Research on Intelligent Fault Diagnosis Method for Rolling Bearings Based on Multi-Feature Fusion of CWT-2D-CNN

Due to the issue of fault recognition, in order to achieve a higher matching rate between samples and labels, it is necessary to be able to extract richer fault feature information. Therefore, a CWT-2D-CNN bearing fault intelligent diagnosis method is proposed. Firstly, the one-dimensional vibration signal collected is transformed by CWT to construct a two-dimensional image with rich time-frequency information, which is used as the input of the 2D-CNN network, enabling the network to learn as much as possible about the fault time-frequency feature information. Secondly, through deep convolutional pooling layers, further dimensionality reduction of the results from the previous step is a necessary step, which can reduce the difficulty and parameters of optimization and further reduce information redundancy. In the last, the network structure is optimized through backpropagation algorithm to construct an integrated framework model from the extracting feature, decreasing dimensionality to classifying of fault signal. Through fault classification experiments under different states, the model achieves an identification accuracy of 97.98% for faults and can accurately discriminate fault states, verifying the effectiveness of the method, which can serve as one of the inspiration sources for future intelligent diagnosis directions.