一种新颖的基于混合故障字典方法的模拟故障诊断策略  

A Novel Strategy of Analog Fault Diagnosis Based on Hybrid Fault Dictionaries

作　　者：崔江[1] 王友仁[1] 

机构地区：[1]南京航空航天大学自动化学院,南京210016

出　　处：《电工技术学报》2013年第4期272-278,共7页Transactions of China Electrotechnical Society

基　　金：航空科学基金(2009ZD52045);南京航空航天大学基本科研业务费专项科研(NN2012005)资助项目

摘　　要：针对模拟电子电路的故障诊断和定位问题,提出了一种基于one-against-rest支持向量机分类器(SVC)的混合故障字典决策策略。首先,借助于欧式距离计算对待测样本进行粗略定位;其次,利用SVC构造基于符号分析的故障决策机制,并根据算法对样本进行准确定位。相对于常规的one-against-rest SVC方法而言,新方法简单且减少了冗余计算,因而测试时间显著减少,但诊断精度与常规one-against-rest SVC和one-against-one SVC等方法接近,甚至更优。仿真和物理实验均验证了方法的有效性。Focusing on the problem of analog electronic circuit fault diagnosis and location, a novel hybrid fault dictionaries（FDs） based on the one-against-rest support vector machines classifier （SVC） is presented. Firstly, the Euclidean distances between the testing sample and all fault centroids are calculated. Secondly, the signal analysis mechanism is employed for the SVC and the fault can be located accurately by the designed algorithm. The presented method avoids some unnecessary calculations; hence the testing time needed is reduced significantly. Compared to the conventional SVC, such as one-against-rest SVC and one-against-one SVC, the proposed method is simple and fast, also, the performance of this classifier is near or even superior to its counterparts. The simulated and physical experiments validate the presented method.

关 键 词：模拟电子电路 故障诊断 混合故障字典 支持向量机分类器 符号分析 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]