Application of the CMOR-EfficientNetV2 Network Method in Bearing Fault Diagnosis

To solve the problems of large parameters, slow training speed and low accuracy of deep neural network fault diagnosis method, a CMOR-EfficientNetV2 network fault diagnosis model is proposed for bearing fault diagnosis. The vibration signal of rolling bearing is processed by continuous wavelet transform to obtain time-frequency diagram, and then the time-frequency diagram is compressed to an appropriate size; The compressed time-frequency graph is used as the input of feature graph to establish EfficientNetV2 classifier model to realize the fault diagnosis of rolling bearings. Then, the CMOR-EfficientNetV2 network is tested using the bearing data set of Case Western Reserve University. The comparison results show that the CMOR-EfficientNetV2 network model has a classification accuracy of more than 99.5% for the bearing dataset of CWRU University, and the recognition rate and generalization performance are better than other lightweight neural networks.