Tobacco machinery bearing fault diagnosis method and diagnosis system

The invention discloses a tobacco machine bearing fault diagnosis method and diagnosis system, relates to the technical field of mechanical fault diagnosis, and solves the problems of low model convergence speed and low diagnosis precision in the prior art. According to the method, the vibration signals are converted into the SDP patterns through the SDP method for optimizing the parameters through the Pearson image correlation coefficient method, and fault signal features are effectively, visually and fully represented on polar coordinates. And the difference embodied by the SDP pattern is large, so that the diagnosis precision can be improved. In addition, the improved residual neural network ResNeXt50 has a self-adaptive fallback mechanism, so that negative effects caused by network overfitting can be effectively avoided. Model parameters are trained by introducing transfer learning, the convergence speed of an optimization function is increased, the training cost of the model is reduced, and the diagnosis