The reliable operation of power supply is very important to the operational effectiveness of surface-to-air missile weapon system. In order to realize fault diagnosis and fault tolerant operation of surface-to-air missile power supply inverter, the wavelet packet analysis and Elman neural network were introduced. Wavelet packet analysis combined with Elman neural network were applied to the power failure diagnosis of missile power supply system for the fault feature extraction and fault identification. On the basis of accurate fault diagnosis, fault isolation circuit is used to isolate fault components. The backup bridge arm is put into use to replace fault components so that the system can continue to operate normally. The results of the fault diagnosis and fault reconfiguration simulation verify the effectiveness of the method.