电子测量技术ELECTRONICMEASUREMENTTECHNOLOGY第45卷第23期2022年12月DOI:10.19651/j.cnki.emt.2209997基于ICEEMDAN-MPE和AO-LSSVM的滚动轴承故障诊断*李铭何毅斌马东唐权胡明涛(武汉工程大学机电工程学院武汉430205)摘要:针对滚动轴承故障诊断中特征提取困难和故障类型识别准确率偏低等情况,提出一种基于改进型自适应噪声完整集成经验模态分解(ICEEMDAN)与多尺度排列熵(MPE)结合天鹰算法(AO)优化最小二乘支持向量机(LSSVM)正则化参数和核参数的故障诊断方法。首先通过ICEEMDAN对轴承原始振动信号进行分解,其次根据相关系数和方差贡献率双原则选取符合标准的本征模态分量(IMF),并计算对应分量的MPE,以全面获取故障特征信息;最后将其构成多维特征向量,利用AO-LSSVM辨识模型实现对轴承故障诊断。同时进行多组对比实验,研究结果表明了所提方法在滚动轴承故障诊断中的优越性且识别准确率可达98.95%。关键词:故障诊断;ICEEMDAN;多尺度排列熵;天鹰算法;最小二乘支持向量机中图分类号:TH133.33文献标识码:A国家标准学科分类代码:0803RollingbearingfaultdiagnosisbasedonICEEMDAN-MPEandAO-LSSVMLiMingHeYibinMaDongTangQuanHuMingtao(SchoolofMechanicalandElectricalEngineering,WuhanInstituteofTechnology,Wuhan430205,China)Abstract:Inviewofthedifficultyoffeatureextractionandthelowaccuracyoffaulttyperecognitioninrollingbearingfaultdiagnosis,afaultdiagnosismethodbasedonImprovedCompleteEnsembleEmpiricalModeDecompositionwithadaptivenoise(ICEEMDAN)andMulti-scalePermutationEntropy(MPE)combinedwithAquilaOptimizer(AO)tooptimizetheregularizationparametersandkernelparametersofLeastSquaresSupportVectorMachine(LSSVM)isproposed.Firstly,theoriginalvibrationsignalofthebearingisdecomposedbyICEEMDAN.Secondly,accordingtothedoubleprinciplesofcorrelationcoefficientandvariancecontributionrate,theeigenmodecomponent(IMF)thatmeetsthestandardisselected,andtheMPEofthecorrespondingcomponentiscalculatedtocomprehensivelyobtainthefaultcharacteristicinformation;Finally,themulti-dimensionalfeaturevectorisformed,andthebearingfaultdiagnosisisrealizedbyusingAO-LSSVMidentificationmodel.Atthesametime,severalgroupsofcomparativeexperimentsarecarriedout.Theresultsshowthesuperiorityoftheproposedmeth...
