Multi-Fault Diagnosis for Autonomous Underwater Vehicle Based on FuzzyWeighted Support Vector Domain Description

在线阅读 下载PDF 导出详情
摘要 Thispaperaddressesthemulti-faultdiagnosisproblemofthrustersandsensorsforautonomousunderwatervehicles(AUVs).Traditionalsupportvectordomaindescription(SVDD)haslowclassificationaccuracyintheprocessofAUVmulti-faultpatternclassificationbecauseoftheeffectofsamplesparsedensityandtheunevendistributionofsamples,andsoon.Thus,afuzzyweightedsupportvectordomaindescription(FWSVDD)methodbasedonpositiveandnegativeclasssamplesisproposed.Inthismethod,thenegativeclasssampleisintroducedduringclassifiertraining,andthelocaldensityandtheclassweightareintroducedforeachsample.Toimprovethemulti-faultpatternclassifiertrainingspeedandfaultdiagnosisaccuracyofFWSVDD,amulti-faultmodeclassificationmethodbasedonahierarchicalstrategyisproposed.Thismethodaddsfaultcontaindetectionsurfaceforeachthrusterandsensortoisolatefaultcomponentsduringfaultdiagnosis.Byconsideringtheproblemofpatternclassificationforafuzzysample,whichmaybelocatedintheoverlappingareaofhyper-spheresormaynotbelongtoanyhyper-sphereintheprocessofmulti-faultclassificationbasedonFWSVDD,arelativedistancejudgmentmethodisgiven.Theeffectivenessoftheproposedmulti-faultdiagnosisapproachisdemonstratedthroughwatertankexperimentswithanexperimentalAUVprototype.
机构地区 不详
出版日期 2014年05月15日(中国期刊网平台首次上网日期,不代表论文的发表时间)
  • 相关文献