Rolling bearing fault diagnosis method

A rolling bearing fault diagnosis method comprises a model training process and a diagnosis process, in the process, rolling bearing vibration information is obtained, continuous wavelet transform processing is carried out on the rolling bearing vibration information to obtain a two-dimensional time-frequency domain image, and the two-dimensional time-frequency domain image is sent to a fault diagnosis model for processing. A main body part of the fault diagnosis model takes a RegNet network as a basis, introduces ConvLSTM to extract spatial-temporal characteristics of an input image, and feeds the spatial-temporal characteristics to a RegNet residual network. The ConvLSTM takes the feature information in the residual network as input and generates a hidden state, the hidden state is connected with the feature information in the residual network and then is used in subsequent convolution of the residual network, and the hidden state output by the ConvLSTM of the previous layer is transmitted to the ConvLSTM o