外文翻译--建模与优化为20 - H的冷连轧.doc
附录-外文翻译Modelingandoptimizationfora20-hcoldrollingmillQUALITYanditsreproducibilityaredominantcriteriaforcoldrolledproducts.Inparticular,highstripsurfacequalitycanbeachievedwithspecialmillarrangementssuchasthe20-hmill.Thistypeofmillusessmallworkrollsincontactwiththestrip,thatarekeptinplacewithavarietyofintermediateandbackuprolls.Theuseofdifferentactuatorswhich,inpart,onlyactindirectlytoaffecttherollbitegeometry,makesthepresettingofthemillwithregardtostripthicknessandflatnessacomplextask.Thisarticledescribesamodeltheobjectiveofwhichisoptimizingtheentirerollingprocessina20-hmill.Resultsobtainedfromseveralon-lineapplicationsarediscussed.Aclosedsendzimirmillarrangement,showninFig.1,illustratesthemainactuatorsthataffectrollbitegeometrywithregardtostripthicknessandglatness.Sideeccentricslocatedatthebackuprollsareusedtoadjusttheoverallpositionofthecorrespondingrollaxisoverawiderangewhich,indirectly,adjuststherollgapgeometrywithregardtothemillpasslineandstripthickness.Sideeccentricsmaybemechanicallyorelectricallycoupled.Crowneccentricsareavailableatseverallocationsoverthebarrellength.Those,typicallyonupperbackuprolls,arecapableofprovidingspecialrollgapcontours.Theymatchthegaptotheprofileofthestripenteringthemill.Crowneccentricsarethemajoractuatorsforachievingstripflatness.Shiftable,firstintermediaterollsarealsoshapeactuators;theymainlyserveformodificationsinthestripedgeareausingataperedrollprofile.Measurementofmillgeometryisavailableonlyindirectlythroughtherotationofthesideandcrowneccentricsandthroughthepositionofthefirstintermediaterolls.Considerationofmillspringandelasticdeformationeffectsinthestackleadstotherollgapgeometry.Accountingformillspringandelasticdeformationrequiresknowledgeoftherollseparatingforcewhich,inaclosed20-hmill,ismeasuredindirectlythroughtheadjustmentpressureneededforthemainsideeccentrics.Apartfromhysteresiseffects,theeffectsofthevariablegeometrymakethisindirectmeasurementcritical.Besidesrollgapgeometry,thetaskofpresettingthemillalsoincludesthedesignofpassschedulestailoredtomeetrequirementsofaproductandthecurrentmillcondition.Whileoptimalutilizationofthemillisamajorobjective,thepassschedulemustachievetherequiredproducequality.Generationofpassschedulestocoverthestatisticalaverageandstoringthemindatabasesrelatedtosteelgrade,surfaceandcoilgeometryisstateofthearttechnology,Inparticular,millparameterssuchasrollgeometryorthethermalconditionoftheworkrollsrequiredynamiccorrectionofthepassschedulestoobtainareproduciblefinalproduct.Thesameappliestovariationsinthematerialcharacteristicsofthecoilsrolled.Becauseofthecomplexityof20-hmills,achievingreproducibilityofthefinalproductqualityandtheoptimumuseofavailablemillresourcestoincreaseproductivityrepresentsanextremelydifficulttask.Thistaskcanbeaccomplishedwithacomprehensivemodelapproachthattakesallrelevantmillandprocessparametersintoaccount.Tooptimizetheporcess,variousmathematicalmodelsareneededtodescribetheelasticstandbehaviorandtheelastic/plasticcharacteristicsofthematerialtotherolledbecauseneitherdirectgeometricalinformationnoraccuraterollforcemeasurementsexist.1、Force,torqueandpowerTherollforce,rolltorqueanddrivepowernecessarytoformthematerialaresomeofthemostimportantitemsofprocessinformation.Whilepowerrequirementsaffectthedesignofapassscheduleforoptimaluseoftheavailablemillresources,rollforceismandatoryforpresettingthegeometricalactuators.Bothforceandtorque,ontheotherhand,needtobeknownformillpresettingsothatmechanicalorpracticallimitsarenotexceeded.Theapproachselectedtodescribetheeffectsintherollgapwithregardtopower,torqueandforce,isbasedonastripfibermodelusingthebasictheorydevelopedbyKarmanandSiebel.Therollgapmodelprovidesbothverticalandtangentialstresscomponentsactingontheworkroll.Therollseparatingforceresultsfromtheintegrationoftheverticalpressurecomponents.Torqueanddrivepowerarederivedfromthetangentialstress.Therollgapmodelsimultaneouslyprovidesaccurateinformationabouttheverticalandtangentialstresscomponentsactingontherolland,thus,thedrivepowerandrollforce.Theabilitytoevaluatetherollingprocess,basedonaccuratecalculationoftherollseparatingforceandmaindrivepower,enhances,inparticular,thematerialyieldstressevaluation.Thisisbeneficialsincetherollforcemeasurementisaffected,toalargeextent,bymeasurementhysteresispresentinaclosed20-hmill.2、MaterialyieldstressadaptionMaterialyieldstressadaptionisrequiredinanycasewherethereistheneedtorollawiderangeofsteelgrades.Also,thedemandforself-learningmodelalgorithmsforcestheuseofadaptivemethodswithregardtotheyieldstress.Theyieldstressofthematerialisinitiallyevaluatedinoff-linetestsusingtorsionbarsamples.Whileoff-linetestsprovidegoodinitialinformation,eachprocessandproducthasitsownpersonality.Thismayresultfromtheannealingpracticesorvariationsinthechemicalcompositionofthesteelgrades.Theyieldstressadaptionisbrokendownintoashort-termadaptiontorapidlyadjusttheyieldstresscurve,andalong-termadaption,wherecomplexrelationshipsbetweenstrain,strainrateandtemperatureareevaluatedandrepresented.Statisticalyieldstressinformationisavailablebygradeandalsoonanindividualcoilbasisifneeded,whichimprovesqualityassurance.3、FrictionrepresentationBesidesobtainingarepresentationofthematerialyieldstress,itisalsomandatorytodescribethefrictionintherollgap.Inavarietyofapplications,thefrictioncoefficientisadjustedsothatduringlong-termanalysisthemostappropriatefrictioncoefficient;ie,thecoefficientthatprovidesthebestmatchbetweencalculationandmeasurement,isapplied.Anotherapproachistocarryoutrollingtestsandanalyzetheresults.Whilerollingtestsaffectproduction,theanalysismethodistime-consumingandmayoftenhavethedisadvantagethatnotallrelevantfactorsaffectingfrictionareadequatelyconsidered.Theapproachselectedinthecurrentstudyisbasedonanartificialneuralnetwork.Theentrylayeroftheneuralnetworkreceivesallrelevantinformationasithasbeengatheredandmayaffectfriction.Thisinformationisprocessedthroughthemultilayerperceptronfeedforwardnetworkinanoff-lineinvestigationusingthebackpropagationmethodfortrainingthat,finally,leadstothefrictioncoefficient.Witharepresentativework,evenphysicalrelationshipsbetweenthefrictioncoefficientandprocessinformationcanbeevaluated.Theresultsderivedfromtheneuralnetworkhavebeenusedasthebasisforananalyticalmodel,whichwasimplementedon-line.Theaccuracyoftherepresentationhasbeenevaluatedinseveralon-linerollingtestsinindustrialfacilities.Sincemillspeedisoneofthemainvariablesaffectingfriction,onepasswasmadeduringthecommissioningphaseofthemodelwithdifferentmillspeeds.Boththemeasuredandcalculatedrollforcewererecorded.Apartfromthefrictioncoefficient,boththetemperatureofthestripapproachingtherollbiteandthestrainvariedinthetest.4、ElasticmillstandbehaviorInadditiontorollforce,powerandtorque,theelasticbehaviorofthemillstandmustalsobedescribedtoallowpropagationfromthemeasuredeccentricadjustmentstotherollbitecontour,whichisthetargetforfurtheroptimizationsteps.Onerequirementintheelasticmillstandmodelwasitsabilitytocoveravarietyofdifferentmillconfigurations,rollprofilesandrollmaterials.Thesevariableswere