基于LMM和NARNN的车轮踏面退化状态预测

作为列车的关键走行部件,车轮的退化状态对列车的安全具有重要影响。以车轮踏面磨耗量为研究对象,将历史车轮踏面磨耗数据作为输入,分别采用线性混合模型(LMM)和非线性自回归神经网络(NARNN)对车轮踏面磨耗进行建模。首先,对比不同随机效应的LMM,选择随机系数相关的LMM,进而预测车轮踏面磨耗量;其次,使用随机搜索算法优化NARNN中的参数。结果显示,基于LMM的踏面磨耗值的预测精度更高。

Prediction of Wheel Tread Degradation Based on LMM and NARNN

As the key running part of the train,the degradation of the wheel has an important impact on the safety of the train.Taking the wheel tread wear as the research object and taking the historical wheel tread wear as the input,the wheel tread wear is modeled by linear mixed model (LMM) and nonlinear autoregressive neural network (NARNN).Firstly,the LMM with different random effects are compared,and the LMM with random coefficient is selected to predict the wheel tread wear.Secondly,the random search algorithm is used to optimize the parameters of NARNN.The results show that the prediction accuracy of wheel thread wear based on LMM is higher.