地震作用下结构响应实时预测方法研究  

Real-time prediction method of structural responses under earthquake action

作　　者：李正农[1] 顾珂泽 黄斌 吴红华[1] LI Zhengnong;GU Keze;HUANG Bin;WU Honghua(Key Laboratory of Building Safety and Energy Efficiency of Ministry of Education,Hunan University,Changsha 410082,China;School of Civil Engineering and Architecture,Hainan University,Haikou 570228,China)

机构地区：[1]湖南大学建筑安全与节能教育部重点实验室,长沙410082 [2]海南大学土木建筑工程学院,海口570228

出　　处：《振动与冲击》2024年第22期62-69,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金(52178476)。

摘　　要：该研究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。该研究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,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,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,引入了一种EraquseqNet模型。相较于其他神经网络方法,该研究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。In order to make up the shortcomings in previous literatures,the paper puts forward the modified theory and method of real-time prediction of structural responses under earthquake,and improves the defects of the previously established neural network model.The effectiveness of real-time prediction of structural responses under earthquake by using neural network was demonstrated,and the omissions in previous theories were pointed out.The problems faced in the practical application were focused on,including the preprocessing of the training set and the application method of the model after training.On the basis of improving the deficiency of the theory,the method of using the training model to predict structural responses,was deeply discussed and a practical method for predicting the real-time responses of the structure under earthquake was provided.The preprocessing method of the data set was improved to ensure that the explosive accumulation of errors would not occur in the model prediction.In order to improve the accuracy and efficiency of structural response prediction,an EraquseqNet model was introduced,which was based on codec neural network model,bi-directional neural network module and attention mechanism.Compared with other neural network methods,the problem of accuracy decreasing 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 

分 类 号：TH212[机械工程—机械制造及自动化] TH213.3