Research on Motor Fault Diagnosis Methods Based on Machine Learning

Abstract

The motor system is a complex system and widely applied in industrial production process. In this paper, the mechanism model and common faults of the motor are analyzed firstly, and the motor model is established by the LabVIEW software. Secondly, the methods used to realize motor fault diagnosis are introduced. Based on the diagnosis result of various fault diagnosis methods, the advantages and disadvantages of various fault diagnosis methods are analyzed. Considering that random forest (RF) has strong classification ability and convolutional neural network (CNN) has strong feature extraction ability, a fault diagnosis model of CNN-RF is proposed combining the CNN and RF algorithm. The proposed CNN-RF model can realize the fault diagnosis of motor effectively. Finally, the motor fault diagnosis system based on C# language is built. The experiments show that the proposed method has good performance in fault diagnosis.