Dynamic Weighted Slow Feature Analysis-based Fault Detection for Running Gear Systems of High-speed Trains

The running gear system provides the safety guarantee for the normal operation of high-speed trains. The massive historical data in the system can be used for fault detection and diagnosis. This data inevitably exists redundancy, which makes the valuable data not fully utilized in the process of extracting latent variables. In this paper, to make full and effective use of historical data, a dynamic weighted slow feature analysis (DWSFA) method is proposed, which can detect slow-change faults in the running gear system of high-speed trains. The proposed method based on basis functions can reduce the amount of time lags required for the process of extracting latent variables, and it obtains the better fault detection (FD) performance. The effectiveness of the proposed method is verified via a running gear system of high-speed train.