Abstract: A radial basis function probabilistic neural network (RBPNN) fault diagnosis method based on particle swarm optimization (PSO) algorithm is proposed for the stator winding fault problem of brushless DC motor. Establish a mathematical model for field circuit coupling of brushless DC motor, calculate the stator three-phase current after turn to turn short circuit, phase to phase short circuit, and ground fault of the motor, and construct a sample library using the mean, variance, root mean square, and third harmonic components of the current as fault characteristic quantities. Using PSO algorithm to optimize the structure and initial values of RBPNN, establish a PSO-RBPNN fault diagnosis and classification model. Using the sample library as input for PSO-RBPNN, complete the diagnosis of stator winding faults in brushless DC motors. The results show that the proposed method has a diagnostic accuracy of over 96% and can reliably diagnose different types of stator winding faults.