Iowa State University Capstones, Theses and Retrospective Theses and Dissertations Dissertations 2000 Development of a new comprehensive predictive modeling and control framework for multiple- input, multiple-output processes Nidhi Bhandari Iowa State University Follow this and additional works at: https://lib.dr.iastate.edu/rtd Part of the Chemical Engineering Commons, and the Systems Engineering Commons Recommended Citation Bhandari, Nidhi, "Development of a new comprehensive predictive modeling and control framework for multiple-input, multiple- output processes " (2000). Retrospective Theses and Dissertations. 12309. https://lib.dr.iastate.edu/rtd/12309 This Dissertation is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Retrospective Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. INFORMATION TO USERS This manuscript has been reproduced from the microfiim master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistnct print colored or poor quaRty illustratons and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorced copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g.. maps, drawings, charts) are reproduced by sectioning ttie original, beginning at the upper left-hand comer and continuing from left to right in equal sections witti small overiaps. Photographs included in the original manuscript have been reproduced xerographically in tiiis copy. Higher qualify 6' x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an add'rtional charge. Contact UMI directiy to order. Bell & Howell Information and Learning 300 North Zeeb Road. Ann Arbor, MI 48106-1346 USA 800-521-0000 Development of a new comprehensive predictive modeling and control framework for multiple-input, multiple-output processes fay NidM Bhandaii A dissertation submitted to the graduate faculty in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Major: Chemical Engineering Major Professon Derrick K. Rollins Iowa State University Ames, Iowa 2000 Copyright © Nidhi Bhandari, 2000. All rights reserved. UMl Number. 9990433 Copyright 2000 by Bhandari. Nidhi All rights reserved. UMI' UMl Microform9990433 Copyright 2001 by Bell & Howell Information and Learning Company. Ail rights reserved. This microform edition is protected against unauthorized copying under TtUe 17, United States Code. Bell & Howell Information and Leaming Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Mt 48106-1346 ii Graduate College Iowa State University This is to certify that the Doctoral dissertation of Nidhi Bhandari has met the dissertation requirements of Iowa State University Signature was redacted for privacy. Major Professor Signature was redacted for privacy. For the Major Pro] Signature was redacted for privacy. For the G] ite College iii TABLE OF CONTENTS CHAPTER L INTRODUCTION 1 L General Introduction 1 2- Model Predictive Control 2 3. Discussion of Related Literature for Modeling and MPC 5 4. Dissertation Organization 8 5. References 9 CHAPTER 2. BACKGROUND 12 1. Review of Modeling Approaches 12 2. Empirical Modeling Methods 13 3. SET 18 4. References 22 CHAPTER 3. ACCURATE PREDICTIVE MODELING OF RESPONSE VARIABLES UNDER DYNAMIC CONDITION WITHOUT THE USE OF PAST RESPONSE DATA 24 Abstract 24 1. Introduction 25 2. The Process Model 26 3. SET Model and Algorithm 27 4. The Study 29 5. Closing Remarks 32 6. References 33 CHAPTER 4. SUPERIOR SEMI-EMPIRICAL DYNAMIC PREDICTIVE MODELING THAT ADDRESSES INTERACTIONS 34 Abstract 34 1. Introduction 34 2. The Process 36 3. Model Identification 38 4. SET Algorithm 41 5. Current Approach 42 6. The Study 44 7. Closing Remarks 47 8. References 49 CHAPTER 5. APPUCATION OF A NEW DYNAMIC PREDICTIVE MODELING APPROACH 50 Abstract 50 Nomenclature 50 1. Introduction 51 2. Proposed Approach 54 iv 3. The Experimental Setup 58 4. The Experimental Design and Model Development 61 5. SET Algorithm 70 6. The Study 71 7. Qosing Remarks 76 8. Acknowledgements 76 9. References 76 CHAPTER 6. NONLINEAR DWAiVflC PREDICTIVE MODELING FULLY UTILIZING STATISTICAL EXPERIMENTAL DESIGN 78 Abstract 78 1. Introduction 79 2. Proposed Approach 83 3. The Process 90 4. SET Model Development 93 5. Current Approach 102 6. Results of he Study 111 7. Concluding Remarks 115 8. Acknowledgements 116 9. Notation 116 10. References 117 CHAPTER 7. CONCLUSIONS AND FUTURE WORK 119 L Conclusions 119 2. Future Work 121 APPENDIX A. DYNAMIC PARAMETERS FOR THE DRYER PROCESS 123 APPENDIX B. DYNAMIC PARAMETERS FOR THE MIMO CSTR DESCRIBED IN CHAPTER 6 126 1 CHAPTER 1. INTRODUCTION 1. General Introduction The competitive nature of the chemical process industry today dictates the need to respond to rapidly changing marketplace conditions. This requires synergism between the many system technologies that are involved in the decision-making process: measurement, control, optimization, and logistics. The control technology thus should be able to encompass not only traditional regulatory aspects but also take into accotmt these other technologies in an automated and integrated manner. The application of digital computers to process control has helped to unify these. The two major areas of application have been acquisition of the process data (passive) and manipulation of the process (active). The growth of digital technology in the last few decades has also represented a challenge to researchers in the field of automatic control in the sense that it made them wonder if the basic approach to control system design and application should not be reconsidered? Model-based predictive control or simply model predictive control (MFC) is one of the advanced control concepts conceived as a part of the answer. MFC is a family of controllers where explicit modeling and digital computation play a major role. MFC is not a specific control strategy but more of an ample range of control methods developed around certain common ideas, which appear, to a greater or lesser degree, in all of them. Various versions of MFC have demonstrated their effectiveness in only a few industrial applications. 2 The most advantageous feature of this strategy is that it is ideally suited, for multivariable control operations. It combines optimization with feedback/feed forward control and provides constraint-handling capabilities. As a result, for multivariable processes with equal numbers of inputs and outputs, the strategy leads to excellent servo and regulatory control. For a process with more outputs than inputs, the user may have tighter control of some outputs relative to others. Other advantages of MPC include the fact that the resulting control law is linear and is especially useful when future references are known. Some of the limitations of MPC are that the derivation of the control law is more complex, there is lack of some theoretical results on stability and robustness, and benefits greatly deteriorate when discrepancies arise between the real process and the model. However, a careful weighing of the strengths and the weaknesses of MPC shows that it still offers a lot of promise in improving control as compared to conventional controllers. 2. Model Predictive Control MPC has emerged as a powerful tool for dynamic control and optimization. Although difE^nt in form, the underlying idea of all MPC schemes is the same and can be stated as follows: 1. A dynamic model and online measurements are used to build a prediction of future output behavior expressed in terms of current and foture manipulated moves. 2. On the basis of the prediction, optimization is performed to find a sequence of input moves that minimizes a chosen measure of the output deviation fix>m its respective reference values while satisfying all constraints. 3 3. Since the quality of the prediction may improve as more measurements are collected, only the first of the calculated input sequences is implemented and the whole optimization is repeated at the next sampling time. This so-called "receding horizon strategy" makes MPC a feedback control algorithm. The basic components of any MPC method are stated below: i. The Predictive model The predictive model is the core of any MPC system, and the success of the algorithm depends heavily on the quality of this model. It therefore becomes imperative to select a model structure and a set of model parameters that yield a model with a high degree of accuracy. Practical experience with the development of MPC and other types of model-based control has shown that, once an adequate dynamic model has been obtained, 80-90% of the work is done [I], Several types of model representations have been used for MPC purposes, and they can be broadly classified into non-parametric and parametric models. Non- parametric models are represented by a curve, a fimction, or a table. Examples of these are impulse and step response models. Parametric models are characterized by a parameter vector of finite dimension, examples being state-space models, transfer fimction models, and the like. Another classification of model representation is linear or non-linear. Most of the models used in MPC have been linear, but many extensions and modifications have been proposed to incorporate non-linear process models as well. Another classification of modeling is empirical, theoretical, or semi-empirical. The models used primarily in MPC are empirical in nature, in that they require past sampled output data of the process for prediction.