摘要
支持向量机(Support Vector Machines,SVM)故障分类器,在不易取得训练样本的情况下,实现较高准确率的故障诊断,并且具有较强的通用性和实用性。提出三种支持向量多类分类器(一对一算法、一对多算法,以及改进型一对多算法),通过将其应用到实际电路进行故障诊断当中对其性能进行比较,得出串行支持向量机无论在分类速度上还是在分类精度上都好于其它两种方法,核函数的选择对故障诊断的性能也存在着一定的影响。
Under uneasy achievement of training samples, fault classifier based on support vector machine （SVM） can diagnose faults with high accuracy and strong versatility and practicabillity. We proposed three multi-class classifiers including one-to-one algorithm, one-to-many algorithm, and improved one-to-many algorithm. Compared with their performances by using in practical circuit fault diagnosis, the results show the serial SVM is better than the other two methods in some aspects such as classification speed and accuracy, and the selection of Kernel function has influence on the performance of fault diagnosis.
出处
《兵工自动化》
2009年第4期79-81,共3页
Ordnance Industry Automation
关键词
支持向量机
多类分类
模拟电路
故障诊断
SVM
Multi-class classification
Analog circuit
Fault diagnosis
