Pulsed power supply is the most vulnerable part of the electromagnetic launch system, and the failure of the pulsed power supply will make the whole system performance decline. Thus, in view of the pulse power supply failure, a model search support vector machine fault diagnosis method based on multi-layer wavelet analysis for fault information extraction was proposed. By establishing the soft fault model of pulse power supply, the current fault data samples were obtained through simulation analysis, and the discrete wavelet decomposition of the fault samples was carried out, then the wavelet coefficients of the specified layer details were obtained as the fault feature quantities. The principal component analysis of the fault feature was carried out, and the dimension of wavelet coefficients was reduced, so that the fault diagnosis was effectively carried out. Through the experiment, the proposed fault diagnosis method was compared with the other three kinds of fault diagnosis results, and the effectiveness of the method was verified.