Electronic Circuit Fault Diagnosis Based on SCA-SVM

With an eye to solving the problems on lack of the sample and low accuracy in analog circuit fault diagnosis, one effective method based on the Sine Cosine Algorithm(SCA) is proposed for soft fault diagnosis in analog circuit with tolerance. Primarily, monte carlo analysis is performed on the experimental circuit and then the output voltage signal collected as dataset. Then the wavelet entropy is calculated from dataset transformed by the wavelet transform, which forms the fault feature dataset. The principal component analysis(PCA) is used for lowering the feature dimension. Finally, support vector machine(SVM) optimized by SCA is employed for classification of the fault dataset. Experiment on the Sallen-Key band-pass filter circuit analysis shows that the SCA-SVM classifier has higher classification accuracy and faster iteration speed than GridSearch, Genetic Algorithm and Particle Swarm Optimization. Verification experiment on the four-opamp biquad highpass filter circuit indicates excellent adaptive capacity in analog circuit fault diagnosis.