Fault Diagnosis of Analog Circuits Based on Multi Classification SVDD Aliasing Region Identification

Abstract

In order to solve the problem of poor diagnostic accuracy of aliasing samples in multi class support vector data description (SVDD) method, a multi classification SVDD algorithm with heterogeneous samples is proposed. The method is based on the common SVDD hyper sphere model, in the presence of aliasing in the regional category, all the samples for the target class, other classes with aliasing samples were heterogeneous, using SVDD algorithm with heterogeneous sample re training, until all the hyper-sphere after optimization. Simulation results show that the algorithm can eliminate aliasing and improve the accuracy of the algorithm, and the algorithm is applied to analog circuit fault diagnosis. Compared with SVDD multi classification algorithm, one to one and one to many SVM algorithms, this method has higher diagnostic accuracy in analog circuit fault diagnosis.