Neural-network based analog-circuit fault diagnosis using wavelet transform as preprocessor

We have developed an analog-circuit fault diagnostic system based on backpr opagation neural networks using wavelet decomposition, principal component analysis, and data normalization as preprocessors. The proposed system has the capability to detect and identify faulty components in an analog electr onic circuit by analyzing its impulse response. Using-wavelet decomposition to preprocess the impulse response drastically reduces the number of input s to the neural network, simplifying its architecture and minimizing its tr aining and processing time. The second preprocessing by principal component analysis can further reduce the dimensionality of the input space and/or s elect input features that minimize diagnostic errors. Input normalization r emoves large dynamic variances over one or more dimensions in input space, which tend to obscure the relevant data fed to the neural network. A compar ison of our work with [1], which also employs backpropagation neural networ ks, reveals that our system requires a much smaller network and performs si gnificantly better in fault diagnosis of analog circuits due to our propose d preprocessing techniques.