An Approach for Feature Extraction and Diagnosis of Motor Rotor Bearing Based on Convolution Neural Network

The traditional rotor bearing fault diagnosis and analysis method is difficult to get the prior knowledge and experience, resulting in the low accuracy of fault diagnosis. In this paper, a method of fault feature extraction and diagnosis of rotor bearing based on convolution neural network is proposed. This method uses the chaotic characteristic of the vibration signal of the rotor bearing, uses the phase space reconstruction method to obtain the embedding dimension as the scale of the convolution neural network input composition, avoid the limitation of traditional frequency analysis method in the process of decomposition and transformation, the fault information can be extracted more comprehensively. In order to make full use of the advantages of the convolution neural network in the field of two-dimensional image analysis and improve the accuracy of the fault diagnosis model, a method of learning input form neural network based on convolution neural network for grayscale graph is proposed. The results of the simulation show the effectiveness of the method.