The invention discloses a rolling bearing fault intelligent diagnosis method based on improved S-transformation and deep learning. According to the method, a window width adjustment factor is added toS-transformation, so that the window width of a Gaussian window function can be changed, and therefore, the time-frequency resolution of the S-transformation can be improved, and the S-transformationcan accurately detect an impact component in vibration signals, and thus, the fault characteristics of the vibration signals of a rolling bearing can be better extracted. Improved S-transformation isperformed on the vibration signals of the fault of the bearing, so that the feature matrix of the signals can be obtained; the feature matrix is subjected to column-based expansion so as be convertedinto a feature vector, and the feature vector is inputted into a sparse autoencoder model; on the basis of the characteristics of the encoder, the deep features of the data are further extracted, sothat some important hidden information that is not recognized by humans can be mined; the extracted features are accurately classified. With the method of the invention adopted, the accuracy of the fault diagnosis of the rolling bearing can be effectively improved.