The Research On Fault Diagnosis Methods In The Analog Circuit Based On Genetic Algorithm And BP Neural Network

With the continuous improvement of the electron components’manufacturing process, the gradually increase of the varieties of circuits’ integration, especially, the complexity of the structure of the analog circuit increases, thus, the relevant research corresponding to the fault diagnosis technique of the analog circuit becomes more important. In the face of the lots of disadvantages of the BP neural network used for circuit diagnosis, including lower diagnosis efficiency and short of effective convergence, all kinds of modified BP neural networks algorithm emerge, for example, additional momentum algorithm、statistical algorithm and competition BP algorithm, etc. Generally speaking, the above mentioned algorithms improve the learning rate by means of randomly selecting BP network weights and threshold, and it gains certain diagnosis effect, however, it does not optimize the network structure and adjust the weights and threshold momentarily to make the network best. Therefore, in this article, we do the work from the view of optimizing the weights and threshold of the BP neural network, and make fault diagnosis of the analogy circuit by using the genetic algorithm to optimize the BP neural network, further, to reach the high efficiency and high precision of diagnosis. The main research contexts are listed as follows:(1) We use the circuit simulation software OrCAD to draw the schematic diagram of the analog circuit. Then, we make the direct current (DC) sensitivity analysis of the circuit and construct the fault type according to the result of the DC sensitivity analysis on the electronic component.(2) We make the Monte Carlo (MC) analysis to the analog circuits, and gain the circuit parameter information in all kinds of fault condition. Besides, the information has been integrated and used, and is divided into the training sample set and the testing sample set.(3) We make the fault diagnosis to the circuits by using the traditional BP neural network. In addition, we view the training step and training curve error of the traditional BP neural network, and record its fault diagnosis rate.(4) We optimize the BP neural networks by means of the simple generic algorithm (SGA), to make the weights and threshold of the BP neural network optimization and the networks to achieve the global optimal solution. Then, we use the network to make the fault diagnosis of the circuits, observe and apply the algorithm to diagnose the circuit, and record the fault diagnosis effect.(5) We modify the coding pattern of the SGA, use real number codes and a series of the relevant genetic operator to optimize the BP neural networks. Besides, we apply it to make the circuits’fault diagnosis, and compare the diagnosis effect between the modified algorithm and SGA to obtain the merits of the modified algorithm. Finally, we put forward the research direction of the further work.