A Naive-Bayes-Based Fault Diagnosis Approach for Analog Circuit by Using Image-Oriented Feature Extraction and Selection Technique

Analog circuit is one of the most commonly used components in industrial equipment, and circuit failure may lead to significant causalities and even enormous financial losses. To address this problem, a novel scheme based on the wavelet spectrum features, feature selection, and Naive Bayes classifier is presented for the fault location of an analog system in this paper. The scheme mainly consists of three stages. First, the cross-wavelet transform (XWT) method is utilized to obtain the time-frequency representations of the raw signals of analog circuits. Second, the local optimal-oriented pattern is applied to all the XWT spectrum images to generate the original high-dimensional feature set. Then, an integration feature selection approach via joint Hilbert-Schmidt independence criterion and kernel Fisher linear discriminant analysis is proposed and utilized to obtain low-dimensional fault features, which are uncorrelated and distinctive. Finally, the training samples set is imported into the Naive Bayes classifier, and the fault diagnosis results can be drawn through inputting the testing samples set into the trained Naive Bayes classifier. The simulation results on two typical circuits have demonstrated that the proposed method is a promising means to detect and classify most analog circuit faults, achieving a better diagnosis accuracy than that of the other published works.