Aiming at the problems such as slow convergence rate and low fault state recognition rate in the training process of permanent magnet synchronous motor fault diagnosis method, a fault identification method combining complete EEMD with Adaptive Noise (CEEMDAN) and convolutional neural network (CNN) was proposed. The method imported the current, voltage and dynamic magnetic field signals of the motor Full noise assisted empirical mode decomposition of polymerization using Spearman correlation coefficient (Spearman correlation) signals are reconstructed, and then the reconstructed signals are fused to fully extract signal features, and then the convolutional neural network model is established for the fused signals. Finally, the fault diagnosis of permanent magnet synchronous motor is completed based on the results obtained from the convolutional neural network. Data analysis shows that the proposed model has small error and high accuracy compared with other models, and can effectively diagnose faults.