基于节点优化的决策树支持向量机及在故障诊断中的应用  

Support Vector Machine Based on Optimal Decision Tree and Use in Fault Diagnosis

作　　者：王晴晴[1] 汪正东[1] 黄衍法[2] 赵小虎[1] 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221008 [2]山东省邹城市兖矿集团东滩煤矿供电工区,山东邹城273512

出　　处：《煤矿机械》2010年第6期241-243,共3页Coal Mine Machinery

摘　　要：针对目前故障诊断中,难以获得大量的故障数据样本以及诊断知识获取困难等不足,提出了专门针对有限样本的新一代机器学习的算法—支持向量机(SVM),它在样本很少的情况下具有较好的泛化能力,比较适合解决故障诊断小样本情况的实际问题。在多故障诊断时,必须先进行多分类扩展.决策树是一种性能优秀的多分类扩展策略,但该方法的决策结果与结点的排部密切相关,结点的排部影响了诊断的正确率.提出一种根据故障数据的空间分布来优化结点排部的方法,它能够提高支持向量机诊断的正确率.采用该方法扩展的多分类支持向量机在故障诊断中获得良好效果.At present, how to gain a lot of sample and data is difficult in fault diagnosis. Thinking about this, a new generation of machine learning algorithm for finite sample is put forward -- support vector machine （SVM）.It has good generalization ability and suitable for fault diagnosis of small sample solution in practical problems. In fault diagnosis, classification must be expanded first. The decision tree is an excellent performance of expansion strategy, but this diagnosis accuracy based on the classification method is closely related to the node platoon. Optimized node on the basis of fault data of spatial distribution is proposed. The diagnostic accuracy can be improved. The results of the method in the fault diagnosis are satisfactory.

关 键 词：支持向量机 故障诊断 多分类 决策树 结点优化 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] TP306[自动化与计算机技术—控制科学与工程]