Sensor Fault Diagnosis for High-Speed Traction Converter System Based on Bayesian Network

Abstract

Because of the harsh operation environment of the traction converter system, sensor faults occasionally occur, which threaten the stable operation of the high-speed train. Therefore, the research on sensor fault diagnosis technology is of great significance to improve the stability and reliability of the operation of high-speed trains. In order to avoid the uncertainty of sensor fault caused by various conditions, such as the service life of inverter is reduced or the inverter is damaged, the paper proposes a sensor fault diagnosis method based on signal processing technology and Bayesian network (BN). The proposed method consists of two stages. In the first stage, short-time Fourier transformation (STFT) is used to extract features of sensor measurements, and then principal component analysis (PCA) is used to reduce the dimension of extracted features. In the second stage, BN is used as a classifier for sensor fault detection and diagnosis. The experiments in hardware-in-the-loop simulation platform of a high-speed train traction system shows that the proposed method can diagnose sensor fault correctly and effectively, and it is superior to other fault diagnosis methods.