Analog Circuit Fault Diagnosis Based on Support Vector Machine Classifier and Fuzzy Feature Selection

In analog circuit, the component parameters have tolerances and fault present a wide distribution, which brings obstacle to classification diagnosis. To tackle this problem, article proposes soft diagnosis method combining improved barnacles mating optimizer(BMO) algorithm with support vector machine (SVM) classifier, can achieve minimum redundancy maximum relevance for feature dimension reduction fuzzy mutual information. be concrete, first, optimizer is used optimize learning classification. We adopt six test functions that are on three data sets from University of California, Irvine (UCI) repository performance SVM classifier five different optimization algorithms. The results show combined characterized high accuracy in Second, information, enhanced redundancy, principle applied reduce vector. Finally, circuit experiment carried out verify proposed effectively when both fixed distributed. 92.9% distributed, 1.8% higher than other classifiers average. When fixed, rate 99.07%, 0.7%