Research on Three Common Fault Diagnosis Methods for AC Asynchronous Motors Based on Deep Learning

In response to the problem that traditional fault diagnosis methods mainly rely on manual search, this paper proposes an improved convolutional neural network based three item asynchronous motor fault diagnosis method. Taking the motor rotor bar fault as the research object, in the early stage of the fault, the characteristic signal is easily mixed with the motor fundamental frequency signal. Therefore, first, the current characteristics of the motor rotor bar fault are analyzed, and then the motor vibration signal is converted into a time-frequency map using wavelet analysis method. Then, based on the superpixel segmentation method, the image is generated into a superpixel block. Finally, the image information is input into an improved neural network, The improved neural network can adaptively extract fault features. The experimental results show that the method described in this article can improve the diagnostic ability for rotor bar breaking faults, and has a higher fault recognition accuracy compared to traditional methods.