WVPMCD及其在滚动轴承故障诊断中的应用  

Weighted Least Square-VPMCD and Its Application in Roller Bearing Fault Diagnosis

作　　者：杨宇[1] 李永国[1] 程军圣[1] 

机构地区：[1]湖南大学汽车车身先进设计制造国家重点实验室,湖南长沙410082

出　　处：《湖南大学学报（自然科学版）》2014年第2期52-57,共6页Journal of Hunan University:Natural Sciences

基　　金：国家自然科学基金资助项目(51175158;51075131);湖南省自然科学基金资助项目(11JJ2026);中央高校基本科研业务费专项基金资助项目

摘　　要：多变量预测模型(Variable predictive model based class discriminate,简称VPMCD)分类方法是建立在回归模型为同方差性基础上的,而当模型出现异方差性时,会导致预测精度降低.基于此,本文提出了WVPMCD(WLS-Variable predictive model based class discriminate,简称WVPMCD)方法,即用加权最小二乘法(WLS)代替原方法中的最小二乘法(OLS)进行参数估计,消除异方差性,从而提高了模式识别的精度.采用局部特征尺度分解(Local characteristic-scale decomposition,简称LCD)方法对滚动轴承振动信号进行分解,提取分量矩阵的奇异值组成故障特征向量作为WVPMCD的输入,并对正常状态、滚动体故障、内圈故障和外圈故障4种不同工作状态和故障类型下的滚动轴承振动信号进行分析,结果表明,在模型存在异方差性时,WVPMCD比原VPMCD具有更好的分类效果和识别率.Variable predictive model-based class discriminate （VPMCD） classification method was built on homoscedastic regression model. When the regression model is heteroscedastic, it will lower the predic-tion accuracy. So a WLS-variable predictive mode-based class discriminate （WVPMCD） pattern recogni-tion method was presented. Its parameter estimation approach uses the weighted least square method to re-place ordinary least square method to eliminate homoscedasticity, thus raising the accuracy of pattern rec-ognition. In this paper, LCD （Local characteristic-scale decomposition） approach was adopted to decom-pose the roller bearing vibration signal. Then, the singular values are abstracted from the component ma- trix and formed into fault feature vector which will act as the input in WVPMCD. The analysis results from the roller bearing vibration signals of normal, roller fault, inner race fault and outer race fault dem-onstrate that WVPMCD has a higher recognition rate when the regression model is heteroscedastic.

关 键 词：WVPMCD 局部特征尺度分解 加权最小二乘 滚动轴承 故障诊断 

分 类 号：TH113[机械工程—机械设计及理论]