基于LSTM神经网络车辆零部件载荷预测方法研究  

Research on Vehicle Component Load Prediction Method based on LSTM Neural Network

作　　者：陈沛 张桂明[1] 何海[1] CHEN Pei;ZHANG Guiming;HE Hai

机构地区：[1]泛亚汽车技术中心有限公司,上海201208

出　　处：《上海汽车》2023年第7期21-26,共6页Shanghai Auto

摘　　要：载荷信号是对车辆零部件结构的可靠性和疲劳耐久分析的基础。文章针对当前工程应用中载荷谱获取方法的不足,提出了一种基于LSTM神经网络来预测车辆零部件载荷的方法。将实采信号经过预处理后,选取整车六分力参数为输入,实现了减振器弹簧、稳定杆等底盘悬架零部件在比利时路面下载荷的预测,并从时域信号、功率谱密度、损伤方面将预测结果与实际测量结果进行了对比。结果表明,LSTM模型有足够高的精度,为整车结构载荷获取提供了一种新的途径。The load signals are the basis for reliability and fatigue durability analysis of ve⁃hicle components.Aiming at the shortcomings of current engineering application in load spectrum ac⁃quisition methods,a method based on LSTM neural network is proposed in order to predict the load of vehicle components.After pre-processing the actual acquisition signals,the six-component force parameters of the whole vehicle are selected as inputs,and the prediction of chassis suspension com⁃ponents such as damper spring and stabilizer bar under the load of Belgian road surface is realized.The predicted results are compared with the actual measurement results from the time domain signal,power spectral density and damage aspects.The results show that the LSTM model has sufficiently high accuracy and provides a new way for obtaining the load of the entire vehicle structure.

关 键 词：LSTM 神经网络 六分力 信号处理 载荷预测 

分 类 号：U463[机械工程—车辆工程]