Fault diagnosis based semi-supervised global LSSVM for analog circuit

Focusing on the issue of fault diagnosis method in existing analog circuit, the nonlinear of circuit and the global and local underlying information of fault should be considered simultaneously. Global and local preserving of fault information is proposed for semi-unsupervised learning algorithms of analog circuit faults diagnosis, the within-class scatter of linear discriminant analysis (LDA) combines the locality preserving projections (LPP) to fully consider the global and local geometric structure between samples and make the method in succession on the basis of the characteristics of the traditional SVM methods in this paper. Experiment takes the pass filter as the diagnosis circle to prove that proposed method in this paper can highly recognize the known and unknown faults comparing with fault diagnosis method based on common SVM.