基于LSTM网络的车辆动态荷载识别方法  

Dynamic vehicle load identification method based on LSTM network

作　　者：张璐[1,2] 冯东明[1,2] 吴刚[1,2] Zhang Lu;Feng Dongming;Wu Gang(Key Laboratory of Concrete and Prestressed Concrete Structures of Ministry of Education,Southeast University,Nanjing 211189,China;National and Local Joint Engineering Research Center for Intelligent Construction and Maintenance,Southeast University,Nanjing 211189,China)

机构地区：[1]东南大学混凝土及预应力混凝土结构教育部重点实验室,南京211189 [2]东南大学智慧建造与运维国家地方联合工程研究中心,南京211189

出　　处：《东南大学学报（自然科学版）》2023年第2期187-192,共6页Journal of Southeast University：Natural Science Edition

基　　金：国家重大科研仪器研制资助项目(52127813);中央高校基本科研业务费专项资金资助项目(3205002203A1)。

摘　　要：为了识别车辆的动态荷载,提出了一种基于长短时记忆网络的方法.该方法以桥梁加速度响应为输入,采用有限的数据集,实现车辆动态荷载的识别.建立了车桥耦合模型进行验证,以60组桥梁加速度响应为输入,以相应的车辆动态荷载为输出,通过训练长短时记忆网络来反演车辆动态荷载,并讨论了环境噪声及路面粗糙度对识别效果的影响.结果表明:测试集的车辆动态荷载识别误差平均值均小于5%;车辆动态荷载识别误差不随噪声水平变化,且平均误差小于5%;车辆动态荷载识别误差随着路面粗糙度等级的增加呈现略微增加的趋势,平均误差小于5%.在不同噪声及粗糙度水平下,长短时记忆网络均可用于车辆动态荷载的识别.To identify the dynamic vehicle load,a method based on long short-term memory(LSTM)network was proposed.The bridge acceleration response was used as the input in the method,and the dynamic vehicle load can be identified based on the finite data sets.The verification was carried out based on a vehicle-bridge interaction model.Taking 60 groups of bridge acceleration responses as the input and the corresponding vehicle dynamic load as the output,the vehicle dynamic load inversion was realized by training the LSTM network.The influence of ambient noise and road roughness on the identification was discussed.The results show that the average vehicle dynamic load identification error of the test set is less than 5%.The vehicle dynamic load identification error does not change with the noise level,and the average error is less than 5%.The vehicle dynamic identification error increases slightly with the increase of road roughness,and the average error is less than 5%.The LSTM network can be used to identify vehicle dynamic loads under different noise and roughness levels.

关 键 词：长短时记忆网络 结构健康监测 车桥耦合系统 加速度响应 车辆动态荷载识别 

分 类 号：U441[建筑科学—桥梁与隧道工程]