A fault diagnosis method based on wavelet leaders multifractal analysis and support vector machine is proposed,
aiming at complex and nonlinear in nonlinear analog circuit faults. Firstly, circuit is excited by artificial multifractal Brownian motion signal. Then, output voltage signals from the test node are obtained, and the multifractal spectrum fault feature vectors, as input samples in support vector machine classifier, are extracted by using wavelet leaders multifractal analysis method. Finally, circuit work modes are classified by support vector machine. Fault diagnosis simulation experiment on video amplifier circuit shows the effectiveness of the proposed method.