Research On Fault Diagnosis Of Analog Circuits Based On Support Vector Machine And Wavelet Transform

Fault diagnosis of analog circuits play an important role in the design of circuit and theproduction and maintenance of equipment. It is very important to do some research on thesubject. Electronic technology is developing rapidly in recent decades, which promptscontinuous growth of the intensity and complexity of the circuit. So we need betterperformance in efficiency and accuracy of fault diagnosis of analog circuits.Firstly, based on the analysis of the basic fault diagnosis idea, we stated the theory offault diagnosis of analog circuits by wavelet analysis and support vector machine. Waveletanalysis has good robustness. It’s a time-frequency joint method with good partial andmulti-resolution characteristics. Secondly, support vector machine does well in dealing withlearning of small sample, samples of nonlinear and recognition of high dimensional pattern, italso can effectively reduce the computational time of fault classification. We researched thefault diagnosis methods based on wavelet and support vector machine in this paper.We chose CTSV filter and Sallen-key band-pass filter. Then, we ran Monte Carloanalysis about the normal and fault condition of the circuit by Pspice software. Then weimported the simulation data into the workspace of MATLAB, did the signal de-noising workwith wavelet packet and extracted wavelet fault characteristics from the response signal bywavelet, which were de-noised or not.Finally, we got the diagnosis by training and testing the multiclass classifiers, whichwere structured by support vector machine. Comparing the two kinds of diagnosis, we provedthat the de-noising processing will help to improve the accuracy significantly. Compared withother papers, it proved that SVM does better than other methods in improving the accuracy offault diagnosis, so SVM is suitable for fault diagnosis of analog circuit.