为弥补以往学者理论中的缺陷，文章提出修正后的地震下结构响应实时预测理论和方法，并改进了先前建立的神经网络模型的缺陷。对利用神经网络进行地震作用下结构响应实时预测的有效性进行了论证，指出了以往理论中存在的遗漏。论文着重阐述了在实际应用中面临的问题，包括训练集预处理和训练后模型应用方法等。在修正理论不足的基础上，文章深入探讨了利用训练后模型进行结构响应预测的方法，提供了可实际应用的地震作用下结构实时响应预测方法。对数据集的预处理方法进行了改进，确保在模型预测时不会发生误差的爆发式累积。为提高结构响应预测的精度和效率，文章引入了一种EraquseqNet模型，基于编解码器神经网络模型、双向神经网络模块和注意力机制。相较于其他神经网络方法，文章利用注意力机制解决了信息冗余导致精度下降的问题，并通过双向神经网络模块解决了长时间地震作用下响应预测精度迅速下降的难题。
Abstract
In order to make up for the shortcomings of previous scholars, this paper puts forward the modified theory and method of real-time prediction of structural response under earthquake, and improves the defects of the previously established neural network model. In this paper, the effectiveness of real-time prediction of structural response under earthquake by using neural network is demonstrated, and the omissions in previous theories are pointed out. This paper focuses on the problems faced in the practical application, including the preprocessing of the training set and the application method of the model after training. On the basis of correcting the deficiency of the theory, this paper deeply discusses the method of using the training model to predict the structural response, and provides a practical method for predicting the real-time response of the structure under earthquake. The preprocessing method of the data set is improved to ensure that the explosive accumulation of errors will not occur in the model prediction. In order to improve the accuracy and efficiency of structural response prediction, an EraquseqNet model is introduced, which is based on codec neural network model, bi-directional neural network module and attention mechanism. Compared with other neural network methods, the problem of decreasing accuracy caused by information redundancy is solved by using attention mechanism, and the problem of rapid decline of response prediction accuracy under long-term earthquake is solved by bi-directional neural network module.
关键词
地震作用 /
结构响应 /
注意力机制 /
双向神经网络
{{custom_keyword}} /
Key words
Earthquake action /
Structural response /
Attention mechanism /
Bidirectional neural network
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 李华玥, 文鑫涛, 陈雅慧, 等. 2020年国外地震灾害及其影响综述[J]. 震灾防御技术, 2021,16(03):583-588.
Li H Y, Wen X T, Chen Y H, et al. Overview of foreign earthquake disasters and their impacts in 2020 [J]. Earthquake Disaster Prevention Technology, 2019,16(03):583-588(in Chinese).
[2] YAMADA K, KOBORI T. LINEAR QUADRATIC REGULATOR FOR STRUCTURE UNDER ON-LINE PREDICTED FUTURE SEISMIC EXCITATION[J]. Earthquake engineering & structural dynamics, 1996,25(6):631-644.
[3] Moaveni B, Conte J P, Hemez F M. Uncertainty and Sensitivity Analysis of Damage Identification Results Obtained Using Finite Element Model Updating[J]. Computer-aided civil and infrastructure engineering, 2009,24(5):320-334.
[4] Abdeljaber O, Avci O, Kiranyaz S, et al. Real-time vibration-based structural damage detection using one-dimensional convolutional neural networks[J]. Journal of Sound and Vibration, 2017,388:154-170.
[5] 王希, 王宪杰, 董艳秋, 等. 基于离散单元法和物理引擎的结构连续倒塌可视化模拟[J]. 振动与冲击, 2020,39(13):267-275.
Visual Simulation of Continuous structural collapse based on discrete element Method and Physics Engine [J]. Journal of Vibration and Shock, 20,39(13):267-275(in Chinese).
[6] Zheng Z, Tian Y, Yang Z, et al. Hybrid Framework for Simulating Building Collapse and Ruin Scenarios Using Finite Element Method and Physics Engine[J]. Applied sciences, 2020,10(12):4408.
[7] Fishwick, P.A..Neural Network Models In Simulation: A Comparison With Traditional Modeling Approaches[C]// 1989 Winter Simulation Conference Proceedings. Washington D C: IEEE, 1989.
[8] Pei J S; Mai E C.Constructing Multilayer Feedforward Neural Networks to Approximate Nonlinear Functions in Engineering Mechanics Applications[J].Journal of Applied Mechanics, 75(6), 061002–.
[9] Pei J, Wright J P, Masri S F, et al. Toward constructive methods for sigmoidal neural networks - function approximation in engineering mechanics applications[C]// The 2011 International Joint Conference on Neural Networks. San Jose:IEEE, 2011.
[10]Wu R,Jahanshahi,Mohammad R.Deep Convolutional Neural Network for Structural Dynamic Response Estimation and System Identification[J].Journal of Engineering Mechanics, 145(1):04018125-.
[11]Zhang R, Chen Z, Chen S, et al. Deep long short-term memory networks for nonlinear structural seismic response prediction[J]. Computers & Structures, 2019,220:55-68.
[12]Zhang R, Liu Y, Sun H. Physics-informed multi-LSTM networks for metamodeling of nonlinear structures[J]. Computer Methods in Applied Mechanics and Engineering, 2020,369:113226.
[13]Bengio Y, Simard P, Frasconi P. Learning long-term dependencies with gradient descent is difficult[J]. IEEE Trans Neural Netw, 1994,5(2):157-166.
[14]高经纬, 张春涛. 基于长短时记忆网络的结构地震响应预测[J]. 工程抗震与加固改造, 2020,42(03):130-136.
Gao Jingwei, Zhang Chuntao. Seismic response prediction of structures based on long and short time memory network [J]. Earthquake Resistance and Reinforcement Engineering, 2019,42(03):130-136(in Chinese).
[15]孟诗乔, 周颖, 张啸天, 等. 地震作用下建筑结构响应时程实时预测算法研究[J]. 建筑结构学报, 2022,43(S1):334-344.
Meng Shiqiao, Zhou Ying, Zhang Xiaotian, et al. Research on real-time prediction algorithm of building structure response time history under earthquake [J]. Journal of Building Structures, 2022,43(S1):334-344(in Chinese).
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(2039 KB)
