基于多尺度时不可逆与t-SNE流形学习的滚动轴承故障诊断  

Rolling bearing fault diagnosis method based on multiscale time irreversibility and t-SNE manifold learning

作　　者：姜战伟 郑近德[1] 潘海洋[1] 潘紫微[1] 

机构地区：[1]安徽工业大学机械工程学院,安徽马鞍山243032

出　　处：《振动与冲击》2017年第17期61-68,84,共9页Journal of Vibration and Shock

基　　金：国家自然科学基金(51505002;51305046);安徽省高校自然科学研究重点资助项目(KJ2015A080);安工大研究生创新研究基金(2016062)

摘　　要：为了精确地提取机械振动信号的非线性故障特征,提出了一种新的振动信号复杂性测量方法——多尺度时不可逆。同时结合t-分布邻域嵌入(t-SNE)流形学习和粒子群优化-支持向量机(PSO-SVM),提出了一种新的滚动轴承故障诊断方法。采用多尺度时不可逆提取复杂振动信号的特征信息;利用t-SNE对高维特征空间进行降维;将低维特征向量输入到基于PSO-SVM多故障模式分类器中进行识别与诊断。将提出的方法应用于试验数据分析,并与现有方法进行了对比,分析结果表明,该方法不仅能够有效地诊断滚动轴承的工作状态和故障类型,而且优于现有方法。In order to accurately extract nonlinear fault features of mechanical vibration signals,a novel method for complexity measurement of vibration signals called the multiscale time irreversibility（ MSTI） was proposed. Meanwhile,combining the t-distributed stochastic neighbor embedding（ t-SNE） and the particle swarm optimization-support vector machine（ PSO-SVM）,a new fault diagnosis method for rolling bearings was proposed. Firstly,MSTI was used to extract the characteristic information of complex vibration signals. Secondly,t-SNE was used to reduce dimensions of the high dimension feature space. Then the selected lower dimensional feature vectors were input to a PSO-SVM-based multi-fault classifier for fault diagnosis. Finally,the proposed method was applied in the test data analysis and compared with the existing methods. The analysis results showed that the proposed method can be used to effectively diagnose the working status and fault types of rolling bearings,it is superior to the existing methods.

关 键 词：多尺度时不可逆 t-分布邻域嵌入 支持向量机 滚动轴承 故障诊断 

分 类 号：TN911.7[电子电信—通信与信息系统] TH165.3[电子电信—信息与通信工程]