Nonlinear Analog Circuit Fault Diagnosis Method Based On Neural Networks

Analog circuit fault diagnosis theory began been studied since the mid1970, up to now many achievements have been made, however, due to difficulties of analog circuits fault diagnosis itself, lead to research in this area in a considerable amount of progress has been slow for some time. Traditional diagnostic methods and theories to single process, single failure of and failure of simple system can play a better role, but for multiple processes, multiple failure and sudden failure of large and complex, highly automated equipment and systems, has a significant limitation. With the development of artificial intelligence techniques such as neural networks, analog circuit fault diagnosis research has opened up a new road; analog circuit fault diagnosis method based on neural networks has become the new hot research.This article on the application of neural network in fault diagnosis in nonlinear analog circuit has conducted in-depth research, to the main object of the study among fuzzy neural network, BP neural network and fuzzy neural networks for analog fault diagnosis compare simulation results, conclusions.This article summarizes the research development of analog circuit fault diagnosis technology, analyzes the advantages and disadvantages of a variety of modern diagnostic techniques. Respectively on the neuron model of BP neural network and fuzzy neural networks, network structure as well as a detailed description of the learning algorithm. This paper applies BP neural network to solve nonlinear mapping relationships between fault characteristics of analog circuits and circuit components on the basis of fault diagnosis for nonlinear problems, solve analog circuits fuzzy due to component tolerances using series fuzzy neural network, gives a kind of tandem-type fuzzy neural network based on frequency characteristics of diagnostic methods. Simulation results indicate that the given method can improve fuzzy neural networks insufficient training cycles for fault diagnosis in nonlinear analog circuit, the convergence rate and diagnosis rates have increased significantly. Then based on the study of wavelet packet transform, we used it to denoise the fault signals and extract the fault eigenvalues, and improve the drawbacks of the way of effective sampling point extraction. The result showed that the method with wavelet packet transform effectively enhanced the convergence speed of fuzzy neural network and the accuracy of fault diagnosis.