重庆大学-过程控制-process-control-中文-翻译-第六章

1ProcessControlCollegeofAutomatonChongqingUniversity重庆大学自动化学院2Design&TuningofSingleLoopSystemsOutlineTuningcontrollerBasedonPRC基于脉冲比控制的调优控制器Cohen-CoonRelation科恩-库恩的关系Ziegler-NicholsRelation,齐格勒-尼科尔斯关系、ITAErelation,Ciancone-MarlinrelationDirectSynthesis直接合成InternalModelControl内模控制3Design&TuningofSingleLoopSystemsPIDTuning协调trial实验P(t)4Design&TuningofSingleLoopSystemsTuningControllerswithEmpiricalRelationsApracticeofdesigningthecontrollersettings设置Makinganopen(orclosed)loopexperimentFittingthesystemresponsetoamodel拟合模型的系统响应Calculating计算thecontrollergainsusingthefittingcurve拟合曲线(empiricalmodel经验模型)FollowingsomeempiricalrelationsEmpiricaltuninggivesaquickstartingpointThoughitdoesn’tdictatesystemdynamicspecifications尽管它不能决定系统动态参数Itisseldomoptimal最佳的,involving包括fieldtuning现场调试1.Cohen-CoonRelation2.Ziegler-NicholsRelation,3.ITAErelation,4.Ciancone-Marlinrelation5Design&TuningofSingleLoopSystemsTwoRemarks评论onEmpiricalTuning经验调整Tuningrelations调整关系arebasedonFittingopenloopdatatothe1storderwithdeadtimemodel拟合一阶开环数据的延迟时间模型Systemsissupposedto被期望beself-regulating自动调节的Deadtimecanapproximatemulti-stageprocess,ormeasuringlag延迟时间可以近似的多阶段过程，或测量滞后Sometuningrelations,e.g.,CohnandCoonbeingdeveloped（intermof在……方面）¼decayratio衰退率inhandlingdisturbancebeingnotsuitablefor适合servoproblems(why?)6Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCObtaining获取processreactioncurve过程响应曲线(PRC)Disablingthecontroller禁用控制器Introducingastepchangetotheactuator执行机构引入阶跃变化Measuringtheopenloopstepresponse,resultingin导致PRC7Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCInthisopenloopsteptest开环阶跃测试Thelumped总的responsecanberepresented表现byUsingGPRCforDesignoftherequiredcontrolleriftheprocessmodelsarenotknowninadvance预先Areducedordermodel降阶模型fortheproduct乘积ofGa,Gp,andGm8Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample:Aself-regulating自调整capacityprocessSteptestisshowninthefigurebelow阶跃测试结果如下表Notethat注意weassumethattheprocessisself-regulating9Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample:Aself-regulatingmulti-capacityprocessSteptestisshowninthefigurebelow一个自我调节的多容量的过程测试步骤如下图所示Openlooptestdata1storderwithdeadtimeTimeconstantdecidedbytheinitialslopeandavisualestimationofdeadtimetd时间常数由最初的斜坡和目测的延迟时间TD决定sigmoidalformS形10Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample(cont.):WhydoesFOPDT纯滞后过程work?TherealtimedatausuallytakesthesigmoidalformS形Processunitoperationsaregenerallyopen-loopstable过程单元操作通常都是开环稳定,thus,theapproximationworksinrealityReminder提示:thereisnounderdampedresponseduetothetakingoutof拿出thecontrollerinthesteptestTheempiricalrelationssuchasZ-N,C-C,C-M,andIAEareusuallybeusedinthiscase这种情况下11Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCCommentson评论theminimumerror最小误差integralcriteria积分标准Calculating计算theerrorbyE=R-C,whereCisbasedGPRCMinimizing极小化的theerrorovertime超时(e.g.,byLagrangeMultiplier)Theformsofintegral积分的errorWhysomanyforms?12Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCCommentsontheminimumerrorintegral积分过程(cont.)Integraloftheabsoluteerror绝对误差Integraltime-weighted加权的absoluteerrorIAEputstheequalweight重要性tolargeandsmallerror1.Itputsaheavypenaltyonerrorspersistingforlongperiodsoftime很大的误差将持续很长一段时间1.Resultingcontrollerallowsfor得到lowsettlingtimes建立时间13Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCCommentsontheminimumerrorintegral(cont.)Integralofthesquareerror平方误差Squaringemakesitevensmallerwhene1,thus1.Suppressing抑制thelarge,initial初始的errors2.Resultingcontrollertendstohaveahighproportionalgain3.Thesystemisveryunderdamped14Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCCommentsontheminimumerrorintegral(cont.)PIDcontrollerwithI=10D=10,15Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample:WhatwouldbethePIDcontrollersettingsforthedyemixingproblem?燃料混合问题Solution:Rememberthatwehavealreadyobtained获得Makingtheopenloopexperiment实验usingGPRC,thewell-fitted合适的modelis16Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample(cont.):ThedyemixingproblemHowever,the1storderapproximationmaylooklikeThePIDcontrollersettingsbydifferenttuningrelationsCohen-CoonTuning17Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample(cont.):ThedyemixingproblemBecausePRCdoesnotrepresenttheprocessexactly,weassumethatthemodelhas25%errorsinprocess1.Gain,2.Timeconstant3.Deadtime0.94-1.560.68-1.133.00-5.0018Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample(cont.):why25%errorsMakingyourmodelrobustforpossibledynamics动力19Design&TuningofSingleLoopSystemsControllersettingsBasedonPRCExample(cont.):Thedyemixingproblem1.ThelargestKCinCohn-Coonresultsin导致themostunderdampedresponse,thusaggressive2.Ciancone-Marlingivesthemostconservative保守的setting,showingacommonindustrypreferencetoPIcontroller3.Otherscanbeconsideredasmoderate适当的Reminder:•Theempiricaltuningisaveryimprecise不精确的science•Tuningparameters参数with2significantdigits有效数字isenough20Design&TuningofSing