In order to reduce downtime caused by power electronic equipment faults and maintain devices in advance, the research of power electronic circuit soft fault diagnosis is of great significance. Therefore, the soft fault diagnosis methods based on back propagation neural network (BPNN) and support vector machine(SVM) are studied . Single output and multiple output BPNN models are used in the fault diagnosis methods, respectively. As for SVM, onevsall(OVA) and onevsone(OVO) algorithms are adopted. Taking DC/DC conversion circuit with feedback control as an example, the effectiveness of these diagnosis algorithms are verified and the fault diagnosis performances are evaluated.