基于TESPAR与GMM的滚动轴承性能退化评估  

Bearing performance degradation assessment based on TESPAR and GMM

作　　者：张龙[1] 黄文艺[1] 熊国良[1] 周建民[1] 周继慧[1] 

机构地区：[1]华东交通大学机电工程学院,南昌330013

出　　处：《仪器仪表学报》2014年第8期1772-1779,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(51205130;51265010);江西省教育厅科技项目(GJJ12318);江西省自然科学基金(2013BAB216029)资助项目

摘　　要：状态维修根据设备当前运行状态制定维修计划,可避免维修不足与维修过剩等问题。性能退化程度量化评估是实现滚动轴承状态维修的基础。提取滚动轴承早期无故障振动信号的TESPAR参数中的S矩阵作为原始特征,利用主分量分析对其进行降维处理后构建特征矢量,并建立无故障轴承高斯混合模型GMM。将轴承后期振动信号的S矩阵经降维处理后输入该GMM模型,得到被测样本与无故障样本之间的量化相似程度,以此建立时间编码对数似然值TELLP作为滚动轴承性能退化定量指标。轴承疲劳试验表明该方法能及时发现轴承早期故障,并且能很好地跟踪故障发展趋势。Condition-based maintenance （CBM）as a new maintenance philosophy can avoid the occurrence of insufficient and excessive maintenance efforts.For the purpose of quantitative assessment of bearing performance degradation underlying CBM,a GMM （Gauss mixture model）and TESPAR （time encoded signal processing and recognition）based approach is formulated.The S matrices in the TESPAR parameters of the bearing signals collected from fault-free stage are extracted as the original features,and dimensionally reduced with principal component analysis （PCA）to construct the feature vectors;and a GMM model for the fault-free bearing is established and trained with the S matrices.The S matrices from subsequent bearing stages are dimensionally reduced,and then fed to the trained GMM model to obtain the quantitive similarity degree between the fault-free sample and the sample under test.Such similarity degree serves as a fault severity index,which is herein called TELLP （time encoded log-likelihood probability）.Experiment results of bearing fatigue test show that the proposed method is able to detect incipient bearing faults and can track the fault progress trend well.

关 键 词：状态维修 滚动轴承 高斯混合模型 性能退化评估 

分 类 号：TH165.3[机械工程—机械制造及自动化] TN91[电子电信—通信与信息系统]