Rolling bearing fault diagnosis method based on ECARsNet

The invention discloses a rolling bearing fault diagnosis method based on ECARsNet, and the method specifically comprises the steps: converting a one-dimensional vibration signal into a three-channel color time-frequency diagram through the combination of short-time Fourier and a pseudo-color processing method, and enabling the time-frequency diagram to serve as an input sample set; secondly, an ECAResNet bearing fault diagnosis model is constructed, the model comprises a convolution layer, a maximum pooling layer, an improved residual module and a full connection layer, shallow edge feature extraction and compression are performed on the data based on the convolution layer and the pooling layer, the improved residual module is constructed to increase the network depth, and Dropout is introduced between the two connection layers to suppress overfitting; and then training the ECAResNet bearing fault diagnosis model, and inputting the sample added with the Gaussian white noise into the trained ECAResNet bearing