【摘要】 基于统计学习理论结构风险最小化原则和VC维理论的支持向量机对小样本决策具有较好的学习推广性。由于基本支持向量机算法最初是针对两分类问题推导出来的,在解决故障诊断这种典型的多类分类问题时存在困难。针对模拟电路故障诊断问题,在分析比较支持向量机"一对多"和"一对一"多分类算法的基础上,构建改进的串行支持向量机多分类方法,并依据该算法建立了多故障分类器。将其应用于典型的电源电路故障诊断,仿真试验结果证明了该方法的有效性。更多还原

【Abstract】 Support vector machine(SVM) based on VC dimension theory and the principle of structural risk minimization(SRM) from statistical learning theory exhibits good learning generalization in small sample decision problem.Since basic SVM algorithm was originally designed for binary classification,some difficulties should be overcome while applying it in fault diagnosis problem that is a typical multi-class classification.Aiming at analogous circuit fault diagnosis problem,an improved serial SVM multi-class classification algorithm was designed,which is based on analyzing SVM "one against one" and "one against all" algorithms.Based on the proposed algorithm,a multi-fault classifier was designed,and applied to typical analogous power supply circuit fault diagnosis.Simulation results demonstrate the effectiveness of the proposed method.更多还原