基于粒子群优化决策树的齿轮箱故障诊断  

Constructed of SVM Decision Tree Based on Particle Swarm Optimization Algorithm for Gear Box Fault Diagnosis

作　　者：程珩[1] 黄超勇[1] 张永刚[1] 

机构地区：[1]太原理工大学机械电子工程研究所,太原030024

出　　处：《振动．测试与诊断》2013年第1期153-156,172,共4页Journal of Vibration,Measurement & Diagnosis

基　　金：山西省自然科学基金资助项目(2011011026-3)

摘　　要：针对现有支持向量机(support vector machines,简称SVM)在构造多类分类器的过程中存在计算费时、搜索率不高的问题,提出了一种新的SVM决策树设计算法。引入具有优良的全局搜索性能的粒子群算法,将其应用于优化决策树,构造出一种自适应性强、识别率高的多元分类器,实现SVM的有效多值分类。将其结果应用于齿轮箱故障诊断中,试验结果证明改进后的SVM构造方法的有效性和准确性。In order to solve the problem of consuming too much calculation time and low rate of searching in the construction process of multi-class classification of support vector machines(SVM),the fine global search performance of particle swarm is applied to the optimization of decision tree in order to construct a diverse classifier and ultimately achieve more effective multi-value classification for SVM.The improved SVM constructing method is proved more effective and accurate while being applied to gearbox fault diagnosis.

关 键 词：粒子群 决策树 支持向量机 故障诊断 

分 类 号：TH165.3[机械工程—机械制造及自动化]