Process Modelling for Control: A Unified Framework Using Standard Black-box Techniques PDF
Preview Process Modelling for Control: A Unified Framework Using Standard Black-box Techniques
Advances in Industrial Control Other titles published in this Series: NonlinearIdentificationand Control GuopingLiu DigitalControllerImplementationandFragility RobertS.H.IstepanianandJamesF.Whidborne(Eds.) OptimisationofIndustrialProcessesatSupervisoryLevel DorisSa´ez,AldoCiprianoandAndrzejW.Ordys AppliedPredictiveControl HuangSunan,TanKokKiongandLeeTongHeng HardDiskDriveServoSystems BenM.Chen,TongH.LeeandVenkatakrishnanVenkataramanan RobustControlofDieselShipPropulsion NikolaosXiros HydraulicServo-systems MoheiddineJelaliandAndreasKroll Model-basedFaultDiagnosisinDynamicSystemsUsingIdentificationTechniques SilvioSimani,CesareFantuzziandRonJ.Patton StrategiesforFeedbackLinearisation:ADynamicNeuralNetworkApproach FreddyGarces,VictorM.Becerra,ChandrasekharKambhampatiand KevinWarwick RobustAutonomousGuidance AlbertoIsidori,LorenzoMarconiandAndreaSerrani DynamicModellingofGasTurbines GennadyG.KulikovandHaydnA.Thompson(Eds.) ControlofFuelCellPowerSystems JayT.Pukrushpan,AnnaG.StefanopoulouandHueiPeng FuzzyLogic,IdentificationandPredictiveControl JairoEspinosa,JoosVanderwalleandVincentWertz OptimalReal-timeControlofSewerNetworks MagdaleneMarinakiandMarkosPapageorgiou ComputationalIntelligenceinTimeSeriesForecasting AjoyK.PalitandDobrivojePopovic ModellingandControlofmini-FlyingMachines PedroCastillo,RogelioLozanoandAlejandroDzul RudderandFinShipRollStabilization TristanPerez ControlofPassengerTrafficSystemsinBuildings SandorMarkon PublicationdueJanuary2006 NonlinearH2/H∞Feedbackcontrol MuradAbu-Khalaf,FrankL.LewisandJieJuang PublicationdueJanuary2006 Benoˆıt Codrons Process Modelling for Control A Unified Framework Using Standard Black-box Techniques With 74 Figures Benoˆıt Codrons, Dr.Eng. Laborelec, 125 Rue de Rhode, B-1630 Linkebeek, Belgium BritishLibraryCataloguinginPublicationData Codrons,Benoˆıt Processmodellingforcontrol:aunifiedframeworkusing standardblack-boxtechniques.—(Advancesinindustrial control) 1.Processcontrol—Mathematicalmodels I.Title 629.8′312 ISBN1852339187 LibraryofCongressControlNumber2005923269 Apartfromanyfairdealingforthepurposesofresearchorprivatestudy,orcriticismorreview,as permittedundertheCopyright,DesignsandPatentsAct1988,thispublicationmayonlyberepro- duced,storedortransmitted,inanyformorbyanymeans,withthepriorpermissioninwritingof thepublishers,orinthecaseofreprographicreproductioninaccordancewiththetermsoflicences issuedbytheCopyrightLicensingAgency.Enquiriesconcerningreproductionoutsidethoseterms shouldbesenttothepublishers. AdvancesinIndustrialControlseriesISSN1430-9491 ISBN1-85233-918-7SpringerLondonBerlinHeidelberg SpringerScience+BusinessMedia springeronline.com ©Springer-VerlagLondonLimited2005 PrintedintheUnitedStatesofAmerica Theuseofregisterednames,trademarks,etc.inthispublicationdoesnotimply,evenintheabsence of a specific statement, that such names are exempt from the relevant laws and regulations and thereforefreeforgeneraluse. Thepublishermakesnorepresentation,expressorimplied,withregardtotheaccuracyoftheinfor- mationcontainedinthisbookandcannotacceptanylegalresponsibilityorliabilityforanyerrors oromissionsthatmaybemade. Typesetting:Electronictextfilespreparedbyauthor 69/3830-543210 Printedonacid-freepaper SPIN11009207 Advances in Industrial Control SeriesEditors ProfessorMichaelJ.Grimble,ProfessorofIndustrialSystemsandDirector ProfessorMichaelA.Johnson,EmeritusProfessorofControlSystemsand DeputyDirector IndustrialControlCentre DepartmentofElectronicandElectricalEngineering UniversityofStrathclyde GrahamHillsBuilding 50GeorgeStreet GlasgowG11QE UnitedKingdom SeriesAdvisoryBoard ProfessorE.F.Camacho EscuelaSuperiordeIngenieros UniversidaddeSevilla CaminodelosDescobrimientoss/n 41092Sevilla Spain ProfessorS.Engell LehrstuhlfrAnlagensteuerungstechnik FachbereichChemietechnik Universita¨tDortmund 44221Dortmund Germany ProfessorG.Goodwin DepartmentofElectricalandComputerEngineering TheUniversityofNewcastle Callaghan NSW2308 Australia ProfessorT.J.Harris DepartmentofChemicalEngineering Queen’sUniversity Kingston,Ontario K7L3N6 Canada ProfessorT.H.Lee DepartmentofElectricalEngineering NationalUniversityofSingapore 4EngineeringDrive3 Singapore117576 ProfessorEmeritusO.P.Malik DepartmentofElectricalandComputerEngineering UniversityofCalgary 2500,UniversityDrive,NW Calgary Alberta T2N1N4 Canada ProfessorK.-F.Man ElectronicEngineeringDepartment CityUniversityofHongKong TatCheeAvenue Kowloon HongKong ProfessorG.Olsson DepartmentofIndustrialElectricalEngineeringandAutomation LundInstituteofTechnology Box118 S-22100Lund Sweden ProfessorA.Ray PennsylvaniaStateUniversity DepartmentofMechanicalEngineering 0329ReberBuilding UniversityPark PA16802 USA ProfessorD.E.Seborg ChemicalEngineering 3335EngineeringII UniversityofCaliforniaSantaBarbara SantaBarbara CA93106 USA DoctorI.Yamamoto TechnicalHeadquarters NagasakiResearch&DevelopmentCenter MitsubishiHeavyIndustriesLtd 5-717-1,Fukahori-Machi Nagasaki851-0392 Japan Series Editors’ Foreword The series Advances in Industrial Control aims to report and encourage technology transfer in control engineering. The rapid development of control technology has an impact on all areas of the control discipline. New theory, new controllers, actuators, sensors, new industrial processes, computer meth- ods, new applications, new philosophies..., new challenges. Much of this development work resides in industrial reports, feasibility study papers and the reports of advanced collaborative projects. The series offers an opportu- nityfor researcherstopresent anextended expositionofsuch newworkin all aspects of industrial control for wider and rapid dissemination. Thestandardframeworkforprocessmodellingandcontroldesignhasbeen: Determine nominal model – Validate model – Progress control design – Im- plement controller. Benoıˆt Codrons brings new depth and insight to these standard steps in this new Advances in Industrial Control monograph Pro- cess Modelling for Control. His approach is to pose a searching set of ques- tionsforeachcomponentofthemodelling – controlcycle.Thefirsttwochap- ters set the scene for model identification including a short survey of transfer function system representation, prediction-error identification and balanced model truncation methods. This is followed by four investigative chapters each constructed around answering some questions very pertinent to the use of model identification for control design. As an example, look at Chapter 3 where the motivating question is simply: “How should we identify a model (Gˆ, Hˆ) in such a way that it be good for control design?” Dr. Codrons then presents a careful theoretical analysis and supporting examples to demon- strate that “it is generally preferable to identify a (possible low-order) model in closed loop when the purpose is to use the model for control design”. This type of presentation is used to examine similar questions for the other stages in the model identification – control design cycle. The mix of theory and illustrative examples is used by Dr. Codrons to lead on to recommendations andsetsofguidelinesthattheindustrialcontrolengineerwillfindinvaluable. vii viii SeriesEditors’Foreword This new volume in the Advances in Industrial Control series has thor- oughtheoreticalmaterialthatwillappealtotheacademiccontrolcommunity lecturersandpostgraduatestudentsalike.Therhetoricalquestion-and-answer style along with the provision of recommendations and guidelines will give the industrial engineer and practitioner the motivation to consider and try new approaches in the model identification field in industrial control. Hence thevolumeisaverywelcomeadditiontotheAdvancesinIndustrialControl monographserieswhichwehopewillencouragetheindustrialcontrolcommu- nity to look anew at the current model identification paradigm. M.J. Grimble and M.A. Johnson Industrial Control Centre Glasgow, Scotland, U.K. Preface The contents of this book are, for the most part, the result of the experience I gained during the five years I spent as a researcher at the Centre for Sys- tems Engineering and Applied Mechanics (CESAME), Universite´ catholique de Louvain, Belgium, from September 1995 to June 2000. I am very grateful to Michel Gevers, Brian D.O. Anderson, Xavier Bombois and FrankyDeBruyneforthegoodtimeswespent,thegreatideaswehadandthe good work we did together. I also want to give credit to Pascale Bendotti and Cl´ement-Marc Falinower of E´lectricite´ de France for their support and collaboration in the elaboration of the case study in Chapter 6. After completion of my Ph.D. thesis (Codrons, 2000), I landed in an indus- trial R&D laboratory, fully loaded with scientific knowledge but with almost no practical experience. This was the real beginning of the genesis of this book. As I have spent the last four years discovering the industrial reality and trying to put my theoretical knowledge into practice, I have also discovered the gap that often separates both worlds. For instance, when the issue is the realisation of identification tests on an industrial system in order to design or tunesomecontroller, in aratheramusingway, both thescientistandtheplant operator will recommend making the tests in closed loop. The latter, because it is often more comfortable for him and less risky for plant operation than open-loop tests; the former, for the reasons exposed in this book. In spite of this, however, the usual practice consists of opening the loop, applying a step on the input on the process, and measuring the output response from which essentialvalueslikethedominanttimeconstantandthegainoftheprocesscan be determined. What is the problem, then? Who is right? Is it the scientist and his closed-loop... whim? Or is it the control or process engineer, with his rules of good practice or her pragmatic approach? To be fair, both are prob- ably right. The open-loop approach is satisfactory most of the time when the objectiveistotunePIDcontrollers. grossomodo,thoseonlyrequireknowledge ix x Preface of the processtime constant and gain for their tuning. However, this approach deprives the engineer of the necessary knowledge of the process dynamics that would help him to design more sophisticated control structures, especially in thecaseofcoupledsubsystemsthatwouldrequireaglobalhandlingwithmulti- variablecontrolsolutionsinlieuofmonovariablebutinteracting(andmutually disturbing) PID loops. The problem is even worse when the objective is to design a controller that has to be simultaneously optimal (with respect to some performance indica- tor, e.g., the variance of the output or an LQG criterion), robust (with respect to modelling errors, disturbances, changes in operating conditions, etc.) and implementable in an industrial control system (which will generally put a con- straintonitsorderand/orcomplexity). Mostofthetime,suboptimalsolutions are worked out on the basis of, again, monovariable control loops and some- timesindecipherablelogicprogrammingaimedatmanagingtheworkingstatus of the various PID controllers in function of the process conditions. This re- sults, generally, from a lack of knowledge of the existing modelling and control design tools, or of the way they can be used in industrial practice. Actually,oneofthemostcriticalandbadlytunedloopsIhavebeenfacedwith isprobablythelearningloopbetweenindustryandtheacademicworld. With- out pretentiousness, my intention when starting to write of this book was to make some of the most recent results in process modelling for control available to the industrial community. The objective is twofold: firstly, it is to provide the control engineer with the necessary theory on modelling for control using somechosenspecificblack-boxtechniquesinalinear,time-invariantframework (itsoundsareasonablefirststepbeforenonlineartechniquescanbeaddressed); secondly,andthisisperhapsthemostimportantissue,itistoinitiateachange is the way modelling and control problems are perceived and tackled in indus- trial practice. This is just a small part of what a feed-forward action from University to Industry might be. Also very important is the feedback from Industry to Academia that happens essentially through collaborations. Benoˆıt Codrons Brussels, 1st September, 2004
The list of books you might like
