PID控制器及智能控制英文文献_第1页
PID控制器及智能控制英文文献_第2页
PID控制器及智能控制英文文献_第3页
PID控制器及智能控制英文文献_第4页
PID控制器及智能控制英文文献_第5页
已阅读5页,还剩10页未读 继续免费阅读

付费下载

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

PID控制器及智能控制英文文献ArticleID:1000-8152(2000)06-0861-07IntroductionstoPIDControllersandIntelligentControlTANYonghongandDANGXuanju(SchoolofComputerSciences,GuilinInstituteofElectronicTechnologyGuilin,541004,P.R.China)AchielR.VanCauwenberghe(DepartmentofControlEngineeringandAutomation,UniversityofGhentGhent,Belgium)MehrdadSaif(SchoolofEngineeringSciences,SimonFraserUniversityBurnaby,BC,V5A1S6,Canada)Abstract:Industrialautomationlevelhasbecomeameasureofprofessionsmodernizationlevelisanimportantsymbol.Atthesametime,thedevelopmentofthetheoryofcontrolalsoexperiencedtheclassicalcontroltheoryandmoderncontroltheoryandintelligentcontroltheoryofthreephaseautomaticcontrolsystemcanbedividedintotheopenloopcontrolsystemandclosed-loopcontrolsystem.Acontrolsystemincludingthecontroller,sensors,actuators,andtransmitterinputandoutputinterface.Theoutputafteroutputinterface,actuators,addtothecontrolsystem,Controlsystem,theaccused,thetransmitter,throughaftersensorinputinterfacetocontroller.Differentcontrolsystem,itssensorsandactuators,thetransmitterisnotthesame.Atpresent,thePIDcontrolandcontrollerorintelligentinstrumenthasmanyPIDcontrollerintheengineeringpractice,theproductshavebeenwidelyapplied,thereareallsortsofPIDcontrollerproducts,companiesaredevelopedwithPIDparameterself-settingfunctionofintelligentregulator(intelligentregulator),includingthePIDcontrollerisadjustedbyintelligentautomaticadjustmentorcalibration,theadaptivealgorithmisproposedtorealize.HaverealizedusingPIDcontrolpressure,temperature,flow,liquidlevelcontroller,canrealizePIDcontrolfunctionoftheprogrammablelogiccontroller(PLC),andcanrealizePIDcontrolPCsystem,etc.Keywords:intelligentcontrol,PIDcontrolPIDcontrollerscanbestand-alonecontrollers(alsocalledsingleloopcontrollers),controllersinPLCs,embeddedcontrollers,orsoftwareinVisualBasicorC#computerprograms.PIDcontrollersareprocesscontrollerswiththefollowingcharacteristics:_Continuousprocesscontrol_Analoginput(alsoknownas“measurement”or“ProcessVariable”or“PV”)_Analogoutput(referredtosimplyas“output”)_Setpoint(SP)_Proportional(P),Integral(I),and/orDerivative(D)constantsExamplesof“continuousprocesscontrol”aretemperature,pressure,flow,andlevelcontrol.forexample,controllingtheheatingofatank.Forsimplecontrol,youhavetwotemperaturelimitsensors(onelowandonehigh)andthenswitchtheheateronwhenthelowtemperaturelimitsensorturnsonandthenturntheheateroffwhenthetemperaturerisestothehightemperaturelimitsensor.Thisissimilartomosthomeairconditioningheatingthermostats.Incontrast,thePIDcontrollerwouldreceiveinputastheactualtemperatureandcontrolavalvethatregulatestheflowofgastotheheater.ThePIDcontrollerautomaticallyfindsthecorrect(constant)flowofgastotheheaterthatkeepsthetemperaturesteadyatthesetpoint.Insteadofthetemperaturebouncingbackandforthbetweentwopoints,thetemperatureisheldsteady.Ifthesetpointislowered,thenthePIDcontrollerautomaticallyreducestheamountofgasflowingtotheheater.Ifthesetpointisraised,thenthePIDcontrollerautomaticallyincreasestheamountofgasflowingtotheheater.LikewisethePIDcontrollerwouldautomaticallyforhot,sunnydays(whenitishotteroutsidetheheater)andforcold,cloudydays.Theanaloginput(measurement)iscalledthe“processvariable”or“PV”.YouwantthePVtobeahighlyaccurateindicationoftheprocessparameteryouaretryingtocontrol.Forexample,ifyouwanttomaintainatemperatureof+or-onedegreethenwetypicallystriveforatleasttentimesthatorone-tenthofadegree.Iftheanaloginputisa12bitanaloginputandthetemperaturerangeforthesensoris0to400degreesthenour“theoretical”accuracyiscalculatedtobe400degreesdividedby4,096(12bits)=0.09765625degrees.[1]Wesay“theoretical”becauseitwouldassumetherewasnonoiseanderrorinourtemperaturesensor,wiring,andanalogconverter.Thereareotherassumptionssuchaslinearity,etc..Thepointbeing-with1/10ofadegree“theoretical”accuracy-evenwiththeusualamountofnoiseandotherproblemsonedegreeofaccuracyshouldeasilybeattainable.Theanalogoutputisoftensimplyreferredtoas“output”.Oftenthisisgivenas0~100percent.Inthisheatingexample,itwouldmeanthevalveistotallyclosed(0%)ortotallyopen(100%).Thesetpoint(SP)issimply-whatprocessvaluedoyouwant.Inthisexample-whattemperaturedoyouwanttheprocessat?ThePIDcontroller’sjobistomaintaintheoutputatalevelsothatthereisnodifference(error)betweentheprocessvariable(PV)andthesetpoint(SP).InFig.16.1,thevalvecouldbecontrollingthegasgoingtoaheater,thechillingofacooler,thepressureinapipe,theflowthroughapipe,thelevelinatank,oranyotherprocesscontrolsystem.PIDcontrollerWhatthePIDcontrollerislookingatisthedifference(or“error”)betweenthePVandtheSP.Itlooksattheabsoluteerrorandtherateofchangeoferror.Absoluteerrormeans-isthereabigdifferenceinthePVandSPoralittledifference?Rateofchangeoferrormeans-isthedifferencebetweenthePVorSPgettingsmallerorlargerastimegoeson.Whenthereisa“processupset”,meaning,whentheprocessvariableorthesetpointquicklychanges-thePIDcontrollerhastoquicklychangetheoutputtogettheprocessvariablebackequaltothesetpoint.[2]Ifyouhaveawalk-incoolerwithaPIDcontrollerandsomeoneopensthedoorandwalksin,thetemperature(processvariable)couldriseveryquickly.ThereforethePIDcontrollerhastoincreasethecooling(output)tocompensateforthisriseintemperature.OncethePIDcontrollerhastheprocessvariableequaltothesetpoint,agoodPIDcontrollerwillnotvarytheoutput.Youwanttheoutputtobeverysteady(notchanging).Ifthevalve(motor,orothercontrolelement)isconstantlychanging,insteadofmaintainingaconstantvalue,thiscouldcausemorewearonthecontrolelement.Sotherearethesetwocontradictorygoals.Fastresponse(fastchangeinoutput)whenthereisa“processupset”,butslowresponse(steadyoutput)whenthePVisclosetothesetpoint.Notethattheoutputoftengoespast(overshoots)thesteady-stateoutputtogettheprocessbacktothesetpoint.Forexample,acoolermaynormallyhaveitscoolingvalveopen34%tomaintainzerodegrees(afterthecoolerhasbeenclosedupandthetemperaturesettleddown).Ifsomeoneopensthecooler,walksin,walksaroundtofindsomething,thenwalksbackout,andthenclosesthecoolerdoo-rthePIDcontrollerisfreakingoutbecausethetemperaturemayhaveraised20degrees!Soitmaycrankthecoolingvalveopento50,75,oreven100percent-tohurryupandcoolthecoolerbackdown-beforeslowlyclosingthecoolingvalvebackdownto34percent.[3]Let’sthinkabouthowtodesignaPIDcontroller.Wefocusonthedifference(error)betweentheprocessvariable(PV)andthesetpoint(SP).Therearethreewayswecanviewtheerror.ThismeanshowbigisthedifferencebetweenthePVandSP.IfthereisasmalldifferencebetweenthePVandtheSP-thenlet’smakeasmallchangeintheoutput.IfthereisalargedifferenceinthePVandSP-thenlet’smakealargechangeintheoutput.Absoluteerroristhe“proportional”(P)componentofthePIDcontroller.ThesumoferrorsovertimeGiveusaminuteandwewillshowwhysimplylookingattheabsoluteerror(proportional)onlyisaproblem.Thesumoferrorsovertimeisimportantandiscalledthe“integral”(I)componentofthePIDcontroller.EverytimewerunthePIDalgorithmweaddthelatesterrortothesumoferrors.InotherwordsSumofErrorS=Error1+Error2+Error3+Error4+....ThedeadtimeDeadtimereferstothedelaybetweenmakingachangeintheoutputandseeingthechangereflectedinthePV.Theclassicalexampleisgettingyourovenattherighttemperature.Whenyoufirstturnontheheat,ittakesawhilefortheovento“heatup”.Thisisthedeadtime.Ifyousetaninitialtemperature,waitfortheoventoreachtheinitialtemperature,andthenyoudeterminethatyousetthewrongtemperature-thenitwilltakeawhilefortheoventoreachthenewtemperaturesetpoint.Thisisalsoreferredtoasthe“derivative”(D)componentofthePIDcontroller.Thisholdssomefuturechangesbackbecausethechangesintheoutputhavebeenmadebutarenotreflectedintheprocessvariableyet.AbsoluteError/ProportionalOneofthefirstideaspeopleusuallyhaveaboutdesigninganautomaticprocesscontrolleriswhatwecall“proportional”.Meaning,ifthedifferencebetweenthePVandSPissmall-thenlet’smakeasmallcorrectiontotheoutput.IfthedifferencebetweenthePVandSPislargethenlet’smakealargercorrectiontotheoutput.Thisideacertainlymakessense.WesimulatedaproportionalonlycontrollerinMicrosoftExcel.Fig.16.2isthechartshowingtheresultsofthefirstsimulation(DEADTIME=0,proportionalonly):ProportionalandIntegralControllersTheintegralportionofthePIDcontrolleraccountsfortheoffsetprobleminaproportionalonlycontroller.WehaveanotherExcelspreadsheetthatsimulatesaPIDcontrollerwithproportionalandintegralcontrol.Here(Fig.16.3)isachartofthefirstsimulationwithproportionalandintegral(DEADTIME=0,proportional=0.4).Asyoucantell,thePIcontrollerismuchbetterthanjustthePcontroller.However,deadtimeofzero(asshowninthegraph)isnotcommon.ThesimulationchartDerivativeControlDerivativecontroltakesintoconsiderationthatifyouchangetheoutput,thenittakestimeforthatchangetobereflectedintheinput(PV).Forexample,let’stakeheatingoftheoven.ThesimulationchartIfwestartturningupthegasflow,itwilltaketimefortheheattobeproduced,theheattoflowaroundtheoven,andforthetemperaturesensortodetecttheincreasedheat.Derivativecontrolsortof“holdsback”thePIDcontrollerbecausesomeincreaseintemperaturewilloccurwithoutneedingtoincreasetheoutputfurther.SettingthederivativeconstantcorrectlyallowsyoutobecomemoreaggressivewiththePIconstants.Intelligenceandintelligentsystemscanbecharacterizedinanumberofwaysandalonganumberofdimensions.Therearecertainattributesofintelligentsystems,commoninmanydefinitions,whichareofparticularinteresttothecontrolcommunity.Inthefollowing,severalalternativedefinitionsandcertainessentialcharacteristicsofntelligentsystemsarefirstdiscussed.Abriefworkingdefinitionofintelligentsystemsthatapturestheircommoncharacteristicsisthenpresented.Inmoredetail,westartwitharathereneraldefinitionofintelligentsystems,wediscusslevelsofintelligence,andweexplaintherolefcontrolinintelligentsystemsandoutlineseveralalternativedefinitions.[1]Wethendiscussdaptationandlearning,autonomyandthenecessityforefficientcomputationalstructuresinntelligentsystems,todealwithcomplexity.Weconcludewithabriefworkingcharacterizationfintelligent(control)systems.Westartwithageneralcharacterizationofintelligentsystems:Anintelligentsystemhastheabilitytoactappropriatelyinanuncertainenvironment,wherenappropriateactionisthatwhichincreasestheprobabilityofsuccess,andsuccessisthechievementofbehavioralsubgoalsthatsupportthesystem’sultimategoal.[2]Inorderforaman-madeintelligentsystemtoactappropriately,itmayemulatefunctionsofivingcreaturesandultimatelyhumanmentalfaculties.Anintelligentsystemcanbeharacterizedalonganumberofdimensions.Therearedegreesorlevelsofintelligencethatcanemeasuredalongthevariousdimensionsofintelligence.Ataminimum,intelligencerequiresheabilitytosensetheenvironment,tomakedecisionsandtocontrolaction.Higherlevelsofntelligencemayincludetheabilitytorecognizeobjectsandevents,torepresentknowledgeinaorldmodel,andtoreasonaboutandplanforthefuture.Inadvancedforms,intelligencerovidesthecapacitytoperceiveandunderstand,tochoosewisely,andtoactsuccessfullyunderlargevarietyofcircumstancessoastosurviveandprosperinacomplexandoftenhostilenvironment.[3]Intelligencecanbeobservedtogrowandevolve,boththroughgrowthinomputationalpowerandthroughaccumulationofknowledgeofhowtosense,decideandactincomplexandchangingworld.Theabovecharacterizationofanintelligentsystemisrathergeneral.Accordingtothis,areatnumberofsystemscanbeconsideredintelligent.Infact,accordingtothisdefinition,evenathermostatmaybeconsideredtobeanintelligentsystem,althoughoflowlevelofintelligence.Itiscommon,however,tocallasystemintelligentwheninfactithasaratherhighlevelofintelligence.Thereexistanumberofalternativebutrelateddefinitionsofintelligentsystemsandinthefollowingwementionseveral.Theyprovidealternative,butrelatedcharacterizationsofintelligentsystemswithemphasisonsystemswithhighdegreesofintelligence.Thefollowingdefinitionemphasizesthefactthatthesysteminquestionprocessesinformation,anditfocusesonman-madesystemsandintelligentmachines:Machineintelligenceistheprocessofanalyzing,organizingandconvertingdataintoknowledge;where(machine)knowledgeisdefinedtobethestructuredinformationacquiredandappliedtoremoveignoranceoruncertaintyaboutaspecifictaskpertainingtotheintelligentmachine.Thisdefinitionleadstotheprincipleofincreasingprecisionwithdecreasingintelligence,whichclaimsthat:applyingmachineintelligencetoadatabasegeneratesaflowofknowledge,lendingananalyticformtofacilitatemodelingoftheprocess.Next,anintelligentsystemischaracterizedbyitsabilitytodynamicallyassignsubgoalsandcontrolactionsinaninternalorautonomousfashion:Manyadaptiveorlearningcontrolsystemscanbethoughtofasdesigningacontrollawtomeetwell-definedcontrolobjectives.Thisactivityrepresentsthesystem’sattempttoorganizeororderits“knowledge”ofitsowndynamicalbehavior,sotomeetacontrolobjective.Theorganizationofknowledgecanbeseenasoneimportantattributeofintelligence.Ifthisorganizationisdoneautonomouslybythesystem,thenintelligencebecomesapropertyofthesystem,ratherthanofthesystem’sdesigner.Thisimpliesthatsystemswhichautonomously(self)-organizecontrollerswithrespecttoaninternallyrealizedorganizationalprincipleareintelligentcontrolsystems.[5]Aproceduralcharacterizationofintelligentsystemsisgivennext:Intelligenceisapropertyofthesystemthatemergeswhentheproceduresoffocusingattention,combinatorialsearch,andgeneralizationareappliedtotheinputinformationinordertoproducetheoutput.Onecaneasilydeducethatonceastringoftheaboveproceduresisdefined,theotherlevelsofresolutionofthestructureofintelligencearegrowingasaresultoftherecursion.Havingonlyonelevelstructureleadstoarudimentaryintelligencethatisimplicitinthethermostat,ortoavariable-structureslidingmodecontroller.Theconceptsofintelligenceandcontrolarecloselyrelatedandtheterm“IntelligentControl”hasauniqueanddistinguishablemeaning.Anintelligentsystemmustdefineandusegoals.Controlisthenrequiredtomovethesystemtothesegoalsandtodefinesuchgoals.Consequently,anyintelligentsystemwillbeacontrolsystem.Conversely,intelligenceisnecessarytoprovidedesirablefunctioningofsystemsunderchangingconditions,anditisnecessarytoachieveahighdegreeofautonomousbehaviorinacontrolsystem.Sincecontrolisanessentialpartofanyintelligentsystem,theterm“IntelligentControlSystems”issometimesusedinengineeringliteratureinsteadof“IntelligentSystems”or“IntelligentMachines”.Theterm“IntelligentControlSystem”simplystressesthecontrolaspectoftheintelligentsystem.Below,onemorealternativecharacterizationofintelligent(control)systemsisincluded.Accordingtothisview,acontrolsystemconsistsofdatastructuresorobjects(theplantmodelsandthecontrolgoals)andprocessingunitsormethods(thecontrollaws):Anintelligentcontrolsystemisdesignedsothatitcanautonomouslyachieveahighlevelgoal,whileitscomponents,controlgoals,plantmodelsandcontrollawsarenotcompletelydefined,eitherbecausetheywerenotknownatthedesigntimeorbecausetheychangedunexpectedly.Thereareseveralessentialpropertiespresentindifferentdegreesinintelligentsystems.Onecanperceivethemasintelligentsystemcharacteristicsordimensionsalongwhichdifferentdegreesorlevelsofintelligencecanbemeasured.[6]Belowwediscussthreesuchcharacteristicsthatappeartoberatherfundamentalinintelligentcontrolsystems.AdaptationandLearning.Theabilitytoadapttochangingconditionsisnecessaryinanintelligentsystem.Althoughadaptationdoesnotnecessarilyrequiretheabilitytolearn,forsystemstobeabletoadapttoawidevarietyofunexpectedchangeslearningisessential.Sotheabilitytolearnisanimportantcharacteristicof(highly)intelligentsystems.AutonomyandIntelligence.Autonomyinsettingandachievinggoalsisanimportantcharacteristicofintelligentcontrolsystems.Whenasystemhastheabilitytoactappropriatelyinanuncertainenvironmentforextendedperiodsoftimewithoutexternalintervention,itisconsideredtobehighlyautono

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论