Train wheel size prediction based on EEMD-LSTM model

Abstract

Predicting the wear state of the wheels and predicting the possible problems of the wheels in advance can help guide and optimize the maintenance and repair work of the vehicles, and reduce the economic losses caused by the untimely vehicle maintenance scheduling. In this paper, the long short-term memory (LSTM) recurrent network is used as the prediction model. In order to improve the prediction accuracy of the network model, the Ensemble Empirical Mode Decomposition (EEMD) algorithm is introduced to stabilize the non-stationary wear sequence and improve the training effect neural network. Firstly, the non-stationary original wheel diameter sequence is decomposed into multiple stationary sub-sequences by the EEMD algorithm, and then the training and prediction of the LSTM model are completed separately based on each component. The experimental results show that after EEMD processing, the prediction effect of tread wear has been improved, which verifies the improvement effect of EEMD processing on the prediction accuracy of neural network.