Bearing Feature Extraction and Fault Diagnosis Algorithm Based on Convolutional Neural Networks

The first step of the traditional fault diagnosis method is to process the signal and then put the signal into the classifier for recognition. The feature extraction process depends on the experimenter's experience, and the recognition rate of the shallow diagnostic model does not achieve satisfactory results. The traditional diagnosis algorithms are based on a single object, so that their versatility can not be guaranteed, and they fail to meet the requirements of fault diagnosis in the big data era. In view of this problem, this paper presents an intelligent diagnosis algorithm based on convolution neural network, which can automatically complete the feature extraction and fault identification of the signal. The validity of the method is validated by using bearing data, and tests were performed using different sample sizes to analyze their impact on the diagnostic ability of CNN. The test results show that the proposed method has an accuracy rate of 99.82% for bearing fault diagnosis, which achieves the highest recognition rate and can meet the timeliness of fault diagnosis.