基于改进二叉树支持向量机的多故障分类算法  

A multi-fault Classification Algorithm Based on Improved Binary Tree Support Vector Machines

作　　者：李伟伟[1] 王莉[1] 张琳[1] 刘进 

机构地区：[1]空军工程大学防空反导学院 [2]解放军95425部队

出　　处：《探测与控制学报》2015年第3期34-39,共6页Journal of Detection & Control

摘　　要：针对二叉树支持向量机(SVM)的结构设计会影响多类分类器精度的问题,提出了将层次分析法(AHP)和二叉树支持向量机相结合的多故障分类算法。该算法首先运用层次分析法建立评价体系模型,综合衡量多个影响因子确定各类故障的权重,然后根据权重大小对故障进行排序,由故障排列顺序设计二叉树支持向量机的结构,最后进行故障诊断分析。仿真验证表明:该算法适合进行多类故障诊断,相比其他算法诊断效率更高,诊断精度更好,推广应用前景较广。Concerning that the structure design of binary tree support vector machines（SVM）had a great influence on multi-class classifier＇s accuracy,a multi-fault classification algorithm that combines analytic hierarchy process（AHP）and binary tree support vector machines was proposed.First,the model of an assessment system was established based on analytic hierarchy process and the weight of faults was confirmed by making a comprehensive survey on several factors,then the faults were put in the right order on the basis of the weight and the structure of binary tree support vector machines was designed;finally,the algorithm was used for fault diagnosis analysis.Simulation verification showed that the algorithm performed well in multi-class fault diagnosis,the efficiency and accuracy of diagnosis were better than others method.

关 键 词：二叉树 支持向量机 层次分析法 轴承故障诊断 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]