基于改进半监督LTSA与BA-SVM的滚动轴承故障诊断  

Fault Diagnosis for Rolling Bearings Based on Improved Semi-Supervised LTSA and BA-SVM

作　　者：吴保林 戚晓利[1] 王振亚[1] 叶绪丹 郑近德[1] WU Baolin;QI Xiaoli;WANG Zhenya;YE Xudan;ZHENG Jinde(School of Mechanical Engineering,Anhui University of Technology,Ma′anshan 243032,China)

机构地区：[1]安徽工业大学机械工程学院

出　　处：《轴承》2020年第1期48-54,共7页Bearing

基　　金：国家自然科学基金项目(51505002);安徽省自然科学基金项目(1808085ME152)

摘　　要：为有效利用振动信号进行故障诊断,提出了一种基于复合多尺度排列熵(CMPE)、改进距离度量公式的半监督局部切空间排列算法(SS-LTSA)与蝙蝠算法优化支持向量机(BA-SVM)的滚动轴承故障诊断新方法。首先,计算振动信号的CMPE值,构成原始高维特征集;然后,利用改进距离度量公式的SS-LTSA进行降维;最后,将降维后的低维特征集输入BA-SVM完成故障的分类。2类轴承试验分析结果表明:SS-LTSA的降维效果优于主成分分析算法(PCA)和LTSA;BA-SVM的故障识别率明显高于遗传算法优化支持向量机(GA-SVM)和粒子群优化支持向量机(PSO-SVM),SS-LTSA与BA-SVM相结合可以获取更高的识别精度。In order to use vibration signal for fault diagnosis effectively,a new fault diagnosis method for rolling bearings is proposed based on composite multi-scale permutation entropy(CMPE),semi-supervised local tangent spatial algorithm(SS-LTSA)with improved distance measurement formula and bat algorithm optimization support vector machine(BA-SVM).Firstly,the CMPE value of vibration signal is calculated to constitute original high-dimensional feature set.Then,the dimensionality reduction is carried out by using SS-LTSA with improved distance measurement formula.Finally,the low dimensional feature sets after dimensionality reduction are input into BA-SVM to complete classification of faults.The test analysis results of two kinds of bearings show that the dimensionality reduction effect of SS-LTSA is superior to that of PCA and LTSA,the fault recognition rate of BA-SVM is higher than that of GA-SVM and PSO-SVM.The combination of SS-LTSA with BA-SVM achieves higher recognition accuracy.

关 键 词：滚动轴承 故障诊断 多尺度排列熵 降维 蝙蝠算法 支持向量机 

分 类 号：TH133.33[机械工程—机械制造及自动化] TN911.7[电子电信—通信与信息系统]