Deep long short-term memory networks for nonlinear structural seismic response prediction

被引:379
作者
Zhang, Ruiyang [1 ]
Chen, Zhao [1 ]
Chen, Su [2 ]
Zheng, Jingwei [3 ]
Buyukozturk, Oral [4 ]
Sun, Hao [1 ,4 ]
机构
[1] Northeastern Univ, Dept Civil & Environm Engn, Boston, MA 02115 USA
[2] China Earthquake Adm, Inst Geophys, Beijing 100124, Peoples R China
[3] Elect Power Planning & Engn Inst, Beijing 100120, Peoples R China
[4] MIT, Dept Civil & Environm Engn, 77 Massachusetts Ave, Cambridge, MA 02139 USA
关键词
Deep learning; Long short-term memory; LSTM; Nonlinear dynamic analysis; Seismic response prediction; Time series clustering; NEURAL-NETWORKS; SYSTEM-IDENTIFICATION; MODELS; FUNCTIONALS; SIMULATION; DIAGNOSIS;
D O I
10.1016/j.compstruc.2019.05.006
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
This paper presents a comprehensive study on developing advanced deep learning approaches for nonlinear structural response modeling and prediction. Two schemes of the long short-term memory (LSTM) network are proposed for data-driven structural seismic response modeling. The proposed deep learning model, trained on available datasets, is capable of accurately predicting both elastic and inelastic response of building structures in a data-driven fashion as opposed to the classical physics-based nonlinear time history analysis using numerical methods. In addition, an unsupervised learning algorithm based on a proposed dynamic K-means clustering approach is established to cluster the seismic inputs in order to (1) generate the least but the most informative datasets for training the LSTM and (2) improve the prediction accuracy and robustness of the model trained with limited data. The performance of the proposed approach is successfully demonstrated through three proof-of-concept studies that include a nonlinear hysteretic system, a real-world building with field sensing data, and a steel moment resisting frame. The results show that the proposed LSTM network is a promising, reliable and computationally efficient approach for nonlinear structural response prediction, and offers significant potential in seismic fragility analysis of buildings for reliability assessment. (C) 2019 Elsevier Ltd. All rights reserved.
引用
收藏
页码:55 / 68
页数:14
相关论文
共 66 条
[1]  
Abadi Martin, 2016, Proceedings of OSDI '16: 12th USENIX Symposium on Operating Systems Design and Implementation. OSDI '16, P265
[2]   FITTING AUTOREGRESSIVE MODELS FOR PREDICTION [J].
AKAIKE, H .
ANNALS OF THE INSTITUTE OF STATISTICAL MATHEMATICS, 1969, 21 (02) :243-&
[3]  
[Anonymous], 2017, ARXIV170910432
[4]   Damage assessment through structural identification of a three-story large-scale precast concrete structure [J].
Belleri, Andrea ;
Moaveni, Babak ;
Restrepo, Jose I. .
EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS, 2014, 43 (01) :61-76
[5]   LEARNING LONG-TERM DEPENDENCIES WITH GRADIENT DESCENT IS DIFFICULT [J].
BENGIO, Y ;
SIMARD, P ;
FRASCONI, P .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 1994, 5 (02) :157-166
[6]   An inverse gray-box model for transient building load prediction [J].
Braun, JE ;
Chaturvedi, N .
HVAC&R RESEARCH, 2002, 8 (01) :73-99
[7]   Dynamic assessment of curved cable-stayed bridge by model updating [J].
Brownjohn, JMW ;
Xia, PQ .
JOURNAL OF STRUCTURAL ENGINEERING-ASCE, 2000, 126 (02) :252-260
[8]  
Castaneda Aguilar Nestor E., 2012, DEV VALIDATION REAL
[9]   Development of direct integration algorithms for structural dynamics using discrete control theory [J].
Chen, Cheng ;
Ricles, James M. .
JOURNAL OF ENGINEERING MECHANICS, 2008, 134 (08) :676-683
[10]  
CHEN S, 1992, INT J CONTROL, V56, P319, DOI [10.1080/00207179208934317, 10.1080/00207179008934126]