多模态数据驱动的智能故障诊断方法

打开文本图片集

关键词：多模态；滚动轴承；故障诊断；加权融合；GraphSAGE网络；数据驱动中图分类号：TN911.23-34；TP18 文献标识码：A 文章编号：1004-373X（2026）06-0184-05

Method of multi-modal data-driven intelligentfaultdiagnosis

BAOYiguo，WANLanjun，NIWei （School of Computer ScienceandArtificial Intellgence，Hunan Universityof Technology，Zhuzhou412oo7，China）

Abstract：Themulti-modaldata can provide morecomprehensiveandmulti-dimensional operation status informationof mechanicalequipmentthanthesingle-modaldatainthedata-driveninteligentrotatingmachineryfaultdiagnosis（RMFD）.The methodof multi-modaldata-drivenintellgentfault diagnosiscansignificantlyimprove theaccuracyandrobustnessof RMFD. However，themulti-modaldatacolectedbydiferenttypesofsensors intheoperationofrotating machineryequipmentarelarge scaleandhavesignificantheterogeneityandomplementarities.Howtoefectivelyextractandfusethefaultfeaturesofdierent modalitiesisakeyproblemtobesolvedinmulti-modaldata-drivenfaultdiagnosis.Onthisbasis，amethodofmulti-modaldatadrivenintellgentfaultdiagnosisisproposed.Themultimodaldataconsistingofvibrationsignalsandcurentsignalsare constructedintomultiplemultimodalradiusgraphscontaining multimodalfaultfeaturesbasedontheradiusneighboralgorithm， sothatthemodelcanefectivelylearnandextractdeep-levelinformationofmultimodalfaultfeatures.Theinputandoutputof eachlayerinthegraphsampleandaggregate（GraphSAGE）networkare weightedandfusedtofullcapturethepotentialassociationsinmulti-modaldataandimprovetheexpressionabilityof themodel.Aseriesofexperimentsarecarredouttoverifythe ffctivenessof theproposed method，and theresultsshowthatthemethod hashighaccuracyin faultdiagnosis.

Keywords：multi-modal; roling bearing;fault diagnosis;weighted fusion; GraphSAGE network;data driver

0 引言

轴承是旋转机械中应用最广泛且最容易损坏的部件之一，其健康状况对旋转机械的运行状态有着至关重要的影响。（剩余7091字）