Research Of Analog Circuits Fault Diagnosis Optimization Theory And Method

With the urgent need of modern industrial applications, and electronic technology especially the rapid development of microelectronics tech-nology, highly complex and integrated analog circuit and the number of systems has also exponentially grown. It become an urgent need to solve problems of practical engineering that study how to use modern diagnostic techniques to accurately diagnose fault sub-circuits and components of circuits, it is also key steps of analog circuit fault diagnosis theory and methods to the practical application.The main contributions of this paper include as follows:(1)A wavelet neural-network analog fault diagnosis method is studied in detail. Firstly, From the neural network, wavelet analysis basic theory and methods of analysis, two methods based on wavelet neural network analog circuit fault diagnosis are presented,(a), wavelet transform is used to fault signal pre-processor in analog circuit,(b), a layered, multi-resolution neural networks is constructed based on the wavelet theory, That is, wavelet function is used instead of the usual nonlinear Sigmoid function or common weights of network. Finally, The two methods are used to identify the circuit fault diagnosis of analog circuits.(2) A wavelet neural-network analog fault diagnosis method based on GA is studied in detail. Based on analog circuit fault diagnosis WNN method, the genetic algorithm is used to optimize BP neural network, in order to find the most appropriate neural network parameters. The method combined the advantages of genetic algorithms, neural networks and wavelet transform, the search space of genetic algorithms is reduced, search efficiency is improved, the structure of neural network is simplified, and the network training time is reduced, so it can avoid shortcomings of relying on experience and testing to determine the structure and parameters of neural networks, and more easily converge to the optimal solution. The diagnosis steps are given in the paper. It indicate the superiority of the proposed method that example is diagnosed to further verify the accuracy of the method. (3) The basic theory of the artificial immune algorithm and fuzzy sets, and several common methods are introduced in the paper. The general process of fuzzy immune algorithm is studied in detail. And the application of analog circuit fault diagnosis combined artificial immune algorithm with fuzzy algorithm is proposed. The artificial immune algorithm has played on the role of learning samples, looking into cluster centers of each sample group; and the fuzzy classification algorithm is complete accurately classification task of samples, so it is an area worthy of study How to effectively combine artificial immune algorithm and fuzzy classification techniques and the application in analog circuit fault diagnosis. Finally, the method is proposed in the paper and BP neural network are compared to diagnose analog circuit fault, the compares results prove that the method is proposed in the paper has obvious advantages.(4) The method based on symbolic analysis and modeling techniques is studied in the paper. In the method the transfer function of the CUT is used to fault diagnosis equation, a specific frequency excitation is applied to nodes can be measured, detection and localization of fault are achieved through measuring the gain and phase response of the circuit. Compared with the traditional fault diagnosis method, the The method has the following characteristics:fast diagnostic speed, high fault identification rate, small amount of calculation, and the diagnostic process is simple and intuitive, diagnostics is implement easyly.This method is faster diagnosis, fault identification rate, small amount of calculation, and the diagnostic process is simple and intuitive, easy to implement diagnosis.(5) In the paper, large-scale circuit fault diagnosis tear rules and how to apply neural network to hierarchical diagnose the various sub-circuits the large-scale circuit are studied. In the method, firstly, the best incentive test circuit, information of node voltages are extracted. Then wavelet packet coefficients of these information are extracted, their feature vector are obtained. Fusion feature vectors are used as input sample of neural network, the output of the neural network is circuit fault information. Finally, these sub-networks are consolidated; faults of the circuit are diagnosed.