CWDCGAN-GP Fault Diagnosis Method for Rolling Bearing Under Imbalanced Sample Conditions

Abstract

Difficulty in fault diagnosis of bearings under imbalanced sample conditions. Aiming at the problem of bearing fault diagnosis under unbalanced sample conditions, a rolling bearing fault diagnosis method based on the gradient penalty condition Wasserstein deep convolution generative adversarial network (CWDCGAN-GP) with convolutional block attention module (CBAM) attention mechanism and wide convolution kernel deep convolution neural network (WDCNN) is proposed. Firstly, fast fourier transform (FFT) is used for signal processing; Then, the preprocessed data set is input into the CWDCGAN-GP network for parameter adjustment, model training and signal generation; Finally, the original signal and the generated signal are input into the WDCNN network for parameter adjustment, model training and fault diagnosis. The experimental results show that the proposed model can significantly improve the prediction accuracy under the conditions of multiple sample imbalance ratios, and verify the superiority of CWDCGAN-GP model and WDCNN model in bearing fault diagnosis under sample imbalance conditions.