Research on Substation Auxiliary Motor Bearing Fault Diagnosis Based on Voiceprint Recognition and INOA-SVM

For motors that have been put into operation, the traditional motor fault diagnosis method of deploying vibration sensors is inconvenient and susceptible to high voltage interference on the primary side. The voiceprint recognition method is a noncontact measurement. It is easy to install and deploy without damaging the motor’s original working structure. This paper applies voiceprint recognition to realize substation auxiliary motor bearing fault monitoring. To enhance the accuracy of fault diagnosis, a motor fault diagnosis technology based on voiceprint recognition and an improved Nutcracker optimizer algorithm optimized support vector machine (INOA-SVM) is proposed. Firstly, variational modal decomposition combined wavelet noise reduction is used for noise reduction of the base acoustic signal. Acoustic features are extracted by Mel frequency cepstral coefficients(MFCC). Then, the Nutcracker optimizer algorithm was improved using the good point set strategy, random opposition-based learning strategy, and elite selection strategy. Finally, a motor fault diagnosis method by INOA optimized SVM is proposed. The method’s superiority was experimentally verified with a recognition accuracy of 98.6088%.