Theory And Methods For Fault Diagnosis Of Analog Circuits Based On Neural Network And Wavelet Transform

Analog circuit fault diagnosis began in the 70's, have obtained fruitful production all over the world, and gradually formed systemic theory, established it's status in the network theory, and become the third branch following the net analysis and the net synthesis. However, the approaches are limited to deal with fault diagnosis because of the variety and complexity of the analog circuits, especially the large-scale analog circuits with the tolerance or the soft fault. With the development of electronic technology, especially VLSI and mixed signal circuits, it brings up new challenge to fault dianosis of the analog circuits, and requires the new theory and approaches for fault diagnosis of the analog circuits.Artifical Neural Network(ANN) have gained quick development and have been widely applied in many areas. This paper main study the Neural Network based approach for fault diagnosiso of analog circuits with tolerance, especially discussed the error back propogation Neural Network(BPNN) in detail, give examples for analog circuits fault diagnosis using improved error back propogation Neural Network. As for multiple and soft fault in analog circuits, the paper attempt using Self-Organizing Feature Map(SOFM) to diagnosis. The example indicate that Artifical Neural Network has associational and memorial abilities, strong study ability, strong robust and non-linear mapping ability, which make this diagnosis method better than traditional fault dictionary methods.This paper combines Wavelet Transform(WT) with ANN and presents a Wavelet and Neural Network method through unite loose for fault diagnosis. Due to WT have better part characteristic in time-domain and frequency-domain and stronger character extracting function for signal, using Wavelet Transform, decompose fault signal in levels and acquire signal component in various frequency segment, using the ingredient that could reflect the fault signal character as circuits fault character, and input ANN, by this method, reduce the ANN's input number,simplify the ANN's structure and reduce the ANN's training time, enhance the ability of identifying fault sort.This paper researchs deeply on modified nodal decomposition approach for large-scale analog circuits. The suitable self-testing conditions and mutual-testing conditions are developed for all kinds of decomposition cases in which subnetworks satisfied solvability conditions. The diagnosis example is given. The approach provides an even more efficient method for locating fault subnetworks in large analog circuits.