Least squares support vector machine based Analog-Circuit Fault Diagnosis using wavelet transform as preprocessor

Analog fault diagnosis has been an active area of research since the mid-1970s, now many diagnosis methods use neural networks. But it needs lots of fault samples and it is also not easy to train the neural network. We have presented a analog-circuit fault diagnosis method based on LS-SVM. To reduce the fault feature vectors to train LS-SVM, we use the energy of high frequency of wavelet transform coefficients (detail signals) of various levels as the fault feature vectors as fault features of analog circuits. The simulation experiment results show that it need less fault samples, and produce higher class correct rate, and computation time is less than Neural Networks.