基于LSTM网络的连续箱梁桥地震响应预测  

Seismic Response Prediction of a Continuous Girder Bridge Based on LSTM Network

作　　者：程湧 龚伟 赵梦阳 秦泽翔 李怡冰 程方昱 王丹生[3] CHENG Yong;GONG Wei;ZHAO Mengyang;QIN Zexiang;LI Yibing;CHENG Fangyu;WANG Dansheng(Wuhan Airport Road Development Co Ltd,Wuhan 430014,China;Central&Southem China Municipal Engineering Design and Research Institute Co Ltd,Wuhan 430010,China;School of Civil and Hydraulic Engineering,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]武汉机场路发展有限公司,湖北武汉430014 [2]中国市政工程中南设计研究总院有限公司,湖北武汉430010 [3]华中科技大学土木与水利工程学院,湖北武汉430074

出　　处：《土木工程与管理学报》2024年第5期26-33,40,共9页Journal of Civil Engineering and Management

基　　金：国家重点研发计划(2021YFF0501001)。

摘　　要：现有的桥梁地震响应分析方法存在计算成本高、计算效率低等问题。快速有效地预测桥梁地震响应对桥梁结构的抗震性能评估和地震安全性有着重要的意义。建立了一种基于长短时记忆神经网络的混凝土连续箱梁桥地震响应预测模型。根据特征反应谱筛选出80条在峰值、持时和平均波速上有较大差异的地震波,利用ABAQUS建立桥梁有限元模型并生成加速度响应样本数据,以此对网络进行训练和测试。研究结果表明:基于LSTM网络的地震响应预测模型对差异较大的地震动均能够快速、有效地预测桥梁的加速度响应,表现出良好的预测性能和鲁棒性;与堆栈序列LSTM网络模型(LSTM-s)相比,该模型具有更高的预测精度。The existing seismic response analysis methods for bridge have the problems of high computational cost and low computational efficiency.Rapid and effective prediction of seismic response of bridges is of great significance to seismic performance evaluation and seismic safety of bridge structures.A seismic response prediction model for continuous girder bridge was established based on long short-term memory(LSTM)neural network.Based on the characteristic response spectrum,80 seismic waves with large differences in peak value,duration and average wave speed were screened out.ABAQUS was used to build a bridge model to generate acceleration response sample data,and then they were used to train and test the network.The research results show that the seismic response prediction model based on LSTM network can predict the acceleration response of bridge quickly and effectively for the ground motions with large differences,and shows good prediction performance and robustness.Compared with stacked sequence to sequence LSTM(LSTM-s)network,LSTM-based model holds higher prediction accuracy.

关 键 词：长短时记忆神经网络 地震响应预测 混凝土连续梁桥 加速度 

分 类 号：U442.55[建筑科学—桥梁与隧道工程] TP183[交通运输工程—道路与铁道工程]