A novel linear ridgelet network approach for analog fault diagnosis using wavelet-based fractal analysis and kernel PCA as preprocessors

We have developed a novel fault diagnosis approach of analog circuits based on linear ridgelet network using wavelet-based fractal analysis, kernel principal components analysis (kernel PCA) as preprocessors. The proposed approach can detect and identify faulty components in the analog circuits by analyzing their time responses. First, using wavelet-based fractal analysis to preprocess the time responses obtains the essential and reduced candidate features of the corresponding response signals. Then, the second preprocessing by kernel PCA further reduces the dimensionality of candidate features so as to obtain the optimal features as inputs to linear ridgelet networks. Meanwhile, we also adopt the kernel PCA to select the proper numbers of hidden ridgelet neurons of the linear ridgelet networks. The simulation results show that the resulting diagnostic system using these techniques can not only simplify the architectures (including input nodes and hidden neurons) and minimize the training and processing time of these networks considerably, but also diagnose single and multiple faults effectively in classifying faulty components of example circuits to improve the accuracy and efficiency of fault diagnosis with a highly correct classification rate.

► We develop a novel fault diagnosis approach of analog circuits.
► We perform wavelet-based fractal analysis to get the candidate features.
► Kernel PCA is used to further extract the optimal features.
► Adopting kernel PCA selects the structure of a linear ridgelet network.