毕业设计英文翻译.docx
《毕业设计英文翻译.docx》由会员分享,可在线阅读,更多相关《毕业设计英文翻译.docx(12页珍藏版)》请在冰豆网上搜索。
毕业设计英文翻译
英文翻译
系别机电工程系
专业自动化
班级163803
学生姓名田琨
学号081470
指导教师张科、王敏
DesignOfFuzzyControllers
1.1Introduction
WhileitisrelativelyeasytodesignaPIDcontroller,andtheinclusionofthefuzzyrulescreatesalotofextradesignproblems,andalthoughalotofintroductorytextbooksexplainfuzzycontroltherearefewgeneralguidelinesforsettingtheparametersofasimplefuzzycontroller.
Theapproachhereisbasedonathreestepdesignprocedure,thatbuildsonPIDcontrol:
1.StartwithaPIDcontroller.
2.Insertanequivalent,linearfuzzycontroller.
3.Makeitgraduallynonlinear.
Guidelinesrelatedtothedifferentcomponentsofthefuzzycontrollerwillbeintroducedshortly.Inthenextthreesectionsthreesimplerealisationsoffuzzycontrollersaredescribed:
atable-basedcontroller,aninput-outputmappingandaTakagi-Sugenotypecontroller.Ashortsectionsummarisesthemaindesignchoicesinasimplefuzzycontrollerbyintroducingachecklist.Theterminologyisbasedonaninternationalstandardwhichisunderway(IEC,1996).Fuzzycontrollersareusedtocontrolconsumerproducts,suchaswashingmachines,videocameras,andricecookers,aswellasindustrialprocesses,suchascementkilns,undergroundtrains,androbots.Fuzzycontrolisacontrolmethodwhichisbasedonfuzzylogic.Justasfuzzylogiccanbedescribedsimplyas“computingwithwordsratherthannumbers”,fuzzycontrolcanbedescribedsimplyas“controlwithsentencesratherthanequations”.
Afuzzycontrollercanincludetheempiricalrules,andthatisespeciallyusefulinoperatorcontrolledplants.
Takeforinstanceatypicalfuzzycontroller
1.IferrorisNegandchangeinerrorisNegthenoutputisNB
2.IferrorisNegandchangeinerrorisZerothenoutputisNM
……
Thecollectionofrulesiscalledarulebase.Therulesareinthefamiliarif-thenformat,andformallytheif-sideiscalledtheconditionandthethen-sideiscalledtheconclusion(moreoften,perhaps,thepairiscalledtheantecedent-consequentorpremise-conclusion).Theinputvalue“Neg”isalinguisticshortforthewordNegative,theoutputvalue“NB”standsforNegativeBigand“NM”forNegabiveMeduim.Thecomputerisabletoexecutetherulesandcomputeacontrolsignaldependingonthemeasuredinputserrorandchangeinerror.
Theobjectivehereistoidentifyandexplainthevariousdesignchoicesforengineers.
Inarulebasedcontrollerthecontrolstrategyisstoredinamoreorlessnaturallanguage.Thecontrolstrategyisisolatedinarulebaseopposedtoanequationbaseddescription.Arulebasedcontrolleriseasytounderstandandeasytomaintainforanon-specialistend-user.Anequivalentcontrollercouldbeimplementedusingconventionaltechniquesinfact,anyrulebasedcontrollercouldbeemulatedin,say,Fortran-itisjustthatitisconvenienttoisolatethecontrolstrategyinarulebaseforoperatorcontrolledsystems.
Fuzzycontrollersarebeingusedinvariouscontrolschemes(IEC,1996).Themostobviousoneisdirectcontrol,wherethefuzzycontrollerisintheforwardpathinafeedbackcontrolsystem.Theprocessoutputiscomparedwithareference,andifthereisadeviation,thecontrollertakesactionaccordingtothecontrolstrategy.Inthefigure,thearrowsmaybeunderstoodashyper-arrowscontainingseveralsignalsatatimeformultiloopcontrol.Thesub-componentsinthefigurewillbeexplainedshortly.Thecontrollerishereafuzzycontroller,anditreplacesaconventionalcontroller,say,aPID(proportional-integral-derivative)controller.
Infeedforwardcontrol(Fig.1)ameasurabledisturbanceisbeingcompensated.Itrequiresagoodmodel,butifamathematicalmodelisdifficultorexpensivetoobtain,afuzzymodelmaybeuseful.Figure1showsacontrollerandthefuzzycompensator,theprocessandthefeedbackloopareomittedforclarity.Thescheme,disregardingthedisturbanceinput,canbeviewedasacollaborationoflinearandnonlinearcontrolactions;thecontrollerCmaybealinearPIDcontroller,whilethefuzzycontrollerFisasupplementarynonlinearcontroller.
Fuzzyrulesarealsousedtocorrecttuningparametersinparameteradaptivecontrolschemes(Fig.2).Ifanonlinearplantchangesoperatingpoint,itmaybepossibletochangetheparametersofthecontrolleraccordingtoeachoperatingpoint.Thisiscalledgainschedulingsinceitwasoriginallyusedtochangeprocessgains.Againschedulingcontrollercontainsalinearcontrollerwhoseparametersarechangedasafunctionoftheoperatingpointinapreprogrammedway.Itrequiresthoroughknowledgeoftheplant,butitisoftenagoodwaytocompensatefornonlinearitiesandparametervariations.Sensormeasurementsareusedasschedulingvariablesthatgovernthechangeofthecontrollerparameters,oftenbymeansofatablelook-up.
Whetherafuzzycontroldesignwillbestableisasomewhatopenquestion.Stablityconcernsthesystem’sabilitytoconvergeorstayclosetoanequilibrium.Astablelinearsystemwillconvergetotheequilibriumasymptoticallynomatterwherethesystemstatevariablesstartfrom.Itisrelativelystraightforwardtocheckforstabilityinlinearsystems,forexamplebycheckingthatalleigenvaluesareinthelefthalfofthecomplexplane.Fornonlinearsystems,andcheckingthatalleigenvaluesareinthelefthalfofthecomplexplane.Fornonlinearsystems,andfuzzysystemsaremostoftennonlinear,thestabilityconceptismorecomplex.Anonlinearsystemissaidtobeasymptoticallystableif,whenitstartsclosetoanequilibrium,itwillconvergetoit.Evenifitjuststaysclosetotheequilibrium,withoutconvergingtoit,itissaidtobestable(inthesenseofLyapunov).Tocheckconditionsforstabilityismuchmoredifficultwithnonlinearsystems,partlybecausethesystembehaviourisalsoinfluencedbythesignalamplitudesapartfromthefrequencies.TheliteratureissomewhattheoreticalandinterestedreadersarereferredtoDriankov,Hellendoorn&Reinfrank(1993)orPassino&Yurkovich(1998).Theyreportonfourmethods(
Lyapunovfunctions,Popov,circle,andconicity),andtheygiveseveralreferencestoscientificpapers.
Figure1:
Feedforwardcontrol
Itischaracteristic,however,thatthemethodsgiveratherconservativeresults,whichtranslateintounrealisticallysmallmagnitudesofthegainfactorsinordertoguaranteestability.Anotherpossibilityistoapproximatethefuzzycontrollerwithalinearcontroller,andthenapplytheconventionallinearanalysisanddesignproceduresontheapproximation.Itseemslikelythatthestabilitymarginsofthenonlinearsystemwouldbecloseinsomesensetothestabilitymarginsofthelinearapproximationdependingonhowclosetheapproximationis.Thispapershowshowtobuildsuchalinearapproximation,butthetheoreticalbackgroundisstillunexplored.
Thereareatleastfourmainsourcesforfindingcontrolrules(Takagi&SugenoinLee,1990).
1.Expertexperienceandcontrolengineeringknowledge.Oneclassicalexampleistheoperator’shandbookforacementkiln(Holmblad&Ostergaard,1982).Themostcommonapproachtoestablishingsuchacollectionofrulesofthumb,istoquestionexpertsoroperatorsusingacarefullyorganizedquestionnaire.
2.Basedontheoperator’scontrolactionrulescanbededucedfromobservationsofanoperator’scontrolactionsoralogbook.Thisrulesexpressinput-outputrelationships.
Figure2:
Fuzzyparameteradaptivecontrol
3.Basedonafuzzymodeloftheprocess.Alinguisticrulebasemaybeviewedasaninversemodelofthecontrolledprocess.Thusthefuzzycontrolrulesmightbeobtainedbyinvertingafuzzymodeloftheprocess.Thismethodisrestrictedtorelativelylowordersystems,butitprovidesanexplicitsolutionassumingthatfuzzymodelsoftheopenandclosedloopsystemsareavailable(Braae&RutherfordinLee,1990).Anotherapproachisfuzzyidentification(Tong;Takagi&Sugeno;Sugeno–allinLee,1990;Pedrycz,1993)orfuzzymodel-basedcontrol(seelater).
4.Basedonlearning.Theself-organisingcontrollerisanexampleofacontrollerthatfindstherulesitself.Neuralnetworksisanotherpossibility.
Thereisnodesignprocedureinfuzzycontrolsuchasroot-locusdesignandthe
frequencyresponsedesign,poleplacementdesign,orstabilitymargins,becausetherulesareoftennonlinear.Thereforewewillsettlefordescribingthebasiccomponentsandfunctionsoffuzzycontrollers,inordertorecogniseandunderstandthevariousoptionsincommercialsoftwarepackagesforfuzzycontrollerdesign.
Thereismuchliteratureonfuzzycontrolandmanycommercialsoftwaretools(MIT,1995),butthereisnoagreementontheterminology,whichisconfusing.Thereareefforts,however,tostandardisetheterminology,andthefollowingmakesuseofadraftofastandardfromtheInternationalElectrotechnicalCommittee(IEC,1996).Throughout,lettersdenotingmatricesareinbolduppercase,forexampleA;vectorsareinboldlowercase,forexampleX;scalarsareinitalics,forexamplen;andoperationsareinbold,forexamplemin.
1.2Structureofafuzzycontroller
Therearespecificcomponentscharacteristicofafuzzycontrollertosupportadesignprocedure.IntheblockdiagraminFig.3,thecontrollerisbetweenapreprocessingblockandapost-processingblock.Thefollowingexplainsthediagramblockbyblock.
Figure3:
Blocksofafuzzycontroller.
1.3Preprocessing
Theinputsaremostoftenhardorcrispmeasurementsfromsomemeasuringequipment,ratherthanlinguistic.Apreprocessor,thefirstblockinFig.3,conditionsthemeasurementsbeforetheyenterthecontroller.Examplesofpreprocessingare:
1.Quanti