中文摘要：

支持向量机用于模拟电路多种故障诊断时,其多分类扩展策略与诊断的效率和正确率密切相关。本文提出模糊聚类与支持向量机集成的算法,通过分析电路故障特征数据的空间分布特性,以多级二叉树结构的SVM实现故障的分级诊断。通过对各故障模式两种小波特征的逐次聚类二分获得二叉树,根据F测度为每个节点的SVM选择具有最大分类间隔的故障子类及特征,避免了不可分故障区域的出现,从而优化了SVM的组合策略。采用该方法组建的SVM结构简单,在滤波器电路的故障诊断中获得良好的效果。与几种常用的SVM方法相比,本文方法有效地提高了故障诊断的精度和效率。

英文摘要：

When Support Vector Machine （SVM） is used for multi-fault diagnosis in analog circuits, the diagnosis accuracy and efficiency largely depend on the extension strategies for multi-classification. A hierarchical decision approach for fault diagnosis is proposed in this paper, which integrates fuzzy clustering and SVM algorithms to form multi-level binary tree SVM classifiers according to the distribution property of fault data sets. The structure of the tree is determined by iteratively dividing faulty class patterns into two clusters with two aspects of wavelet features. During the clustering process, F-measure is adopted for deciding fault subclass partition and feature selection on each tree node, which leads to maximize the separation margin between fault subclasses for SVM on that node so as to eliminate unclassifiable fault regions. As a result, the combination strategy of SVM is optimized. The presented method has been employed for filter circuit diagnosis with simplified classifier structure and satisfactory results have been achieved. The diagnosis precision and efficiency are greatly improved by our method in comparison with other conventional SVM approaches.