Remaining Useful Life Prediction of Wheel of Heavy-duty Railway Train based on Dual Channel Multi-scale Deep convolution Multi-scale Deep Long Short-Term Memory network

Abstract

The running conditions of wheels of Heavy-duty Railway Train are complex, and the real-time running state data is Multi-Dimension and Time-Sequence. Aiming at the problems that the traditional deep learning models have weak learning ability, cannot extract different scale information and gradient explosion in the prediction of remaining useful life (RUL), this paper proposes a multi-scale deep long short-term memory (MDLSTM) network model, which extracts time-series features of different scales through different number of hidden layer units of LSTM networks. In order to obtain more robust features under the premise of reducing the loss of original information and better to predict RUL of wheels, A Dual Channel Multi-scale Deep convolutional Multi-scale Deep long short-term memory (DC-MDCNN-MDLSTM) is proposed which combined the CNN and LSTM to extract multi-scale feature of wheels under different conditions and extract the different time step features of wheels from time series data. Using the actual wheels data to experiments. The results show that DC-MDCNN-MDLSTM network model is effective in predicting the degradation state of the wheels and provides technical support for repairing on condition of Heavy- duty Railway Train.