Abstract: In order to improve the accuracy of motor bearing fault identification, a rolling bearing fault diagnosis method based on Variational Mode Decomposition（VMD）, Multi scale Fuzzy Entropy（MFE）, and Probabilistic Neural Network（PNN） is proposed. The two important parameters of VMD are optimized using the optimized genetic algorithm; VMD is used to decompose various bearing vibration signals, and the optimal modal components containing more fault characteristics are selected based on the kurtosis correlation criterion; The multi-scale fuzzy entropy of the component is calculated, and a certain scale fuzzy entropy value is selected as the feature vector, which is input into PNN for fault identification. Compared to the VMD-PE-PNN, VMD-FE-PNN, and VMD-MPE-PNN methods, the motor bearing diagnosis method based on VMD-MFE-PNN can more accurately identify the fault types of rolling bearings.