基于TWSVM的煤矿井下钻机轴承故障诊断  

Based on TWSVM Drill Bearing in Coal Mine Fautt Diagnosis

作　　者：董洪波 申中杰 姚亚峰 

机构地区：[1]中国煤炭科工集团西安研究院有限公司,西安710077

出　　处：《煤矿机械》2015年第5期298-300,共3页Coal Mine Machinery

基　　金："十二五"国家科技重大专项项目(2011ZX05041-002-002);中煤科工集团西安研究院技术创新项目(2012XAYCX010)

摘　　要：在轴承的监测数据中,故障数据所占比例很小。在这种不平衡样本集中,如何准确地诊断轴承故障成为了一大难题。为此,提出运用孪生支持向量机(TWSVM)解决该难题。TWSVM可以构造2个不平行的超平面,每个超平面都要尽可能靠近与其对应的一类而远离另一类。TWSVM的超平面只关注其对应类别的样本分布,而不受整个数据集中各类样本数量比例、其他样本分布的影响。运用轴承故障试验对TWSVM进行检验,结果表明:与支持向量机(SVM)相比,TWSVM拥有更高的识别精度和更快的计算速度,能较好地诊断轴承故障。The fault samples are only the small portion in all monitoring samples of bearing. How to diagnose the roiling bearing fault accurately becomes a challenge in the unbalance sample set. Twin Support Vector Machine （TWSVM） is applied into the bearing fault diagnosis in the study. It aims at generating two nonparallel planes such that each plane is closer to one of the two classes and is as far as possible from the other. The nonparallel plane of TWSVM only focusing on its corresponding sample distribution, without the influence of the other sample distribution and proportion in the whole data set. The fault diagnosis experiments verify that TWSVM has higher accuracy and faster speed than SVM, and identify the bearing fault well.

关 键 词：故障诊断 TWSVM 非平行超平面 滚动轴承 

分 类 号：TD421.4[矿业工程—矿山机电] TH133.3[机械工程—机械制造及自动化]