Realization Of Fault Diagnosis Of Analog Circuits Based On Neural Networks

Fault diagnosis of analog circuit and neural network are currently two hot academic issues. The paper aims to explore a new approach to solving the problem of failure diagnosis of analog electric circuits through combining the latest results of failure diagnosis and neural network fileds.Firstly, the author introduces the basic fault diagnosis techniques and its development, the research results of analog fault diagnosis field in recent years are also be summarized in the paper. The paper expatiates on the general process of the traditional dictionary method for the fault diagnosis of analog circuits, researches the methods of failure diagnosis on analog electric circuits based on the neural network. The paper also introduces the basic algorithm of the BP neural network, Hopfield neural network and SOFM neural network, and the normal way and steps of the failure diagnosis on analog electric circuit. The paper describes in details about the principles, the algorithm model and algorithm realization of self-organizing feature map neural network, analyses its learning regulations, presents the questions encountered in algorithm realization and gives corresponding resolutions to them. The author also discusses the solution to fuzzy-fault-set that may arise in actual application of analog fault diagnosis. The paper applies the SOFM neural network to the fault diagnosis of analog circuit, proposes the fault dictionary technique based on SOFM neural network for fault diagnosis of analog circuit. The associational memory function, error-correction, robustness and the well non-linear reflection ability make the proposed method superior to the traditional dictionary method apparently. Simulating for the fault diagnosis example by the neural network toolbox of Matlab and OrCAD/Pspice9.1 software, the obtained results show that the proposed method in the paper is feasible.