Fault Diagnosis Algorithms for Power Devices of Traction Inverters in High-Speed Train

Abstract

Open-circuit fault is one of the main faults in traction inverters. If an open-circuit fault occurs in power devices and is not resolved in time, it will lead to secondary faults and system crashes. In order to solve these problems, this paper studies two fault diagnosis methods for traction inverters, including mixed logic dynamic (MLD) model and convolutional neural network (CNN) methods. The MLD model-based diagnosis method analyzes current model, which can realize fault diagnosis and location quickly and accurately. The CNN-based diagnosis method firstly preprocesses the signal of three phase current, then carries out identification and fault processing, which can realize the diagnosis of open-circuit fault online. Simulation results show that both methods can accurately achieve fault diagnosis and location online. The MLD model-based diagnosis method is faster and more accurate. The CNN based diagnosis method presents good generalization performance.