基于递归定量分析与高斯混合模型的齿轮故障识别  

Gear fault recognition based on recurrence quantification analysis and Gaussian mixture model

作　　者：肖涵[1] 李友荣[1] 吕勇[1] 

机构地区：[1]武汉科技大学机械自动化学院冶金装备及其控制教育部重点实验室,湖北武汉430081

出　　处：《振动工程学报》2011年第1期84-88,共5页Journal of Vibration Engineering

基　　金：国家自然科学基金(50705069);湖北省教育厅科研计划资助项目(Q20100003)

摘　　要：针对递归图只能对信号进行定性分析,不利于其深入应用的缺点,应用递归定量分析方法对各种故障模式振动信号进行定量分析。采用确定率和层流率组成齿轮故障识别的特征向量,并结合高斯混合模型实现齿轮故障模式识别。以齿轮故障实验台上所测取的实验数据为对象,分别采用Re-substitution检验法,Jackknife检验法和Inde-pendent dataset检验法对提出的方法和RBF人工神经网络分类算法进行检验。结果表明,递归定量分析与高斯混合模型相结合应用于齿轮故障模式识别具有更高的识别率。In order to overcome the shortcoming of recurrence plot that can only supply the qualitative analysis to signals,the recurrence quantification analysis is used to analysis different fault modes gear＇s vibration signal.The gear fault pattern recognition method that combined the gaussian mixture model with the feature vector that consists of determinism and laminarity is proposed.Based on the signals that acquired form gear fault experiment table,the proposed method is compared with RBF artificial neural network classification method by Re-substitution test,Jackknife test and Independent data set test respectively.The classification results show that the higher discrimination can be achieved by the proposed method.

关 键 词：递归图 递归定量分析 高斯混合模型 故障识别 

分 类 号：TH17[机械工程—机械制造及自动化] O322[理学—一般力学与力学基础]