Research On Fault Diagnosis Of Nonlinear Analog Cirecuit Based On Volterra Series

Since the1960’s, Analog circuit fault diagnosis has been the emphasis in the fieldof circuit and system and gets large mount of theory achievements. Because of thecomplexity of the analog circuit, the development of the analog circuit fault diagnosis isstill very slow. Especially for fault diagnosis of nonlinear analog circuit, it does not exista generally applicable theoretical model for diagnosis so that there is not the unifiedmethod of analysis. As for the rapid development of Volterra series theory andincreasingly perfect of the classification theory with support vector machine (SVM), ithas become a hot spot of research direction to analyze the nonlinear analog circuits withVolterra series and make the faults classification with SVM. The paper analyzes theapplying of the Volterra series theory into the nonlinear analog circuits response, putsforward the methods of circuits faults testing with SVM. The major contents:1. The paper studies Volterra time domain nuclear identification method based onthe recursive least squares (RLS), avoid the shortcomings of the large counting amountof common least square method and large occupation of data storage space. Thesimulation experience shows that the method of nonlinear system identification Volterratime domain nuclear has a high precision and good antinoise ability.2. Through in-depth study the feature of SVM model identification in linearseparation and linear nonseperation, it adopt a SVM fault classifier design methodincluding the formation of classifier topological structure, choosing of nuclear function,confirmation methods of parameters and classification progress of multi faults.3. In order to solve nonlinear analog circuit classification issues, the paper adoptsnonlinear analog fault diagnosis method based on Volterra time domain nuclear andSVM. It gets the diagnosis results through identifying the Volterra time domain nuclearwith RLS to complete fault feature extraction, classifying fault models with SVMmultiple fault classifier.An example test showed that this method is efficient.