Fault diagnosis method of AT-ICNN based on fusion attention mechanism and enhanced convolutional layer

The invention discloses an AT-ICNN fault diagnosis method based on fusion of an attention mechanism and an enhanced convolutional layer, and relates to the field of mechanical fault diagnosis. The bearing fault diagnosis method is regarded as solving the problem that an existing bearing fault diagnosis method neglects the importance of signal feature prominence and global information capturing capability. The method comprises the following steps: obtaining a to-be-diagnosed cross-domain bearing original vibration signal, and inputting the to-be-diagnosed bearing original vibration signal into a bearing fault diagnosis network to obtain a bearing fault severity diagnosis result; the bearing fault diagnosis network is obtained through the following steps: obtaining original vibration signals under different working conditions, and dividing the original vibration signals of the bearing into a training set and a test set; constructing an AT-ICNN network, wherein the AT-ICNN network comprises an enhanced convoluti