关键词: 局部均值分解, 分段幂函数插值, 端点效应, 最小二乘支持向量机, 模拟电路, 故障特征提取

Abstract: To solve the problem of few test points and hard to get fault features,an improved Local Mean Decomposition(LMD)algorithm was proposed to extract fault features of analog circuits.Aiming at the disadvantages of low speed and large errors in smoothing process by using moving averaging method in traditional LMD,a method of piecewise power function interpolation was used to process the extreme points and to obtain the upper envelope and the lower envelop.According to the upper envelope and the lower envelop,the local mean and envelop function were solved.Least Square Support Vector Machine(LS-SVM)was used as an extending method to decrease the end effect.The outputs of test points in circuit were decomposed by using improved LMD and the fault feature vectors were obtained by computing the energy of Product Functions(PF).The results of simulation and experiment showed that the method could extract the fault feature vectors of the circuit quickly and accurately.

Key words: local mean decomposition, piecewise power function interpolation, end effect, least square support vector machine, analog circuits, fault feature extraction