Fault Diagnosis for IGBTs Open-Circuit Faults in High-Speed Trains Based on Convolutional Neural Network

Three-phase voltage source inverter is an important part of high-speed trains (HSTs) drive system. The stability of inverter directly affects the stability of motor speed control system and the safety of motor. IGBTs open-circuit faults are an important reason for the electrical faults of the inverters. Fault diagnosis is helpful to discover the cause of the fault in time and improve the safety and maintenance efficiency of HSTs. Most of the existing fault diagnosis methods need to build complex mathematical models or extract features from sensor signals manually. In order to solve these problems, the paper proposes a new method for IGBTs open-circuit fault diagnosis named Convolutional Neural Network for Extracting Comprehensive Information (CI-CNN) and a current signal to gray image conversion method named k-gray. In this paper, the three-phase current signals of traction motor stator are synchronously resampled by angle increment, and then the angle domain signals are transformed into gray image. CI-CNN extracts fault features from gray images to diagnose IGBTs open-circuit faults. The method does not need to extract fault features manually and achieves end-to-end fault diagnosis. Experiments show that the method has excellent adaptability in speed domains and load domains. At the same time, the combination of CI-CNN and k-gray has good diagnostic accuracy in noisy environment.