For more accurate diagnosis and location of faults in switched current circuits, a new faultdiagnosis approach is proposed based on wavelet transform and particle swarm optimization（PSO）support vector machine(SVM). Monte Carlo analysis is applied to node current signals, followed by wavelet decom position, fractal dimension calculation, and kernel principal component analysis （KPCA）, aiming at abstracting optimal fault features and reducing signal redundancy. Finally, the classification of various failure modes is accomplished by PSO-SVM. A 100% accuracy of fault diagnosis is obtained in simulation experiments for verification done with a sixth-order Chebyshev low-pass filter. Compared with other approaches, the proposed approach is superior with the support of the experimental results.