Online Stator Interturn Fault Detection Using Hilbert Transform and Neural Network for Traction Motors of Urban Rail Vehicles

In order to avoid catastrophic accidents, it is necessary to detect the stator interturn fault at the very early stage. However, the complex operating conditions of traction motors of urban rail vehicles bring challenges to real-time fault feature extraction and highly sensitive fault detection. In this article, a new online stator interturn fault detection method is proposed. First, fault feature extraction from transient current signals is realized by a modified Hilbert transform. In addition, considering the difficulty of threshold calibration under complicated operating conditions and the lack of actual fault samples in the application, an automatic fault identification approach based on an artificial neural network (ANN) model which is trained only utilizing healthy data is proposed. In this article, a 210-kW traction induction motor of metro vehicle is taken as the research object, and the existing traction control unit is used as the implementation platform of the algorithm. The results illustrate the effectiveness and feasibility of the proposed approach in application.