电动汽车混沌振动信号的小波神经网络预测

Wavelet neural network prediction of electric vehicle chaotic vibration signals

对某款电动汽车进行了实车试验,研究电动汽车的混沌动力学特性,并建立了小波神经网络预测模型对混沌时间序列进行预测。在中等比利时路面上对电动汽车进行实车试验,得到右前轮心垂向和电池底部中心垂向的振动加速度信号。对采集到的信号进行时频分析,计算得到三维相图和庞加莱截面,利用互信息法计算时间延迟,Cao法计算嵌入维,并利用Wolf方法计算得到最大李雅普诺夫指数,发现振动加速度信号存在混沌运动。利用小波神经网络对右前轮心垂向的混沌时间序列进行预测,表明利用小波神经网络对混沌时间序列进行预测能取得较好的预测效果。

A vehicle experiment was carried out to study the chaotic dynamics of electric vehicles,and chaotic time series were predicted by using wavelet neural network.First,the experiment of electric vehicle on medium-Belgian road was carried out,the vertical vibration acceleration signal of the right front wheel center and battery bottom center were obtained.Second,time-frequency analysis,three-dimensional phase diagrams and the Poincarésections of the signals were obtained.The time delay was calculated by using the mutual information method,and also minimum embedding dimension was got with the Cao method,the largest Lyapunov exponent was got with Wolf method.The presence of chaotic motions in the vertical acceleration signal was found.Finally,chaotic time series of the right front wheel center vertical signal was predicted using wavelet neural network.It is found that the use of wavelet neural network to predict chaotic time series can achieve better results.