Analog Circuit Fault Diagnosis with Hybrid PSO-SVM

Recent analog circuit fault diagnosis literature has shown that support vector machine (SVM) methods generally outperform traditional statistical and neural methods. However, there are still open issues that, if suitably addressed, could allow further improvement of their performances in terms of classification accuracy. Two especially critical issues are: 1) the determination of the most appropriate feature subspace and 2) model selection. In this paper, these two issues are addressed through a classification system that optimizes the SVM classifier accuracy for analog circuit fault diagnosis. This system is based on a hybrid PSO optimization framework formulated in such a way as to detect the best discriminative features and to estimate the best SVM parameters in a completely automatic way. The effectiveness of the proposed classification system is assessed by fault diagnosis experiment on leapfrog filter circuit.