CNN rolling bearing fault diagnosis method based on improved GAF and SA

The invention discloses a CNN rolling bearing fault diagnosis method based on improved GAF and SA, relates to the technical field of bearing fault diagnosis, and solves the technical problem that the bearing fault diagnosis performance is reduced in a high-noise environment. The key points of the technical scheme are as follows: firstly, the improved GAF is utilized to encode a collected vibration signal, and a corresponding feature map is generated; and then the rolling bearing fault features are extracted by inputting the rolling bearing fault features into a convolutional neural network, an SA attention module suitable for the convolutional neural network is introduced to realize adaptive weighting of the features, and finally the rolling bearing fault features are input into a softmax layer to complete rolling bearing fault classification. The model is good in robustness and higher in diagnosis accuracy, and meanwhile a good diagnosis result can be kept in a complex environment. 本发明公开了一种基于改进GAF及SA的CNN滚动轴承