and Qifang Xie

Chapter 9 Validation of Proposed SHM Model Based on Inverse Dynamic Approach


with Limited Noisy Dynamic Responses by Experimental Study............................. 89
D. Bandyopadhyay and J. S. Ali

Chapter 10 Grouting Compactness Assessment in Post-Tensioning Tendon Ducts Using


Piezoceramic Transducers and Wavelet Packet Analysis........................................... 99
X. T. Sun, D. Li, and W. X. Ren

v
vi Contents

Chapter 11 The Analysis of the Temperature Effect on Frequencies of a Footbridge ............... 107
De-hui Tang, Jun Teng, and Wei-hua Hu

Chapter 12 Numerical Simulation of Precast Concrete Structure with Cast-In-Situ


Monolithic Joint........................................................................................................ 115
Du Yongfeng, Chi Peihong, and Fang Dengjia

Chapter 13 Simulation Analysis of a Bridge with a Nonlinear Tuned Mass Damper


Using Incremental Harmonic Balance Method........................................................ 123
Feng Gu and Chiu Jen Ku

Chapter 14 Analysis on the Internal Explosion Effects of Single-Layer Spherical


Reticulated Shell....................................................................................................... 135
Shiqi Fu and Xuanneng Gao

Chapter 15 Damage Features from Direct Modal Strain Measurements ................................... 145
G. De Roeck, E. Reynders, and D. Anastasopoulos

Chapter 16 Displacement Estimation by Multi-Rate Data Fusion of Strain and


Acceleration Data ..................................................................................................... 159
K. Gao, S. Weng, H.P. Zhu, Y. Xia, X.S. Yu, G.Z. Qu, P.C. Yin, and J.F. Yan

Chapter 17 Concrete Crack Image Recognition Based on DBSCAN and KPCA ...................... 169
Mingxin Gao and Yang Liu

Chapter 18 Pedestrian Induced Vibration of Slab....................................................................... 175


Yan-an Gao

Chapter 19 Piezoelectric Admittance-Based Damage Detection via Data Compression


and Reconstruction ................................................................................................... 183
H. Li, D. Ai, H. Zhu, H. Ge, and H. Luo

Chapter 20 Several Damage Indices Based on Transmissibility for Application in


Structural Damage Detection ................................................................................... 193
Jianping Han, Hongyu Zhang, and Zhiqiang Hou

Chapter 21 Substructuring-Based Damage Assessment of a Steel Railway Bridge


Using Operational Modal Data ................................................................................ 203
L. He, E. Reynders, V. Zabel, G. C. Marano, B. Briseghella, and G. De Roeck
Contents vii

Chapter 22 Time-Frequency Features of Continuous Metro Bridge under Various


Excitations ........................................................................................................... 211
J. Yang, X. H. He, and Y. F. Zou

Chapter 23 Experimental System and Damage Identification of Small-Scale Wind


Turbine Blades .......................................................................................................... 219
J. M. Gutierrez, R. Astroza, J. Abell, Soto, C., F. Jaramillo, M. Guarini,
and M. Orchard

Chapter 24 Estimation of Nonlinear System States with Unknown Loading and


Limited Measurements ............................................................................................. 229
J. He, X. X. Zhang, and M. C. Qi

Chapter 25 Performance Analysis of Shear Connectors in Demountable Composite


Bridge Deck with Steel Sheets ................................................................................. 239
K. Ye and J. Zhang

Chapter 26 Bridge Damage Classification and Detection Using Fully Convolutional


Neural Network Based on Images from UAVs......................................................... 247
Jiyuan Shi, Rongzhi Zuo, and Ji Dang

Chapter 27 Quarter Car Parameter Estimation with Application to Road Profile


Evaluation Using a Smartphone ............................................................................... 255
K. Xue, T. Nagayama, and D. Su

Chapter 28 Highway Bridge Weigh-in-Motion via Moving Load Identification ........................ 263
Xiangang Lai, A. Emin Aktan, Kirk Grimmelsman, Matteo Mazzotti, and
Ivan Bartoli

Chapter 29 Experimental and Numerical Studies on Dynamic Performance of


a Steel Stair under Human Action............................................................................ 271
Li Junxing and Zhang Wenyuan

Chapter 30 Study on Seismic Mitigation of Elastic Cables on Long-Span Cable-Stayed


Bridges by Shaking Table Test ................................................................................. 281
Li Xu, Qunjun Huang, Junhao Zheng, and Kang Liu

Chapter 31 Study on Vibration Control of Double-Tower Cable-Stayed Bridges by


Adding Longitudinal Constraint System.................................................................. 301
Li Xu, Yang Xu, Junhao Zheng, and Wenqian Chen
viii Contents

Chapter 32 Automatic Modal Parameter Identification of Cable-Stayed Bridge Based on


the Stochastic Subspace Identification Method ........................................................ 321
Liao Yuchen, Zhang Kun, Zong Zhouhong, and Lin Dinan

Chapter 33 Statistical Analysis of Random Dynamic Responses of Bridge under Dense


Traffic Flow .............................................................................................................. 331
Linfeng Hu, Jiayan Lei, Wei Shi, and Zihao Wang

Chapter 34 Influence of Joint Reinforcement Anchorage Detailing on the Seismic


Performance of Double-Column Bridge Piers ......................................................... 343
Liyuan Wang, Chaoyang Tang, Yongjian Chen, Siyang Ma, and Tianyu Zheng

Chapter 35 Pseudo Damage Training for Seismic Fracture Detection Machine ........................ 355
Luyao Wang, Ji Dang, and Xin Wang

Chapter 36 A Review of Experimental Studies on Laboratory Grandstands ............................. 363


Mohammad Almutairi, Onur Avci, and Nikolaos Nikitas

Chapter 37 Experimental Analysis to Find Factor of Safety in Case of Dams and


Slopes Using Finite Element and Limit Equilibrium Methods ................................ 373
Muhammad Israr Khan, Shuhong Wang, and Pengyu Wang

Chapter 38 Development of Tuned Mass Damper Using Multi-Stage Steel Laminated


Rubber Bearings ....................................................................................................... 389
P. Chansukho and P. Warnitchai

Chapter 39 On the Performance of Unscented Kalman Filters in Parameter Estimation


of Nonlinear Finite Element Models ........................................................................ 399
R. Astroza and N. Barrientos

Chapter 40 Uncertainty Handling in Structural Damage Detection Using a


Non-Probabilistic Meta-Model................................................................................. 409
R. Ghiasi, M. R. Ghasemi, Mohammad Noori, and Wael A. Altabey

Chapter 41 A Non-Parametric Approach toward Structural Health Monitoring for


Processing Big Data Collected from the Sensor Network ....................................... 419
R. Ghiasi, M. R. Ghasemi, Mohammad Noori, and Wael A. Altabey

Chapter 42 An Investigation into the Active Vibration Control of Three Coupled


Oscillators Using the Twin Rotor Dampers ............................................................. 427
R. Terrill and U. Starossek
Contents ix

Chapter 43 Development and Experimental Verification of IoT Sensing-Based Structural


Seismic Monitoring System ..................................................................................... 435
Rongzhi Zuo, J. Dang, and C.S. Goit

Chapter 44 Non-Probabilistic Damage Detection Using Classic and Modal Interval Analysis ... 445
J.Y. Huang and S.E. Fang

Chapter 45 A 3D Isolation Device with Vertical Variable Stiffness for Long-Span Spatial
Structures ................................................................................................................. 451
Y. Shi, H. Liu, Z. Chen, and Y. Ding

Chapter 46 Detection of Crack in Euler-Bernoulli Beam Using Bayesian Inference ................. 461
T. Wang, Mohammad Noori, Wael A. Altabey, Z. Ying, and R. Ghiasi

Chapter 47 Seismic Response of Counterweight-Roller-Rail Coupled System for


High-Speed Traction Elevator .................................................................................. 473
Wen Wang and Yan Jiang

Chapter 48 Modal Shape Estimation Based on a Parked Vehicle Induced Frequency


Variation ................................................................................................................... 481
W. Y. He and W. X. Ren

Chapter 49 Wind Speed and Wind Direction Joint Distribution Model Based on
Structural Health Monitoring ................................................................................... 487
X. W. Ye and Y. Ding

Chapter 50 Application of a 3D Base Isolation and Overturn Resistance Device on the


Large Height-Width Ratio Structure ........................................................................ 505
X. Yan, S. Shi, C. Zhang, H. Mao, and L. Zhou

Chapter 51 Non-Contact Video-Based Identification for Dynamic Behaviors of


Beam Structures ....................................................................................................... 515
Yu Cheng and Wen Xiong

Chapter 52 Bridge Damage Localization Using Moving Embedded Principal Component


Analysis with a Single Sensor .................................................................................. 525
Z. F. Shen, Y. K. Xie, and Z. H. Nie

Chapter 53 Damage Detection of RC Bridges Based on Comprehensive Information of


Influence Lines ......................................................................................................... 531
Zhiwei Chen, Weibiao Yang, and Jun Li
x Contents

Chapter 54 Hybrid DIC-Meshless Method for Evaluating Strain Field around the Crack ......... 541
Zhu Zhihui, Luo Sihui, Wang Fan, Feng Qianshuo, and Jiang Lizhong

Chapter 55 Identification of Moving Forces in the Case of Unknown Bridge Structural


Parameters Using Truncated Generalized Singular Value Decomposition
Algorithm ................................................................................................................. 551
Zhuhong Ouyang and Chiu Jen Ku

Chapter 56 Estimation of Repeated Slip Surface in Cut Slope Stability Analysis ..................... 561
Zulkifl Ahmed, Shuhong Wang, and Dong Furui
Index .............................................................................................................................................. 571
Preface
This book is a collection of selected papers presented at the International Conference on
Experimental Vibration Analysis for Civil Engineering Structures (EVACES), held in Nanjing,
China in September 2019.
EVACES is a premier venue where recent progress in the field of civil infrastructure systems,
especially the latest developments in experimental vibration analysis research in civil engineer-
ing structures, is presented and discussed by experts from the global community. Over the past
decade, due to emergence of new technologies and computational tools, the field of civil engineer-
ing and infrastructure systems has witnessed a significant advancement in the aforementioned
areas. EVACES provides a forum for the global research community to share and exchange the
latest achievements in these areas. After the first seven successful EVACES conferences which
took place in Bordeaux, France (2005), Porto, Portugal (2007), Wroclaw, Poland (2009), Varenna,
Italy (2011), Ouro Preto, Brazil (2013), Dubendorf, Switzerland (2015), and San Diego, United
States of America (2017), EVACES 2019 was organized by the Southeast University (SEU)
at Nanjing, held from September 5–8, 2019, on the main campus of SEU. The main topics of
EVACES 2019 included (1) damage identification and structural health monitoring, (2) testing,
sensing, and modeling, (3) vibration isolation and control, (4) system and model identification,
(5) coupled dynamical systems (including human-structure, vehicle structure, and soil-structure
interaction), and (6) application of big data and artificial intelligence techniques. The key objec-
tive of this meeting was to present and discuss the latest developments in the aforementioned
areas in order to better understand how to assess the state of health of structural, geo-structural,
and complex infrastructure systems and how to predict their remaining useful life using vibration
data collected from these systems when subjected to operational and extreme loads. It is notewor-
thy to indicate this was the first time that the importance of advanced data analysis (e.g., system
and damage identification and artificial intelligence) methods in experimental vibration analysis
of critical infrastructure, as an important and emerging area, was addressed to support decision
making related to maintenance and inspection, retrofit, upgrade, and rehabilitation of these sys-
tems especially in case of emergency response.
This book contains selected papers prepared by the leading researchers that were presented
at EVACES 2019. These contributions can provide strategic roadmaps for research in the area of
experimental vibration analysis for civil engineering structures and for building more resilient and
sustainable civil infrastructure systems.
We express our sincere thanks to the members of the Scientific Committee and, in particular, to
all the authors and participants, for their essential and valuable contributions.
Workshop Co-Chairs and the Organizing Committee:

Jian Zhang, Southeast University, China (Co-Chair)


Zhishen Wu, Southeast University, China (Co-Chair)
A. Emin Aktan, Drexel University, USA (Co-Chair)
James Brownjohn, University of Exeter, UK (Co-Chair)
Gang Wu, Southeast University, China (Organizing Committee)
Jingquan Wang, Southeast University, China (Organizing Committee)
Joel P. Conte, University of California San Diego, USA (Organizing Committee)
Yong Li, Southeast University (Secretary General)
Zheng Liu, University of British Columbia, Canada (Organizing Committee)
Maosen Cao, Hehai University, China (Organizing Committee)
Yufeng Zhang, Jiangsu Transportation Institute (JSTI), China (Organizing Committee)

xi
xii Preface

Yan Xu, Southeast University, China (Organizing Committee)


Mohammad Noori, Cal Poly, San Luis Obispo, USA (Organizing Committee)

MATLAB® is a registered trademark of The MathWorks, Inc. For product information,


please contact:
The MathWorks, Inc.
3 Apple Hill Drive
Natick, MA 01760-2098 USA
Tel: 508-647-7000
Fax: 508-647-7001
E-mail: info@mathworks.com
Web: www.mathworks.com
Editors
Professor Jian Zhang is a professor and vice dean of the School of Civil Engineering, SEU, China.
He received his PhD from Kyoto University, Japan, and worked at the University of California at
San Diego and Drexel University, USA. In the area of structural health monitoring, he has published
4 books and over 50 first/corresponding-author Science Citation Index journal papers. His research
results have been applied on over 20 long-span bridges including the Sutong Yangzi-River Bridge and
the Second Humen Bridge. He is the editor board member of Journal of Computer-Aided Civil and
Infrastructure Engineering and Journal of Structural Control and Monitoring. He serves as co-chair
of the EVACES-8 conference and as the International Society for Structural Health Monitoring of
Intelligent Infrastructure (ISHMII council member. He was awarded the first prize of Science and
Technology Award of Jiangsu Province (twice), the first prize of Science and Technology of China
Highway Society, and the second prize of National Award for Technological Invention.

Professor Zhishen Wu is a distinguished professor at Ibaraki University, Japan and at SEU, China.
His research expertise includes structural/concrete/maintenance engineering and advanced com-
posite materials. He is the author or co-author of over 600 refereed papers including over 200
journal articles and has given 50 keynote or invited papers. He also holds 50 patents. Dr Wu
was awarded the JSCE Research Prize by Japan Society of Civil Engineering in 1990, the JSCM
Technology Award by the Japan Society for Composite Materials in 2005 and 2009, SHM Person
of the Year Award by International Journal of Structural Health Monitoring, and the National
Prize for Progress in Science and Technology (2nd) of China in 2012. He is also a member of Japan
National Academy of Engineering. He chairs China Chemical Fibers Association Committee on
Basalt Fibers and is the president of the ISHMII. He is also an elected follow of American Society of
Civil Engineering (ASCE), Japan Society of Civil Engineers (JSCS), ISHMII, and the International
Institute of Fiber Reinforced Polymers (FRP) in Construction. He also serves as an editor, associ-
ate editor, and editorial board member for more than ten international journals. He is the founding
chief editor of International Journal of Sustainable Materials and Structural Systems. Professor
Wu has supervised a large number of national research projects in China and was the founder of the
International of Institute for Urban Systems Engineering, at SEU.

Professor Mohammad Noori is a professor of mechanical engineering at Cal Poly, San Luis
Obispo, a Fellow of the American Society of Mechanical Engineering, and a recipient of the Japan
Society for Promotion of Science Fellowship. Dr Noori’s work in nonlinear random vibrations, seis-
mic isolation, and the application of artificial intelligence methods for structural health monitoring
is widely cited. He has authored over 250 refereed papers, including over 100 journal articles and
6 scientific books, and has edited 25 technical and special journal volumes. Noori has supervised
over 90 graduate students and post-doc scholars, and has presented over 100 keynote, plenary, and
invited talks. He is the founding executive editor of an international journal and has served on the
editorial boards of over 10 other journals and as a member of numerous scientific and advisory
boards. He has been a Distinguished Visiting Professor at several highly ranked global universities,
and directed the Sensors Program at the National Science Foundation in 2014. He has been a found-
ing director or co-founder of three industry-university research centers and held Chair professor-
ships at two major universities. He served as the dean of engineering at Cal Poly for five years, has
also served as the Chair of the national committee of mechanical engineering department heads,
and was one of 7 co-founders of the National Institute of Aerospace, in partnership with NASA
Langley Research Center. Noori also serves as the Chief Technical Advisor for several scientific
organizations and industries.

xiii
xiv Editors

Dr. Yong Li is an assistant professor in the School of Civil Engineering, SEU, China. He received
his undergraduate degree in 2012 and his PhD in 2017, both degrees in aerospace engineering and
applied mechanics, at Tongji University, in Shanghai, China. His research focused on the coupling
relationship between mechanics and chemical reaction. He has published 1 book and over 20
journal papers in leading journals. He is also a reviewer for several journals in the fields of mechan-
ics and physics. Dr Li is the recipient of an Outstanding Young Investigator Award supported by
the National Natural Science Foundation of China. He is also engaged in several other national
key research and development programs as an investigator. He served as secretary general of the
International Conference on EVACES-8.
Contributors
J. Abell R. Astroza
Faculty of Engineering and Applied Sciences Faculty of Engineering and Applied Sciences
Universidad de los Andes Universidad de los Andes
Santiago, Chile Santiago, Chile

Zulkifl Ahmed Onur Avci


School of Resource and Civil Engineering Department of Civil Engineering
Northeastern University Qatar University
Shenyang, China Doha, Qatar

D. Ai Bai Wen
School of Civil Engineering and Mechanics Institute of Engineering Mechanics
Huazhong University of Science and China Earthquake Administration
Technology Beijing, China
Wuhan, China and
Key Laboratory of Earthquake Engineering
and Engineering Vibration of China
J. S. Ali Earthquake Administration
Department of Civil Engineering Harbin, China
Aliah University
Kolkata, India A. Bakhtiar
Department of Engineering
Mohammad Almutairi Sepehr Institute of Higher Education
School of Civil Engineering Isfahan, Iran
University of Leeds
Leeds, United Kingdom D. Bandyopadhyay
and Department of Construction Engineering
Sustainability and Infrastructure Programme, Jadavpur University
Energy and Building Research Center Kolkata, India
Kuwait Institute for Scientific Research
Safat, Kuwait Baozhuang Zhang
State Key Laboratory of Green Building in
Wael A. Altabey Western China
International Institute for Urban Systems Xi’an University of Architecture &
Engineering Technology
Southeast University Xi’an, China
Nanjing, Jiangsu, China.
and N. Barrientos
Faculty of Engineering, Department of Faculty of Engineering and Applied Sciences
Mechanical Engineering Universidad de los Andes
Alexandria University Santiago, Chile
Alexandria, Egypt
Ivan Bartoli
D. Anastasopoulos Department of Civil, Architectural and
Department of Civil Engineering Environmental Engineering
University of Leuven (KU Leuven) Drexel University
Leuven, Belgium Philadelphia, Pennsylvania

xv
xvi Contributors

Farhad Behnamfar Chuan-Zhi Dong


Department of Civil Engineering Department of Civil, Environmental, and
Isfahan University of Technology Construction Engineering
Isfahan, Iran University of Central Florida
Orlando, Florida
Binbin Li
ZJU-UIUC Institute Chunyu Qian
Zhejiang University China Jikan Research Institute of Engineering
Haining, China Investigations and Design, Co., Ltd
Xi’an, Shaanxi 710043, China
B. Briseghella
Sustainable and Innovative Bridge Engineering Dai Junwu
Research Center (SIBERC) Institute of Engineering Mechanics
College of Civil Engineering, Fuzhou China Earthquake Administration
University Beijing, China
Fuzhou, P. R. China and
Key Laboratory of Earthquake Engineering
P. Chansukho and Engineering Vibration of China
Field of Structural Engineering Earthquake Administration
Asian Institute of Technology Harbin, China
Pathumthani, Thailand
Ji Dang
Chaoyang Tang Department of Civil and Environmental
School of Civil Engineering Engineering
Fuzhou University University of Saitama
Fuzhou, China Saitama, Japan
and
Fujian Highway Administrative Bureau De-Hui Tang
Fuzhou, China Harbin Institute of Technology (Shenzhen)
School of Civil Engineering
Z. Chen Shenzhen, China
School of Civil Engineering
Tianjin University G. De Roeck
Tianjin, China Department of Civil Engineering
University of Leuven (KU Leuven)
Zhiwei Chen Leuven, Belgium
Department of Civil Engineering and
Xiamen University Structural Mechanics Division
Xiamen, China University of Leuven (KU Leuven)
Leuven, Belgium
Chi Peihong and
Institute of Earthquake Protection and Disaster Sustainable and Innovative Bridge Engineering
Mitigation Research Center (SIBERC)
Lanzhou University of Technology College of Civil Engineering, Fuzhou
Lanzhou, China University
Fuzhou, P. R. China
Chiu Jen Ku
Department of Civil and Environmental
Engineering
Shantou University
Guangdong, China
Contributors xvii

Y. Ding Feng Qianshuo


Key Laboratory of Coast Civil Structures School of Civil Engineering
Safety of Ministry of Education Central South University
Tianjin University Changsha, China
Tianjin, China
K. Gao
and School of Civil Engineering and Mechanics
Department of Civil Engineering Huazhong University of Science and
Zhejiang University Technology
Hangzhou, China Wuhan, P. R. China

Dong Furui H. Ge
School of Resource and Civil Engineering Department of Civil Engineering
Northeastern University Meijo University
Shenyang, China Nagoya, Japan

Du Yongfeng M. R. Ghasemi
Institute of Earthquake Protection and Faculty of Engineering, Department of
Disaster Mitigation Civil Engineering
Lanzhou University of Technology University of Sistan and Baluchestan
Lanzhou, China Zahedan, Iran

A. Emin Aktan R. Ghiasi


Department of Civil, Architectural and Faculty of Engineering, Department of
Environmental Engineering Civil Engineering
Drexel University University of Sistan and Baluchestan
Philadelphia, Pennsylvania Zahedan, Iran

Fang Dengjia C. S. Goit


Institute of Earthquake Protection and Department of Civil and Environmental
Disaster Mitigation Engineering
Lanzhou University of Technology Saitama University
Lanzhou, China Saitama, Japan

S. E. Fang Kirk Grimmelsman


School of Civil Engineering Intelligent Infrastructure Systems
Fuzhou University Philadelphia, Pennsylvania
Fuzhou, Fujian Province, China
and M. Guarini
National and Local United Research Center Faculty of Engineering and Applied Sciences
for Seismic and Disaster Informatization of Universidad de los Andes
Civil Engineering Santiago, Chile
Fuzhou University
Fuzhou 350108, China J. M. Gutierrez
Faculty of Engineering and Applied Sciences
Feng Gu Universidad de los Andes
Department of Civil and Environmental Santiago, Chile
Engineering
Shantou University
Guangdong, China
xviii Contributors

Jianping Han F. Jaramillo


Key Laboratory of Disaster Prevention Department of Electrical Engineering
and Mitigation in Civil Engineering of University of Chile
Gansu Province Santiago, Chile
Lanzhou University of Technology
Lanzhou, China Jiang Lizhong
School of Civil Engineering
J. He Central South University
Key Laboratory of Wind and Bridge Changsha, China
Engineering of Hunan Province
College of Civil Engineering, Hunan Jiayan Lei
University School of Architecture and Civil Engineering
Changsha, China Xiamen University
Xiamen, China
L. He
Sustainable and Innovative Bridge Engineering Jiyuan Shi
Research Center (SIBERC) Department of Civil and Environmental
College of Civil Engineering, Fuzhou Engineering
University University of Saitama
Fuzhou, P. R. China Saitama, Japan

W. Y He Jun Li
Department of Civil Engineering Centre for Infrastructure Monitoring and
Hefei University of Technology Protection, School of Civil and Mechanical
Hefei, Anhui Province, China Engineering
Curtin University
X. H. He Bentley, Australia
School of Civil Engineering
Central South University Jun Teng
Changsha, China Harbin Institute of Technology (Shenzhen)
and Shenzhen, China
National Engineering Laboratory for High
Speed Railway Construction Junhao Zheng
Changsha, China College of Civil Engineering
and Fuzhou University
Joint International Research Laboratory of Fuzhou, Fujian, China
Key Technology for Rail Traffic Safety
Changsha, China Junlong Zhu
Shanghai Baitong Project Management
Zhiqiang Hou Consulting, Co., Ltd
The Fifth Construction Co., Ltd. of China Shanghai, China
TIESIJU Civil Engineering Group
Jiujiang, China Kang Liu
College of Civil Engineering
J. Y. Huang Fuzhou University
School of Civil Engineering Fuzhou, Fujian, China
Fuzhou University
Fuzhou, Fujian Province, China Muhammad Israr Khan
School of Resources and Civil Engineering
Northeastern University
Shenyang city, Liaoning Province, China
Contributors xix

D. Li H. Luo
School of Civil Engineering School of Civil Engineering and Mechanics
Hefei University of Technology Huazhong University of Science and
Hefei, Anhui, China Technology
Wuhan, China
H. Li
School of Civil Engineering and Luo Sihui
Mechanics School of Civil Engineering
Huazhong University of Science and Central South University
Technology Changsha, China
Wuhan, China
H. Mao
Li Junxing School of Ecological Environment and
Department of Civil Engineering Urban Construction
Harbin Institute of Technology Fujian University of Technology
Harbin, China Fuzhou, China

Li Xu G. C. Marano
College of Civil Engineering Sustainable and Innovative Bridge Engineering
Fuzhou University Research Center (SIBERC)
Fuzhou, Fujian, China College of Civil Engineering, Fuzhou
University
Liao Yuchen Fuzhou, P. R. China
School of Civil Engineering
Southeast University Matteo Mazzotti
Nanjing, China Department of Civil, Architectural and
Environmental Engineering
Lin Dinan Drexel University
Fujian Academy of Bulding Research Philadelphia, Pennsylvania
Fujian Key Laboratory of Green Building
Technology Mingxin Gao
Fuzhou, China Department of Bridge and Tunnel Engineering
Harbin Institute of Technology
Linfeng Hu Harbin, China
School of Architecture and Civil
Engineering T. Nagayama
Xiamen University Department of Civil Engineering
Xiamen, China The University of Tokyo
Tokyo, Japan
H. Liu
School of Civil Engineering Mohammadreza Najar
Tianjin University Department of Civil Engineering
Tianjin, China University of Isfahan
Isfahan, Iran
Liyuan Wang
College of Civil Engineering F. Necati Catbas
Fuzhou University Department of Civil, Environmental, and
Fuzhou, China Construction Engineering
University of Central Florida
Orlando, Florida
xx Contributors

Z. H. Nie G. Z. Qu
College of Science & Engineering China Railway Construction Bridge Design and
Jinan University Research Branch
Guangzhou, Guangdong, China China Railway Siyuan Survey and Design
and Group Co., LTD.
The Key Laboratory of Disaster Forecast and Wuhan, Hubei, China
Control in Engineering
Guangdong, Guangzhou, China Qunjun Huang
College of Civil Engineering
Nikolaos Nikitas Fuzhou University
School of Civil Engineering Fuzhou, Fujian, China
University of Leeds
Leeds, United Kingdom W. X. Ren
Department of Civil Engineering
Mohammad Noori Hefei University of Technology
International Institute for Urban Systems Hefei, Anhui Province, China
Engineering (IIUSE)
Southeast University E. Reynders
Nanjing, Jiangsu, China Department of Civil Engineering
and University of Leuven (KU Leuven)
Department of Mechanical Engineering Leuven, Belgium
California Polytechnic State University and
San Luis Obispo, California Structural Mechanics Division
University of Leuven (KU Leuven)
M. Orchard Leuven, Belgium
Department of Electrical Engineering
University of Chile Z. F. Shen
Santiago, Chile College of Science & Engineering
Jinan University
Pengyu Wang Guangzhou, Guangdong, China
School of Resources and Civil Engineering
Northeastern University S. Shi
Shenyang City, Liaoning Province, China College of Civil Engineering
Fuzhou University
M. C. Qi Fuzhou, China
Key Laboratory of Wind and Bridge
Engineering of Hunan Province Y. Shi
College of Civil Engineering, Hunan Key Laboratory of Coast Civil Structures
University Safety of Ministry of Education
Changsha, China Tianjin University
Tianjin, China
Qifang Xie
State Key Laboratory of Green Building in Shiqi Fu
Western China College of Civil Engineering
Xi’an University of Architecture & Huaqiao University
Technology Xiamen, P. R. China
Xi’an, China
Shuhong Wang
School of Resource and Civil Engineering
Northeastern University
Shenyang, China
Contributors xxi

A. I. Silik Luyao Wang


International Institute for Urban Systems Department of Civil and Environmental
Engineering (IIUSE) Engineering
Southeast University Saitama University
Nanjing, China Saitama, Japan
and
Faculty of Engineering Sciences, T. Wang
Department of Civil Engineering International Institute for Urban Systems
Nyala University Engineering
Nyala, Sudan Southeast University
Nanjing, Jiangsu, China
Siyang Ma
College of civil Engineering Xin Wang
Fuzhou University Department of Architecture and Building
Fuzhou, China Engineering
Tokyo University of Science
Soto, C. Tokyo, Japan
Faculty of Engineering and Applied Sciences
Universidad de los Andes P. Warnitchai
Santiago, Chile Department of Structural Engineering
Asian Institute of Technology
U. Starossek Pathumthani, Thailand
Head of Institute, Structural Analysis
Institute Wei Shi
Hamburg University of Technology School of Architecture and Civil Engineering
Hamburg, Germany Xiamen University
Xiamen, China
D. Su
Department of Civil Engineering Weibiao Yang
The University of Tokyo CSCEC Strait Construction and Development
Tokyo, Japan Co., Ltd
Fuzhou, China
X. T. Sun
School of Civil Engineering Wei-Hua Hu
Hefei University of Technology School of Civil Engineering
Hefei, Anhui, China Harbin Institute of Technology (Shenzhen)
Shenzhen, China
R. Terrill
Structural Analysis Institute Wen Wang
Hamburg University of Technology School of Civil Engineering and Architecture
Hamburg, Germany Guangxi University of Science and Technology
Guangxi, China
Tianyu Zheng and
College of Civil Engineering State Key Laboratory of Disaster Reduction in
Fuzhou University Civil Engineering
Fuzhou, China Tongji University
Shanghai, China
Wang Fan
School of Civil Engineering
Central South University
Changsha, China
xxii Contributors

Wen Xiong K. Xue


Department of Bridge Engineering, Department of Civil Engineering
School of Transportation The University of Tokyo
Southeast University Tokyo, Japan
Nanjing, China
J. F. Yan
S. Weng China Railway Construction Bridge Design and
School of Civil Engineering and Research Branch
Mechanics China Railway Siyuan Survey and Design
Huazhong University of Science and Group Co., LTD.
Technology Wuhan, Hubei, China
Wuhan, P. R. China
Yan Jiang
Wenqian Chen School of Civil Engineering and Architecture
College of Civil Engineering Guangxi University of Science and Technology
Fuzhou University Guangxi, China
Fuzhou, Fujian, China
X. Yan
Z. Wu College of Civil Engineering
International Institute for Urban Systems Fuzhou University
Engineering Fuzhou, China
Southeast University
Nanjing, Jiangsu, China Yan-An Gao
Faculty of Architecture and Civil Engineering
Y. Xia Huaiyin Institute of Technology
Department of Civil and Environmental Huaian, P. R. China
Engineering
The Hong Kong Polytechnic University J. Yang
Hong Kong, P. R. China School of Civil Engineering
Central South University
Xiangang Lai Changsha, China
Department of Civil, Architectural and and
Environmental Engineering National Engineering Laboratory for
Drexel University High Speed Railway Construction
Philadelphia, Pennsylvania Changsha, China
and
Y. K. Xie Joint International Research Laboratory of
College of Science & Engineering Key Technology for Rail Traffic Safety
Jinan University Changsha, China
Guangzhou, Guangdong, China
and Yang Liu
The Key Laboratory of Disaster Forecast and Department of Bridge and Tunnel Engineering
Control in Engineering Harbin Institute of Technology
Guangdong, Guangzhou, China Harbin, China

Xuanneng Gao Yang Xu


College of Civil Engineering College of Civil Engineering
Huaqiao University Fuzhou University
Xiamen, P. R. China Fuzhou, Fujian, China
Contributors xxiii

Yan-Long Xie V. Zabel


ZJU-UIUC Institute Institute of Structural Mechanics
Zhejiang University Bauhaus-Universität Weimar
Haining, China Weimar, Germany

A. A. Yazdani C. Zhang
Department of civil engineering College of Civil Engineering
University of Isfahan Fuzhou University
Isfahan, Iran Fuzhou, China

K. Ye Hongyu Zhang
Department of Civil Engineering Key Laboratory of Disaster Prevention and
Hefei University of Technology Mitigation in Civil Engineering of Gansu
Hefei, China Province
Lanzhou University of Technology
X. W. Ye Lanzhou, China
Department of Civil Engineering
Zhejiang University J. Zhang
Hangzhou, China Department of Civil Engineering
Hefei University of Technology
P. C. Yin Hefei, China
China Railway Construction Bridge Design and
Research Branch Zhang Kun
China Railway Siyuan Survey and School of Civil Engineering
Design Group Co., LTD. Southeast University
Wuhan, Hubei, China Nanjing, China

Z. Ying Zhang Wenyuan


International Institute for Urban Systems Department of Civil Engineering
Engineering Harbin Institute of Technology
Southeast University Harbin, China
Nanjing, Jiangsu, China
X. X. Zhang
Yongjian Chen Key laboratory of wind and bridge engineering
College of civil Engineering of Hunan Province
Fuzhou University College of Civil Engineering, Hunan University
Fuzhou, China Changsha, China

Yu Cheng L. Zhou
Department of Bridge Engineering, College of Civil Engineering
School of Transportation Fuzhou University
Southeast University Fuzhou, China
Nanjing, China
H. P. Zhu
X. S. Yu School of Civil Engineering and Mechanics
China Railway Construction Bridge Design and Huazhong University of Science and
Research Branch Technology
China Railway Siyuan Survey and Design Wuhan, P. R. China
Group Co., LTD.
Wuhan, Hubei, China
xxiv Contributors

Zhu Zhihui Y. F. Zou


School of Civil Engineering School of Civil Engineering
Central South University Central South University
Changsha, China Changsha, China
and
Zhuhong Ouyang National Engineering Laboratory for
Department of Civil and Environmental High Speed Railway Construction
Engineering Changsha, China
Shantou University and
Shantou, China Joint International Research Laboratory of
Key Technology for Rail Traffic Safety
Zihao Wang Changsha, China
School of Architecture and Civil Engineering
Xiamen University Rongzhi Zuo
Xiamen, China Department of Civil and Environmental
Engineering
Zong Zhouhong University of Saitama
School of Civil Engineering Saitama, Japan
Southeast University
Nanjing, China
1 Wavelet-Based Damage-
Sensitive Features Extraction
A. I. Silik
Southeast University
Nyala University

Mohammad Noori
California Polytechnic State University

W. A. Altabey
Southeast University
Alexandria University

CONTENTS
1.1 Introduction .............................................................................................................................. 1
1.2 Data Preprocessing ................................................................................................................... 2
1.2.1 Feature Detection and Extraction ................................................................................. 2
1.2.2 Statistical Model Formulation ...................................................................................... 4
1.2.3 Scale Selection .............................................................................................................. 4
1.2.4 Feature Damage Index Identification Based Wavelet ................................................... 5
1.3 Numerical Simulation and Results: Discussion ........................................................................ 6
1.4 Conclusions ............................................................................................................................... 6
References ........................................................................................................................................ 10

1.1 INTRODUCTION
Signal-based damage detection techniques have been introduced to extract features to represent
the signal characteristics which are used to answer the questions associated to damage levels [1,2].
Over the last decades, various methods have been developed to improve the feature extraction (FE)
procedure [3–10]. Yoon et al. [11] and Staszewski [12] discussed the signal processing role for FE in
composite materials and stated that the advances in damage detection attribute to the development
of signal processing techniques such as denoising, FE, and optimal sensor location procedures.
FE-based signal processing has been classified into time, frequency, and time-frequency methods
[13]. It aims to transform data into a lower dimension by removing redundant information and pre-
serving the useful elements [14]. In time domain, features are extracted from the signal using linear
and nonlinear functions [15]. In frequency and time-frequency domain wavelet transform (wt) have
been used for FE. Pittner and Kamarthi [16] explored the changes in wavelet coefficients for FE for
pattern recognition tasks. Balafas and Kiremidjian [17] proposed damage features for earthquake
damage estimation based on continuous wavelet transform (CWT), using ground motion and its
structure responses. Patel et al. [18] used complex Gaussian wavelet to extract the discontinuity of
the acceleration response of a reinforced concrete (RC) building. The results showed that wavelet
weights are directly affected by the change in physical properties and can identify damage to a rea-
sonable extent. The modified wavelet energy rate-based damage detection methods are proposed by

1
2 Experimental Vibration Analysis for Civil Structures

several researchers [19–22]. The method based on Approximate Entropy [23] is proposed to assess
the damaged condition from the vibration signals generated by excitation [24–26] on structures.
It is able to capture the deterioration of beams with more sensitive characteristics. Although many
works have been done using wt for damage detection [27,28], however, it cannot always identify the
damage extent or its location. Most of them are not suited when the signal is nonstationary and is
mainly restricted to small structures, simulated models, or laboratory tests. Moreover, some of them
refer to specific load scenarios that are not common in real structures; their applicability is limited
and the ground motion recording is neglected. Further, only one feature was considered to resemble
the dynamic characteristic, and the base wavelet and resolution level are chosen in an arbitrary man-
ner, which results in a negative effect in results. It’s clear that there is a lack of a truly intelligent
and comprehensive approach that can detect the slight damage in a large, complex system and in a
noisy environment that may include a large number of relevant features, and not just a few dynamic
features, to detect the location and the damage extent. Thus, an advanced intelligent algorithm for
sensitive features extraction of large complex data is required. Other crucial issue in damage detec-
tion is how to select the correct features that describe the damage and improve the AI-algorithm
accuracy. In this study, given the modulation characteristics of structural dynamic responses (SDR),
the wavelet is used for damage-sensitive parameter (DSF) extraction which cannot be obtained
either by short Fourier transform or Wigner-Ville distribution. Here, FE is used to describe how to
find the best features which represent various damage conditions, whereas feature selection is used
to define how to select features for damage detection based wavelet DSFs. The proposed algorithm
is able to de-noise and extract the features that define the structure’s true state. The algorithm’s nov-
elty is to identify and extract several key features that can determine the key dynamic characteristics
of the structure and correlate to the damages.

1.2 DATA PREPROCESSING


First, the raw data are visualized as in Figure 1.1 to get an intuitive feeling for the data. Some
transitory characteristics are quite evident in the time domain, while the frequency content is quite
clear and the user can specify a range of frequencies for further analysis. Hence, it seems neces-
sary to do some form of preprocessing prior to any statistical model development. Preprocessing
involves normalization, trend isolation, and denoising. Noise presence may affect the detection of
high-frequency (HF) components related to damage, so, signal de-noising is necessary, and in some
situations, the frequency band of an unwanted signal lies within the same signal frequency range
which is required to remove and preserve the useful information. Figure 1.1 visualizes the excitation
in various domains. The upper right plot shows the discrete Fourier transform (DFT) of the signal.
It indicates that the signal has frequencies between 0.0 and 10 Hz, but neither gives any indication of
their evolution or the presence of the discontinuity. The lower left plot is the time-frequency plot of
the CWT showing the localization in time and frequency to identify changes in the signal frequency
as time evolves, which is a contour map.

1.2.1 FEATURE DETECTION AND EXTRACTION


FE is a key aspect of damage diagnosis using a statistical model for civil structures [15]. DSFs are
used to track the structure change to detect the damage. However, it is hard to extract useful features
from SDR just by observing them in the time domain, so, several techniques have been developed
for this purpose. The specific technique choice is usually tied to the application under study and
specific requirements. Due to the modulation characteristics of SDR, a wavelet-based method has
been developed to analyze nonstationary data in time and frequency domain simultaneously and
then extract the features. The input and output variations that reflect changes in the structure due to
damage are considered. Based on wavelet weights, DSFs are derived through statistical processing
to indicate the true structure state. Due to variations in the input and output, DSFs are normalized
Damage-Sensitive Features Extraction

FIGURE 1.1 Excitation visualization in va r ious domains.


3
4 Experimental Vibration Analysis for Civil Structures

to account for variations in loading and environmental conditions. To distinguish between the useful
and poor features, the relationship between DSFs and the structure’s physical parameters should be
investigated. FE procedures are as follows:

1. Load and visualize the data obtained from sensor i, denoted by xi(t)
2. Data are standardized prior to any subsequent analyses
3. Prepossess the signal to remove the artifacts such as all trends and environmental conditions
4. Choose a wavelet function and a set of scales to analyze.
5. Decompose the measured signal into various frequency bands
6. Study every frequency band component with a resolution matched to its scale.
7. Extract DSFs that define the key structure dynamic characteristics that correlated well
with the damage.
8. Choose the features exposing the best classification accuracy of these features.

1.2.2 STATISTICAL MODEL FORMULATION


Once an analyzing wavelet ψ(t) has been chosen, two basic parameters should be chosen and manip-
ulations should be done so as to make the ψ(t) more flexible. The ψ(t) is stretched or squeezed
through varying its dilation parameter s and moved through its translation parameter τ (i.e. along
the localized time index τ).
t −τ ⎞
ψ ⎜⎛
1
ψ s,τ (t ) =
⎝ s ⎟⎠
(1.1)
s
Let x(t) be the system acceleration response, where t denotes time. CWT of a function x(t) ∈ L2 (ℜ),
where L2 (ℜ) is the set of square integrable functions denoted as Ws,τ and defined as
−∞
⎛ t − τ ⎞ dt
∫ x(t) ⋅ψ
1
Ws,τ = *
⎜⎝ ⎟ (1.2)
s s ⎠

The basic analyzed wavelet ϕ(𝑡) is a square integrable function, and it meets the following
relationship:

2
ψ (ω )
CΨ =
∫ R ω
dω < ∞ (1.3)

where s and τ are used to adjust the frequency and time location. Ws,τ shows how closely ψ s,τ (t ) cor-
related with x(t). High scale gives low frequency (LF) and low scale offers HF. HF in a given scale is
an indicator to the signal frequency content at that time [29]. At each scale, there is a corresponding
characteristic frequency Fa that relates to s, time rate Δt, and wavelet center frequency Fc as

Fa = Fc ( s ⋅ Δt ) (1.4)

1.2.3 SCALE SELECTION


Multiplier matrix is affected by the scale at which wt is evaluated. So, the procedures proposed by
Balafas and Kiremidjian are followed to determine the scales at which wt can be evaluated.

1. The Fourier transforms for the output and input signals are calculated.
2. The frequencies that correspond to 10% and 90% cumulative Fourier amplitudes are determined.
Damage-Sensitive Features Extraction 5

3. The maximum and minimum frequencies determined above are taken as the outer bound
frequencies.
4. The frequency boundaries are converted to wavelet scales to determine the scale bounds.
5. The scales at which the wavelet transform will be evaluated are linearly spaced between
the scale bounds.

1.2.4 FEATURE DAMAGE INDEX IDENTIFICATION BASED WAVELET


Ws,τ is used to extract the feature vectors. To reduce its dimensionality, statistics over Ws,τ is used as

1. Maximum of the wavelet coefficients in each sub band.


2. Minimum of the wavelet coefficients in each sub band.
3. Mean of the wavelet coefficients in each sub band.
4. Standard deviation of the wavelet coefficients in each sub band.
5. Wavelet Entropy, Wavelet Energy, and Wavelet Power Spectrum
6. Residual Error between reconstructed signal and the actual measured signal

Wavelet Energy at a given scale can indicate damages such as loss of energy and how slowly the
energy decays. Nair and Kiremidjian [15] stated that Ws,τ shift in a given scale is described by the
energy as

∑ Wx ( a,b)
2
Ea = (1.5)
b=1

For a multidegree system with mass (M), damping (C), and stiffness K, subjected to force g(t),
damping ratio in each mode is ξ. The transfer response function of the displacement at the kth
degrees of freedom is derived as

N
φkrφlr
H kl (s) = ∑ (ω
r =1
2
r − s 2 + 2 jξω r s )
(1.6)

And for acceleration at the kth degrees of freedom as

N
− s 2φkrφlr Gl (s)
X k (s ) = ∑(
r =1
ω r2 − s 2 + 2 jξω r s )
(1.7)

Ws,τ of acceleration response (x (t )) with respect to the Morlet wavelet is derived as

∞ N
− s 2φkrφlr Gl (s)
∫∑( exp ( jsb ) exp ⎢⎡ − ( − as + ω 0 ) ⎥⎤ ds
a 1 2
WM x k (a,b) = (1.8)

−∞ r =1
ω r − s + 2 jξω r s
2 2
) ⎣ 2 ⎦

Instantons Energy is a feature that provides the distribution energy in each frequency band and is
given as
Ns

IE = log10 1/N j ⋅ ∑(W


s=1
s,τ )2 (1.9)
6 Experimental Vibration Analysis for Civil Structures

Teager Energy is a nonlinear operator used to track the signal energy fluctuations and is defined as

⎛ 1 N s −1 ⎞
TE = log10 ⎜
⎜⎝ N s ∑ (W
s=1
s,τ )2 − (Ws,τ (t − 1) ⋅Ws,τ (t + 1)) ⎟⎟

(1.10)

Wavelet Entropy: Entropy change rate CR(t) can be used as a damage index to describe the struc-
tural state changes and defined as

swt (t ) − swt (0)


CR(t) = (1.11)
swt (0)

where swt (0) is undamaged entropy and swt (t ) for damaged. Approximate Entropy which is a mea-
sure of the irregularity and nonlinear time series complexity can be used as a damage feature. It is
computed as

ApEn = Φ m − Φm + 1 (1.12)

N − m +1

where Φm = ( N − m + 1) −1
∑ log( N ), where N is the number within range points.
i =1
i i

1.3 NUMERICAL SIMULATION AND RESULTS: DISCUSSION


To validate the effectiveness of the proposed method, a simulated two-story frame is done with lumped
mass 100 kips/g at each floor, and same story stiffness 31.54 kips/in for all stories and the story height is
12 ft. The damping ratio for all natural modes is = 0.05. The structure is subjected to El Centro ground
motion. The SDR of first floor is selected to illustrate the concept of the process. Figure 1.2 shows the
response at first story and its filtered and decay versions representation in time and frequency domains.
Figures 1.3 and 1.4 show the various features in wavelet domain based on amplitude, energy, and
entropy. The magnitude gives the time-frequency behavior for the data. By looking at wt and the scales
giving large Ws, τ the signal frequencies can be estimated. By visual correlation of time-frequency rep-
resentation plots, it is possible to discriminate several time regions with various frequency behaviors,
as well to define the instant of time separating those regions. An abrupt change in SDR means local
maxima of wt modulus. Therefore, the ridges are recognized by seeking out the points in scalogram
plot where the time-frequency transform coefficients take on local maximum values. The ridges can be
observed as modal parameters which are based on finding the location of max Ws, τ . In summary, from
the wavelet results it is possible to extract the sensitive features which are used for further analysis.

1.4 CONCLUSIONS
In this study a sequence of DSFs for civil structures under arbitrary excitation has been proposed.
DSFs are derived from the acceleration signals by using CWT and both the input ground motion
and the output are taken into account. Here, various algorithms are identified and used in order to
track the change of the structures and to extract the damage features. The procedures for extracting
these features are completely data-driven; there are no assumptions about the structure and there is
minimal user-input besides the input and output signals. These features could be able to distinguish
not only a few dynamic characteristics for detecting the location and the damage extent but also the
damage types in a structure. Further, it can identify the slightest damage in a large, complex system
in a noisy environment subjected to arbitrary dynamic loads.a
Damage-Sensitive Features Extraction

FIGURE 1.2 Single representations: (a) time domain and (b) frequency.
7
8 Experimental Vibration Analysis for Civil Structures

FIGURE 1.3 Visual representations of damage features.


Damage-Sensitive Features Extraction

FIGURE 1.4 Additional visual representations of damage features.


9
10 Experimental Vibration Analysis for Civil Structures

REFERENCES
1. W. A. Altabey, An exact solution for mechanical behavior of BFRP nano-thin films embedded in NEMS,
Advances in Nano Research, vol. 5, no. 4, 2017, pp. 337–357. doi: 10.12989/anr.2017.5.4.337.
2. W. A. Altabey, A study on thermo-mechanical behavior of MCD through bulge test analysis, Advances
in Computational Design, vol. 2, no. 2, 2017, pp. 107–119. doi: 10.12989/acd.2017.2.2.107.
3. Z. Ying, N. Mohammad, A. W. A. Altabey, Z. Wu, Fatigue damage identification for composite pipe-
line systems using electrical capacitance sensors, Smart Materials and Structures, vol. 27, no. 8, 2018,
p. 085023. doi: 10.1088/1361-665x/aacc99.
4. Z. Ying, N. Mohammad, W. A. Altabey, R. Ghiasi, Z. Wu, A fatigue damage model for FRP composite
laminate systems based on stiffness reduction, Structural Durability and Health Monitoring, vol. 13,
no. 1, 2019, pp. 85–103, doi: 10.32604/sdhm.2019.04695.
5. Z. Ying, N. Mohammad, W. A. Altabey, R. Ghiasi, Z. Wu, Deep learning-based damage, load and sup-
port identification for a composite pipeline by extracting modal macro strains from dynamic excitations,
Applied Sciences, vol. 8, no. 12, 2018, p. 2564. doi: 10.3390/app8122564.
6. W. A. Altabey, N. Mohammad, Detection of fatigue crack in basalt FRP laminate composite pipe using
electrical potential change method, Journal of Physics Conference Series, vol. 842, 2017, p. 012079.
doi: 10.1088/1742-6596/842/1/012079.
7. W. A. Altabey, Delamination evaluation on basalt FRP composite pipe by electrical potential
change, Advances in Aircraft and Spacecraft Science, vol. 4, no. 5, 2017, pp. 515–528. doi: 10.12989/
aas.2017.4.5.515.
8. W. A. Altabey, EPC method for delamination assessment of basalt FRP pipe: Electrodes number effect,
Structural Monitoring and Maintenance, vol. 4, no. 1, 2017, pp. 69–84. doi: 10.12989/smm.2017.4.1.069.
9. W. A. Altabey, N. Mohammad, Monitoring the water absorption in GFRE pipes via an electrical
capacitance sensors, Advances in Aircraft and Spacecraft Science, vol. 5, no. 4, 2018, pp. 411–434. doi:
10.12989/aas.2018.5.4.499.
10. W. A. Altabey, FE and ANN model of ECS to simulate the pipelines suffer from internal corrosion,
Structural Monitoring and Maintenance, vol. 3, no. 3, 2016, pp. 297–314. doi: 10.12989/smm.2016.3.3.297.
11. H. Yoon, K. Yang, C. Shahabi, Feature subset selection and feature ranking for multivariate time series,
IEEE Transactions on Knowledge and Data Engineering, vol. 17, no. 9, 2005, pp. 1186–1198.
12. W. J. Staszewski, Intelligent signal processing for damage detection in composite materials, Composites
Science and Technology, vol. 62, no. 7–8, 2002, pp. 941–950.
13. P. W. P. Man, M. H. Wong, Efficient and robust feature extraction and pattern matching of time series by
a lattice structure. In Proceedings of the Tenth International Conference on Information and Knowledge
Management, Atlanta, GA 2001, pp. 271–278, ACM.
14. K. P. Chan, A. W. C. Fu, Efficient time series matching by wavelets. In Proceedings 15th International
Conference on Data Engineering (Cat. No. 99CB36337), Sydney, Australia, 1999, pp. 126–133, IEEE.
15. K. K. Nair, A. S. Kiremidjian, Time series based structural damage detection algorithm using Gaussian
mixtures modeling, Journal of Dynamic Systems, Measurement, and Control, vol. 129, no. 3, 2007,
pp. 285–293.
16. S. Pittner, S. V. Kamarthi, Feature extraction from wavelet coefficients for pattern recognition tasks,
IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 21, no. 1, 1999, pp. 83–88.
17. K. Balafas, A. S. Kiremidjian, Extraction of a series of novel damage sensitive features derived from
the continuous wavelet transform of input and output acceleration measurements, In Sensors and Smart
Structures Technologies for Civil, Mechanical, and Aerospace Systems, vol. 9061, 2014, p. 90611L.
International Society for Optics and Photonics.
18. S. S. Patel, A. P. Chourasia, S. K. Panigrahi, J. Parashar, N. Parvez, M. Kumar, Damage identification
of RC structures using wavelet transformation, Procedia Engineering, vol. 144, 2016, pp. 336–342.
19. M. Noori, H. Wang, W. A. Altabey, A. I. Silik, A modified wavelet energy rate-based damage identifica-
tion method for steel bridges, Scientia Iranica, vol. 25, 2018, pp. 3210–3230.
20. Z. Ying, N. Mohammad, B. B. Seyed, W. A. Altabey, Mode shape based damage identification for a
reinforced concrete beam using wavelet coefficient differences and multi-resolution analysis, Structural
Control and Health Monitoring, vol. 25, no. 3, 2017, pp. 1–41. doi: 10.1002/stc.2041.
21. Z. Ying, N. Mohammad, W. A. Altabey, Damage detection for a beam under transient excitation via
three different algorithms, Structural Engineering and Mechanics, vol. 63, no. 6, 2017, pp. 803–817. doi:
10.12989/sem.2017.64.6.803.
Damage-Sensitive Features Extraction 11

22. Z. Ying, N. Mohammad, W. A. Altabey, T. Awad, A comparison of three different methods for the iden-
tification of hysterically degrading structures using BWBN model, Frontiers in Built Environment, vol.
4, 2019, p. 80. doi: 10.3389/fbuil.2018.00080.
23. Z. K. Xie, G. H. Liu, Z. H. Zhang, Damage detection of concrete structure based on approximate
entropy, Applied Mechanics and Materials, vol. 226, 2012, pp. 920–925. Trans Tech Publications.
24. W. A. Altabey, Free vibration of basalt fiber reinforced polymer (FRP) laminated variable thick-
ness plates with intermediate elastic support using finite strip transition matrix (FSTM) method,
Vibroengineering, vol. 19, no. 4, 2017, pp. 2873–2885. doi: 10.21595/jve.2017.18154.
25. W. A. Altabey, Prediction of natural frequency of basalt fiber reinforced polymer (FRP) laminated vari-
able thickness plates with intermediate elastic support using artificial neural networks (ANNs) method,
Vibroengineering, vol. 19, no. 5, 2017, pp. 3668–3678. doi: 10.21595/jve.2017.18209.
26. W. A. Altabey, High performance estimations of natural frequency of basalt FRP laminated plates with
intermediate elastic support using response surfaces method, Vibroengineering, vol. 20, no. 2, 2018,
pp. 1099–1107. doi: 10.21595/jve.2017.18456.
27. Z. Hou, A. Hera, M. Noori, Wavelet-Based Techniques for Structural Health Monitoring. In: Health
Assessment of Engineered Structures: Bridges, Buildings and Other Infrastructures, ed., Achintya
Haldar (University of Arizona), World Scientific, Singapore, 2013, pp. 179–202.
28. Z. Hou, M. Noori, R. S. Amand, Wavelet-based approach for structural damage detection, Journal of
Engineering Mechanics, vol. 126, no. 7, 2000, pp. 677–683.
29. D. Hester, A. González, A wavelet-based damage detection algorithm based on bridge acceleration
response to a vehicle, Mechanical Systems and Signal Processing, vol. 28, 2012, pp. 145–166.
2 Deep Learning for Automated
Damage Detection
A Novel Algorithm in CNN
Family for Faster and Accurate
Damage Identification
Wael A. Altabey
Southeast University
Alexandria University

Mohammad Noori
California Polytechnic State University

CONTENTS
2.1 Introduction ............................................................................................................................ 13
2.2 Deep Learning Based Approaches for Image Classification and Object Detection............... 15
2.2.1 Deep Learning for Image Classification..................................................................... 15
2.3 Methodology ........................................................................................................................... 17
2.4 Conclusions ............................................................................................................................. 23
Acknowledgements .......................................................................................................................... 23
References ........................................................................................................................................ 23

2.1 INTRODUCTION
For industrial and civil structures, structural health monitoring is more complicated, especially
monitoring in a real time when the types of structures are complicated as well as when the measure-
ment signal from sensors is corrupted with environmental noise. Therefore, structural health moni-
toring needs other techniques to identify structural damage. Deep Learning can solve this problem
satisfactorily due to its superior adaptive learning of datasets [1–12].
Krizhevsky et al. [13] classify 1.2 million high-resolution images by using a large and deep convo-
lutional neural network (CNN). Bouvrie [14] presented the derivation and implementation of CNNs.
The detection and localization of damage based on real-time vibrations have been proposed using 1D
CNNs [15–19]. A sparse coding algorithm was applied to a large number of untagged examples to
train a feature extractor, and the features were then used to pass to a neural network classifier to dis-
tinguish various damage states from bridges [20]. A type of deep architecture CNN has been used to
detect cracks in concrete without directly extracting the features of the defects, to overcome challenges
such as lighting and shadow changes [21]. The damage in steel structures was indicated by using the
differences in wavelet coefficients to analyze the responses of a healthy and loose connection structure
[22,23]. The wavelets were used to detect strong earthquake response waves of framed structures indi-
cating the occurrence of damage [24]. A new signal processing algorithm was proposed by the deep
learning model, which consists of combined wavelets, neural networks, and Hilbert transform [25,26].

13
14 Experimental Vibration Analysis for Civil Structures

In addition, three types of deep neural network models, such as deep Boltzmann machines, deep belief
networks, and stacked auto-encoders, were investigated to identify rolling bearing failure conditions
[27]. Compared to a traditional support vector machine (SVM) and backpropagation neural network,
planetary gearbox conditions could be effectively detected with the best diagnostic accuracy using
deep CNNs [28]. The benefits of generalization and memorization ability for recommending systems
were combined by using the wide and deep common learning by applying wide linear models and
deep neural networks [29]. A multi-scale structural health monitoring system has been constructed
to monitor health conditions and assess the usefulness of large-scale bridges using the Hadoop eco-
system (MS-SHM-Hadoop). A Bayesian network has been studied to assess the reliability of specific
components as a function of ease of maintenance and inter-component correlations [30] and a hybrid
response surface method [31]. Unlike buildings or bridges, for pipeline networks system, damage usu-
ally begins on the internal surface, especially when the internal temperature is higher than the external
temperature. Micro damage, cracking of the matrix [32], delamination [33,34], Webepage [35], and
fiber failure are the damage cases that result in ultimate failure of fiber reinforced polymers (FRP)
composite structure, and damages may probably initiate and concentrate inside the pipes rather than
outside the surface. Previous studies have shown that by acquiring images from structural damage
areas (mostly outside surfaces of structures), it may be good to distinguish different health states of
structures. Images captured from the external surface of pipelines may not be an appropriate way to
identify damage. The specific purpose of current work is to identify and locate closed-circuit television
(CCTV) image pipeline failures through deep learning to overcome the limitations of conventional
computer vision techniques. Automatic detection of image faults and binary segmentation (mapping)
with very high precision were established; the current disk format can have a high degree of accuracy
in measuring the type and location of image errors, especially in mini image errors. The current
approach provides pointers for others to apply in-depth learning techniques that address similar topics,
such as identifying and locating civic infrastructure failures in building and property management.
Practical implementation of the proposed approach in industry is expected to significantly reduce
inspection time and resources and improve the efficiency of pipeline status assessment.
The present work generally relates to a new algorithm in Region-based CNN (R-CNN) family
architecture and is called Faster Dual/Multi R-CNN (Faster D/M-R-CNN) since it provides a fast
and accurate damage detection and classification for various structures and especially pipelines at real
time. This method has the potential to be used as the fastest system for damage identification in various
structures such as different infrastructures (bridges, high buildings, dams, pipelines, tanks, etc.), for
traffic control system, and transportation systems. It can also be used for image analysis and processes
occurring in smart cities and traffic control and transportation systems. Figure 2.1 shows the block dia-
gram depicting the new algorithm (Faster D/M-R-CNN) for damage detection and image classification.

FIGURE 2.1 A block diagram depicting the new algorithm (Faster D/M-R-CNN) for damage detection and
classification.
Deep Learning for Automated Damage Detection 15

2.2 DEEP LEARNING BASED APPROACHES FOR IMAGE


CLASSIFICATION AND OBJECT DETECTION
2.2.1 DEEP LEARNING FOR IMAGE CLASSIFICATION
The damage in structures has several important signs that indicate structural degradation and even
the inception of catastrophic failure. Image-based damage detection has been attempted in research
communities that bear the potential of replacing human-based inspection.
Deep learning has been widely developed and applied in various areas such as computer vision,
speech recognition, and natural language processing through various deep learning architectures,
among which CNNs are commonly applied.
As shown in Figure 2.2 a CNN model typically consists of feature extraction through a stack of
layers on the input image such as convolution, activation and poling, and classification through fully
connected (FC) layers for outputting the scores for each class.
Each layer is responsible for different functions and uses the result from the previous layer as
the input.
For supervised computer vision tasks, CNNs (1) extract features from raw images, (2) feed the
features forward using filters assigned with initial random weights and bias to predict the classes,
(3) calculate the loss between predicted scores and the ground truth, and (4) apply backpropagation
to adjust the filter weights and bias continuously to finally obtain an optimized model.
Compared with conventional approaches, CNNs require less image preprocessing, and image
features are extracted through learning. Therefore, there is no requirement of expertise for manual
design of complex feature extractors.
The region-based CNN (R-CNN) is one typical deep learning approach for object detection. As
shown in Figure 2.3, region proposals are generated through an external method called selective search
for the input image. Each warped region proposal image is forwarded into a CNN model to compute
the features, which is then fed into an SVM classifier to calculate the classification scores. Bounding
box (bbox) regression is then conducted for the classified image such that the location of each object

FIGURE 2.2 Architecture of a CNN model.

FIGURE 2.3 Example architecture of R-CNN.


16 Experimental Vibration Analysis for Civil Structures

can be predicted. One limitation of R-CNN is that the multi-staged training process which is time
consuming requires a high computation cost. In addition, the detection speed is quite slow for each
image as convolution, classification, and regression need to be implemented for each region’s proposal.
Among the many methodologies, deep learning based damage detection is actively explored in
recent years. However, how to automatically extract damage quickly and accurately at a pixel level, i.e.
damage delineation (including both detection and segmentation) at real time, is a challenging issue.
This work proposes a new algorithm in R-CNN family-based architecture that is Faster D/M-R-
CNN for real-time object (damage) detection and classification with high accuracy.
Let’s quickly summarize the different algorithms in the R-CNN family (R-CNN, Fast R-CNN,
and Faster R-CNN).
R-CNN extracts a bunch of regions from the given image using selective search, and then checks
if any of these boxes contains an object. We first extract these regions, and for each region, CNN is
used to extract specific features. Finally, these features are then used to detect objects. Unfortunately,
R-CNN becomes rather slow due to the multiple steps involved in the process.
Fast R-CNN, on the other hand, passes the entire image to ConvNet which generates regions of
interest (instead of passing the extracted regions from the image). Also, instead of using three differ-
ent models (as we saw in R-CNN), it uses a single model which extracts features from the regions,
classifies them into different classes, and returns the bboxs.
All these steps are done simultaneously, thus making it execute faster as compared to R-CNN.
Fast R-CNN is, however, not fast enough when applied on a large dataset as it also uses selective
search for extracting the regions.
Faster R-CNN fixes the problem of selective search by replacing it with Region Proposal Network
(RPN). We first extract feature maps from the input image using ConvNet and then pass those maps
through an RPN which returns object proposals. Finally, these maps are classified and the bboxs
are predicted.
A block diagram depicting the new algorithm (Faster D/M-R-CNN) for object (damage) detec-
tion and classification is shown in Figure 2.1. As shown in the figure, apply Dual/Multi RPN (D/M-
RPN) {3} on these feature maps {2}, get Dual/Multi object (damage) proposals for each candidate
object (damage) in the image, and compare between these proposals to create a confidence score
representing a confidence that the each candidate object (damage) detected in the bboxs {3C} is the
desired object (damage) meticulously. This algorithm increases the detection accuracy in short time,
without following the traditional method followed in other networks in the R-CNN family by sup-
plementing more images into the database to reduce overfitting and to improve detection accuracy.
Faster D/M-R-CNN has high precision and recall value and very high speed for extracting all
object (damage) features from images at real time, which is important for accurately detecting dam-
ages from images collected that increased previous damage detection systems ability to achieve
real-time detection.
Therefore, the Faster D/M-R-CNN will be investigated and applied in this invented product
for automated structure damage and object detection from image. We expected that the proposed
approach will be demonstrated to be applicable for detecting structure damages and object detection
with high accuracy and speed for extracting all damage features from images.
We expected this innovative product will increase the speed and accuracy of object (damage)
detection and improves dataset size, training mode, and network hyper-parameters, which have
influence on model performance.
Specifically, when an increase of dataset size and convolutional layers can improve the model
speed and accuracy, it can achieve a mean average precision (mAP) of up to 98%–99% approximately.
The new designed network will lay the foundation for applying a new generation of deep learning
techniques in structural damage detection systems as well as addressing deficiencies in the previous
structural damage detection systems based on deep learning.
The comparison between the algorithms in R-CNN family and the new algorithm described
herein is shown in Table 2.1.
Deep Learning for Automated Damage Detection 17

TABLE 2.1
Comparison between the Algorithms in R-CNN Family and the New Algorithm
(Faster D/M-R-CNN) Described Herein Using Damage Detection and Classification
Prediction
Algorithm Features Time/Image Limitations
CNN Divides the image into multiple regions and then – Needs a lot of regions to predict
classifies each region into various classes accurately and hence high
computation time
R-CNN Uses selective search to generate regions. Extracts 40–50 High computation time as each
around 2,000 regions from each image seconds region is passed to the CNN
separately. Also, it uses three
different models for making
predictions
Faster Each image is passed only once to the CNN and feature 2 seconds Selective search is slow and hence
R-CNN maps are extracted. Selective search is used on these computation time is still high
maps to generate predictions. Combines all the three
models used in R-CNN together
Faster Replaces the selective search method with RPN, which 0.2 seconds Object proposal takes time and as
R-CNN makes the algorithm much faster there are different systems
working one after the other, the
performance of systems depends
on how the previous system has
performed
Faster Applied D/M-RPN to get Dual/Multi object (damage) 0.1–0.2 Needs to obtain an optimum
D/M-R- proposals for each candidate object in the same image seconds number of RPN to investigate
CNN and compare between these proposals to result in the the target
desired object, which makes the algorithm with the
highest accuracy and the fastest

2.3 METHODOLOGY
As shown in Figure 2.1, the steps followed by a Faster D/M-R-CNN algorithm to detect objects
(damage) in various structure image are summarized below:

1. Take an input image {1} and pass it to the Deep-CNN {2A} which returns feature maps {2}
for the image.
2. Apply D/M-RPN {3} on these feature maps {2} and get Dual/Multi object (damage) pro-
posals for each candidate object (damage) in the image.
3. Compare between the Dual/Multi object (damage) proposals to create a confidence score
representing a confidence that the object (damage) detected in the bboxs {3C} is the desired
object (damage).
4. Pass these proposals to an FC {3B} layer in order to classify and predict the bboxs {3C} for
the image.
5. Finally, receive the final image results {4} and compute a confidence score associated with
it. The computing device can then provide a requestor with an output including object clas-
sification and/or confidence score.

The flowchart of the new algorithm process steps of the Faster D/M-R-CNN is introduced in Figure 2.4.
To ensure the accuracy of the object (damage) detection and classification network, the D/M-
RPN {3} can be trained with training images. A discussion of the object (damage) detection and
classification network training will be discussed in greater detail with regard to Figure 2.5.
18 Experimental Vibration Analysis for Civil Structures

FIGURE 2.4 A flow diagram that illustrates a process flow for damage detection and classification network.

The Faster D/M-R-CNN training process consists of four steps. The first step is training a D/M-
RPN {3} initialized from pre-trained Deep-CNN {2A}, in this case generating the Dual/Multi object
(damage) proposals using the trained D/M-RPN {3}. The second step is training a Faster D/M-R-
CNN initialized from the pre-trained Deep-CNN on the Dual/Multi object (damage) obtained by
the first step. In the third step, the D/M-RPN {3} is trained again using the weights from step two as
shown in Figure 2.4, without changing the Deep-CNN {2A} layers.
After generating new Dual/Multi object (damage) proposals from the trained D/M-RPN {3},
Faster D/M-R-CNN is trained again in step four with the parameters trained in the previous step.
The classification process is presented in Figure 2.6. A series of images from the image of a
single target (i.e., as a sequence of temporally consecutive frames of the structures damage types)
are fed to the D/M-CNN that is applied to feature extraction.
The two-step learning method is applied, i.e., the D/M-CNN is trained with the first N − 1 layers
viewed as feature maps, and these maps are used to train a Dual/Multi SVM (D/M-SVM) classi-
fier. The output of the SVM from each CNN is compared with them and all damage features from
images with high accuracy are collected and presented as P-Tensors as follows:

⎡ c(1,1) c(1, j )nc ⎤


⎢ 1

P(i , j )k = ⎢ ⎥, k = 1, , n, (2.1)
⎢ c i,1 c(i , j )nc ⎥
⎢⎣ ( )1 ⎥⎦
Deep Learning for Automated Damage Detection 19

FIGURE 2.5 A flow diagram that illustrates a process flow of training one or more parameters of an object
(damage) detection network.
20 Experimental Vibration Analysis for Civil Structures

FIGURE 2.6 Data flow diagram depicting an example architecture of the process of the damage
classification network described in Figure 2.1.

where ci , j is the probability for class ( i, j ), nc is the number of classes, and n is the number of images
of training examples; thus there will be one P-Tensor for each image in any given image series.
The composition of the tensor is the following: calculate the average of the size estimates of the
image series, check from the size-look-up table all the classes that contain the average size, e, and
turn those elements to one and set the others to zero, yielding

⎡ e(1,1) e(1, j )nc ⎤


⎢ 1

E(i , j ) = ⎢ ⎥ (2.2)
⎢ e i,1 e(i , j )nc ⎥
⎢⎣ ( )1 ⎥⎦

With elements:

⎧⎪ 1 if e fitsclass i, j,
e(i , j ) = ⎨ (2.3)
⎩⎪ 0 otherwise.

The velocity of the target damage type is composed in a similar way as the E-Tensor in Size Estimate
(2.2), i.e., check from the velocity-look-up table all the classes that contain the provided velocity, v,
and turn those elements to one and the others to zero.

⎡ v(1,1) v(1, j )nc ⎤


⎢ 1

V(i , j ) = ⎢ ⎥ (2.4)
⎢ v i ,1 v(i , j )nc ⎥
( )1
⎣⎢ ⎥⎦
Another random document with
no related content on Scribd:
Vater Assurrisisi, wie Andrae nachweisen konnte, den Grundstein zu
dem Tempel gelegt hat.

Rekonstruktion des Tempels des Assurrisisi (nach Andrae).


Rekonstruktion des Tempels Salmanassars II. (nach Andrae).
Von dem jüngeren Tempel ist fast die ganze Nordseite
verschwunden, von andern Teilen aber sogar der Überbau erhalten.
Zweihundertundfünfzig Jahre nach Tiglat-Pileser I. ließ Salmanassar
II. die jedenfalls noch bedeutenden Reste des alten Tempels bis auf
fünf Meter über dem Felsengrund abtragen und errichtete nun auf
diesem eingeebneten Grund den Neubau, der an Umfang kleiner,
aber noch immer sehr groß war. Eine hier gefundene Urkunde
berichtet, das Dach habe aus Zedernholzbalken bestanden; eine
andere ist in Basalt eingehauen und lautet: „Salmanassar, der
mächtige König, der König des Alls, der König des Landes Assur,
der Sohn Assurnasirpals, des Königs des Landes Assur, des Sohnes
von Tukulti Ninib, dem Könige des Landes Assur, der Erbauer des
Tempels des Gottes Anu, des Tempels des Gottes Adad“.
Salmanassars Brunnen auf dem Hofe ist von einer Ringmauer
umgeben und bis zum Grundwasser 29,5 Meter tief. Unter den
Tempelmauern fand man Beile und Schwerter, die wahrscheinlich
symbolische Bedeutung hatten. Nach dem goldenen Blitz zu urteilen,
den Adad in der Hand hielt, müssen die Götterbilder prächtig
ausgestattet gewesen sein.

Phot.: Andrae.
Der Tigris bei Assur.
Neun Jahre verwandten Andrae und seine Mitarbeiter auf die
Freilegung der Festungswerke Assurs (vgl. „Die Festungswerke von
Assur“ von Walter Andrae. 2 Bde., 1913), denn ihre Bestimmung war
nicht nur wichtig für die Erkenntnis der ganzen Stadtanlage, für
Ermittlung ihrer Zugänge und zugleich des Verlaufs der wichtigsten
Handelsstraßen, sondern auch für die Geschichte der
Befestigungskunde überhaupt, da man assyrische Festungen bis
dahin nur ungenügend kannte. Die Arbeit war um so schwerer, als
die Mauern am Rande des Stadthügels naturgemäß am meisten der
Vernichtung ausgesetzt waren, und obendrein der Tigris den größten
Teil der Ostfront zerstört hatte.
Assur liegt auf der Spitze eines Ausläufers der Chanukekette,
und der Platz war für eine Festung wie geschaffen. Im Osten
bespülte ihn der schnellfließende, das ganze Jahr über tiefe Tigris,
ein Angriff von dort war also unmöglich. Im Norden fiel der Fels
(weicher Sandstein und Kieselkonglomerat) jäh nach einem
Stromarm ab, der trefflich als Festungsgraben diente. Am Rande
dieses noch erkennbaren Flußbettes hatten wir unser Lager
aufgeschlagen. Vor der Westfront erleichterten zwei kleine Täler die
Anlage von Gräben, die nur da zugeschüttet waren, wo Straßen zu
den Toren führten. Im Süden war eine Geländesenkung. Der einzige
Nachteil war, daß man von dem Hügelplateau im Westen aus in die
Stadt hineinsehen konnte. Deshalb baute man die Westmauer am
höchsten.
Das Alter der Festungsbauten Assurs ist sehr verschieden.
Andrae unterscheidet die archaische Zeit bis zur Mitte des 2.
vorchristlichen Jahrtausends, die altassyrische bis Ende des 2.
Jahrtausends, die jungassyrische vom Anfang des 1. Jahrtausends
bis Sargon, die spätassyrische unter Sargon und den Sargoniten bis
zur Zerstörung des assyrischen Reiches im Jahre 606 v. Chr., die
nachassyrische der Wiedereinwanderung unter den Neubabyloniern
und Cyrus (6. Jahrhundert) und die parthische Zeit, die ersten zwei
Jahrhunderte vor und nach Christus.
Phot.: Andrae.
Der Strand von Assur.
Aus vorgeschichtlicher Zeit haben sich keine Befestigungen
gefunden, nur Grundmauern von Häusern, Feuerstätten und
Kanälen. An der Ostseite führte man schon zu Anfang des 2.
Jahrtausends Mauern auf, um die Stromfahrt zu beherrschen und
das Ufer gegen die Erosion zu schützen. Diese Mauern befestigte
Adadnirari I. in altassyrischer Zeit. Davon ist noch vieles erhalten.
Auch legte man Landeplätze und Treppen am Ufer an. An der
Nordostecke der Stadt lag der Assurtempel mit der Front nach
Norden, und an der Nordwestecke der Palast Tukulti-Ninibs I. auf
seiner ungeheuren Plattform.
In jungassyrischer Zeit baute Salmanassar III. im Westen und
Südwesten eine äußere und eine innere senkrechte Mauer, auf
denen je eine Fahrstraße hinlief. Blaugelbe und schwarzweiß
glasierte Ziegel schmückten die Zinnen. Nach seiner Regierung,
aber vor Sargon, verfiel die innere Mauer; an ihrer Stelle entstanden
Wohnhäuser, und davor legte man eine niedrigere Mauer an.
In spätassyrischer Zeit führten Sargon und Sanherib noch
mancherlei Verbesserungen aus. Die Achämeniden dagegen ließen
die Befestigungen unverändert, und auch in der parthischen Zeit
wurde nichts daran getan.
Aus der Zeit Salmanassars III. grub man sieben Tore aus. Jedes
Tor flankierten zwei Türme; nach innen waren Wachtstuben,
Rampen und Treppen, die zur Mauerzinne hinaufführten. Eines der
Tore hieß Abul gurgurri, das Stadttor der Metallarbeiter; die übrigen
sind bisher namenlos. Die Zapfen der gewaltigen Flügel des
Gurgurritores, zylinderförmige Basaltblöcke, sind noch vorhanden.
Verkohlte Zedernholzbalken lassen auf eine Feuersbrunst schließen.
Zwei sargonitische Kalksteinblöcke an diesem Tor tragen Sanheribs
Namen. Eine Bildsäule Salmanassars III., die im Gurgurritor stand,
besitzt jetzt das Ottomanische Museum in Konstantinopel, eine
andere, die den König auf seinem Throne sitzend darstellt, das
Britische Museum. Beide sind in Lebensgröße.
Der ebenfalls ausgegrabene, offenere Zugang von Norden her,
den man nicht als Tor bezeichnen kann, hieß Muschlal und wird
schon Ende des 3. Jahrtausends auf Ziegelinschriften erwähnt.
Adadnirari I. in altassyrischer Zeit ließ ihn erneuern; auch die
jungassyrische Zeit unter Salmanassar III. kennt ihn. Unter Sanherib
heißt es: „Der Palast Muschlal in der Stadt Assur“, und bei
Assarhaddon: „Bît muslalu, das am Palast der Stadt Assur liegt, ließ
ich aufs neue erbauen als Ein- und Ausgang“.
Straßen an der inneren Mauer stammen aus spätassyrischer
Zeit; sie erinnern an die Straßen Pompejis und der heutigen Städte
des Orients.
Professor Andrae beschreibt ausführlich alle Einzelfunde, die an
den Mauern gemacht wurden, Ziegelkanäle, Straßen und Häuser,
Abflußtrommeln, Poternen, Wehrgänge, Turmtreppen, Bastionen,
emaillierte Terrakottareliefs, Kupferbecken, Konsolen, Haken und
Pfeilspitzen aus Bronze, Gräber und Ziegel mit Inschriften, von
denen folgende aus der Zeit Salmanassars III. als Probe angeführt
sei: „Salmanassar, der König des Alls, König des Landes Assur, der
Sohn des Assurnasirpal, des Königs des Landes Assur. Erobernd
herrschte ich vom großen Meer beim Lande Amurru gegen
Sonnenuntergang bis zum Meer beim Lande Kaldu, genannt Marratu
(d. h. der Salzstrom). Da brach ich die Ruinen der früheren
Festungsmauer meiner Stadt Assur nieder, die Tukulti-Ninib,
Salmanassars Sohn, ehedem gebaut hatte; ich erreichte ihren
Grund; von ihrem Fundament bis zu ihrer Brustwehr fügte und
vollendete ich sie; prächtiger und gewaltiger als zuvor machte ich
sie. Meine Tafeln und Urkunden brachte ich an. Ein zukünftiger Fürst
soll ihre Ruinen wieder aufrichten und meinem Namen wieder seinen
Platz einräumen, dann wird Assur seine Gebete erhören.“
Salmanassar gedachte also der kommenden Jahrtausende, die
seinen Namen vergessen könnten. Dann sollten die Steine für ihn
reden!
Andraes Grabungsmethode in den Ruinen Assurs war eine
andere als die Koldeweys in Babylon. Er zog 5 Meter breite
„Suchgräben“ quer über das ganze Stadtgebiet; sie laufen je 100
Meter voneinander entfernt parallel von der Westmauer bis nach
dem Tigrisufer im Osten. Stieß solch ein Graben auf Reste von
Palästen, Mauern, Toren, Häusern, Kanälen usw., so grub man
seitwärts weiter, bis der ganze Fund bloßgelegt war. Manchmal
zwangen Bodengestaltung oder neuere mohammedanische
Grabstellen zur Aufgabe des 100-Meter-Zwischenraums. Solch ein
unregelmäßiger Graben führte in den Jahren 1909–1911 zur
Entdeckung der merkwürdigen Königspfeiler im Winkel zwischen
dem breiten Nordteil und dem schmalen Südteil der Stadt. (Vgl. „Die
Stelenreihen in Assur“ von Walter Andrae, Leipzig, 1913).
Diese Pfeiler stammen aus der Zeit zwischen dem 14. und 7.
Jahrhundert v. Chr. Sie sind flach, oben abgerundet und tragen eine
Inschrift oder ein Reliefbild der Personen, zu deren Gedächtnis sie
errichtet wurden. Die größten sind aus Basalt und nennen Tukulti-
Ninib I., Semiramis und Assurnasirpal III.; kleinere sind mit den
Namen anderer Könige und hoher Beamten bezeichnet. Einer aus
körnigem, gelbgrauem Kalkstein zeigt das Bild einer Palastdame
Sardanapals. Sie sitzt, nach rechts gewendet, auf einem Thron, ist
mit Armbändern und Ohrringen geschmückt, trägt Rosetten auf den
Schultern und auf ihren üppigen, den Rücken herabwallenden
Locken eine Königskrone; in der Linken hält sie eine Blume, die
Rechte streckt sie nach oben. Das Gesicht entspricht dem
Schönheitsideal des Orients: volle runde Wangen, kräftiges Kinn,
gerade, scharf gezeichnete Nase, schön geschwungene, breite
Augenbrauen und lachende Lippen.
Phot.: Schölvinck.
Ein Suchgraben in Assur.

Eines dieser Denkmäler nennt Andrae den „Kalksteinpfeiler der


Semiramis“. Auf ihm stehen die Worte: „Pfeiler für Sammuramat, die
Palastdame Samsi Adads, des Königs des Alls, des Königs von
Assur, die Mutter des Adadnirari, des Königs des Alls, des Königs
von Assur; die Schwiegertochter Salmanassars, des Königs der vier
Weltgegenden.“ Die in dieser Inschrift genannten Könige sind
Salmanassar III., Samsi Adad V. und Adadnirari IV., die zwischen
858 und 781 regierten. Nach einem Schlummer, der vor der
Gründung Roms begann, hat also der Forscher diese Semiramis von
den Toten erweckt und ihre Verwandtschaft mit drei Königen
festgestellt. Von ihrem sonstigen Schicksal aber wußte der Stein
nichts zu melden. —
So weit war Andraes Werk über die Ausgrabungsergebnisse von
Assur gediehen, als der Weltkrieg auch dieses stolze Denkmal
deutscher wissenschaftlicher Forschung einstweilen zu einem Torso
machte.

Phot.: Schölvinck.
Das deutsche Expeditionshaus in Assur.

Wir besuchten auch das Haus der deutschen Archäologen am


Ufer, ein großes, festes Gebäude, auf dessen Innenhof, wie üblich,
eine offene Galerie hinausging. Am Tor trat bei unserer Ankunft die
Wache ins Gewehr. Dann empfing uns ein alter Türke, erzählte von
der Zeit, als Andrae, Jordan, Lührs und Bachmann hier arbeiteten,
und zeigte uns die Zimmer, die jeder von ihnen bewohnt hatte. Jetzt
lagen darin vierzehn englische Soldaten und eine Anzahl
schwerkranker Inder, die auf dem Wege in die Gefangenschaft
zusammengebrochen waren.
Später unterhielt ich mich in der Nähe des deutschen Hauses mit
einem dieser Patienten, der sich bereits wieder völlig erholt hatte. Es
war ein Hindusoldat, der fließend Englisch sprach und den Anflug
europäischer Bildung sehr geschickt mit seiner ursprünglichen
orientalischen Weisheit zu verbinden wußte. Auch er hatte anfangs
den Versicherungen der englischen Offiziere geglaubt, Deutschland
habe aus Raubsucht den Krieg begonnen und die Türkei
gezwungen, mitzumachen; die Deutschen seien Barbaren und die
Feinde der Menschheit, die Kosaken aber die Herolde der
Zivilisation; Deutschland sei schon so gut wie vernichtet, und die
Türkei verdiene eine exemplarische Strafe, da sie die Geschäfte der
Deutschen besorge. Das alte Lied! Seit dem Fall von Kut-el-Amara
war aber dem Hindusoldaten ein Licht aufgegangen. Von der
Barbarei der Deutschen und Türken hatte er nichts gemerkt, er war
im Gegenteil als Gefangener von den Türken gut behandelt worden.
Auch die Bedrohung Indiens durch Deutschland hatte er als
Schwindel erkannt, und er wußte nun, daß man ihn gezwungen
hatte, für ein Land zu kämpfen, dessen Geschick ihn nichts anging
und für das sich zu opfern er und alle seine Landsleute wahrlich
nicht die geringste Veranlassung hatten.
Phot.: Schölvinck.
Assur von Norden gesehen. Unser Lager bei Kalaat-Schergat.
Der Inder bat mich noch, ihn in meinen Dienst zu nehmen, in der
Hoffnung, dann früher wieder in seine Heimat kommen zu können.
Diesen Wunsch mußte ich ihm natürlich versagen, und er kehrte
wieder zu den türkischen Wächtern zurück, die ihre Schützlinge frei
umherstreifen ließen. Ich mußte dieser Begegnung noch oft
gedenken. Die angeblichen Vertreter der Zivilisation und des
Christentums führten die Orientalen gegeneinander ins Feld und
brachten sie, was noch schlimmer ist, nach Europa, um gegen
Christen zu fechten. Diese Saat Englands wird noch einmal furchtbar
in die Halme schießen! Dann werden Männer wie dieser Hindusoldat
und viele Tausende seiner Leidensgefährten, die nun wissen, wie
man sie betrog, ihren Landsleuten vieles zu erzählen haben!
Am Abend lagen die Hügel von Assur in bleichem Mondschein.
Im Lager verklang der Lärm, die Unterhaltung verstummte, und die
Steppe schlief. Die lautlose Stille war unheimlich. Leise Schritte
umschlichen unsere Betten — waren es Gespenster, die die
Mitternachtsstunde aus den tausendjährigen Gräbern der Ruinen
ringsum heraufbeschwor? Sollten all der Glanz und die Herrlichkeit
der alten Königsstadt doch noch nicht so ganz verschollen sein und
nächtlicher Weile eine geisterhafte Wiederauferstehung feiern?
In den Ruinen regte sich nichts. Was mich aus dem Halbschlaf
emporschreckte, waren hungrige Hunde aus der Nachbarschaft, die
in unserm Lager nach Beute suchten.
Phot.: Schölvinck.
Unser großes Zelt bei Schergat-hauesi.

Einundzwanzigstes Kapitel.
Erlebnisse auf einer
Etappenstraße.

E s war eine sehr stattliche Karawane, die am Morgen des 7.


Junis, in dünne Staubwolken gehüllt, von Assur nordwärts zog:
an der Spitze Konsul Schünemann mit den persischen Reitern, dann
unsere Wagen mit türkischen, arabischen und armenischen
Kutschern, und als Nachhut 25 Esel, die wir von Arabern gemietet
hatten, um unsere Gepäckwagen zu entlasten. Doch bedurften
unsere Zugtiere so sehr der Schonung, daß wir beschlossen, heute
nur bis Schergat-hauesi zu marschieren, das kaum eine Stunde
entfernt war. Wir hatten also reichlich Zeit und folgten daher der
Einladung eines Araberhäuptlings vom Hedschadschstamm, der
eine Strecke weiter unten am Ufer des Tigris sein Lager hatte.
Die Dörfer der Hedschadscharaber in der Nähe von Schergat-
hauesi zählten gegen 100 Zelte mit je 8–10 Bewohnern. Auch bei
Kalaat-Schergat und an einigen andern Plätzen bildeten sie kleine
Gemeinden Sie gelten schon als Fellachen oder Ackerbauer, haben
aber die Liebe der Nomaden zu ihren schwarzen Zelten noch nicht
abgestreift.
An Ort und Stelle angelangt, fanden wir ein prächtiges schwarzes
Zelt mit Teppichen für uns hergerichtet, und in der luftigen Wohnung
des Häuptlings wurden wir feierlich empfangen, mußten auf dem
Ehrenplatz niedersitzen und uns von mindestens hundert seiner
Stammesgenossen anstaunen lassen, der Frauen und Kinder nicht
zu gedenken, die von draußen hereinsahen. Der Zweck der
Einladung war hauptsächlich ein Geschäft, das der Araber mit dem
Herzog machen wollte; er hatte einen weißen Hengst zu verkaufen,
der uns als Ersatz für die gehabten Verluste sehr willkommen
erschien.
Das Tier wurde vorgeführt und war prächtig anzusehen: ein
herrlich gerundeter Hals, sprühende schwarze Augen, eine zarte
Nase mit schnaubenden Nüstern — ganz das Urbild des feurigen
Arabers, eine Freude für jeden Pferdekenner, und der Hengst selbst
schien sich seines verführerischen Reizes vollkommen bewußt, so
elastisch-kokett tänzelte er daher und schien den Boden kaum mit
den Hufen zu berühren. Warum mochte nur der Besitzer sich von
solch einem Prachtpferd trennen wollen? Dieses Übermaß von
Liebenswürdigkeit gegen uns Fremde schien verdächtig.
Der Herzog bat also den Mann, aufzusitzen und in Schritt, Trab
und Galopp die Künste des Tieres zu zeigen. Dazu zeigte unser
freundlicher Wirt aber keine Lust, und auch seine Stammesbrüder
drückten sich einer nach dem andern mit Redensarten beiseite.
Nach langem Hin und Her fand sich endlich ein junger Kerl für ein
Trinkgeld zu einem Versuch bereit. Aber kaum saß er auf, als der
Hengst ihn auch schon zu Boden geschleudert hatte, und mehreren
andern, die der Backschisch reizte, erging es nicht besser ... Nun
wagte sich einer unserer Perser heran, und er schien die Kunst zu
verstehen, diesen Bucephalus zu zähmen. Zur höchsten
Verwunderung der Araber trabte er auf dem widerspenstigen Gaul in
eleganten Kurven zwischen den Zelten kreuz und quer umher und
kehrte dann im Galopp zu uns zurück. Aber in diesem Augenblick
stand das Pferd ganz plötzlich, stemmte die Vorderbeine in die Erde,
senkte den Kopf und warf seinen Reiter in großem Bogen mitten in
die Zuschauer hinein. Soviel war nun sicher, daß sich das schöne
Tier zu Karawanenfahrten nicht eignete, und aus dem Handel wurde
infolgedessen nichts.

Zu Besuch bei den Arabern von Nalaat-Schergat.


In der Mitte der Herzog.
Gleichwohl blieben wir den Tag über bei den Arabern. Als wir am
Abend im Mondschein vor unserm Zelt saßen, ließ sich einer von
ihnen bei uns nieder, um uns mit seiner Maultrommel zu unterhalten.
Die Töne des einfachen Instruments erinnerten an die des
Dudelsacks, aber die Kunstfertigkeit des Mannes war
bewundernswert, und nicht weniger seine Ausdauer, er schien gar
nicht Atem holen zu müssen. Von Melodie konnte man kaum reden.
Melancholisch und einförmig, wie immer im Orient, quollen zwischen
seinen glockenförmig gehaltenen Händen langgezogene,
wimmernde Töne hervor, die ein träumerisches Behagen erweckten.
Man hörte in dieser Musik die trippelnden Schritte der Schafe über
die Steppe, den Hufschlag der Beduinenpferde, das traurige Flüstern
des Windes im Grase und das rieselnde Rauschen des Tigris gegen
eine Landspitze. Das ganze einförmige Leben der Araber in der
Wüste, in der ein Tag verläuft wie der andere, schien in dieser
Naturmusik lebendig zu werden.

Der edle arabische Hengst.

Die Töne der Maultrommel hatten zahlreiche Zuschauer


angelockt. In dichten, dunkeln Gruppen ließen sie sich mit Anstand
und Würde auf dem Boden nieder und hörten lautlos zu, höchstens
flüsterten sie leise oder rauchten Zigaretten. Als dann aber der
Musikant zum Tanz aufspielte, kam Leben in die Masse; etwa fünfzig
Araber sprangen auf, faßten sich in einer langen Kette, aber immer
eng aneinandergedrängt, an den Händen und begannen sich in
einer bärenmäßig trottenden Gangart zu bewegen, erst einige
Schritte nach rechts, dann nach links, ganz im Takt mit dem
wimmernden Rhythmus der Flöte. Allmählich weitete sich der Kreis,
die Schnelligkeit nahm zu und wurde immer stürmischer, wobei der
Mondschein die hohen, dunkeln Gestalten, die flatternden braunen
Burnusse und die weißen und bunten Kopftücher noch
phantastischer erscheinen ließ als sonst.
Mein Diener Sale hatte beim Tanz das Kommando übernommen.
Er hieß nun die Tänzer sich in großem Kreise niederhocken, und in
die Mitte des freien Platzes traten zwei Solotänzer, die sich in einer
Art Bauchtanz, nicht eben schön anzusehen, aufeinanderzu
bewegten. Der eine stellte den Verfolger dar, der andere den
Verfolgten. So jagten sie sich mehrmals im Kreise herum, und die
Zuschauer verfolgten die Pantomime mit größter Aufmerksamkeit
und mit taktmäßigem Händeklatschen. Als schließlich der eine
Tänzer den andern einholte und sich über ihn warf, brach
allgemeiner Jubel los.
Die nächste Programmnummer war der Schwertertanz, wobei die
Klingen dumpf aufeinander rasselten. Auch dieses Spiel endete
damit, daß einer den andern übermannte; der Sieger setzte einen
Fuß auf den Besiegten und die Spitze des Schwertes auf seine
Brust. Hinter dem Kreis der Sitzenden hatten sich dichte Reihen
stehender Zuschauer gesammelt, und in diesem Rahmen gewann
das Schauspiel noch an Ursprünglichkeit und phantastischem Reiz.
Am nächsten Morgen verließen wir die Nomaden. Als wir unsere
Straße erreicht hatten, war schon wieder ein Lastwagen in Stücke
gegangen. Von links trat nun ein Ausläufer der großen
Kalksteinplatte Mesopotamiens an den Tigris heran und fiel steil zum
Wasser ab. Die Straße wand sich daher auf das Plateau hinauf,
dessen harter, ebener Boden unsern Pferden eine willkommene
Erleichterung brachte. Dann aber zwang uns ein Wadi mit einem
rieselnden Salzwasserbach, wieder in das Flachland
zurückzukehren, wo große Schaf-, Ziegen- und Rinderherden
weideten und ausgedehnte Zeltlager den blauen und weißen Rauch
ihrer Feuer in die Morgenluft emporsandten. Wieder stieg die Straße
an. Wir kreuzten einen Zug Heuschrecken, der einen schmalen
Randstreifen besetzt hielt, fuhren an einer Karawane von 400
Kamelen vorüber, deren Last nach Aussage ihrer Führer aus
Uniformen bestand, näherten uns wieder dem Tigrisufer und hielten
bei der Station Giara oder Tell-Kaischara, wo uns ein starker Geruch
von Naphtha und Asphalt entgegenströmte. Hier tritt das Erdpech
offen zutage. Vor mehreren Jahren arbeitete hier eine belgische
Gesellschaft mit gutem Erfolg. Nach einiger Zeit meinten aber die
Türken, den Gewinn besser selbst einstreichen zu sollen, und
kündigten die Konzession. Nun fehlte die geschäftliche Erfahrung,
und das Unternehmen geriet ins Stocken; Häuser und Maschinen
verfielen, und das Erdpech stank in Tümpeln stagnierenden Wassers
zum Himmel.

Giara.
Das Bahnhofsgebäude von Giara hatte nur einen bewohnbaren
Raum, eine ungewöhnlich kühle, gewölbte Kammer, in der der
Stationsvorsteher unter einem von Fliegen umschwirrten
Mückennetz an Ruhr erkrankt darniederlag und aus einem primitiven
Filtrierapparat, einem großen Lehmkrug mit porösem Boden, Wasser
tropfen ließ. Hier mußten wir die heißesten Tagesstunden abwarten,
denn die Temperatur draußen war allmählich unerträglich geworden.
Schon morgens um 7 Uhr hatte sie 31 Grad betragen, um 1 Uhr
stieg sie auf 41,2 und anderthalb Stunden später auf 42,6 Grad.
Konsul Schünemanns persischer Schimmel hatte einen Hitzschlag
und Kolik und außerdem Blutegel in Gaumen und Hals. Noch am
Morgen war das Tier ganz frisch gewesen; jetzt legte es sich im
Schatten des Stalles nieder und verendete. Auch im Schlund der
andern Pferde hatten sich beim Trinken Blutegel festgebissen, und
unsere Kutscher befreiten sie mit vieler Mühe von diesen
Plagegeistern.
In der Kranken- und Fliegenstube von Giara zu übernachten, war
unmöglich. Am Spätnachmittag machten wir uns daher zur nächsten
Station Schura auf, die fünf Stunden entfernt sein sollte. Nahe bei
Giara hatten wir ein ziemlich tief und steil eingeschnittenes Tal zu
passieren, auf dessen nackter Sohle Salzkristalle schimmerten und
Erdpechquellen zutage traten. Der Herzog und Busse ritten voraus;
Schölvinck und ich folgten in der Droschke und fuhren in einer
Morastrinne fest. Die Pferde mußten ausgespannt, der Wagen
zurückgeschoben und ein anderer Weg versucht werden. Nicht
besser erging es dem vorausfahrenden Automobil, das weiter vorn in
einem Graben saß und nicht weiter konnte. Wir luden das Gepäck
ab, aber der Wagen rührte sich nicht vom Fleck, und wir mußten
warten, bis die ganze übrige Kolonne nachgekommen war. Darüber
wurde es dunkel, und im Westen erhob sich drohend eine
Wolkenwand, die den Mond verdeckte. Nach langem Warten kamen
die andern, und mit vereinten Kräften machten wir erst das Auto
wieder flott, das nunmehr jeden einzelnen Wagen über die
schwierige Stelle hinüberziehen mußte; die müden Tiere allein
hätten das nicht fertiggebracht. Drei Stunden kostete uns dieser
Graben — eine schöne Etappenstraße!
Dann ging es weiter, Stunde auf Stunde in stockfinsterer Nacht;
die Lampen des Autos wiesen den Weg. Endlich leuchtete vor uns
der Schein eines Feuers auf: es war Schura, aber noch in weiter
Ferne. Ein neuer Graben hielt die Wagen auf; unsere Droschke kam
glücklich hinüber, und endlich tauchte die hohe Mauer des
Stationsgebäudes aus dem Dunkel hervor. Hastig aßen wir auf dem

You might also like