The fault diagnosis of the analog part is the bottle-neck of the fault diagnosis in the mixed-signal IC, which restrict the development of the whole fault diagnosis in the mixed-signal system. Thus, the fault diagnosis of analog circuits is always a hot subject, and it has a very important practical significance.Develope on the basic of statistical learning theory, Support Vector Machine (SVM) is one of the advanced machine learning algorithm in the area of pattern identification. Due to using the principle of structural risk minimization instead of empirical risk minimization, SVM can handle the problem of small sample size better. As a result of using the idea of kernel function, it can turn the problem from nonlinear space to linear space in order to reduce the complex rate of the problem.The fault diagnosis method of circuits based on SVM with composite kernel function researched in this paper is just the application of SVM----an artificial intelligence and machine learning method.Kernel Principal Component Analysis (KPCA) maps the input vectors into a higher dimension feature space to get the nonlinear principle components though a nonlinear mapping which has been selected. KPCA is used to extract the feature of the fault circuits in order to reduce the dimension of the feature vectors.The classification of nonlinear problem has been insloved effectively by SVM with kernel functions, while different kinds of kernel functions can lead to different performance of classification results. Three kinds of novel composite kernel functions have been construced in the paper, on that basis, the method of fault diagnosis in analog circuits based on SVM with composite kernel functions has been gived out.A fault diagnosis platform based on SVM with composite kernel function has been developed. which realizes fault diagnosis in analog circuits based on SVM with composite kernel functions. Four kinds of single kernel functions compare with composite kernel functions which are constructed in the paper through putting four different circuits into the fault diagnosis platform to simulate:the result proves that the method of fault diagnosis based on SVM with composite kernel functions is effective and feasible. |