摘要
针对模拟电路的故障诊断和支持向量机分类器的设计问题,讨论了一种基于有向无环图支持向量机分类器(DAGSVC)的故障字典新方法,并比较了几种支持向量机故障分类器的平均测试复杂度指标。通过对2个实际模拟滤波器的实际测试和验证表明:该方法性能要优于"1-v-r"SVC,"1-v-1"SVC等常规的故障分类器,并和聚类二叉树SVC的诊断性能接近,适合模拟电路的故障分类和诊断。
Focusing on the design of problem of fault diagnosis of analog circuit and classifier with support vector machines(SVMs),a new method of fault dictionary based on directed acyclic graph SVMs classifier(DAGSVC) is presented,and a specification for estimating the average test complexity of the support vector machine classifier(SVC) is also compared.Two actual analog filter are tested to validate the proposed method,whose performance is proven to be superior to the traditional methods,such as "1-v-r"SVC and "1-v-1"SVC.The proposed method,being proper to perform analog circuit diagnosis and faults isolation,could also achieve almost the same diagnosis rate as the clustering binary tree SVC,whose test structure is not unique.
出处
《传感器与微系统》
CSCD
北大核心
2011年第4期12-16,共5页
Transducer and Microsystem Technologies
基金
安徽省高校自然科学基金资助项目(KJ2010B308)
南京航空航天大学基本科研业务费专项科研计划资助项目(NS2010063)
关键词
模拟电路
故障诊断
故障字典
平均测试复杂度
有向无环图支持向量机分类器
analog circuit
fault diagnosis
fault dictionary
average test complexity
directed acyclic graph support vector machines classifier(DAGSVC)