关键词: 永磁同步电机, 电气故障, 卷积神经网络, 灰度图像, 电网侧电流

Abstract: It is not convenient for P MSM to use the signal on the motor side for fault diagnosis in practical applications. To solve the problem, a P MSM fault diagnosis method based on the root mean square value of grid side current and improved CNN was proposed. A fast differentiation between normal and faulty states of the P MSM was realized by calculating the root mean square value of the grid side current for each cycle. The current signals were collected in different fault state modes, transformed into grayscale images, and the samples were expanded by using the sliding window sampling method. These images were inputted to the CNN for automatic fault type identification. Simulation and experimental validation were carried out, the classification accuracy of this method in detecting
different electrical fault states of P MSM can reach over 95 % , which proves its efficiency and reliability in practical applications.

Key words: permanent magnet synchronous motor ( PMSM), electrical fault, convolutional neural networks ( CNN), grayscale image, grid side current