Fusion of Information for Fault Diagnosis in Analog Circuits

This work proposes a method of analog circuit fault diagnosis by fusion of information from four different domains, time, frequency, wavelet and statistical domain. Fusion of features has been performed by Canonical correlation analysis (CCA), an efficient method of finding statistical dependencies. The problem of information fusion of analog circuits has been converted as generalized eigenvalue formulation with the help of CCA. The fused features constructed by CCA are used to train SVM classifier for fault diagnosis of analog circuits. The proposed method is illustrated with the example of band pass filter circuit and leapfrog low pass filter circuit. The accuracy of fault classification of the proposed method with fused features is found considerably higher than that with individual domain features.