Diagnosing demagnetization failures in mining permanent magnet motors using residual neural network and transformer on cyclic spectrum images

This study focuses on mining permanent magnet motors' normal and demagnetization failure states. The back electromotive force signals are collected and processed into images by establishing an experimental platform. Signal spectrum analysis is conducted using Short-Time Fourier Transform and cyclic spectrum methods to adapt to the noise and interference in the mine environment. Furthermore, a method combining Residual Neural Network and Transformer is used to train and recognise the cyclic spectrum images, improving the accuracy and recall rate of fault identification. The results show that this method is effective in diagnosing demagnetization failures of mining permanent magnet motors, providing new insights for enhancing the reliability and safety of mining motors.