基于AMI-SGMD和MC-1DCNN-GRU-Attention的 电机故障诊断研究

打开文本图片集

中图分类号：TN911.23-34；TP18 文献标识码：A 文章编号：1004-373X（2026）06-0126-07

Research on motor fault diagnosis based on AMI-SGMD MC-1DCNN-GRU-Attention

1，1， 1，² （1.LabratoryEtremeEnviontOptoelectroicDamicestingTchologIsument，TaianOa; 2. Limited，O3oo27，China）

Abstract：Inalusion tothelow diagnosticaccuracycaused bythenon-stationarycharacteristics motor vibration signals thedificultyinextractingfaultfeatures，amethodfaultdiagnosiscombiningadjustedmutual information-optimized sympletic geometricmodaldecomposition（AM-SGMD）withmulti-channeldeeplearningisproposed.TheAM-SGMDisused todenoisemotor vibrationsignalsadaptivelydecompose themintoseveral improved symplecticgeometriccomponents （ISGC）.Acomprehensiveevaluationindexscreeningcriterionareconstructedbyintegratingkurtosis，permutationentropy， correlationcoeficintstoselectISGCcomponntsthatcansensitivelyeflectfaultcharacteristics.Furthermre，aulti-ael 1Dconvolutionalneuralnetwork-gatedrecurrentunit-attentionmechanism（MC-1DCNN-GRU-Atention）hybridmodelisdesigned，thesoweltalgritissedfortheparameerotimzationtovoidtrainingfalingintolocaltialsolutios.By takingthemeasureddatafromDCmotorsassamples，thetestingresultsdiferentdataprocessing methodsfaultdiagnosis modelsare compared.Theresults show that theaccuracy the fault diagnosis the proposed method can reach 98.50% ，can accurately identify motor faults has good robustness.

Keywords：motor fault diagnosis；symplecticgeometrymodedecomposition;adjusting mutual information；snowablatiol optimization algorithm; multi-channel1D convolutional neural network;gated recurrent unit;attention mechanism

0 引言

电机作为最常见的驱动装置，在长时间运行过程中，受操作不当、环境恶劣、材料老化等因素影响，不可避免地会发生故障，进而影响电机性能[1-2]。（剩余7471字）