中英文翻译对轮副直径参数的联机测量办法与测量系统大学论文

ONLINEMEASURINGMETHODANDSYSTEMFORDIAMETERPARAMETERSOFWHEELSETWUKaihua,ZHUFeng,ZHUANGFei,YANKuangSchoolofAutomation,HangzhouDianziUniversity,Hangzhou310018,P.R.China,Email:wukaihua@Keywords:Wheelsetdiameter,onlinemeasurement,optoelectronicmeasurement,automaticmeasuringsystem.AbstractThemeasuringofdiameterparametersofwheelsetisanimportantstepforthesafetyoftrainvehiclerunning.Theweardegreeofwheelincreaseswiththespeedupoftrain.Thepaperintroducedanonlinemeasuringmethodofdiameterparametersofrunningwheelsetbasedonoptoelectronicdetectingtechniquewhilethetrainisrunninginlow-speedwithin5-10km/h.Themethodusedprecisionlaserdisplace-mentsensor,highspeedandhighresolutionCCDanddigitalimageprocessingtechnologytorealizethenon-contactautomaticmeasuringofwheelsetparameters.Theseparametersincludethediameter,roundnessanddiameterdifference.Themaininfluencefactorsformeasuringaccuracy,includingmovement,vibrationandtriggererrorwereanalyzed.Theautomaticmeasuringsystemwasdesigned.Thesystemincludedwheeldetector,laserdisplacementsensor,semiconductorlasersource,CCD,onlineimageacquisitionandprocessingcircuits.Theoreticandexperimentalresultsshowedthatthemeasuringaccuracyofdiameterparameterswaswithin1.0mm.Theaccuracymeetsthedemandofonlinediametermeasurement.1IntroductionInordertoensurethesafetyofarunningtrain,it'sveryimportanttodetectthestatusofthewheelsetautomaticallyandregularly.Thewearingdegreeofthewheelsetisoneofthemainfactorsthatinfluencethesafetyandstabilityofrunningtrain.InChina,themeasurementofwheelsetwearisstillstaticandbyhandworkmostlywhichlimitstheaccuracyandreliability.Thedifferencebetweendifferentoperatorsisoftentoobig.Thedetectingefficiencyisalsoverylow.Inrecentyears,therunningspeedhasincreasedto160-200km/h.Theweardegreeandspeedaremuchquickerthanbefore.Thestaticandhandworkmeasuringmethodcannotmeetthedevelopmentofhigh-speedtrain.Researchingtheautomaticandonlinemethodisanurgentneedoftransportationandmaintainingdepartments.Thewearparametersofwheelsetincludegeometricalparametersandsurfacedefects.Themaingeometricalparametersincludeflangethickness,flangeheight,wheeldiameter,roundnessanddiameterdifferenceetc.Thesurfacedefectsincludescotchingandflakingonwheeltread.Theseparametersarethemainfactorsthatinfluencethesafetyandstabilityofvehicle.Thepapermainlyintroducedtheonlinemeasuringmethodandsystemofdiameterparameters.Uptonow,mostofthemaintainingfactorieswerestillusingspecialmechanictoolstomeasurethegeometricalparametersbymanualwork,especiallyinChina.Thesetools,includingspecialverniercaliper,wheeldiameterrulerandriminsidedistanceruler,willweardown.Thestatusofthesetoolswillinfluencethemeasuringaccuracy.Themanualworkmannerhadstronglaborintensityandlowefficiency.Theman-madefactorsproducedbydifferentoperatorsperhapsbringdifferentandvariableerror.Themeasuringaccuracy,reliabilityandrepeatabilitycannotmeetthedemandofwheelsetmaintainingdepartments.Theautomaticmeasuringmethodsandsystemshadbeenresearchedinrecentyears[1,2,3,4].Yetthesemethodswereonlysuitableformaintainingfactoryandthevehiclemustbedisassembled.Severalcountries,suchasGermany[5],Japan,ItalyandtheUnitedStates,havedevelopedonlineautomaticdetectingsystemsbasedondifferentprinciples.Thesesystemsareveryexpensiveandhavenotbeenwidelyused.InChina,therearealsoresearchonthisfield[6],butnopracticalequipmenthadbeendeveloped.Regularmeasurementandrepaircannotfilltherequirementsofmaintenance.Theonlinemeasuringofprofileparametersisimportanttothedistinguishingandforecastoffailureswhilethetrainisrunninginlow-speed.Thepreventiverepairwillreducethemaintenancecostgreatly.Thepaperintroducedanewnon-contactonlinemeasuringmethodandsystemofdiameterparameters.2Measuringmethod2.1DefinitionofdiameterparametersTrainvehiclewheelsetisarollingpartsproducedbypressingtwowheelsintoanaxle.Fig.1istheprofileofacargowheelsection.Treadisthecontactingpartwiththerail.Therightplaneisriminsideplane.Theplanewon’tcontactwiththerailandhasnowear.Thepartbetweenriminsideplaneandtreadiscalledflange.Thetapingpointisabasepointonthetreadandhas70mmdistancetoriminsideplane[7].Therollingcircleisacirclepassingthroughthetapingpointonthetread.Theaveragediameterofrollingcircleequalswheeldiameter.Theroundnesscanbederivedfromdiametersatdifferentpositionsontherollingcircle.Thediameterdifferenceisthedifferencebetweenleftandrightwheel.Theseparametersbelongtogeometricparameters.2.2MeasuringprincipleThemeasuringprincipleofdiameterparametersisbasedontheprojectionimagingofstructuredlightbeam.Fig.3istheschematicdiagram.Accordingtothedefinitionofdiameterparameters,thediametercanbededucediftherollingcirclewasobtained.Theleft-rightmovementwasunavoidablewhilethevehiclewasrunning.Sothepositionofrollingcirclewasuncertain.Alaserdisplacementsensorwasusedtoacquirethepositionofriminsideplaneinrealtime.Sothepositionofrollingcirclewasalsoobtained.Twolaserlinesfromsemiconductorsourceilluminatedonthesurfaceoftreadalongthedirectionofrollingcircleplane.Therollingcirclewasbetweenthetwolines.Thespacebetweentwolineswasdesignedtoensuretherollingcirclewasalwaysbetweenthem.Theincidentopticalplaneoflightsourceparalleledwiththerollingcircleplaneandtwobrightlightstripswereformedonthetreadsurface.Thenarrowstripsrecordedthediameterinformationofwheel.IfaCCDsensorcapturedthestripsimageatacertainanglerelativetotheincidentopticalplane,theprofileinformationwasrecordedintheimage.Thenthegeometricparameterscanbecalculatedbydigitalimageprocessingmethod.2.3PositiontheriminsideplaneThepositionoftapingpointshouldbeobtainedbeforemeasuringthediameter.ItmeansthatthepositionofriminsideplaneshouldbedeterminedfirstlyaccordingtoFig.1.Thepositionofriminsideplanewasvariedbecausedifferentwheelsethaddifferentriminsidedistance.Themeasuringaccuracyofdiameterwasdirectlyinfluencedbythepositionaccuracyofriminsideplane.Aprecisionlaserdisplacementsensorwasusedtoacquirethepositionofriminsideplaneaccuratelybasedonthetriangulationprinciple.Thesensorhadcolorcompensationfunctiontoadaptthechangeofsurfacecolor.Thepositionaccuracywas0.03mmandresponsetimewaslessthan1ms.Thecharacteristiccanmeettheonlinemeasuringneed.2.4ImageacquisitionandprocessingForimageacquiring,thekeyproblemishowtoobtainthecleartreadprofileatcertainposition.Sotherelativepositionoflightsource,CCDandwheeldetectingsensorisveryimportant.Becausethemovementofthetrainwillresultsinoverlapofconsecutivetwofieldimages,theCCDcannotworkinmodeofframe.OnlythesinglefieldimagewillbeacquiredortheCCDsupportstheprogressivescanmode.Inordertoobtainclearimage,theexposuretimemustverysmallandtheCCDandimageacquisitioncardmustsupportasynchronousresetfunction.Themomentofacquiringiscontrolledbywheeldetectingsensor.Theoutputofwheeldetectortriggerstheacquisitionofimageinrealtime.Theimagesareacquiredbyhigh-speedimageacquisitioncardwithDSPfunction.TreadFlangeFig.1:theprofileofwheelsection.70mmRiminsideplaneTapingpointLeftwheelRightwheelWheelsetAxleRollingcircleRollingcircleRiminsideplanethewheelset.Thenaturallighthadinfluenceonthecapturedoriginalimage.InordertoimprovetheoriginalimagequalityandobtainhighimageSNR,asetofnarrowband-passopticalfilterswhichwavelengthmatchedwithlasersourcewereaddedinfrontofCCDlens.Besides,theelectronicexposuretimeofCCDwascontrolledtoavoidtheimageblur.TheresolutionofCCDwas1024×768.Thecentralwavelengthoflasersourcewas650nm.Thebandwidthwas15nm.Theelectronicexposuretimewassetto1/10000s.Theoriginalprofilecurveimagewasdiffusedbecausethelaserlinehad1.0mmwidth.Theonepixelwidthcentrallineoftheprofilecurvewasobtainedbyapplying(1)geometriccorrection,(2)windowadjacentaveragesmoothing,(3)thresholdsegmentation,(4)thinning,(5)edgetracing,and(6)curvefittingalgorithm.Fig.4showedthedigitalimageprocessingresults(photographycondition:speed:5.6km/h,poweroflasersource:20mw,resolutionofCCD:1024x768,shuttertime:1/10000s,lightwavelength:650nm,bandwidth:15nm,distancebetween2lines:5mm,widthofline:1.0mm).2.5MeasurementofwheeldiameterBasedonFig.4andthepositionofriminsideplanedecidedbylaserdisplacementsensor,arollingcirclewasfittedaccordingtotherelativepositionbetweentwolinesandbasepoint.Theobject-imagerelationandmagnifyingratioweredeterminedbytheopticalimagingsystemandwillbeusedinthecalculationprocess.Diameterofacirclewascertainif3points’coordinateswereknownonanarc.AserialofdiametersatdifferentpositionscouldbededucedbasedonFig.5.Ifimagewasacquiredwhilevehiclerolledtodifferentposition,manyotherdiametervaluescouldbeconcluded,thenwheeldiameterwastheaverageofthesediameters.Theroundnesscouldalsobededucedaccordingtothedistributionofthesediameters.Thediameterdifferenceisthemaximumdifferencebetweenleftandrightwheeldiameters.3MeasuringsystemTheschematicdiagramofmeasuringsystembasedontheabovedetectingprincipleshowedinFig.Thesystemconsistedofwheeldetector,precisionlaserdisplacementsensor,laserlinesource,CCDandlens,opticalnarrow-bandfilter,signalprocessingandisolatedamplifyingcircuits,imageacquisitioncircuits,interfacecircuits,controlcircuitsandindustrialcontrolledcomputer.Thearrangementofmeasuringunitswassymmetrical.Twotothreemeasuringunitseverysidewerenecessaryinordertoensuretheprecisionandreliability.AlloftheinitialdataandmeasuredparametersweretransferredtodatabaseandcouldbetransmittedtocentralcontrolcomputerbyInternetforfurtherprocessingandanalyzing.Inthedetectingprocess,thelasersource,CCDandlaserdisplacementsensordidn’tcontactwiththewheelset.Thismethodhadhighpositionaccuracyandsimplifiedtheimageacquisitionequipment.Thenon-contactmeasuringmannerwaseasytomaintainingandhadcharacteristicsofhighreliability,highaccuracyandrepeatability.4ErroranalysisThemeasuringprecisionwasinfluencedbymanyfactors,suchaslensaberrance,CCDresolution,triggererrorproducedbywheeldetector,movementerror,locationerrorofriminsideplane,errorproducedbythevibration,imageacquisitionandprocessingerroretc.Themeasuringunitswerefixedandvibratedwiththerailwayandwheel,sothevibrationerrorcanbeignored.Inthesefactors,themovementandtriggererrorwerethemainerrorsource.4.1MovementerrorIftherunningspeedwas10km/handtheexposuretimewassetto1/10000second,themovementerrorewillbe0.278mm.4.2TriggererrorThetriggermomentmaydelayoraheadofthedesignedtimeandposition.Sotheimageacquisitionwasalsodifferenttothescheduledtime.Iftheresponsespeedofwheeldetectorwas100kHz,thepositiontriggererrorwillbelessthan0.1mm.4.3GeneralerrorThepositionaccuracyoflaserdisplacementsensorwas0.03mm.Thealgorithmerrorofimageprocessingwasoneelement.Movementandtriggererrorwere0.278mmand0.1mmrelatively.Experimentalresultsshowedthegeneralerrorofdiametermeasuringwaswithin1.0mm.5ConclusionAnewmethodforwheelsetdiametermeasuringwasintroducedinrunningcondition.Themethodcombinedoptoelectronicmeasuringanddigitalimageprocessingtechnologytogetherandrealizedthenon-contactautomaticmeasuringofdiameterparameters.Thedynamicmeasuringsystemwasdesigned.Themeasuringaccuracyofdiameterparameterswaslessthan1.0mm.Therepeatabilityandaccuracyofthesystemcanmeetthedemandofonlinewheelsetmaintaining.AcknowledgementsThepaperwassupportedby(1)ZhejiangProvincialNaturalScienceFoundationofChina(Y104578),(2)EducationDepartmentofZhejiangProvinceofChina(20040446)and(3)ScienceandTechnologyDepartmentofZhejiangProvinceofChina(2005C31064).References[1]ZuoJianyong,ZhouWenxiang,ZengJing,etal,“Experimentalresearchonmeasuringwheelrimprofileusinglasersensor”,RailwayVehicle,40(2):11-13,2002.[2]ZhengFenfang,LiuJi,FanPeixin,“Measuringwheelsettreadprofileusingdigitalcamera”,RailwayVehicle,40(1):19-22,2002.[3]WuKaihua,YanKuang,“Researchonthemethodofmeasuringdefectsofwheelsettreadusingoptoelectronictechnique”,OpticalTechnique,31(3):465-467,2005.[4]WuKaihua,ZhangJianhua,YanKuang,JiangPeng,“OptoelectronicAutomaticMeasuringSystemforWheelSetParameters”,ChineseJournalofScientificInstrument,27(3):298-301,306,2006.[5]WangHao,WangLi,GaoXiaorong,“Applicationofelectro-magneticandultrasonictechnologyinthedetectionofwheelandtheprocessingofdetectionsignals”,Locomotive&RollingStockTechnology,6,34-36,2004.[6]YanKuang,WuKaihua,WangRuirong,JiangPeng,“Theoreticalresearchonthemeasuringmethodoftherunningwheelsettreaddefectsbasedonoptoelectronictechnique”,Proc.ofSPIE,6150,61502w-1-61502w-7,2006.[7]TrainVehicle[1998]No.2,“AssembleandRegulationofCargoWheelsetandRollingAxle”,Beijing:ChinaRailingPress,1998.Fig.7:measurementflowchart.I/OcontrolImageacquisitionDataacquisitionIndustrialComputerDatabaseImageprocessingmeasuringunitTriggercontrolDataprocessing

对轮副直径参数的联机测量办法与测量系统WUKaihua,ZHUFeng,ZHUANGFei,YANKuang杭州电子大学自动化学院中国杭州310018Email:wukaihua@关键词：轮辐直径，联机测量，光电测量，自动测量系统摘要：轮辐直径参数的测量是确保车辆系安全运行的一个重要环节。车轮的磨损程度与列车增加的速度成正比。这篇论文介绍了当列车在以5-10千米/时的低速运行时，运用光电探测技术的一种对运行中的齿轮的直径参数联机测量的方法。此法利用精密激光位移传感器，高速高分解CCD和数字图象处理技术实现轮辐参数的无接触自动测量。这些参数主要包括直径，圆度及直径偏差。测量的精确度的主要影响方面是对机械装置，震动装置和启动装置的误差的分析。自动测量系统因此而生。此系统包含车轮探测器，激光位移传感器，半导体激光器来源，CCD，联机图象搜索器处理电路。理论与实践的结果表明，直径参数的测量精确度在1.0毫米内。此精确度与联机直径测量的要求正好符合。1绪论为了确保运行中的列车的安全，自动地规律地探测轮辐的状况很重要。轮辐的磨损度是影响列车运行的安全性，稳定性的几个主要方面之一。在中国，轮辐磨损度的测量方法仍是静止的，人工的，而这限制了它的准确性与可靠性。不同的操作人员得出的结果往往大不相同，探测效率也很低。近年来，其运行速度已达到160-200千米/时。但其磨损的程度与速度比以前快了很多。静止的人工的测量方法已不能适应高速列车的发展。探索自动联机的方法是运输与维修部门的紧急需要。轮辐的磨损参数包括几何参数与表面缺陷。主要的几何参数包括边缘厚度，边缘高度，车轮直径，圆度，直径差异等等。表面缺陷包括车轮支撑面上的擦伤，发裂。这些参数是影响车辆安全性，稳定性的主要方面。这篇论文主要介绍联机直径参数的测量方法与测量系统。至今为止，尤其在中国，大多数维修工厂仍通过手动使用专门的机械工具来测量几何参数。这些工具包括专业游标弯脚规，车轮直径尺，内边缘距离尺，会磨损坏掉。这些工具的使用程度会影响测量效果。手动人工的工作方式带来的是高强度劳动量和低工作效率。不同操作员引起的人为因素可能会产生各种各样的不同的误差，其测量精密度，可靠性，可重复性不符合轮辐维修部门的要求。自动测量方法与系统已被研究多年[1,2,3,4]。而这些方法仅仅适合于维修工厂，而且车辆要被分解。一些国家，例如德国[5]，日本，意大利和美国，已不同原则地发展了联机检测系统。这些系统非常昂贵，还没有被广泛使用。在中国，这个领域也有研究[6]，但没有实际的设备发展。正常的测量与维修不能满足维修的要求。轮廓参数的联机测量在低速运行的列车对于区别与预测失误很重要。预防的维修会大大降低维修成本。这篇论文介绍一种新的无接触直径参数联机测量方法与系统。2测量方法2.1直径参数的定义车辆系的轮辐是通过挤压两个车轮形成一个轮轴的卷型部分。图.1是载物轮区域的轮廓（PROFILE）。滑动面（TREAD）是指与铁轨接触的部分。右平面是指内边缘平面。平面与铁轨不接触，没有磨损。内边缘平面与滑动面（TREAD）之间的部分叫轨底。绝缘绕阻点是指在滑动面上的一个基本点，距离内边缘平面70毫米[7]。齿轮的滚动圆是指过滑动面上绝缘绕阻点的一个圆圈。滚动圆的平均直径等于车轮直径。圆度得自不同位置滚动圆的直径。直径差异是指左右车轮的差异。这些参数属于几何参数。2.2测量原则直径参数的测量原则是基于由光束组成的图象设计。根据直径参数的定义，如果可以获得滚动圆，直径可以推断出来。当车辆运行时，左右机械装置不可避免的，所以滚动圆的位置是无法确定的。激光位移传感器被用来获取内边缘平面的位置，这样滚动圆的位置就可以确定了。两道来自半导体激光器的光束沿着滚动圆的滚动方向照亮滑动面表面。滚动圆在两道光束之间。两道光束间的空隙的设计是为了确保滚动圆总在这里。外来的视觉光源平面并列滚动圆平面,两道亮光带在滑动面上形成。较窄的光带记录车轮的直径信息。若CCD感应器在与外来视觉平面相关的某一角度获取了图像带，轮廓信息图像中就可以记录轮廓信息。然后几何参数就能够通过数字图像处理方法记录下来。2.3内边缘平面的位置在测量直径之前绝缘绕阻点的位置应该确定下来。这表明内边缘平面的位置应该由图.1先确定下来。内边缘平面的位置是多种多样的，因为不同的轮辐有不同的内边距。内边缘平面的位置的精确度直接影响直径测量的效果。用三角测量原则，用精密激光位移传感器精确获取内边缘平面的位置。感应器拥有色彩补偿功能，适应表面色彩的改变。位置的精确度是0.03毫米，反应时间低于1秒钟。这个特点能够满足联机测量的需要。2.4图象的获取与处理要获得图象，关键问题是如何获得某一位置清晰的滑动面轮廓。所以光源的相关位置，CCD和车轮检测器都非常重要，因为列车的运动会使两个局域图象的连续交搭，CCD不能在结构方式中工作。只有单一局域的图象能被获得，而CCD支持先进的浏览方式。为了获得清晰的图象，曝光时间一定要短，CCD和图象获得卡必须支持异步复位方式。取得的一瞬要由车轮探测感应器控制。用高速图象获取卡通过DSP方式能够获得图象。图2：轮辐自然光对于获取的原始图象是有影响的。为了提高原始图象的质量，获得高质量的图象SNR，我们要在CCD透镜前加上一个和激光源匹配波长的窄带视觉过滤器。除此以外，CCD的电子曝光时间要控制住，以免图象模糊。CCD的分辨率是1024×768。激光光源的中央波长是650纳米，波宽是15纳米。电子曝光时间是1/1000