Health State Prediction of Traction Motor Bearings Based on BiLSTM-Transformer

Abstract

Bearings are core components of the traction system in high-speed trains, and ensuring their safe and reliable operation is essential for the normal functioning of the trains. Accurate health state prediction is critical to the safe operation and maintenance of traction motors in high-speed trains. For this reason, this paper first extracts time domain and frequency domain features from vibration signals, fuses degradation features using a Self-Organizing Map (SOM) neural network, and applies the Cumulative Sum (CUSUM) control chart method to construct a quantitative health indicator (HI). Then, Bidirectional Long Short-Term Memory (BiLSTM) network is employed to extract short-term features, while the Transformer is used to capture long-term dependencies, thereby building a health state prediction model aimed at improving prediction accuracy. Finally, the effectiveness of the proposed model is validated on the XJTU-SY bearing dataset.