第2卷第1期Vol.2No.12023年2月JournalofArmyEngineeringUniversityofPLAFeb.2023采用滑动平均多元多尺度色散熵的液压泵故障诊断方法宫建成1,韩涛1,杨小强1,刘武强2,周付明2(1.陆军工程大学野战工程学院,江苏南京210007;2.海军工程大学,湖北武汉430034)摘要:为了提高色散熵的信息提取能力,在兼顾计算效率和效果的前提下,引入多维嵌入重构理论,借鉴滑动平均的思想,更新了传统多尺度算法的粗粒化方式,提出了滑动平均多元多尺度色散熵(movingaver-agemultivariatemultiscaledispersionentropy,MA_mvMDE)用以提取液压泵故障特征。首先,利用均匀相位经验模态分解(uniformphaseempiricalmodedecomposition,UPEMD)将振动信号分解为多个本征模态分量(intrinsicmodefunctions,IMF),再采用相关系数法筛选敏感分量,将包含大量故障信息的模态分量作为多通道数据计算其MA_mvMDE值来提取故障特征。接着,采用MCFS方法选择故障敏感特征实现降维。最后,通过随机森林分类器完成故障识别。采用液压泵故障振动数据验证了该方法能够准确诊断不同类型和不同程度的故障。关键词:均匀相位经验模态分解;滑动平均多元多尺度色散熵;敏感IMF选择;故障诊断;液压泵中图分类号:TH137;TP206DOI:10.12018/j.issn.2097-0730.20211221003FaultDiagnosisofHydraulicPumpAdoptingMovingAverageMultivariateMultiscaleDispersionEntropyGONGJiancheng1,HANTao1,YANGXiaoqiang1,LIUWuqiang2,ZHOUFuming2(1.CollegeofFieldEngineering,ArmyEngineeringUniversityofPLA,Nanjing210007,China;2.NavalUniversityofEngineering,Wuhan430034,China)Abstract:Inordertoimprovetheinformationextractionofdispersionentropy,thispaperupdatesthecoarse-grainingapproachofthetraditionalmulti-scalealgorithmandproposesanewmethodcalledmovingaveragemultivariatemultiscaledispersionentropy(MA_mvMDE)byintroducingthemultidimensionalembeddingreconstructiontheoryandtheideaofmovingaverage,withdueconsiderationofcomputationalefficiencyandeffects.Firstly,thevibrationsignalisdecomposedintoseveralintrinsicmodefunctions(IMF)byuniformphaseempiricalmodedecomposition(UPEMD),thesensitivecomponentsarescreenedoutbycorrelationcoefficientmethod,andtheselectedcomponentscontainingalargeamountoffaultin-formationarecalculatedtogettheirMA_mvMDEvaluesasmulti-channeldatatoextr...
