Current damage detection methods based on model updating and sensitivity Jacobian matrixes show a low convergence ratio and computational efficiency for online calculations. The aim of this paper is to construct a real-time automated damage detection method by developing a theory-assisted adaptive mutiagent twin delayed deep deterministic (TA2-MATD3) policy gradient algorithm. First, the theoretical framework of reinforcement-learning-driven damage detection is established. To address the disadvantages of traditional mutiagent twin delayed deep deterministic (MATD3) method, the theory-assisted mechanism and the adaptive experience playback mechanism are introduced. Moreover, a historical residential house built in 1889 was taken as an example, using its 12-month structural health monitoring data. TA2-MATD3 was compared with existing damage detection methods in terms of the convergence ratio, online computing efficiency, and damage detection accuracy. The results show that the computational efficiency of TA2-MATD3 is approximately 117–160 times that of the traditional methods. The convergence ratio of damage detection on the training set is approximately 97%, and that on the test set is in the range of 86.2%–91.9%. In addition, the main apparent damages found in the field survey were identified by TA2-MATD3. The results indicate that the proposed method can significantly improve the online computing efficiency and damage detection accuracy. This research can provide novel perspectives for the use of reinforcement learning methods to conduct damage detection in online structural health monitoring.
BaoY, ChenZ, WeiS, XuY, TangZ, LiH. The state of the art of data science and engineering in structural health monitoring. Engineering2019;5(2):234‒42. . 10.1016/j.eng.2018.11.027
WangJ, YouH, QiX, YangN. BIM-based structural health monitoring and early warning for heritage timber structures. Autom Construct2022;144:104618.
[4]
YuY, HoshyarAN, SamaliB, ZhangG, RashidiM, MohammadiM. Corrosion and coating defect assessment of coal handling and preparation plants (CHPP) using an ensemble of deep convolutional neural networks and decision-level data fusion. Neural Comput Appl2023;35(25):18697‒718. . 10.1007/s00521-023-08699-3
[5]
JiangT, LiL, SamaliB, YuY, HuangK, YanW, et al. Lightweight object detection network for multi-damage recognition of concrete bridges in complex environments. Comput Aided Civ Infrastruct Eng2024;39(23):3646‒65. . 10.1111/mice.13219
[6]
WangL, YiS, YuY, GaoC, SamaliB. Automated ultrasonic-based diagnosis of concrete compressive damage amidst temperature variations utilizing deep learning. Mech Syst Signal Process2024;221:111719. . 10.1016/j.ymssp.2024.111719
[7]
LiD, ZhangJ. Finite element model updating through derivative-free optimization algorithm. Mech Syst Signal Process2023;185:109726. . 10.1016/j.ymssp.2022.109726
[8]
JiangY, LiY, WangS, XuM. A novel multistage approach for structural model updating based on sensitivity ranking. Smart Struct Syst2020;25:657‒68.
[9]
ZhengB, YuK, LiuS, ZhaoR. Interval model updating using universal grey mathematics and Gaussian process regression model. Mech Syst Signal Process2020;141:106455. . 10.1016/j.ymssp.2019.106455
[10]
SmithMJ, HuttonSG. Frequency modification using newton method and inverse iteration eigenvector updating. AIAA J 1992;30(7):1886‒91. . 10.2514/3.11151
[11]
ZárateBA, CaicedoJM. Finite element model updating: multiple alternatives. Eng Struct2008;30(12):3724‒30. . 10.1016/j.engstruct.2008.06.012
[12]
LiXY, LawSS. Adaptive Tikhonov regularization for damage detection based on nonlinear model updating. Mech Syst Signal Process2010;24(6):1646‒64. . 10.1016/j.ymssp.2010.02.006
[13]
ZhangCW, ChunQ, LengJW, LinYJ, QianYC, CaoG, et al. Optimal placement method of multi-objective and multi-type sensors for courtyard-style timber historical buildings based on meta-genetic algorithm. Struct Health Monit2024;23(3):1468‒97. . 10.1177/14759217231181724
[14]
Rezaiee-PajandM, EntezamiA, SarmadiH. A sensitivity-based finite element model updating based on unconstrained optimization problem and regularized solution methods. Struct Contr Health Monit2020;27(5):e2481. . 10.1002/stc.2481
[15]
CaoZ, FeiQ, JiangD, KapaniaRK, WuS, JinH. A sensitivity-based nonlinear finite element model updating method for nonlinear engineering structures. Appl Math Model2021;100:632‒55. . 10.1016/j.apm.2021.07.034
[16]
CaoZ, WeiH, LiangD, JiaZ, YaoJ, JiangD. A non-intrusive dynamic sensitivity-based substructure model updating method for nonlinear systems. Int J Mech Sci2023;248:108218. . 10.1016/j.ijmecsci.2023.108218
[17]
EsfandiariA, SanayeiM. More insight on function-weighted frequency response function sensitivity method for analytical model updating. J Sound Vibrat2021;509:116143. . 10.1016/j.jsv.2021.116143
[18]
ZhuH, LiJ, TianW, WengS, PengY, ZhangZ, et al. An enhanced substructure-based response sensitivity method for finite element model updating of large-scale structures. Mech Syst Signal Process2021;154:107359. . 10.1016/j.ymssp.2020.107359
[19]
ŞimşekS, KahyaV, AdıyamanG, ToğanV. Damage detection in anisotropic-laminated composite beams based on incomplete modal data and teaching‒learning-based optimization. Struct Multidiscipl Optim2022;65 (11):332. . 10.1007/s00158-022-03421-8
[20]
MousaviZ, VarahramS, EttefaghMM, SadeghiMH, RazaviSN. Deep neural networks-based damage detection using vibration signals of finite element model and real intact state: an evaluation via a lab-scale offshore jacket structure. Struct Health Monit2021;20(1):379‒405. . 10.1177/1475921720932614
[21]
ZhaoY, PengZ. Frequency response function-based finite element model updating using extreme learning machine model. Shock Vib2020;2020:1‒10. . 10.1155/2020/8526933
[22]
BaiY, PengZ, WangZ. A finite element model updating method based on the trust region and adaptive surrogate model. J Sound Vibrat2023;555:117701. . 10.1016/j.jsv.2023.117701
[23]
SarmadiH, YuenKV. Early damage detection by an innovative unsupervised learning method based on kernel null space and peak-over-threshold. Comput Aided Civ Infrastruct Eng2021;36(9):1150‒67. . 10.1111/mice.12635
[24]
LevinRI, LievenNAJ. Dynamic finite element model updating using neural networks. J Sound Vibrat1998;210(5):593‒607. . 10.1006/jsvi.1997.1364
[25]
YangN, ZhangY. Finite element model updating of Tibetan structure based on artificial neural network. J Vibrat Shock2013;32:125‒9.
[26]
DengZ, ZhangX, ZhaoY. Transfer learning based method for frequency response model updating with insufficient data. Sensors2020;20(19):5615. . 10.3390/s20195615
[27]
ZhangY, LiZ, HaoR, LinW, LiL, SuD. High-fidelity time-series data synthesis based on finite element simulation and data space mapping. Mech Syst Signal Process2023;200:110630. . 10.1016/j.ymssp.2023.110630
[28]
HuaH, SolH, WildeDWP. On a statistical optimization method used in finite element model updating. J Sound Vibrat2000;231(4):1071‒8. . 10.1006/jsvi.1999.2721
[29]
BartilsonDT, JangJ, SmythAW. Sensitivity-based singular value decomposition parametrization and optimal regularization in finite element model updating. Struct Contr Health Monit2020;27(6):e2539. . 10.1002/stc.2539
[30]
VegaMA, ToddMD. A variational Bayesian neural network for structural health monitoring and cost-informed decision-making in miter gates. Struct Health Monit2022;21(1):4‒18. . 10.1177/1475921720904543
[31]
WangZ, LiangD, DingS, ZhangW, HeH. A feature map of frequency response functions based model updating method using the Bayesian convolutional neural network. Mech Syst Signal Process2023;204:110818. . 10.1016/j.ymssp.2023.110818
[32]
SeventekidisP, GiagopoulosD. A combined finite element and hierarchical deep learning approach for structural health monitoring: test on a pin-joint composite truss structure. Mech Syst Signal Process2021;157:107735. . 10.1016/j.ymssp.2021.107735
[33]
SeventekidisP, GiagopoulosD, ArailopoulosA, MarkogiannakiO. Structural health monitoring using deep learning with optimal finite element model generated data. Mech Syst Signal Process2020;145:106972. . 10.1016/j.ymssp.2020.106972
[34]
ZhangZ, SunC, GuoB. Transfer-learning guided Bayesian model updating for damage identification considering modeling uncertainty. Mech Syst Signal Process2022;166:108426. . 10.1016/j.ymssp.2021.108426
[35]
WangQA, DaiY, MaZG, NiYQ, TangJQ, XuXQ, et al. Towards probabilistic data-driven damage detection in SHM using sparse Bayesian learning scheme. Struct Contr Health Monit2022;29(11):e3070. . 10.1002/stc.3070
[36]
XiangY, PanB, LuoL. A new model updating strategy with physics-based and data-driven models. Struct Multidiscipl Optim2021;64(1):163‒76. . 10.1007/s00158-021-02868-5
[37]
SedehiO, KosikovaAM, PapadimitriouC, KatafygiotisLS. On the integration of physics-based machine learning with hierarchical Bayesian modeling techniques. Mech Syst Signal Process2024;208:111021. . 10.1016/j.ymssp.2023.111021
[38]
Di LorenzoD, ChampaneyV, MarzinJY, FarhatC, ChinestaF. Physics informed and data-based augmented learning in structural health diagnosis. Comput Methods Appl Mech Eng2023;414:116186. . 10.1016/j.cma.2023.116186
[39]
YamaguchiT, MizutaniT. A physics-informed neural network for the nonlinear damage identification in a reinforced concrete bridge pier using seismic responses. Struct Contr Health Monit2024;2024(1):5532909. . 10.1155/2024/5532909
[40]
ZhangZ, SunC. Structural damage identification via physics-guided machine learning: a methodology integrating pattern recognition with finite element model updating. Struct Health Monit2021;20(4):1675‒88. . 10.1177/1475921720927488
[41]
ZhouW, XuYF. Damage identification for plate structures using physics-informed neural networks. Mech Syst Signal Process2024;209:111111. . 10.1016/j.ymssp.2024.111111
[42]
YuY, LiuY. Physics-guided generative adversarial network for probabilistic structural system identification. Expert Syst Appl2024;239:122339. . 10.1016/j.eswa.2023.122339
[43]
XuS, NohHY. PhyMDAN: physics-informed knowledge transfer between buildings for seismic damage diagnosis through adversarial learning. Mech Syst Signal Process2021;151:107374. . 10.1016/j.ymssp.2020.107374
[44]
SuttonRS, McAllesterD, SinghS, MansourY. Policy gradient methods for reinforcement learning with function approximation. Adv Neur In2000;12:1057‒63.
[45]
FujimotoS, van HoofH, MegerD. Addressing function approximation error in actor-critic methods. PMLR2018;80:1587‒96.
[46]
LoweR, WuY, TamarA, HarbJ, AbbeelP, MordatchI. Multi-agent actor-critic for mixed cooperative‒competitive environments. In: Proceedings of the 31st International Conference on Neural Information Processing Systems; 2017 Dec 4‒9; Long Beach, CA, USA. Red Hook: Curran Associates Inc.; 2017. p. 6382‒93.
[47]
MnihV, KavukcuogluK, SilverD, RusuAA, VenessJ, BellemareMG, et al. Human-level control through deep reinforcement learning. Nature2015;518(7540):529‒33. . 10.1038/nature14236
[48]
LinYJ, ChunQ, ZhangCW, HanYD, FuH. Research on seismic performance of traditional Chinese hall-style timber buildings in the Song and Yuan dynasties (960‒1368 AD): a case study of the main hall of Baoguo Temple. J Wood Sci2022;68:1. . 10.1186/s10086-021-02009-y
[49]
SohnH, DzwonczykM, StraserEG, KiremidjianAS, LawKH, MengT. An experimental study of temperature effect on modal parameters of the Alamosa Canyon Bridge. Earthquake Eng Struct Dynam1999;28(8):879‒97. . 10.1002/(sici)1096-9845(199908)28:8<879::aid-eqe845>3.0.co;2-v
[50]
XiaY, HaoH, ZanardoG, DeeksA. Long term vibration monitoring of an RC slab: temperature and humidity effect. Eng Struct2006;28(3):441‒52. . 10.1016/j.engstruct.2005.09.001
[51]
MoserP, MoaveniB. Environmental effects on the identified natural frequencies of the Dowling Hall Footbridge. Mech Syst Signal Process2011;25(7):2336‒57. . 10.1016/j.ymssp.2011.03.005
[52]
HanQH, MaQ, LiuM, XuJ. Damage diagnosis of space grid structure based on natural frequency clustering analysis under varying temperature effects. J East China Jiaotong Univ2021;38:9‒17. Chinese.
[53]
AnastasopoulosD, ReyndersEPB. Modal strain monitoring of the old Nieuwebrugstraat Bridge: local damage versus temperature effects. Eng Struct2023;296:116854. . 10.1016/j.engstruct.2023.116854
[54]
Lei BaJ, KirosJR, HintonGE. Layer normalization. 2016. arXiv:
[55]
GotmareA, Shirish KeskarN, XiongC, SocherR. A closer look at deep learning heuristics: learning rate restarts, warmup and distillation. 2018. arXiv:
MnihV, KavukcuogluK, SilverD, GravesA, AntonoglouI, WierstraD, et al. Playing Atari with deep reinforcement learning. 2013. arXiv:
[58]
BhattA, PalenicekD, BelousovB, ArgusM, AmiranashviliA, BroxT, et al. CrossQ: batch normalization in deep reinforcement learning for greater sample efficiency and simplicity. 2019. arXiv:
[59]
ZhangC, ChunQ, LinY, HanY, JiaX. Quantitative assessment method of structural safety for complex timber structures with decay diseases. J Build Eng2021;44:103355. . 10.1016/j.jobe.2021.103355
[60]
ZhangC, ChunQ, SunA, LinY, WangH. Improved meta-learning neural network for the prediction of the historical reinforced concrete bond‒slip model using few test specimens. Int J Concr Struct Mater2022;16(1):41. . 10.1186/s40069-022-00530-y
[61]
FinnC, AbbeelP, LevineS. Model-agnostic meta-learning for fast adaptation of deep networks. PMLR2017;70:1126‒35.
[62]
Ministry of Housing and Urban‒Rural Development of the People’s Republic of China, State Administration for Market Regulation. GB/T 50165‒2020: Technical standard for maintenance and strengthening of historic timber building. Chinese standard. Beijing: China Architecture & Building Press; 2020. Chinese.