文档名：基于tSNE多特征融合的JTC轨旁设备故障检测
摘要：无绝缘轨道电路(JointlessTrackCircuit,JTC)的轨旁设备在室外长期运营过程中,其可靠性会逐渐降低,进而给列车行车安全带来严重威胁.以轨道电路读取器(TrackCircuitReader,TCR)感应电压为基础,针对JTC故障诊断研究中轨旁设备故障类型复杂和故障特征提取不充分等问题,提出一种基于t分布随机邻域嵌入(t-distributionStochasticNeighborEmbedding,tSNE)多特征融合的JTC轨旁设备故障检测模型.首先,根据不同轨旁设备故障对TCR感应电压信号的影响,分析各轨旁设备的故障特性.其次,提取TCR感应电压信号的方差、有效值、峰值因子等幅值域特征,以及排列熵、散布熵特征构成原始故障特征集.为了去除其中的冗余信息,得到具有较高判别性的融合流形特征,利用tSNE算法进行特征融合.最后输入深度残差网络(DeepResidualNetwork,DRN)得到故障检测混淆矩阵,实现轨旁设备故障定位.实验结果表明:tSNE算法融合后的特征在异类和同类故障样本之间分别有较大的类间间距和较小的类内间距,相比主成分分析(PrincipalComponentAnalysis,PCA)、随机相似性嵌入(StochasticProximityEmbedding,SPE)、随机邻域嵌入(StochasticNeighborEmbedding,SNE)算法具有更优的融合特征提取效果.此外,结合DRN可以有效识别多种轨旁设备故障,达到98.28%的故障检测准确率.通过现场信号进行实例验证,结果表明该故障检测模型能满足铁路现场对室外设备进行故障定位的实际需求.
Abstract：Thereliabilityofthetracksideequipmentofthejointlesstrackcircuit(JTC)willgraduallydecreaseduringlong-termoutdooroperations,whichposesaseverethreattothesafetyoftrainoperations.AimingattheproblemsofcomplexfaulttypesandinsufficientfaultfeatureextractionoftracksideequipmentinJTCfaultdiagnosisresearch,afaultdetectionmodelofJTCtracksideequipmentbasedont-distributionstochasticneighborembedding(tSNE)multi-featurefusionwasproposed.Firstly,accordingtotheinfluenceofdifferenttracksideequipmentfaultsontrackcircuitreader(TCR)inducedvoltagesignals,thefaultcharacteristicsofeachtracksideequipmentwereanalyzed.Secondly,theamplitudedomainfeatures,suchasthevariance,root-mean-square,andpeakfactoroftheTCRinducedvoltagesignal,wereextractedtoformtheoriginalfaultfeaturesetwiththepermutationentropyanddispersionentropy.Toremovetheredundantinformationandobtainthefusionmanifoldfeatureswithhighdiscrimination,thetSNEalgorithmwasusedforfeaturefusion.Finally,thefaultdetectionconfusionmatrixwasobtainedthroughthedeepresidualnetwork(DRN)torealizethefaultlocationofthetracksideequipment.TheexperimentalresultsshowthatthefeaturesfusedbytSNEhavelargerinter-classdistancesbetweenheterogeneousfaultsamplesandsmallerintra-classdistancesbetweenhomogeneousfaultsamples.Comparedwithprincipalcomponentanalysis(PCA),stochasticproximityembedding(SPE),andstochasticneighborembedding(SNE)algorithms,tSNEhasabetterfeatureextractioneffect.Inaddition,combinedwithDRN,itcaneffectivelyidentifyvarioustracksideequipmentfaults,andthefaultdetectionaccuracycanreach98.28%.Theexampleverificationresultsoffieldsignalsshowthattheproposedfaultdetectionmodelcanmeettheactualneedsoftherailwayfieldforfaultlocationofoutdoorequipment.
作者：武晓春 郜文祥Author：WUXiaochun GAOWenxiang
作者单位：兰州交通大学自动化与电气工程学院,甘肃兰州730070
刊名：铁道科学与工程学报
Journal：JournalofRailwayScienceandEngineering
年，卷(期)：2024, 21(3)
分类号：U284.2
关键词：轨旁设备 幅值域 排列熵 散布熵 多特征融合 故障检测
Keywords：railtracksideequipment amplitudedomain permutationentropy dispersionentropy multi-featurefusion faultdetection
机标分类号：TP391.1TN911U284.238
在线出版日期：2024年4月12日
基金项目：高速铁路基础研究联合基金，国家自然科学基金，甘肃省优秀研究生创新之星项目基于tSNE多特征融合的JTC轨旁设备故障检测[
期刊论文] 铁道科学与工程学报--2024, 21(3)武晓春 郜文祥无绝缘轨道电路(JointlessTrackCircuit,JTC)的轨旁设备在室外长期运营过程中,其可靠性会逐渐降低,进而给列车行车安全带来严重威胁.以轨道电路读取器(TrackCircuitReader,TCR)感应电压为基础,针对JTC故障诊断研究中...参考文献和引证文献
参考文献
引证文献
本文读者也读过
相似文献
相关博文
基于tSNE多特征融合的JTC轨旁设备故障检测 Fault detection of JTC trackside equipment based on tSNE multi-feature fusion
基于tSNE多特征融合的JTC轨旁设备故障检测.pdf
- 文件大小:
- 5.81 MB
- 下载次数:
- 60
-
高速下载
|
