STFT Based CNN for Interturn Short Circuit Fault Diagnosis of Permanent Magnet Synchronous Motor

In view of the difficulties in fault feature extraction under noise interference, the random selection of model hyperparameters, and the low fault recognition rate of traditional machine learning methods in the diagnosis of permanent magnet synchronous motor interturn short circuit fault, this paper proposes a convolutional neural network fault diagnosis method based on STFT. Firstly, the current signal of the motor in fault state is collected, and the short-time Fourier transform is applied to obtain one-dimensional input to extract sufficient fault characteristics. Finally, according to the output results of the model, the fault state of the motor can be obtained, so as to realize the fault diagnosis of the three-phase permanent magnet synchronous motor. The motor current and torque signals with different short-circuit turns ratio were collected on Simulink simulation platform for experimental verification. The experimental results show that the accuracy of fault diagnosis can reach 98%.