A Variable Speed Fault Detection Approach for Electric Motors in EV Applications based on STFT and RegNet

Electric motors play a significant role in the power train system in Electric Vehicles (EV s). Therefore, their online condition monitoring is essential in ensuring the reliable operation of the entire powertrain. In EV applications, the motors work in a non-ideal environment and continuously varying operating conditions. Therefore, the fault diagnosis of the EV motors is challenging, and the fault diagnosis model must work in wide ranges of speeds and loads. In this paper, a short-time Fourier transform with varying window is proposed as the current signal processing of the motor along with a convolutional neural-based network for the detection of the interturn short circuit of permanent magnet synchronous motor. The proposed method is evaluated using a simulation dataset and a benchmark bearing fault dataset. by Matlab/Simulink® and the model is trained with the data in four speeds, and the model tests are carried out for a different operating speed.