Study Of Fault Diagnosis Of Analog Circuit Based On Combinating Support Vector Machine And Genetic Algorithm

The applications of large-scale integrated circuits allow the scale and the structure of network to be more and more functioned and modularized with the development of modern electronic industry. The difficulty of circuit fault diagnosis increases obviously. Once the breakdown happens, it may cause great loss of property. So it is an urgent subject in practical project to study how to identify the fault elements correctly by using modern diagnosis technique for large-scale circuits with tolerance, and it is also a key step for practical application of fault diagnosis for analog networks theory and technology. The fault diagnosis methods based on artificial intelligence and machine learning could construct multi-classification learning models and make fault diagnosis, which is a hot research subject.Based on the Liaoning province nature science fund project 'New methods based on support vector machine study for fault diagnosis and forecast of electric and electron system', the thesis researches the fault diagnosis method of analog circuit based on combinating support vector machines and genetic algorithms.Because analog circuits are usually with tolerances and nonlinear, the voltage and current of different nodes are sensitive to different fault components, this paper simulate the circuit with the software of PSpice. The special functions of PSpice of Monte carlo analysis and the worst case analysis are used to collect different fault information as the input of Support vector machine classifier.Support vector machine is a novel technology of data mining, and new tool of soluting machine learning problem with the method of optimization. It contains the largest marging hyperplane,Merecr kernal,convex programmmg,sparse solution and slack variable techologies. The SVM-based method has advantage of artificial neural network methods in Multi-class classification. But the parameters of the kernel function which influence the result and performance of support vector machine have not been decided. This paper improve the support vector machine in parameters optimization algorithm by introducing genetic algorithms to optimize the choice of the kernel function parameters and the error penalty parameter C, and it applies to Multiclass Classification such as 1-versus-rest,1-versus-1 and decision directed acyclic graph methods for fault diagnosis of analog circuit. The results of simulation show that the generalization performance of the support vector machine can be improved clearly by the proposed method in this paper compared with the grid search method through the experiment.