摘要
提出基于支持向量数据描述(SVDD)的滚动轴承声发射特征的单分类智能诊断方法,适用于故障监测和诊断中缺少故障样本的情况。首先利用谐波小波优良的频域盒形特性,将滚动轴承声发射信号分解到相互独立互不重叠的若干频带内,然后求取主要频带内信号的能量并归一化处理,进而将归一化能量特征作为特征向量输入SVDD分类器中进行故障识别和分类。试验表明,该方法只需要正常轴承声发射特征作为学习样本,不需要其他非目标样本即可实现故障轴承的识别,与支持向量机分类方法比较具有更高的准确率。
The one - class classification intelligent diagnosis method is proposed for rolling bearing acoustic emission feature based on support vector data description（SVDD） , which solves the problem of insufficient fault samples in intel- ligent monitoring and diagnosis for machinery. Firstly, using the excellent box - like spectrum of harmonic wavelet, the acoustic emission signals of beating are decomposed into several independent and non 2 overlapped frequency bands, and the energies in these frequency bands are calculated and normalized. Finally, the normalized energy character is input to the SVDD for fault identification and classification. The results show that this method needs only normal condi- tion signal as target samples, the identification of fault bearings is realized without another target samples. The experi- ment results show the new method has higher accuracy than the support vector machine.
出处
《轴承》
北大核心
2012年第9期43-46,58,共5页
Bearing
基金
国家自然科学基金(51075009)
国家“八六三”计划项目(2009AA04Z417)
北京市优秀人才培养资助计划(2011D005015000006)
关键词
滚动轴承
声发射
谐波小波
支持向量数据描述
单分类
rolling bearing
acoustic emission
harmonic wavelet
support vector data description
one - class classifi- cation
