外文翻译--显微车床的影像反馈控制 英文版.pdf

VisualFeedbackControlofaMicroLatheHirotakaOJIMA1,KatsuhiroSAITO1,LiboZHOU1,JunSHIMIZU1,HiroshiEDA11IbarakiUniversityKeywords:Microlath,Visualfeedback,PositioncontrolAbstractMicromachiningprogressesrapidlyinrecentyears.Inthisresearch,amicrolathewhichisinstallableandoperationalinsideSEMvacuumchamberhasbeendesignedanddeveloped.Asafirststep,visuallyguidedmicrolathesystemisdevelopedwithimageofCCDcameradeviceinsteadofSEMimage.UnliketheconventionalfeedbackcontrolwhichpositionstheX-Ytableonly,thisschemeoffersadirectcontroloftheposition,pathandspeedofthetooltip.Usingproposedmethod,cuttingexperimentwasachieved,anditisconfirmedthatdevelopedmicrolathesystemiseffectivetodocutting.1IntroductionRecently,thesystemcapableofproducingthemicropartsarerequestedalongwiththeminiaturization1.Micromachiningprogressesrapidlyinrecentyears.Theexploratoryresearchhasapproachedtoalevelofaccessingasinglemoleculeoratom.Asadrivingforce,MEMS(microelectronic-mechanicalsystem)hasbeenplayingamajorroleinmakingmicrocomponentsanddevices.However,MEMSisbasedonthephotolithographytechnologyandtherebyapplicableintolimitedmaterialssuchassiliconmonocrystalline.Inordertomeetthedemandsofminiaturizationinelectronicandopticalapplications,alternativemicromachiningtechnologywhichisabletoaccessavarietyofmaterialsina3dimensionalwayisrequired2.Micro-MesoMechanicalManufacturing(M4)offersaccessibilitytodifferentkindsofmaterialaccordingtoeachobjective,andattainshighrepeatabilityandaccuracywiththelatestultraprecisionmeans.Thereare,however,manyscientificandtechnologicalbarriersencounteredinpragmaticimplementationofM4.Oneofthemisthesurfacechemistryeffects.Whenmachiningpartsareatmicroscale,itisrecognizedthatthesurface-area-to-volumeratiowillbeincreasedinbothchipsandtheresultingpartascomparedtoconventional(macro)machiningprocess.Anotherproblemisthedirectmotionandpositioncontrol.Sensorsthatarecapableofdirectlymeasuringtherelativedisplacementbetweenthetoolandworkpiecearenotyetavailable.Inthisresearch,amicrolathewhichisinstallableandoperationalinsideSEMvacuumchamberhasbeendesignedanddeveloped3.Fig.1showstheconceptsofthedevelopedmicrolathe.Atsuchoxygen-freecondition,cuttingtestsareconductedtounderstandsurfacechemistryeffectsonmicromachining.However,sincedevelopedmicrolatheissmallinsize,rigidityofthelatheislow.ThusthepositionofthetoolofthelatheisnotabletobecontrolledaccuratelywithaconventionalmethodwhichcontrolsX-Ytableonly.Therefore,thevisionguidedcontrolmethodisproposed.TheimagefromtheSEM(scanningelectronmicroscope)isdigitizedbyCCDintopixelswith8-bitgrayscale.Sinceeachpixelcontains2Dpositionalinformation,thevisionsystemthusoffersanorthogonalcoordinate(hereafterreferredasthepixelcoordinate)forobjectsinviewtoreferto.ThepixelcoordinateisfreefromthemechanicalinaccuracyandoffersadirectmeasurementofSEMCCDMicrolatheFig.1.Conceptofthedevelopedmicrolathe:H.Ojima,K.Saito,L.Zhou,J.Shimizu,H.Edatherelativepositionoftoolandworkpiece.TheresolutionincreasestogetherwiththemagnificationofthemicroscopeandthenumberofCCDpixels.Inthisresearch,avisioncontrolschemehasbeenproposedandimplementedforfeedbackcontrolofthetoolmovements.UnliketheconventionalfeedbackcontrolwhichpositionstheX-Ytableonly,thisschemeoffersadirectcontroloftheposition,pathandspeedofthetooltip.Asafirststep,visuallyguidedmicrolathesystemisdevelopedwithimageofCCDcameradeviceinsteadofSEMimage.2OverviewofsystemActuatingmoduleSensingmoduleProcessingmoduleImageinformationActuatorsignalCaptureboardMicrolatheXZstageAMPCPUDiamondtoolCCDWorkpieceD/AboardFig.2.BlockdiagramofsystemTable1.SpecificationofsystemSizeofmicrolathe(W×D×H)90×90×42(mm)Spindlerotationalspeed0:*8000(rpm)Depthofcut×Traversefeed10×10(mm)Centerhighadjustment30(Pm)ToolDiamondNoseangle/noseradius40(°)/2(Pm)Scanningrate20(frame/s)Totalpixels0.3megapixelShowninFig.2istheblockdiagramofdevelopedmicrolathesystem,whichconsistsofthreemainmodules；theactuatingmodulethatdrivesmicrolathe,thesensingmodulethatimportsimagesandtheprocessingmodulethatimplementsfeedbackcontrol.Eachmoduleisresponsiblefordifferentfunction.Theactuatingmoduleisthecoreelementwherethecuttingoperationiscarriedout.ThesensingmoduleimportsimagesfromCCDimagedevice,andobtainsthepositionofthetoolandtheworkpiece.Theothertasksincludingtheimageprocessingandfeedbackcontrolareexecutedbytheprocessingmodule.UpperpictureofFig.1showstheoverallappearanceofthesystem.Table.1showsthespecificationsofthesystem.TheactuatingmodulefurtherincorporatesadiamondtoolwithaXZlinearstage,andthesensingmoduleincludesahighresolutionCCDimagedevice.Throughsensingmodule,theappearanceoftheworkingareaisnotonlydisplayedonthemonitortothegivetheoperatorthevisualinformation,butalsoconvertedintodigitalsignalforsubsequentprocessing.AsthecontroldiagramshowinFig.2,themovementsofthediamondtoolaregovernedwiththevisualfeedbackcontrol.Thesensingmodulefirstabstractsthepositionsofthetoolandworkpiecebycomparingthepre-registeredtemplateswiththecapturedvisualinformation.Correspondingtotherelativepositionsoftoolandworkpiece,thetoolpathandspeedarecalculatedandconvertedintoappropriatepulsetrain.egfhefghacbdabcdFig.4.DrivingprincipleofXZ-stageXYZCenterhighadjustmentDCmotorSpindleXZ-stageMicrolatheFig.3.XZ-stageandmicrolathe3ActuatingmoduleThedevelopedmicrolatheisshownrightwardinFig.3.ThislatheconsistsofthemainspindlewiththecolletchuckwiththeDCmotor,thecenterhighadjustmentusingapiezoelectricactuatorandXZ-stagewhichperformsbothdepthofcut(X-axis)andtraversefeed(Z-axis).TheXZ-stageisdrivenbytheinertialsliding,andiscomposedofapiezoelectricactuatorandthelinearguide.XZ-stageisshownleftwardinFig.3.AnaccuratetoolpositioningisachievedbydrivingtheXZ-stageprecisely.ImportantpointsofdrivingtheXZ-stagearethecontrolofthedrivingdirection,distanceandvelocity.Figure4showstheinertialslidingmechanismbythesaw-toothwave.Thedirectionofthemovementisdecidedbytherising/trailingedgeofthesaw-toothwaveasshowninFig.4.Forexample,:VisualFeedbackControlofaMicroLathethemechanismintherightdirection(+)isexplainedasfollows.Thevoltagegraduallyrises,andapiezoelectricactuatorstretchesmostin(1).Theactuatorshrinksbasedonthecentroidin(2)byfallingrapidlyofthevoltage.Onlythesidewherethefrictionalforceissmallmovesastheactuatorstretchesgraduallywiththeascentofthevoltagein(3).Theactuatorisstretchesagainin(4),andadvancestowardtherightdirection.Theactuatorsimilarlyadvancesalsotowardtheleftdirection(-)ifareversepulsetrainisgiven.00.10.20.30.40200400600FrequencyAíHzSpeedAímm/s±80V±40VFig.5.VelocitychangedependingonfrequencyandvoltageNext,thevelocitycontrolofthismechanismisdescribed.AsshowninFig.5,thevelocityisproportionaltobothfrequencyofthepulsetrainanddrivingvoltage.Finally,drivingdistancecanbecontrolledaccordingtothenumberofpulses,becausethedrivingdistancebyoneplusisabout500Pmat±80Vor250Pmat±40V.(500,420)(140,420)(500,60)(140,60)XZ(320,240)4123Fig.8.ExperimentalconditionoflinearpathcontrolXpixelZpixelcount210121011000200300400countFig.7.RecognitionaccuracyoftooltipXZ-stageDiamondtoolCCDWorkpiece(X,Z)ZXFig.6.Visualsensingsystem4SensingmoduleThediamondtoolismountedonXZ-stage,whichusespiezoelectricactuatortodrivetool.Thosemechanicalinaccuracies,mainlycausedbythermalexpansion,hysteresis/driftinactuatorsandmisalignmentoforthogonalaxis,maydirectlydeliveranegativeeffecttothesystemperformance.Tosolvetheseproblems,avisioncontrolschemeasshowninFig.6isdeveloped.TheleftpictureinFig.6showsthemicrolatheandCCDimagedevicelocatedinY-axis.FromtherightpictureinFig.6,theincomingvisualinformationfromtheCCDisdigitizedintopixelswith8-bitgrayscalebythesensingmodule.Aseachpixelbears2Dpositionalinformation,thevisionsystemthusoffersanorthogonalcoordinate(referredasthepixelcoordinate)forobjectsinviewtoreferto.ThepixelcoordinateisfreefromthemechanicalinaccuracyanditsresolutionincreasestogetherwiththemagnificationoftheCCD.Ata480×640pixelframeusedinthecurrentresearch,forexample,theresolutionofthepixelcoordinateisabout6PmwhentheviewoftheCCDistwofoldmagnified.WhentheCCDisalignedalongY-axis,thepositionofthetooltipandworkpieceisprojectedintoa2Dpixelcoordinate(XZ)whichiscommonlysharedbytheXZ-stageandworkpiece.Drivenandcontrolledbythepixelcoordinate,thetoolisabletobepositionedandmovedattheaccuracyofpixelresolutionwithnoeffectbythemechanicalinaccuracy.Inaddition,iftherigiditybetweenXZ-stageandtoolislow,positioningoftooltipisnotachievedbydrivingXZ-stageaccurately.Thus,moreimportantly,thisoperationisaneffectivemethodofpositioningforthemicrolathewithalowrigidity.Figure7showstherecognitionaccuracythatismadebyuseofshapebasedpatternmatching4torecognizetheactualtooltiprepeatedly500times.Wecomprehendfromthegraphthat88.5%reliabilitycanbeachievedwithinthelimesof±1pixel(6Pm).5ProcessingmoduleForthesystemwhichisconsistedoftheactuatingandsensingmoduleinprevioussection3and4,thevisual:H.Ojima,K.Saito,L.Zhou,J.Shimizu,H.Edafeedbackcontrolmethodisdescribedinthissection.ThetooltipisdrivenbyvisualfeedbackcontrolmethodwithpositionsofthetooltipandtargetsfromCCDimagedevice.Asafirststep,weexaminedlinearpathcontrolandcircularpathcontrolofthetooltip.Inthesepathcontrols,drivingfrequencyis300Hz(162Pm/s).Atfirst,linerpathcontroloftooltipisdescribed.AsshowninFig.8,thetargetpositionisdefinedas(320,240)whichisthecenteroftheimagefromCCD,andfourkindsofpathcontrolareexamined.Inthecaseoflinerpathcontrol,theangleformedbythetargetpositionandthepresentpositionofthetooltipisfedbacktoachievethepathcontrol.Figure9(a)showstheresultantpathofthetooltipwithoutfeedbackcontrol,and(b)showsthatwithfeedbackcontrol.Inthecaseofthepathwithoutfeedback,finalerrorsoffourpathsarebetween5pixels(30Pm)and15pixels(90Pm).Ontheotherhand,thepathwithfeedbackfollowsalongthetargetpath,andfinalerroriswithin2pixels(12Pm).Next,thecircularpathcontrolwhichismulti-axialinterpolationisdescribed.TheconditionofthecircularpathcontrolisshowninFig.10.Thecenterofthetargetcircularpathisdefinedas(320,240)whichisthecenteroftheimagefromCCD,andtheradiusofthetargetpathis100pixels(600Pm),moreoverthetooltipisdrivenfromstartingpoint(220,240)alongcounterclockwisedirectionrepeated3times.Inthecaseofcircularpathcontrol,weconsidertofeedbacknotonlytheangleformedthecenterofthetargetcircularpathandthepresenttoolposition,butalsothedeviationoftheradiuswhichistheerrorbetweentheradiusofthetargetcircularpathandthedistancefromthecenterofthetargetpathtothepresenttoolposition.Inthecaseofthedrivingthepathwithoutfeedbackcontrol,thetoolisdrivenbytheanglespreparedinadvance.Figure11(a)showstheresultantpathofthetooltipwithoutfeedbackcontrol,and(b)showsthepathwithfeedbackcontroloftheangleonly,and(c)showsthepathwithfeedbackcontroloftheangleandradius.Figure11(a)showsthattheresultantpathdepartedfromtargetpath,andthecenterandtheradiusofthepatharedeflectedfromthoseofthetargetpath.Figure11(b)showsthecenteroftheresultantpathmatchesthecenterofthetargetpath,butextendstheradiusoftheresultantpathasthepathgoesaround.Moreover,Fig.11(c)showsthattheresultantpath0100200300400500100200300400500600XpixelZpixelgoal1234Zpixel(a)withoutfeedbackcontrol0100200300400500100200300400500600XpixelZpixelgoal1234Zpixel(b)withfeedbackcontrolFig.9.Experimentalresultsoflinearpathcontrol(220,240)(320,240)XZCCWFig.10.Experimentalconditionofcircularpathcontrol0100200300400100200300400500XpixelZpixel0100200300400100200300400500XpixelZpixel0100200300400100200300400500XpixelZpixeltooltargettooltargettooltargetZpixelZpixelZpixelget(a)withoutfeedbackcontrol(b)withfeedbackcontroloftheangle(c)withfeedbackcontroloftheangleandradiusFig.11.Experimentalresultsofcircularpathcontrol: