Advanced Practical Process Control PDF
Preview Advanced Practical Process Control
B.Roffel· B.H.Betlem Advanced Practical Process Control Springer Berlin Heidelberg NewYork HongKong London ONLINELIBRARY Milan Engineering Paris Tokyo http://www.springer.de/engine/ B.Roffel .B.H.Betlem Advanced Practical Process Control With159Figures i Springer Prof.Dr.ir.BrianRoffel Dr.ir.BenH.Betlem UniversityofTwente FacultyofChemicalTechnology 7500AEEnschede The Netherlands E-mail:[email protected] Additional materialtothi..bookcan bedownloadedfromhttp://extras,springer.com. ISBN3-540-40183-0 Springer-Verlag Berlin Heidelberg NewYork Cataloging-in-PublicationDataappliedfor Bibliographic informationpublishedbyDieDeutscheBibliothek. DieDeutsche Bibliothekliststhis publicationintheDeutscheNationalbibliografie; detailed bibliographicdataisavailableintheInternetat<http://dnb.ddb.de> Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialis concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,reproductiononmicrofilmorinotherways,andstorageindatabanks.Duplicationof thispublicationorpartsthereofispermittedonlyundertheprovisionsoftheGermanCopyrightLaw ofSeptember9,1965,initscurrentversion, and permissionforusemust alwaysbeobtainedfrom Springer-Verlag.Violations areliableforprosecutionactunderGermanCopyrightLaw. Springer-VerlagBerlinHeidelbergNewYork amemberofBertelsmannSpringerScience+BusinessMediaGmbH http://www.springer.de e Springer-VeriagBerlinHeidelberg 2004 PrintedinGermany Theuseofgeneraldescriptivenames,registerednames,trademarks,etc.inthispublicationdoesnot imply, eveninthe absence ofaspecificstatement,thatsuch names areexemptfrom the relevant protectivelawsandregulationsand thereforefreeforgeneraluse. Typesetting: Camera readybyauthors Cover-design:medio, Berlin Printedonacid-free paper 62/3020hu-543 210 - Preface In the process industries there is an ongoing need for improvement ofthe opera tion ofthe process. One ofthe disciplines that will help the process engineer to achieve this is process control. There are many industrial automation systems to day that will offer powerful tools to meet the process control needs ofindustries withcontinuous, batch and discrete operations. Advanced control solutions sustain and improve the plant's competitiveness by ensuring: - safeoperations - compliancewith environmentalregulations - effectiveuseofraw materialsand energy - efficient production - manufacturingofhigh quality products - flexible accommodationofchangingprocess requirements This book was written from the perspective ofintroducing advanced control con cepts, which can help the engineer to reach the aforementioned goals. Many ad vanced control techniques have been implemented in industry in recent years, since hardware and software platforms are becoming increasingly powerful. Manufacturersofprocess control equipmentcall this hardware and software envi ronmentgenerally'distributedcontrol system'. The distributedcontrol system equipment offers the engineeran excellentplat form for writing and implementing advanced control solutions. However, most large chemical and petrochemical manufacturers hire control specialists to imple ment these control solutions, while small manufacturers often lack the funds to hire these professionals.Therefore itis ourexperiencethat inthe latter case, proc ess engineers often write the control programs required to improve process opera tion. However, the step fromtextbookortheory tocontrol implementationisamajor one. Simulation can help the engineer to increase his or her understanding ofthe problem and enables him or her to first check whether the proposed solution in deed solves theproblem. Thisbook isset up to help the engineer and student to start from the available theory and build control solutions. The focus ofthe book isnotonprocessdynam icsbut onprocess control andprocess optimization.It isassumedthat the readeris familiar with process dynamics, even though some simple models are developed II Preface and explained. The book uses a very practical approach to problems and gives many examples. If the user isdisappointed because a thorough treatment ofthe underlying the ory is missing, we would like to refer him or her to the many excellent books on process dynamics, control and optimisation. Many ofthese books miss, however, the practical solution and implementation ofthe solution in a simulation environ ment. Rather, many books dwell on theoretical considerations and stability analy sis. Eventhough these areextremely important issues, they areoften oflimited in terest to the practising engineer.The structure followed inthis book isapplication directed and goes from single loop control, advanced control loops, multivariable control toprocess optimisation. The simulation tool that is used and that isespecially suited tosolve identifica tion and control problems is MATLAB. In addition, a special tool was written to solve multivariable control problems, called MCPC. Furthermore, in order to solve optimisation problems, the AMPL (Algebraic Modelling Programming Lan guage) program was used. In all cases, the user needs only little time to master these tools. The book startswith two introductory chapters, inwhich Laplace transform and discrete notations are introduced. Some simple but effective control tools such as dead time compensation and feed-forward control are discussed and simulated. MATLAB Simulink is discussed briefly, in order to enable the user to use this powerful software tool. The exercises that are developed, are all written for ver sion RI3 ofMATLAB.The introductory chapters areconcluded with adiscussion ofDahlin's algorithm. The algorithm isnot discussed because ofitspractical rele vance, it rather provides an excellent opportunity to show a number of control concepts that areimportant tothe practising engineer. The next two chapters are devoted to multivariable empirical model identifica tion. Open loop identification as well as closed loop identification and identifica tion ofdifferent types ofmodels are discussed. For parametric model identifica tion one can use the MATLAB identification toolbox, for non-parametric model identification one can use either the MATLAB model predictive control toolbox orthe MCPCsoftware thatcomes withthisbook. After the chapters on model identification, a number of chapters follow that discuss multivariable process control:Dynamic Matrix Control,MultivariableOp timal Constrained Control Algorithm, Internal Model Control, as examples oflin ear model predictive control. Non-linear multivariable control is discussed in a separate chapter, it is limited to Generic Model Control and some simple but ef fective trajectory following approaches. The practical discussion andexamples are limited to linear model predictive control (MPC) and generic model control (GMC). Both algorithms have been implemented in industry and have therefore practical importance. For other multivariable control techniques the reader will have to consult the literature. The exercises on MPC can be done by using the MATLAB model predictive control toolbox or the MCPC software. Utility files have been written to convert models from MATLAB MPC format to MCPC for mat and viceversa. Supportingsoftware III The exercise on GMC is written in MATLAB Simulink, since the non-linear models used in GMC are case specific and it is difficult to provide a generic framework forthistype ofcontrol. The next three chapters discuss process optimization.One chapterwill give the reader some background with respect to optimization of linear problems, optimization of non-linear problems is not discussed in detail although some practical guidelines aregiven forsolving them.TheAMPL environmentis usedto solve non-linear and linear static models. The next chapter discusses the optimization of eight different problems, starting with a simple separation train andending withthe optimizationofanalkylate plant. The last chapter is devoted to the integration ofmodel predictive control and optimisation and shows that the two tools form an integrated framework for im proved process operation. Supporting software Inorderto support the various identification,control andoptimisationapproaches, numerous examples have been worked out inMATLAB and MATLAB Simulink. To run all the examples, the reader should have MATLAB R13 installed, as well as theIdentification Toolbox and the Model Predictive Control toolbox. Some ex amples inchapter6make useofthe model predictive control graphical user inter face,developed by Ricker.This interface canbedownloaded fromthe internet at: http://depts.washington.eduicontrol/LARRY/GUI/index.html#Introduction In order to run the optimisation examples,the reader should download the stu dentversion ofAMPL fromthe internet at: http://www.ampl.com/cm/cs/what/ampl/index.html The MCPC identification and control software A stand-alone software package was written for the identification ofstep weights or impulse weights models that can subsequentlybe used inmultivariable control lerdesign andcontrollersimulation. The controllerdesign and simulation software was originally written inthe early 90'sby Dr.D.Grant Fisher ofthe University of Alberta. In subsequent years the software was modified and improved by the au thors, and an object-oriented software package was the result. In addition, the software was tested on two industrial installations and proved to run very effec tively. For themodel identificationastandard leastsquares algorithm isused,forsolving constrained control problems a fast non-negative constrained least squares algo rithm is used as published by Bro and de long. [Bro, R. and S. de long, A fast non-negative constrained least squares algorithm, J. Chemometrics, 11(5), 393 401, 1997]. IV Preface Acknowledgements The authors would like to thank Dr. D. Grant Fisher ofthe University ofAlberta for making an initial copy ofthe multivariable control and simulation software available to them. We also like to thank the contributions ofnumerous students who participated in the Advanced Practical Control Course at the University of Twente during the past fewyears. They developed several ofthe control and iden tification examples as part ofthis course. Their overwhelming enthusiasm during the course made the authors approach Springer Verlag,Heidelberg, so as to make the material available toaneven wider audience. Enschede, TheNetherlands Brian Roffel April 2003 Ben. H.I. Betlem Contents 1IntroductiontoAdvanced Process ControlConcepts 1 1.1Process Time Constant. 1 1.2DomainTransformations 3 1.3Laplace Transformation 5 1.4Discrete Approximations 7 1.5z-Transforms 9 1.6Advancedand Modifiedz-Transforms 13 1.7CommonElements inControl 16 1.8The Smith Predictor 18 1.9Feed-forwardControl 21 1.10Feed-forwardControl inaSmith Predictor. 23 1.11Dahlin'sControl Algorithm 26 References 31 2Process Simulation 33 2.1 Simulationusing Matlab Simulink 33 2.2SimulationofFeed-forwardControl 37 2.3Control Simulationofa2x2 System 39 2.4 SimulationofDahlin's Control Algorithm .43 3Process Modeling and Identification 45 3.1Model Applications 45 3.2Types ofModels 46 3.2.1White Box andBlackBox Models .46 3.2.2LinearandNon-linearModels .48 3.2.3 Staticand DynamicModels .48 3.2.4Distributed and Lumped ParameterModels .48 3.2.5 ContinuousandDiscrete Models 49 3.3 Empirical (linear) DynamicModels 50 3.4Model StructureConsiderations 50 3.4.1 ParametricModels 52 3.4.2 Non-parametricModels 54 3.5Model Identification 57 3.5.1 Introduction 57 3.5.2IdentificationofParametricModels 57 3.5.3IdentificationofNon-parametricModels 69 References 70 VI Contents 4 Identification Examples 73 4.1SISOFurnaceParametricModel Identification 73 4.2MISO ParametricModel Identification 79 4.3MISO Non-parametricIdentificationofaNon-integratingProcess 83 404 MIMO Identificationofan Integratingand Non-integratingProcess 85 4.5DesignofPlant Experiments 88 4.5.1Nature ofInputSequence 88 4.5.2PRBSType Input 89 4.5.3Step TypeInput 90 4.504TypeofExperiment. 91 4.6Data File Layout 92 4.7 Conversion ofModelStructures 92 4.8Example and Comparison ofOpen and Closed Loop Identification 97 References 102 5LinearMultivariableControl. 103 5.1InteractioninMultivariable Systems 103 5.1.1The RelativeGain Array 103 5.1.2PropertiesoftheRelativeGainArray 104 5.1.3 SomeExamples 105 5.104The Dynamic Relative Gain Array 107 5.2DynamicMatrixControl... 108 5.2.1 Introduction 108 5.2.2BasicDMC Formulation 108 5.2.3 One Step DMC 112 5.204 Prediction Equationand Unmeasurable DisturbanceEstimation 115 5.2.5 RestrictionofExcessiveMoves 116 5.2.6ExpansionofDMC toMultivariableProblems I18 5.2.7 EqualConcernErrors 119 5.2.8ConstraintHandling 120 5.2.9ConstraintFormulation 121 5.3PropertiesofCommercialMPCPackages 124 References 126 6 MultivariableOptimalConstraintControlAlgorithm 127 6.1GeneralOverview 127 6.2Model Formulation forSystemswith Dead Time 129 6.3Model Formulation forMultivariable Processes 130 604 ModelFormulationfor MultivariableProcesseswithTimeDelays 132 6.5Model Formulation inCase ofaLimitedControlHorizon 132 6.6 Mocca Control Formulation 133 6.7Non-linearTransformations 134 6.8Practical ImplementationGuidelines 135 6.9 CaseStudy 136 6.10 ControlofaFluidizedCatalytic Cracker. 140 6.11 ExamplesofCase StudiesinMATLAB 144
The list of books you might like
