Permanent Magnet Synchronous Motor(PMSM)is finding wider application in electromechanical actuation system, and has higher demands to the fault diagnosis.Therefore, we proposed a new method of fault diagnosis based on Local Mean Decomposition(LMD)and improved frequency-band energy method.This method unfolds the energy distribution of the fault signal in frequency domain by decomposing it into amplitude and frequency modulated signals.Then the feature vector is constructed according to the energy distribution and the correlation between each frequency band and the fault signal, thus to realize feature extraction.Finally, simulation is made for the speed control system of PMSM.The results turn out that the proposed approach can extract the fault feature effectively, while achieving dimension reduction of feature vector.