期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2012
卷号：46
期号：2
页码：0966-0976
出版社：Journal of Theoretical and Applied
摘要：Fault diagnosis of analog circuit is of essential importance for guaranteeing the reliability and maintainability of electronic systems. Taking into account the requirements and characteristics of analog circuit fault diagnosis, two-level diagnostic structure for analog circuit is proposed in this paper. Analog circuit fault diagnosis can be regarded as a pattern recognition issue and addressed by multi-class SVM. Aiming at the uncertainty of the node arrangement and the error accumulation phenomenon, the improved directed acyclic graph support vector machine (DAGSVM) based on fisher separability measure in high dimensional feature space and margin of SVM is proposed. In order to eliminate redundant fault features and the impact of measure noise on the diagnostic accuracy, simultaneously taking into account lightening the workload of the classifier, the separability promotion algorithm based on kernel principal component analysis (KPCA) plus linear discriminant analysis (LDA) is proposed. Then the separability promotion algorithm is used to extract the nonlinear fault features of analog circuit and the two-level structure based on improved DAGSVM is applied to diagnose faults in analog circuit. The unbalanced classification model based on SVM is adopted in the first level. The effectiveness of the proposed method is verified by the experimental results.
关键词：Fault Diagnosis; Analog Circuit; Improved DAGSVM; Separability Promotion Algorithm; Kernel Principal Component Analysis
