An Intelligent Diagnostic Method for Permanent Magnet Synchronous Motors (PMSM) in the Electric Drive of Autonomous Vehicles

In highly automated electric vehicles, the reliability of electrical powertrain system is very important. A critical failure in the powertrain system, e.g. electric machine, would lead to breakdown of the vehicle. To avoid this dangerous situation, the critical faults should be detected at an early stage. This paper focuses on three common faults in the stator of a permanent magnet synchronous machine (PMSM). Based on analytical models, the physical behaviors of the electrical machine within these three faults are analyzed. Then, a data-driven diagnostic method, artificial neural network (ANN), to detect and classify these faults is presented. Simulation data of electric machine under healthy and faulty conditions as well as at different operating points are used to train the ANN model. The three phase currents of PMSM and the inverter input current are selected as input signals of model. Various features in time domain(e.g. average, maximum,...) and frequency domain(e.g. 2nd, 3rd harmonics) are extracted from the selected signals. The result shows that the fault diagnostic model is capable of classifying the faults with near perfect accuracy over 98%, even in case of slight fault. Finally, a driving cycle simulation is used to validate the robustness of the ANN model in dynamic driving situations.