McGraw-HillÕs CHEMICAL ENGINEERING SERIES Key Features: control classesÉthat this is just another mathematics course disguised as an engineering course ¨ ¨ and Excel¨ have been introduced throughout the book. dynamics and control and not get bogged down in the mathematical complexities of each problem The Solutions to the End-of-Chapter Problems are available to Instructors at the textÕs website: www.mhhe.com/coughanowr-leblanc Electronic Textbook Options This text is offered through CourseSmart for both instructors and students. CourseSmart is an online browser where students can purchase access to this and other McGraw-Hill textbooks in a digital Process Systems Analysis and Control Third Edition half the cost of a traditional text. Purchasing the eTextbook also allows students to take advantage of sales representative or visit www.CourseSmart.com. ISBN 978-0-07-339789-4 MHID 0-07-339789-X www.mhhe.com Third Edition Donald R. Coughanowr Steven E. LeBlanc MD DALIM 976649 7/29/08 CYAN MAG YELO BLACK available for the course material Coughanowr LeBlanc Process Systems Analysis and Control Process Systems Analysis and Control, Third Edition retains the clarity of presentation for which this book is well known. It is an ideal teaching and learning tool for a semester-long undergraduate chemical engineering course in process dynamics and control. It avoids the encyclopedic approach of many other texts on this topic. Computer examples using MATLAB¨ and Simulink¨ have been introduced throughout the book to supplement and enhance standard hand-solved examples. These packages allow the easy construction of block diagrams and quick analysis of control concepts to enable the student to explore Òwhat-ifÓ type problems that would be much more difficult and time consuming by hand. New homework problems have been added to each chapter. The new problems are a mixture of hand-solutions and computational-exercises. One-page capsule summaries have been added to the end of each chapter to help students review and study the most important concepts in each chapter. Confirming Pages PROCESS SYSTEMS ANALYSIS AND CONTROL cou9789x_fm_i-xx.indd i 8/25/08 2:48:38 PM Confirming Pages McGraw-Hill Chemical Engineering Series Editorial Advisory Board Eduardo D. Glandt, Dean, School of Engineering and Applied Science, University of Pennsylvania Michael T. Klein, Dean, School of Engineering, Rutgers University Thomas F. Edgar, Professor of Chemical Engineering, University of Texas at Austin Coughanowr and LeBlanc: Davis and Davis: de Nevers: de Nevers: Douglas: Edgar, Himmelblau, and Lasdon: Marlin: McCabe, Smith, and Harriott: Murphy: Perry and Green: Peters, Timmerhaus, and West: Smith, Van Ness, and Abbott: cou9789x_fm_i-xx.indd ii Process Systems Analysis and Control Fundamentals of Chemical Reaction Engineering Air Pollution Control Engineering Fluid Mechanics for Chemical Engineers Conceptual Design of Chemical Processes Optimization of Chemical Processes Process Control Unit Operations of Chemical Engineering Introduction to Chemical Processes Perry’s Chemical Engineers’ Handbook Plant Design and Economics for Chemical Engineers Introduction to Chemical Engineering Thermodynamics 8/25/08 2:48:39 PM Confirming Pages The Founding of a Discipline: The McGraw-Hill Companies, Inc. Series in Chemical Engineering Over 80 years ago, 15 prominent chemical engineers met in New York to plan a continuing literature for their rapidly growing profession. From industry came such pioneer practitioners as Leo H. Baekeland, Arthur D. Little, Charles L. Reese, John V. N. Dorr, M. C. Whitaker, and R. S. McBride. From the universities came such eminent educators as William H. Walker, Alfred H. White, D. D. Jackson, J. H. James, Warren K. Lewis, and Harry A. Curtis. H. C. Parmlee, then editor of Chemical and Metallurgical Engineering, served as chairman and was joined subsequently by S. D. Kirkpatrick as consulting editor. After several meetings, this committee submitted its report to the McGraw-Hill Book Company in September 1925. In the report were detailed specifications for a correlated series of more than a dozen texts and reference books which became the McGraw-Hill Series in Chemical Engineering—and in turn became the cornerstone of the chemical engineering curricula. From this beginning, a series of texts has evolved, surpassing the scope and longevity envisioned by the founding Editorial Board. The McGraw-Hill Series in Chemical Engineering stands as a unique historical record of the development of chemical engineering education and practice. In the series one finds milestones of the subject’s evolution: industrial chemistry, stoichiometry, unit operations and processes, thermodynamics, kinetics, and transfer operations. Textbooks such as McCabe et al., Unit Operations of Chemical Engineering, Smith et al., Introduction to Chemical Engineering Thermodynamics, and Peters et al., Plant Design and Economics for Chemical Engineers have taught to generations of students the principles that are key to success in chemical engineering. Chemical engineering is a dynamic profession, and its literature continues to grow. McGraw-Hill, with its in-house editors and consulting editors Eduardo Glandt (Dean, University of Pennsylvania), Michael Klein (Dean, Rutgers University), and Thomas Edgar (Professor, University of Texas at Austin), remains committed to a publishing policy that will serve the needs of the global chemical engineering profession throughout the years to come. cou9789x_fm_i-xx.indd iii 8/25/08 2:48:39 PM cou9789x_fm_i-xx.indd iv 8/25/08 2:48:39 PM Confirming Pages PROCESS SYSTEMS ANALYSIS AND CONTROL THIRD EDITION Steven E. LeBlanc Chemical Engineering University of Toledo Donald R. Coughanowr Emeritus Professor, Chemical Engineering Drexel University cou9789x_fm_i-xx.indd v 8/25/08 2:48:39 PM PROCESS SYSTEMS ANALYSIS AND CONTROL, THIRD EDITION Published by McGraw-Hill, a business unit of The McGraw-Hill Companies, Inc., 1221 Avenue of the Americas, New York, NY 10020. Copyright © 2009 by The McGraw-Hill Companies, Inc. All rights reserved. Previous editions © 1991 and 1965. No part of this publication may be reproduced or distributed in any form or by any means, or stored in a database or retrieval system, without the prior written consent of The McGraw-Hill Companies, Inc., including, but not limited to, in any network or other electronic storage or transmission, or broadcast for distance learning. Some ancillaries, including electronic and print components, may not be available to customers outside the United States. This book is printed on acid-free paper. 1 2 3 4 5 6 7 8 9 0 DOC/DOC 0 9 8 ISBN 978–0–07–339789–4 MHID 0–07–339789–X Global Publisher: Raghothaman Srinivasan Sponsoring Editor: Debra B. Hash Director of Development: Kristine Tibbetts Developmental Editor: Lorraine K. Buczek Senior Marketing Manager: Curt Reynolds Project Coordinator: Melissa M. Leick Senior Production Supervisor: Laura Fuller Designer: Laurie B. Janssen Compositor: Laserwords Private Limited Typeface: 10/12 Times Roman Printer: R. R. Donnelley Crawfordsville, IN Library of Congress Cataloging-in-Publication Data Coughanowr, Donald R. Process systems analysis and control.—3rd ed. / Donald R. Coughanowr, Steven E. LeBlanc. p. cm.—(Mcgraw-Hill chemical engineering series) Includes index. ISBN 978–0–07–339789–4—ISBN 0–07–339789–X (hard copy : alk. paper) 1. Chemical process control. I. LeBlanc, Steven E. II. Title. TP155.75.C68 2009 660'.2815—dc22 2008018252 www.mhhe.com cou9789x_fm_i-xx.indd vi 8/25/08 2:48:40 PM Confirming Pages Dedication For Molly, my children, and grandchildren . . . cou9789x_fm_i-xx.indd vii 8/25/08 2:48:41 PM cou9789x_fm_i-xx.indd viii 8/25/08 2:48:41 PM Confirming Pages CONTENTS Preface to the Third Edition Chapter 1 1.1 1.2 Introductory Concepts 1 Why Process Control? Control Systems 1 1 PART I MODELING FOR PROCESS DYNAMICS Chapter 2 Modeling Tools for Process Dynamics 2.1 2.2 2.3 Chapter 3 3.1 3.2 Appendix 3A: 9 11 Process Dynamics—A Chemical Mixing Scenario Mathematical Tools for Modeling Solution of Ordinary Differential Equations (ODEs) 11 18 26 Inversion by Partial Fractions 32 Partial Fractions Qualitative Nature of Solutions Further Properties of Transforms and Partial Fractions 32 43 49 PART II LINEAR OPEN-LOOP SYSTEMS Chapter 4 Response of First-Order Systems 4.1 4.2 4.3 4.4 4.5 4.6 4.7 xv Transfer Function Transient Response Forcing Functions Step Response Impulse Response Ramp Response Sinusoidal Response 69 71 71 77 78 79 84 87 87 ix cou9789x_fm_i-xx.indd ix 8/25/08 2:48:41 PM Confirming Pages x CONTENTS Chapter 5 5.1 5.2 Chapter 6 6.1 6.2 6.3 Chapter 7 7.1 7.2 Physical Examples of First-Order Systems Examples of First-Order Systems Linearization 99 109 Response of First-Order Systems in Series 123 Introductory Remarks Noninteracting System Interacting System 123 123 128 Higher-Order Systems: Second-Order and Transportation Lag 137 Second-Order System Transportation Lag 137 153 PART III LINEAR CLOSED-LOOP SYSTEMS Chapter 8 The Control System 163 165 Introduction Components of a Control System Block Diagram Development of Block Diagram 165 165 166 168 Controllers and Final Control Elements 186 9.1 9.2 Appendix 9A: Mechanisms Ideal Transfer Functions Piping and Instrumentation Diagram Symbols 187 190 203 Chapter 10 Block Diagram of a Chemical-Reactor Control System 205 Description of System Reactor Transfer Functions Control Valve Measuring Element Controller Controller Transducer Transportation Lag Block Diagram 206 206 209 210 211 212 212 212 Closed-Loop Transfer Functions 218 Standard Block-Diagram Symbols Overall Transfer Function for Single-Loop Systems Overall Transfer Function for Multiloop Control Systems 218 219 224 8.1 8.2 8.3 8.4 Chapter 9 10.1 10.2 10.3 10.4 10.5 10.6 10.7 10.8 Chapter 11 11.1 11.2 11.3 cou9789x_fm_i-xx.indd x 99 8/25/08 2:48:41 PM Confirming Pages CONTENTS Chapter 12 12.1 12.2 12.3 12.4 12.5 Chapter 13 13.1 13.2 13.3 13.4 Chapter 14 14.1 Transient Response of Simple Control Systems Proportional Control for Set Point Change (Servo Problem—Set Point Tracking) Proportional Control for Load Change (Regulator Problem—Disturbance Rejection) Proportional-Integral Control for Load Change Proportional-Integral Control for Set Point Change Proportional Control of System with Measurement Lag 15.1 15.2 15.3 Chapter 16 16.1 16.2 16.3 16.4 cou9789x_fm_i-xx.indd xi 229 234 236 241 243 252 Concept of Stability Definition of Stability (Linear Systems) Stability Criterion Routh Test for Stability 252 254 254 258 Root Locus 269 Concept of Root Locus 269 285 287 Substitution Rule Bode Diagrams Appendix—Generalization of Substitution Rule 287 300 316 Control System Design by Frequency Response 323 Tank Temperature Control System The Bode Stability Criterion Gain and Phase Margins Ziegler-Nichols Controller Settings 323 326 327 335 PART V PROCESS APPLICATIONS Chapter 17 Advanced Control Strategies 17.1 17.2 17.3 17.4 17.5 228 Stability PART IV FREQUENCY RESPONSE Chapter 15 Introduction to Frequency Response Cascade Control Feedforward Control Ratio Control Dead-Time Compensation (Smith Predictor) Internal Model Control xi 351 353 353 361 370 373 378 8/25/08 2:48:42 PM Confirming Pages xii CONTENTS Chapter 18 Controller Tuning and Process Identification 391 18.1 18.2 18.3 Controller Tuning Tuning Rules Process Identification 391 394 Chapter 19 Control Valves 423 Control Valve Construction Valve Sizing Valve Characteristics Valve Positioner 423 425 427 438 Theoretical Analysis of Complex Processes 443 Control of a Steam-Jacketed Kettle Dynamic Response of a Gas Absorber Distributed-Parameter Systems 443 453 458 19.1 19.2 19.3 19.4 Chapter 20 20.1 20.2 20.3 PART VI STATE-SPACE METHODS Chapter 21 State-Space Representation of Physical Systems 475 21.1 21.2 Appendix 21A: Introduction State Variables Elementary Matrix Algebra 477 477 490 Chapter 22 Transfer Function Matrix 498 Transition Matrix Transfer Function Matrix 499 502 Multivariable Control 512 Control of Interacting Systems Stability of Multivariable Systems 514 525 22.1 22.2 Chapter 23 23.1 23.2 PART VII NONLINEAR CONTROL Chapter 24 Examples of Nonlinear Systems 24.1 24.2 24.3 24.4 24.5 Chapter 25 25.1 25.2 cou9789x_fm_i-xx.indd xii 410 477 531 533 Definition of a Nonlinear System The Phase Plane Phase-Plane Analysis of Damped Oscillator Motion of a Pendulum A Chemical Reactor 533 534 535 543 547 Examples of Phase-Plane Analysis 553 Phase Space Examples of Phase-Plane Analysis 553 561 8/25/08 2:48:42 PM Confirming Pages CONTENTS PART VIII COMPUTERS IN PROCESS CONTROL Chapter 26 Microprocessor-Based Controllers and Distributed Control 26.1 26.2 26.3 26.4 26.5 cou9789x_fm_i-xx.indd xiii xiii 579 581 Historical Background Hardware Components Tasks of a Microprocessor-Based Controller Special Features of Microprocessor-Based Controllers Distributed Control 581 582 583 588 592 Bibliography Index 597 599 8/25/08 2:48:42 PM cou9789x_fm_i-xx.indd xiv 8/25/08 2:48:42 PM Confirming Pages PREFACE TO THE THIRD EDITION I t has been over 17 years since the second edition of this book was published. The second edition, which was written by Dr. Donald R. Coughanowr in 1991, contained many changes and new topics to bring the book up to date at the time of publication. The third edition has been a number of years in the making. I would like to thank Dr. Coughanowr for the opportunity to work on this project and help update this excellent work, which he first published in 1965 with Dr. Lowell B. Koppel. As an undergraduate, I learned process control from the first edition of this text over 30 years ago. It was an excellent book then, and it still is. I’ve used a number of other books over the years as a student and as a professor, but I kept coming back to this one. I felt that it was the best book to learn from. Is it an all-encompassing, totally comprehensive process dynamics and control book? No, but it is not intended to be. It is a clearly written book that is geared toward students in a first process dynamics and control course. Many control books on the market contain more material than one could ever hope to cover in a standard undergraduate semester-long class. They can be overwhelming and difficult to learn from. I have always felt that one of the strengths of this book, from both the student’s and professor’s point of view, was the relatively short, easy-to-read chapters that can be covered in one to two lectures. An additional strength of this text has been its unique ability to be a teaching and learning text. I hope that in this current revision, I have been able to retain that style and flavor, while introducing some new material and examples to update the text. OBJECTIVES AND USES OF THE TEXT This text is intended for use in an introductory one-semester-long undergraduate process dynamics and control course. It is intended to be not a comprehensive treatise on process control, but rather a textbook that provides students with the tools to learn the basic material and be in a position to continue their studies in the area if they so choose. Students are expected to have a background in mathematics through differential equations, material and energy balance concepts, and unit operations. After the first 13 chapters, the instructor may select from the remaining chapters to fit a course of particular duration and scope. A typical one-semester 15-week course, for example, may include Chapters 1 through 19 and 26. Features of the third edition • A capsule summary of the important points at the end of each chapter • Restructuring of the initial chapters to reduce the impression that students frequently have regarding control classes—that this is just another mathematics course disguised as an engineering course • Integration of MATLAB,® Simulink,® and Excel throughout the text: • To reduce the tedium of solving problems so that students may concentrate more on the concepts of dynamics and control and not get bogged down in the mathematical complexities of each problem xv cou9789x_fm_i-xx.indd xv 8/25/08 2:48:42 PM Confirming Pages xvi PREFACE TO THE THIRD EDITION • To give students the tools to be able to ask (and more easily answer) “what if . . .?” type of questions • To allow students to explore more difficult problems than would otherwise be possible in the time available for the course material ACKNOWLEDGMENTS We would like to thank the following reviewers of the third edition for their helpful comments and suggestions: Thomas F. Edgar, University of Texas–Austin; John Erjavec, University of North Dakota; Duane Johnson, University of Alabama; Costas Maranas, Penn State University; Michael Nikolaou, University of Houston; F. Joseph Schork, Georgia Institute of Technology; Delmar Timm, University of Nebraska; and William A. Weigand, University of Maryland. We especially acknowledge the helpful suggestions from Susan Montgomery of the University of Michigan and thank her for her thoroughness and useful comments to help make the text more student-friendly. I would like to thank McGraw-Hill for having confidence in this project and providing the opportunity to revise and update the text. Special thanks go to Lorraine Buczek, Developmental Editor, and Melissa Leick, Project Manager, for their help in the final stages of this revision. I would also like to thank my students and my University of Toledo colleague Sasidhar Varanasi for his help in using manuscript drafts when he taught the Process Control course to “field-test” the revisions. I am also grateful to my friend and colleague Dean Nagi Naganathan, of the College of Engineering at the University of Toledo, for his general support and his willingness to allow me the time required to complete this work. I especially want to thank my wife, Molly, for her love and continuing encouragement and support over the course of the writing and revising. Dr. Steven E. LeBlanc University of Toledo RESOURCES FOR INSTRUCTORS AND STUDENTS: For instructors, the solutions to the end-of-chapter problems are available at the text’s website: www.mhhe.com/coughanowr-leblanc ELECTRONIC TEXTBOOK OPTIONS This text is offered through CourseSmart for both instructors and students. CourseSmart is an online browser where students can purchase access to this and other McGrawHill textbooks in a digital format. Through their browser, students can access the complete text online for one year at almost half the cost of a traditional text. Purchasing the eTextbook also allows students to take advantage of CourseSmart’s Web tools for learning, which include full text search, notes and highlighting, and e-mail tools for sharing notes between classmates. To learn more about CourseSmart options, contact your sales representative or visit www.CourseSmart.com. cou9789x_fm_i-xx.indd xvi 8/25/08 2:48:43 PM Confirming Pages HISTORY OF PROCESS SYSTEMS ANALYSIS AND CONTROL (FROM THE SECOND EDITION PREFACE) Since the first edition of this book was published in 1965, many changes have taken place in process control. Nearly all undergraduate students in chemical engineering are now required to take a course in process dynamics and control. The purpose of this book is to take the student from the basic mathematics to a variety of design applications in a clear, concise manner. The most significant change since the first edition is the use of the digital computer in complex problem solving and in process control instrumentation. However, the fundamentals of process control, which remain the same, must be acquired before one can appreciate the advanced topics of control. In its present form, this book represents a major revision of the first edition. The material for this book evolved from courses taught at Purdue University and Drexel University. The first 17 chapters on fundamentals are quite close to the first 20 chapters of the first edition. The remaining 18 chapters contain many new topics, which were considered very advanced when the first edition was published. Knowledge of calculus, unit operations, and complex numbers is presumed on the part of the student. In certain later chapters, more advanced mathematical preparation is useful. Some examples would include partial differential equations in Chap. 21, linear algebra in Chaps. 28 through 30, and Fourier series in Chap. 33. Analog computation and pneumatic controllers in the first edition have been replaced by digital computation and microprocessor-based controllers in Chaps. 34 and 35. The student should be assigned material from these chapters at the appropriate time in the development of the fundamentals. For example, the transient response for a system containing a transport lag can be obtained easily only with the use of computer simulation of the transport lag. Some of the software now available for solving control problems should be available to the student; such software is described in Chap. 34. To understand the operation of modern microprocessor-based controllers, the student should have hands-on experience with these instruments in a laboratory. Chapter 1 is intended to meet one of the problems consistently faced in presenting this material to chemical engineering students, that is, one of perspective. The methods of analysis used in the control area are so different from the previous experiences of students that the material comes to be regarded as a sequence of special mathematical techniques, rather than an integrated design approach to a class of real and practically significant industrial problems. Therefore, this chapter presents an overall, albeit superficial, look at a simple control system design problem. The body of the text covers the following topics: Laplace transforms, Chaps 2. to 4; transfer functions and responses of open-loop systems, Chaps. 5 to 8; basic techniques of closed-loop control, Chaps. 9 to 13; stability, Chap. 14; root locus methods, Chap. 15; frequency response methods and design, Chaps. 16 and 17; advanced control strategies (cascade, feedforward, Smith predictor, internal model control), Chap. 18; controller tuning and xvii cou9789x_fm_i-xx.indd xvii 8/25/08 2:48:43 PM Confirming Pages xviii HISTORY OF PROCESS SYSTEMS ANALYSIS AND CONTROL (FROM THE SECOND EDITION PREFACE) process identification, Chap. 19; control valves, Chap. 20; advanced process dynamics, Chap. 21; sampled-data control, Chaps. 22 to 27; state-space methods and multivariable control, Chaps. 28 to 30; nonlinear control, Chaps. 31 to 33; digital computer simulation, Chap. 34; microprocessor-based controllers, Chap. 35. It has been my experience that the book covers sufficient material for a onesemester (15-week) undergraduate course and an elective undergraduate course or part of a graduate course. In a lecture course meeting 3 hours per week during a 10-week term, I have covered the following chapters: 1 to 10, 12 to 14, 16, 17, 20, 34, and 35. After the first 14 chapters, the instructor may select the remaining chapters to fit a course of particular duration and scope. The chapters on the more advanced topics are written in a logical order; however, some can be skipped without creating a gap in understanding. I gratefully acknowledge the support and encouragement of the Drexel University Department of Chemical Engineering for fostering the evolution of this text in its curriculum and for providing clerical staff and supplies for several editions of class notes. I want to acknowledge Dr. Lowell B. Koppel’s important contribution as coauthor of the first edition of this book. I also want to thank my colleague Dr. Rajakannu Mutharasan for his most helpful discussions and suggestions and for his sharing of some of the new problems. For her assistance in typing, I want to thank Dorothy Porter. Helpful suggestions were also provided by Drexel students, in particular Russell Anderson, Joseph Hahn, and Barbara Hayden. I also want to thank my wife Effie for helping me check the page proofs by reading to me the manuscript, the subject matter of which is far removed from her specialty of Greek and Latin. McGraw-Hill and I would like to thank Ali Cinar, Illinois Institute of Technology; Joshua S. Dranoff, Northwestern University; H. R. Heichelheim, Texas Tech University; and James H. McMicking, Wayne State University, for their many helpful comments and suggestions in reviewing this second edition. Dr. Donald R. Coughanowr cou9789x_fm_i-xx.indd xviii 8/25/08 2:48:43 PM Confirming Pages ABOUT THE AUTHORS Steven E. LeBlanc is Associate Dean for Academic Affairs and professor of chemical engineering at the University of Toledo. He received a B.S. degree in chemical engineering from the University of Toledo and his M.S. and Ph.D. in chemical engineering from the University of Michigan. He joined the faculty at the University of Toledo in 1980. He served as the department chair for the Department of Chemical and Environmental Engineering from 1993 to 2003, when he became an Associate Dean in the College of Engineering. Dr. LeBlanc’s industrial experience includes power plant process system design and review for Toledo Edison Company (now a division of First Energy). He has taught the Process Dynamics and Control course numerous times, and was responsible for a major revamp of laboratory activities associated with the course. He is a member of the American Institute of Chemical Engineers (AIChE) and the American Society for Engineering Education (ASEE). He has served as an ABET chemical engineering program evaluator for AIChE since 1998. He chaired the national ASEE Chemical Engineering Education Division and cochaired the 2007 ASEE Chemical Engineering Summer School for Faculty. He coauthored and judged the 1992 AIChE Senior Design Project competition. He is also coauthor of a textbook on Strategies for Creative Problem Solving with H. Scott Fogler of the University of Michigan. Donald R. Coughanowr is Emeritus Professor of Chemical Engineering at Drexel University. In 1991 he wrote the second edition of Process Systems Analysis and Control which contained many changes and new topics in order to bring the book up to date at the time of publication. He received a Ph.D. in chemical engineering from the University of Illinois in 1956, an M.S. degree in chemical engineering from the University of Pennsylvania in 1951, and a B.S. degree in chemical engineering from the Rose-Hulman Institute of Technology in 1949. He joined the faculty at Drexel University in 1967 as department head, a position he held until 1988. Before going to Drexel, he was a faculty member of the School of Chemical Engineering at Purdue University for 11 years. At Drexel and Purdue he taught a wide variety of courses, which include material and energy balances, thermodynamics, unit operations, transport phenomena, petroleum refinery engineering, environmental engineering, chemical engineering laboratory, applied mathematics, and process dynamics and control. At Purdue, xix cou9789x_fm_i-xx.indd xix 8/25/08 2:48:43 PM Confirming Pages xx ABOUT THE AUTHOR he developed a new course and laboratory in process control and collaborated with Dr. Lowell B. Koppel on the writing of the first edition of Process Systems Analysis and Control. His research interests included environmental engineering, diffusion with chemical reaction, and process dynamics and control. Much of his research in control emphasized the development and evaluation of new control algorithms for the processes that cannot be controlled easily by conventional control; some of the areas investigated were time-optimal control, adaptive pH control, direct digital control, and batch control of fermentors. He reported on his research in numerous publications and received support for research projects from the National Science Foundation and industry. He spent sabbatical leaves teaching and writing at Case-Western Reserve University, the Swiss Federal Institute, the University of Canterbury, the University of New South Wales, the University of Queensland, and Lehigh University. Dr. Coughanowr’s industrial experience included process design and pilot plant at Standard Oil Co. (Indiana) and summer employment at Electronic Associates and Dow Chemical Company. He is a member of the American Institute of Chemical Engineers. He has served the AIChE by participating in accreditation visits to departments of chemical engineering for ABET and by chairing sessions of the Department Heads Forum at the annual meetings of AIChE. cou9789x_fm_i-xx.indd xx 8/25/08 2:48:43 PM Confirming Pages CHAPTER 1 INTRODUCTORY CONCEPTS I n this chapter we examine the concept of chemical process control and introduce several examples to illustrate the necessity for process modeling as we begin our study of process dynamics and control. 1.1 WHY PROCESS CONTROL? As competition becomes stiffer in the chemical marketplace and processes become more complicated to operate, it is advantageous to make use of some form of automatic control. Automatic control of a process offers many advantages, including • • • • • Enhanced process safety Satisfying environmental constraints Meeting ever-stricter product quality specifications More efficient use of raw materials and energy Increased profitability Considering all the benefits that can be realized through process control, it is well worth the time and effort required to become familiar with the concepts and practices used in the field. 1.2 CONTROL SYSTEMS Control systems are used to maintain process conditions at their desired values by manipulating certain process variables to adjust the variables of interest. A common example of a control system from everyday life is the cruise control on an automobile. The purpose of a cruise control is to maintain the speed of the vehicle (the controlled variable) at the desired value (the set point) despite variations in terrain, hills, etc. 1 cou9789x_ch01_001-008.indd 1 8/21/08 5:14:28 PM Confirming Pages 2 CHAPTER 1 INTRODUCTORY CONCEPTS (disturbances) by adjusting the throttle, or the fuel flow to the engine (the manipulated variable). Another common example is the home hot water heater. The control system on the hot water heater attempts to maintain the temperature in the tank at the desired value by manipulating the fuel flow to the burner (for a gas heater) or the electrical input to the heater in the face of disturbances such as the varying demand on the heater early in the morning, as it is called upon to provide water for the daily showers. A third example is the home thermostat. This control system is designed to maintain the temperature in the home at a comfortable value by manipulating the fuel flow or electrical input to the furnace. The furnace control system must deal with a variety of disturbances to maintain temperature in the house, such as heat losses, doors being opened and hopefully closed, and leaky inefficient windows. The furnace must also be able to respond to a request to raise the desired temperature if necessary. For example, we might desire to raise the temperature by 5⬚, and we’d like the system to respond smoothly and efficiently. From these examples, we can deduce that there are several common attributes of control systems: • The ablity to maintain the process variable at its desired value in spite of disturbances that might be experienced (this is termed disturbance rejection) • The ability to move the process variable from one setting to a new desired setting (this is termed set point tracking) Conceptually we can view the control systems we’ve discussed in the following general manner (Fig. 1–1). The controller compares the measurement signal of the controlled variable to the set point (the desired value of the controlled variable). The difference between the two values is called the error. Error ⫽ (Set point value ) ⫺ ( Measurement sig nal of controlled variable) Depending upon the magnitude and sign of the error, the controller takes appropriate action by sending a signal to the final control element, which provides an input to the process to return the controlled variable to the set point. The concept of using information Disturbances Desired Value (Set Point) Control Signal Controller Measurement Signal Final Control Element Manipulated Variable Controlled Variable Process Measurement Device FIGURE 1–1 Generalized process control system. cou9789x_ch01_001-008.indd 2 8/21/08 5:14:28 PM Confirming Pages CHAPTER 1 INTRODUCTORY CONCEPTS 3 about the deviation of the system from its desired state to control the system is called feedback control. Information about the state of the system is “fed back” to a controller, which utilizes this information to change the system in some way. The type of control system shown in Fig. 1–1 is termed a closed-loop feedback control system. Closed-loop refers to the fact that the controller automatically acts to return the controlled variable to its desired value. In contrast, an open-loop system would have the measurement signal disconnected from the controller, and the controller output would have to be manually adjusted to change the value of the controlled variable. An open-loop system is sometimes said to be in manual mode as opposed to automatic mode (closed-loop). Negative feedback is the most common type of signal feedback. Negative refers to the fact that the error signal is computed from the difference between the set point and the measured signal. The negative value of the measured signal is “fed back” to the controller and added to the set point to compute the error. Example 1.1. Hot water tank control system. As a specific example, let us consider a hot water heater for a home (Fig. 1–2) and examine its control system, using the same type of diagram (Fig. 1–3). The desired hot water temperature is selected by the homeowner, and typically it is in the neighborhood of 120 to 140⬚F. Let us assume that the set point is 130⬚F. The thermocouple measures the temperature of the water in the tank and sends a signal to the thermostat indicating the temperature. The thermostat (controller) determines the error as Error ⫽ Tset point ⫺ Tmeasured Hot water outlet Cold water inlet TPR valve Anode Thermostat Dip tube If the error is positive (⬎ 0), the measured temperature is lower than desired and the thermostat opens the fuel valve to the burner which adds heat to the tank. If the error is zero or negative (ⱕ 0), the thermostat closes the fuel valve and no heat is added to the tank. Disturbances to the system, which decrease the temperature of the water in the tank, include ambient heat losses and hot water demand by the household which is replaced with a cold water feed. Drain valve Types of Controllers FIGURE 1–2 Physical drawing of a hot water heater. cou9789x_ch01_001-008.indd 3 The thermostat on the hot water heater is called an “on/off ” type of controller. Depending on the value of the error signal, the output from the controller is 8/21/08 5:14:28 PM Confirming Pages 4 CHAPTER 1 INTRODUCTORY CONCEPTS Disturbances (heat losses, hot water demand) Desired Hot Water Temperature Manipulated Control Signal Fuel Flow to Hot to Burner Water Tank Fuel Valve Fuel Thermostat Valve Heating Process Actual Hot Water Temperature Indicated Hot Water Temperature Thermocouple FIGURE 1–3 Block diagram of a hot water heater control system. either “full on” or “full off ” and the fuel valve is full open or full closed; there are no intermediate values of the output. Many other types of controllers that we will study can modulate their output based on the magnitude of the error signal, how long the error signal has persisted, and even how rapidly the error appears to be changing. Clearly, the larger the error, the less we are satisfied with the present state of affairs and vice versa. In fact, we are completely satisfied only when the error is exactly zero. Based on these considerations, it is natural to suggest that the controller should change the heat input by an amount proportional to the error. This is called proportional control. In effect, the controller is instructed to maintain the heat input at the steadystate design value as long as the error is zero. If the tank temperature deviates from the set point, causing an error, the controller is to use the magnitude of the error to change the heat input proportionally. We shall reserve the right to vary the proportionality constant to suit our needs. This degree of freedom forms a part of our instructions to the controller. As we will see shortly during the course of our studies, the larger we make the proportionality constant for the proportional controller (called the controller gain), the smaller the steady-state error will become. We will also see that it is impossible to completely eliminate the error through the use of a proportional controller. For example, if the set point is 130⬚F and a disturbance occurs that drops the temperature to 120⬚F, if we use only a proportional controller, then we will never be able to get the tank temperature to exactly 130⬚F. Once the sytem stabilizes again, the temperature will not be exactly 130⬚F, but perhaps 127⬚F or 133⬚F. There will always be some residual steadystate error (called offset). For a home water heater, this is probably good enough; the exact temperature is not that critical. In an industrial process, this may not be adequate, and we have to resort to a bit more complicated controller to drive the error to zero. Considerable improvement may be obtained over proportional control by adding integral control. The controller is now instructed to change the heat input by an additional amount proportional to the time integral of the error. This type of control system has two adjustable parameters: a multiplier for the error and a multiplier for the integral of the error. If this type of controller is used, the steady-state error will be zero. From this standpoint, the response is clearly superior to that of the system with proportional control only. One price we pay for this improvement is the tendency for the system to be more cou9789x_ch01_001-008.indd 4 8/21/08 5:14:28 PM Confirming Pages CHAPTER 1 INTRODUCTORY CONCEPTS 5 oscillatory. The system will tend to overshoot its final steady-state value before slowly settling out at the desired set point. So what is the best control system to use for a particular application? This and related questions will be addressed in subsequent chapters. Some Further Complications At this point, it would appear that the problem has been solved in some sense. A little further probing will shatter this illusion. It has been assumed that the controller receives instantaneous information about the tank temperature. From a physical standpoint, some measuring device such as a thermocouple will be required to measure this temperature. The temperature of a thermocouple inserted in the tank may or may not be the same as the temperature of the fluid in the tank. This can be demonstrated by placing a mercury thermometer in a beaker of hot water. The thermometer does not instantaneously rise to the water temperature. Rather, it takes a bit of time to respond. Since the controller will receive measured values of the temperature, rather than the actual values, it will be acting upon the apparent error, rather than the actual error. The effect of the thermocouple delay in transmission of the temperature to the controller is primarily to make the response of the system somewhat more oscillatory than if the response were instantaneous. If we increase the controller gain (the proportionality constants), the tank temperature will eventually oscillate with increasing amplitude and will continue to do so until the physical limitations of the heating system are reached. In this case, the control system has actually caused a deterioration in performance, and this type of reponse is referred to as an unstable response. This problem of stability of response will be a major concern for obvious reasons. At present, it is sufficient to note that extreme care must be exercised in specifying control systems. In the case considered, the proportional and integral controllers described above will perform satisfactorily if the gain is kept lower than some particular value. However, it is not difficult to construct examples of systems for which the addition of any amount of integral control will cause an unstable response. Since integral control usually has the desirable feature of eliminating steady-state error, it is extremely important that we develop means for predicting the occurrence of unstable response in the design of any control system. Block Diagram A good overall picture of the relationships among variables in the heated-tank control system may be obtained by preparing a block diagram as shown in Fig. 1–1. It indicates the flow of information around the control system and the function of each part of the system. Much more will be said about block diagrams later, but the reader can undoubtedly form a good intuitive notion about them by comparing Fig. 1–1 with the physical description of the process. Particularly significant is the fact that each component of the system is represented by a block, with little regard for the actual physical characteristics of the represented component (e.g., the tank or controller). The major interest is in (1) the relationship between the signals entering and leaving the block and (2) the manner in which information flows around the system. cou9789x_ch01_001-008.indd 5 8/21/08 5:14:29 PM Confirming Pages 6 CHAPTER 1 INTRODUCTORY CONCEPTS SUMMARY We have taken an overall look at a typical control problem and some of its ramifications. At present, the reader has been asked to accept the results on faith and to concentrate on obtaining a physical understanding of the transient behavior of the heated tank. In the forthcoming chapters we develop tools for determining the response of such systems. As this new material is presented, the reader may find it helpful to refer to this chapter to place the material in proper perspective to the overall control problem. PROBLEMS 1.1. Draw a block diagram for the control system generated when a human being steers an automobile. 1.2. Draw a block diagram for the control system generated when a human being shoots a bow and arrow. 1.3. Draw a block diagram for an automobile cruise control system. 1.4. Draw a block diagram for the control system that maintains the water level in a toilet tank. 1.5. Draw a block diagram for a security lighting system that activates at dusk and turns off at dawn. 1.6. Draw a block diagram for the control system for a home oven. cou9789x_ch01_001-008.indd 6 8/21/08 5:14:29 PM Confirming Pages CHAPTER 1 CAPSULE SUMMARY DEFINITIONS Block diagram—Diagram that indicates the flow of information around the control system and the function of each part of the system. Closed loop—In closed loop, the measured value of the controlled variable is fed back to the controller. Controlled variable—The process variable that we want to maintain at a particular value. Controller—A device that outputs a signal to the process based on the magnitude of the error signal. A proportional controller outputs a signal proportional to the error. Disturbance rejection—One goal of a control system, which is to enable the system to “reject” the effect of disturbance changes changes and maintain the controlled variable at the set point. Disturbances—Any process variables that can cause the controlled variable to change. In general, disturbances are variables that we have no control over. Error—The difference between the values of the set point and the measured variable. Manipulated variable—Process variable that is adjusted to bring the controlled variable back to the set point. Negative feedback—In negative feedback, the error is the difference between the set point and the measured variable (this is usually the desired configuration). Offset—The steady-state value of the error. Open loop—In open loop, the measured value of the controlled variable is not fed back to the controller. Positive feedback—In positive feedback, the measured temperature is added to the set point. (This is usually an undesirable situation and frequently leads to instability.) Set point—The desired value of the controlled variable. Set point tracking—One goal of a control system, which is to force the system to follow or “track” requested set point changes. 7 cou9789x_ch01_001-008.indd 7 8/21/08 5:14:29 PM cou9789x_ch01_001-008.indd 8 8/21/08 5:14:29 PM Confirming Pages PART I MODELING FOR PROCESS DYNAMICS cou9789x_ch02_009-031.indd 9 8/14/08 1:55:08 PM Confirming Pages cou9789x_ch02_009-031.indd 10 8/14/08 1:55:09 PM Confirming Pages CHAPTER 2 MODELING TOOLS FOR PROCESS DYNAMICS U nderstanding process dynamics (how process variables change with time) will be very important to our studies of process control. In the examples in Chap.1, we saw some of the implications of process dynamics and their relationship to process control. In this chapter we explore process dynamics further and review some mathematical tools for solving the resulting process models. 2.1 PROCESS DYNAMICS—A CHEMICAL MIXING SCENARIO Consider the following chemical mixing example (Fig. 2–1). Two process streams are mixed to produce one of the feeds for our chemical reactor. After mixing, the blended stream is fed to a heating vessel before being sent to the reactor. Stream 1 Ca1=1 g/L v1=10 L/min Stream 3 Ca3=? g/L Mixing tee v3=30 L/min V=150 L Stream 2 Ca2=4 g/L v2=20 L/min Ca To reactor Heater FIGURE 2–1 Chemical mixing process flow diagram. 11 cou9789x_ch02_009-031.indd 11 8/14/08 1:55:09 PM Confirming Pages 12 PART 1 MODELING FOR PROCESS DYNAMICS The process is running along at steady state. The concentration of A in stream 1 is 1 g/L and in stream 2 is 4 g/L. At 3:00 P.M. the shift changes at the plant. The new operator on our unit misreads the flowmeters for the process and switches the flow rates of the two streams. Stream 1 is switched to 20 L/min, and stream 2 is switched to 10 L/min. At 3:30 P.M. the shift supervisor hurries to the control room to determine the source of the problem now being experiencing with the reactor. Use your knowledge of chemical engineering to determine what has happened to the exit concentration from the heating vessel over the first half-hour of the shift. We can model the mixing tee and the blending tank using an unsteady-state mass balance to predict the behavior of this part of the process since the shift change and the unfortunate error by the new operator. A balance on component A around the mixing tee before and after the change will yield information on how the feed concentration to the heating vessel changes. The component A balance around the mixing tee is Rate of A into Rate of A leaving Rate of A into mixing tee mixing tee mix ing tee in stream 1 (g/min ) in stream 3 (g/min) in strea m 2 (g/min) v1Ca1 v2Ca 2 v3Ca 3 Before the change, we can calculate the original steady-state concentration into the heating vessel: L g L g L g 4 30 10 1 20 Ca 3 min L min L min L g Ca 3 3 L Since the process has been running along at steady state for a long time under these conditions, the concentration in the heating vessel is also the same. After the change, the new feed concentration to the heating vessel is L g L g L g 4 30 1 10 20 Ca 3 min L min L min L g Ca 3 2 L So the net result of the operator error is to decrease the feed concentration to the heating vessel from 3 to 2 g/L. After we analyze the process for a moment, it is apparent that the exit concentration from the heating vessel will eventually also fall from 3 to 2 g/L if the process is left in its current configuration for a long enough time. So, what we know of the situation is shown in Fig. 2–2. To analyze how the exit from the heating vessel (the feed to the reactor) varies with time, we must perform an unsteady mass balance on component A around the heating vessel. cou9789x_ch02_009-031.indd 12 8/14/08 1:55:10 PM Confirming Pages CHAPTER 2 MODELING TOOLS FOR PROCESS DYNAMICS 13 Stream 3 V3=30 L/min Concentration 3 g/L of A in stream 3 V=150 L 2 g/L Ca 3:00 P.M. Time Heater Concentration of A in exit stream To reactor 3 g/L Transient period 2 g/L 3:00 P.M. Time FIGURE 2–2 Chemical mixing process flow diagram showing operator-induced transient. Rate of A into Rate of A leaving Accumulatio n of A in heating vessel (g/min) heating vessel (g/min) heating vessel (g/min) in out v3Ca 3 accum u lation d v3Ca (VCa ) dt Note that the volumetric flow rate v is constant into and out of the heating vessel at v3. Thus the volume of fluid in the tank V is constant. We can rearrange this equation to the following form: V dCa Ca Ca 3 (2.1) v3 dt The coefficient of the derivative term is the residence time of the heating vessel , which in this process is 5 min. Substituting the numbers for this scenario yields 5 dCa Ca 2 dt Ca ( 0 ) 3 g L We can rearrange and solve this equation as follows. Ca ∫ 3 dCa 2 Ca t 1 ∫ 5 dt 0 ln (2 Ca ) C3 a t 5 t 2 Ca ln 23 5 Ca 2 et / 5 cou9789x_ch02_009-031.indd 13 (2.2) 8/14/08 1:55:10 PM Confirming Pages 14 PART 1 MODELING FOR PROCESS DYNAMICS A plot of the exit concentration from the heating vessel is shown in Fig. 2–3. As expected, the concentration starts at the original steady-state concentration of 3 g/L and exponentially decreases to 2 g/L. Modeling the mixing process enables us to determine the concentration of component A in the stream being fed to the reactor. Being able to determine or predict the dynamic behavior of a process is crucial to being able to design a control system for it. As another modeling example, consider the energy balance for the mixing process described above. Prior to 3 P.M. the process conditions are depicted as in Fig. 2–4. Stream 1 (at 25C) mixes with stream 2 (at 55C), producing stream 3, the feed to the heating vessel. The heater adds energy to the vessel to bring the outlet stream to 80C. Before we look at the effect of the disturbance caused by the operator, it is necessary to determine the steady-state process conditions prior to the upset. An energy balance around the mixing tee will enable us to calculate the steady-state feed temperature to the heating vessel T3. Rate of enthalpy into mixing tee with stream 1 Rate of enthalpy into mixing tee with stream 2 Rate of enthalp y leaving mixing tee with stream 3 r v1C p (T1 Tref ) r v2C p (T2 Tref ) r v3C p (T3 Tref ) We have assumed that the stream density (g/L) and specific heat Cp [cal/(g . C)] remain constant, independent of the concentration of component A in the stream. We Heating Vessel Exit Concentration 3 2.5 Ca (g/L) 2 1.5 1 0.5 0 0 5 10 15 Time (min) 20 25 30 FIGURE 2–3 Outlet concentration from the heating vessel as a function of time. cou9789x_ch02_009-031.indd 14 8/14/08 1:55:11 PM Confirming Pages CHAPTER 2 Stream 1 Ca1=1 g/L ν1=10 L/min T1=25°C Mixing tee MODELING TOOLS FOR PROCESS DYNAMICS 15 Stream 3 Ca3=3 g/L ν3=30 L/min T3=?°C V=150 L Ca T=80°C Stream 2 Ca2=4 g/L ν2=20 L/min T2=55°C To reactor Heater FIGURE 2–4 Chemical mixing process flow diagram showing initial temperatures. have also defined a reference temperature Tref for the enthalpy calculation. The energy balance can be simplified to v1T1 v2T2 v3T3 Note that we have made use of the relation (v1 v2)Tref v3Tref to eliminate some terms. Solving for T3 yields T3 v1T1 v2T2 (10)(25) (20)(55) 45 °C v3 30 So, the steady-state inlet temperature to the heating vessel is 45C. We can now determine the steady-state heat input required from the heater by performing a steady-state energy balance around the heating vessel. Rate of enthalpy into heating vessel with streeam 3 Rate of Rate of enthalpy into enthalpy leaving hea ting vessel heating vessel from heater r v3C p (T3 Tref ) Q r v3C p (T Tref ) Solving for the heater input gives g L cal Q v3C p (T T3 ) 1000 30 1 (80°C 45°C) L min g ⋅ °C 1. 05 106 cal 73 . 2 kW min The energy balance for the original steady-state case is summarized in Fig. 2–5. The inlet temperature to the heating vessel after the 3:00 P.M. disturbance can be determined from the steady-state energy balance around the mixing tee using the new flow rates (Fig. 2–6). cou9789x_ch02_009-031.indd 15 8/14/08 1:55:12 PM Confirming Pages 16 PART 1 MODELING FOR PROCESS DYNAMICS Stream 3 ν3=30 L/min T3=45°C V=150 L T=80°C To reactor Heater input 73.2 kW FIGURE 2–5 Chemical mixing process flow diagram summarizing initial temperatures. T3 Temperature of stream 3 v1T1 v2T2 (20)(25) (10)(55) 35ºC v3 30 45°C 35°C 3:00 P.M. Time FIGURE 2–6 Inlet temperature disturbance. To determine the effect of this inlet temperature disturbance on the feed to the reactor, an unsteady-state energy balance on the heating vessel is required. We can logically predict that if the heat input stays constant, as well as the process flow rate, when the inlet temperature falls by 10C, the outlet temperature from the heating vessel will correspondingly decrease by 10 to 70C. The energy balance on the heating vessel is Rate of enthalpy Rate of enthalpy Rate of energy into heating leaving h eating input to vessel vessel heating vessel with stree am 3 with exit stream from heater r v3C p (T3 Tref ) r v3C p (T Tref ) Q Rate of accumulation of enthalpy in heating vessel d ( r VC p (T Tref ) dt Simplifying yields 1 V dT T T3 Q v3 dt r v3C p (2.3) t Substituting values for the scenario we are considering gives cou9789x_ch02_009-031.indd 16 8/14/08 1:55:12 PM Confirming Pages CHAPTER 2 (5 min) MODELING TOOLS FOR PROCESS DYNAMICS 1 dT T 35ºC dt g L cal 1000 L 30 min 1 g ⋅ º C 17 6 cal 1. 05 10 min 7 0 º C T (0) 80 º C Separating and integrating, we have 5 dT 70 T dt T dT ∫ 70 T 80 t ∫ 0 dt 5 70 T t ln 10 5 T 70 10et / 5 (2.4) A plot of the outlet temperature from the heating vessel as a function of time is shown in Fig. 2–7. Outlet Temperature from Heating Vessel as a Function of Time 80 78 76 Temperature (C) 74 72 70 68 66 64 62 60 0 5 10 15 Time (min) 20 25 30 FIGURE 2–7 Outlet temperature transient due to the disturbance. cou9789x_ch02_009-031.indd 17 8/14/08 1:55:13 PM Confirming Pages 18 PART 1 MODELING FOR PROCESS DYNAMICS Notice the shape of the temperature response is the same as the shape of the concentration response that we saw previously. By appropriate modeling of the process, we can predict how the system will respond to changes in the operating conditions. Our ability to model the process will be extremely valuable as we design controllers to automatically control the process variables at their desired settings. 2.2 MATHEMATICAL TOOLS FOR MODELING As we just saw in our analysis of the chemical mixer, the unsteady-state material and energy balance models that we wrote required us to solve differential equations to obtain the concentration and temperature versus time behavior for the process. This will be a common occurrence for us as we continue our studies of process dynamics and control. It would be beneficial to review some additional tools available to us for solving our process models. In Sec. 2.1, we solved the equations by separation and integration. A couple of other useful tools for solving such models are Laplace transforms and MATLAB/Simulink. In the next several sections, we will review the use of these additional tools for solving our model differential equations. Definition of the Laplace Transform The Laplace transform of a function f(t) is defined to be F(s) according to the equation F (s ) ∞ ∫0 f (t )est dt (2.5) We often abbreviate this to F (s) L{ f (t )} where the operator L is defined by Eq. (2.5). Example 2.1. Find the Laplace transform of the function f (t ) 1 According to Eq. (2.5), F (s ) ∞ ∫0 (1)est dt est s t ∞ t 0 1 s Thus, L{1} 1 s There are several facts worth noting at this point: 1. cou9789x_ch02_009-031.indd 18 The Laplace transform F(s) contains no information about the behavior of f(t) for t, 0. This is not a limitation for control system study because t will represent the 8/14/08 1:55:13 PM Confirming Pages CHAPTER 2 2. 3. MODELING TOOLS FOR PROCESS DYNAMICS 19 time variable and we will be interested in the behavior of systems only for positive time. In fact, the variables and systems are usually defined so that f(t) 0 for t , 0. The time we designate as t 0 is arbitrary. We shall generally define t 0 as the time when the process is disturbed from steady state (i.e., when an input is changed). Our usual starting point will be a steady-state system or process, and we will be interested in examining what happens when the system is disturbed. This will become clearer as we study specific examples. Since the Laplace transform is defined in Eq. (2.5) by an improper integral, it will not exist for every function f(t). A rigorous definition of the class of functions possessing Laplace transforms is beyond the scope of this book, but readers will note that every function of interest to us does satisfy the requirements for possession of a transform [see Churchill (1972)]. The Laplace transform is linear. In mathematical notation, this means L{af1 (t ) bf2 (t )} aL{ f1 (t )} bL{ f2 (t )} where a and b are constants and f1 and f2 are two functions of t. Proof. Using the definition, we have L{af1 (t ) bf2 (t )} ∞ st ∫0 [af1 (t ) bf2 (t )]e a∫ ∞ 0 f1 (t )est dt b ∫ ∞ 0 dt f2 (t )est dt aL{ f1 (t )} b L{ f2 (t )} 4. The Laplace transform operator transforms a function of the variable t to a function of the variable s. The t variable is eliminated by the integration. Transforms of Simple Functions We now proceed to derive the transforms of some simple and useful functions. We shall see these common functions repeatedly during our future studies. 1. The step function is f (t ) { 0 t 0 1 t 0 This important function is known as the unit-step function and will henceforth be denoted by u(t). From Example 2.1, it is clear that L{u(t )} 1 s As expected, the behavior of the function for t 0 has no effect on its Laplace transform. Note that as a consequence of linearity, the transform of any constant A, that is, f(t) Au(t), is just F(s) A/s. Notice in the chemical mixing example cou9789x_ch02_009-031.indd 19 8/14/08 1:55:14 PM Confirming Pages 20 PART 1 MODELING FOR PROCESS DYNAMICS 2. that we just discussed that the inlet concentration and temperature are described by a step function initiated at time zero (3 P.M. in the example). The exponential function is { f (t ) 0 t0 e− at t0 u(t )eat where u(t) is the unit-step function. Again proceeding according to definition, we have { L u(t )eat ∞ } ∫0 e(s a)t dt s 1 a e(s a)t 0 ∞ 1 sa provided that s a 0, that is, s a. In this case, the convergence of the integral depends on a suitable choice of s. In case s is a complex number, it may be shown that this condition becomes Re(s) a 3. For problems of interest to us it will always be possible to choose s so that these conditions are satisfied, and the reader uninterested in mathematical niceties can ignore this point. The ramp function is f (t ) { 0t L{tu(t )} t<0 tu(t ) t>0 ∞ st te dt 0 ∫ Integration by parts yields ∞ 1 1 t L{tu(t )} est 2 2 s s 0 s 4. The sine function is f (t ) { sin0 kt L{u(t )sin kt} t<0 t>0 ∞ u(t )sin kt st ∫0 sin kt e dt Integrating by parts, we have ∞ est L{u(t )sin kt} 2 (s sin kt k cos kt ) s k2 0 cou9789x_ch02_009-031.indd 20 k s k2 2 8/14/08 1:55:15 PM Confirming Pages CHAPTER 2 MODELING TOOLS FOR PROCESS DYNAMICS 21 TABLE 2.1 Function Graph Transform 1 u(t) 1 s tu(t) 1 s2 tnu(t) n! s n 1 1 at e u(t) tneatu(t) sin kt u(t) cou9789x_ch02_009-031.indd 21 1 sa n! ( s a )n1 k s2 k 2 8/14/08 1:55:16 PM Confirming Pages 22 PART 1 MODELING FOR PROCESS DYNAMICS TABLE 2.1 (Continued) Function Graph Transform cos kt u(t) s s2 k 2 sinh kt u(t) k s2 k 2 cosh kt u(t) s s2 k 2 1 eat sin kt u(t) k (s a)2 k 2 eat cos kt u(t) sa (s a)2 k 2 Area = 1 (t), unit impulse cou9789x_ch02_009-031.indd 22 1 8/14/08 1:55:18 PM Confirming Pages CHAPTER 2 MODELING TOOLS FOR PROCESS DYNAMICS 23 In a like manner, the transforms of other simple functions may be derived. Table 2.1 is a summary of transforms that will be of use to us. Those which have not been derived here can be easily established by direct integration, except for the transform of (t), which will be discussed in detail in the Appendix at the end of Chap. 3. Using MATLAB for Symbolic Processing—Laplace Transforms MATLAB is capable of symbolic processing. To prepare MATLAB for symbolic operations, some variable names will be declared symbolic (rather than numeric) using the syms command. syms a x y z t k s We can also define u as the Heaviside function (the unit step): u=sym('Heaviside(t)') u= Heaviside(t) Now we can determine the transform of the simple functions we have just discussed: The step function: laplace(u) ans= 1/s The exponential function: laplace(exp(–a*t)) ans= 1/(s+a) The ramp function: laplace(t) ans 1/s^2 The sine function: laplace(sin(k*t)) ans= k/(s^2+k^2) Transforms of Derivatives At this point, the reader may wonder what has been gained by introduction of the Laplace transform. The transform merely changes a function of t into a function of s. The functions of s look no simpler than those of t and, as in the case of A → A/s, may actually be more complex. In the next few paragraphs, the motivation will become clear. It will be shown that the Laplace transform has the remarkable property of transforming cou9789x_ch02_009-031.indd 23 8/14/08 1:55:19 PM Confirming Pages 24 PART 1 MODELING FOR PROCESS DYNAMICS the operation of differentiation with respect to t to that of multiplication by s. Thus, we claim that df (t ) (2.6) L sF (s) f (0) dt { } where F (s) L{ f (t )} and f(0) is f(t) evaluated at t 0. If f(t) is discontinuous at t 0, f(0) should be evaluated at t 0, that is, just to the right of the origin. Since we will seldom want to differentiate functions that are discontinuous at the origin, this detail is not of great importance. However, the reader is cautioned to watch carefully for situations in which such discontinuities occur. Proof L { } df (t ) dt ∞ ∫0 df st e dt dt To integrate this by parts, let u est dv df dt dt Then du sest dt v f (t ) Since ∫ u dv uv ∫ v du we have ∞ df st ∫0 dt e dt f (t )est ∞ 0 ∞ s ∫ f (t )est dt f (0) sF (s) 0 The salient feature of this transformation is that whereas the function of t was to be differentiated with respect to t, the corresponding function of s is merely multiplied by s. We shall find this feature to be extremely useful in the solution of differential equations. To find the transform of the second derivative we make use of the transform of the first derivative twice, as follows: d 2 f L 2 dt d df df df (t ) L sL dt d t dt t 0 dt s[sF (s) f (0)] f ′(0) s 2 F (s) sf (0) f ′(0) cou9789x_ch02_009-031.indd 24 8/14/08 1:55:20 PM Confirming Pages CHAPTER 2 MODELING TOOLS FOR PROCESS DYNAMICS 25 where we have abbreviated df (t ) f ′(0) dt t 0 In a similar manner, the reader can easily establish by induction that repeated application of Eq. (2.6) leads to d n f L n dt n n 1 f (0) s n 2 f (1) (0) ··· sf ( n 2) (0) f ( n 1) (0) s F (s ) s where f i(0) indicates the ith derivative of f(t) with respect to t, evaluated for t 0. Thus, the Laplace transform may be seen to change the operation of differentiation of the function to that of multiplication of the transform by s, the number of multiplications corresponding to the number of differentiations. In addition, some polynomial terms involving the initial values of f(t) and its first n 1 derivatives are involved. In later applications we usually define our variables so that these polynomial terms will vanish. Hence, they are of secondary concern here. Example 2.2. Find the Laplace transform of the function x(t) that satisfies the differential equation and initial conditions d3x d2x dx 5 2x 2 4 dt dt 3 dt 2 x (0) dx (0) d 2 x (0) 0 dt dt 2 It is permissible mathematically to take the Laplace transforms of both sides of a differential equation and equate them, since equality of functions implies equality of their transforms. Doing this, we obtain s 3 x (s) s 2 x (0) sx ′(0) x ′′(0) 4[s 2 x (s) sx (0) x ′(0)] 2 5 [sx (s) x (0)] 2 x (s) s where x(s) L{x(t)}. Use has been made of the linearity property and of the fact that only positive values of t are of interest. Inserting the initial conditions and solving for x(s), we have x (s) ( 2 3 2 s s 4 s 5s 2 ) (2.7) This is the required answer, the Laplace transform of x(t). cou9789x_ch02_009-031.indd 25 8/14/08 1:55:21 PM Confirming Pages 26 PART 1 MODELING FOR PROCESS DYNAMICS 2.3 SOLUTION OF ORDINARY DIFFERENTIAL EQUATIONS (ODES) There are two important points to note regarding this last example. First, application of the transformation resulted in an equation that was solved for the unknown function by purely algebraic means. Second, and most important, if the function x(t), which has the Laplace transform 2/s(s3 4s2 5s 2), were known, we would have the solution to the differential equation and initial conditions. This suggests a procedure for solving differential equations that is analogous to that of using logarithms to multiply or divide. To use logarithms, one transforms the pertinent numbers to their logarithms and then adds or subtracts, which is much easier than multiplying or dividing. The result of the addition or subtraction is the logarithm of the desired answer. The answer is found by reference to a table to find the number having this logarithm. In the Laplace transform method for solution of differential equations, the functions are converted to their transforms, and the resulting equations are algebraically solved for the unknown function. This is much easier than solving a differential equation. However, at the last step the analogy to logarithms is not complete. We obviously cannot hope to construct a table containing the Laplace transform of every function f(t) that possesses a transform. Instead, we will develop methods for expressing complicated transforms, such as x(s) in Example 2.2, in terms of simple transforms that can be found in Table 2.1. For example, it is easily verified that the solution to the differential equation and initial conditions of Example 2.2 is x (t ) 1 2tet e2 t (2.8) The Laplace transform of x, using Eq. (2.8) and Table 2.1, is x (s) 1 1 1 2 2 s s2 (s 1) (2.9) Equation (2.7) is actually the result of placing Eq. (2.9) over a common denominator. Although it is difficult to find x(t) from Eq. (2.7), Eq. (2.9) may be easily inverted to Eq. (2.8) by using Table 2.1. Therefore, what is required is a method for expanding the common-denominator form of Eq. (2.7) to the separated form of Eq. (2.9). This method is provided by the technique of partial fractions, which is developed in Chap. 3. Using MATLAB for Symbolic Processing—Inversion of Laplace Transforms Remember that we have previously declared some variables symbolic (a, k, x, y, z, t and s). Let’s have MATLAB invert Eq. (2.3) for us and determine x(t). x=ilaplace(2/s/(s^3+4*s^2+5*s+2)) x= 1-exp(–2*t)–2*t*exp(–t) which is the same as Eq. (2.8). cou9789x_ch02_009-031.indd 26 8/14/08 1:55:22 PM Confirming Pages CHAPTER 2 MODELING TOOLS FOR PROCESS DYNAMICS 27 Using MATLAB for Symbolic Processing—Solution of Differential Equations MATLAB can solve differential equations symbolically using the DSOLVE command. The derivatives are represented as Dx (first derivative), D2x (second derivative), etc. x=dsolve('D3x+4*D2x+5*Dx+2*x=2','x(0)=0','Dx(0)=0','D2x(0)=0') x= 1–exp(–2*t)–2*t*exp(–t) which, again, is the same as Eq. (2.8). Now, let’s return to our chemical mixing scenario from earlier in the chapter. Example 2.3. Transform the differential equations resulting from the mass and energy balances for the chemical mixer to determine the transform of the exit concentration and temperature. t t dCa Ca Ca 3 dt 1 dT Q T T3 dt r v3C p mass baa lance energy balance For the mass balance, t [sCa (s) Ca (0)] Ca (s) Ca 3 (s) Rearranging and solving for Ca(s), we have Ca ( s ) Ca 3 ( s ) t Ca ( 0 ) ts 1 After the disturbance, Ca3 has a constant value of 2 g/L. Therefore, Ca3 (s) 2/s. Also, from the process description, we know that Ca(0) 3g/L [this is the initial concentration of A in the tank at time 0 (3 P.M.)] and that , the time constant, is 5 min. Substituting these values into the expression for Ca yields Ca ( s ) 2 /s 5(3) 2 15 5s 1 s(5s 1) 5s 1 (2.10) For the energy balance, t [sT (s) T (0)] T (s) T3 (s) cou9789x_ch02_009-031.indd 27 1 Q (s ) r v3C p 8/14/08 1:55:23 PM Confirming Pages 28 PART 1 MODELING FOR PROCESS DYNAMICS Rearranging and solving for T(s), we find T (s ) T3 (s) 1/ ( r v3C p ) Q(s) t T (0) ts1 After the disturbance, T3 is constant at 35C, so T3 (s) 35/s. The initial temperature in the tank T(0), at 3 P.M. is 80C. The heater input Q is constant at 1.05 106 cal/min. The time constant is 5 min. Substituting these values into the expression for T(s) gives 6 cal 1 . 05 10 35 1 min 5 (80) s s g L cal 1000 L 30 min 1 g ⋅ º C T (s ) 5s 1 Simplifying yields T (s ) 70 /s 5(80) 70 400 5s 1 s(5s 1) 5s 1 (2.11) If we can invert these expressions for Ca(s), Eq. (2.10), and T(s), Eq. (2.11), we will obtain the time domain solutions for the exiting concentration and temperature. We will address this topic in Chap. 3. To summarize, we have reviewed a procedure using Laplace transforms for solving linear, ordinary, differential equations (ODEs) with constant coefficients. The procedure is as follows: 1. 2. 3. Take the Laplace transform of both sides of the equation. The initial conditions are incorporated at this step in the transforms of the derivatives. Solve the resulting equation for the Laplace transform of the unknown function algebraically. Find the function of t that has the Laplace transform obtained in step 2. This function satisfies the differential equation and initial conditions and hence is the desired solution. This third step is frequently the most difficult or tedious step and will be developed further in Chap. 3. It is called inversion of the transform. Although there are other techniques available for inversion, the one that we will develop and make consistent use of is that of partial fraction expansion. A simple example will serve to illustrate steps 1 and 2 (we’ll save step 3 until Chap. 3). Example 2.4. Solve dx 3x 0 dt x(0) 2 cou9789x_ch02_009-031.indd 28 8/14/08 1:55:23 PM Confirming Pages CHAPTER 2 MODELING TOOLS FOR PROCESS DYNAMICS 29 We number our steps according to the discussion in the preceding paragraphs: 1. [ sx (s) 2 ] 3 x (s) 0 sx ( s ) x ( 0 ) 2 1 2 s3 s3 2. x (s) 3. x(t) 2e3t SUMMARY In this chapter we discussed the importance of process modeling and worked through a chemical mixing example that led to two differential equations that described the process (one from the mass balance and one from the energy balance). We solved those relatively simple equations by separating and integrating. We also discussed using Laplace transforms for solving differential equations and presented a table of common transforms. We concluded by demonstrating the use of MATLAB for symbolically solving differential equations. In Chap. 3 we will discuss the method of partial fractions for inverting the solutions we obtained by using transforms to the time domain. PROBLEMS 2.1. Transform the following: (a) sin(2t 4 ) (b) et cos 2t (c) Use the formula for the Laplace transform of a derivative to find L{sinh(kt)} if you are given that L{cosh(kt)} s/(s2 k2). 2.2. Invert the following transforms. 3 (a) s 3 (b) s2 (c) 3 (s 2)2 (d) 3 s3 (e) (f) 1 2 2 s 9 3 s 2 4s 8 s4 s 2 4s 8 1 (h) (s 2)2 (g) cou9789x_ch02_009-031.indd 29 8/14/08 1:55:24 PM Confirming Pages 30 PART 1 MODELING FOR PROCESS DYNAMICS 2.3. Find x(s) for the following differential equations. (a) d2x dx 4 3 x u(t ) x (0) x ′(0) 0 dt dt 2 (b) d2x dx 2 x u(t ) x (0) x ′(0) 1 2 dt dt (c) d2x dx 2 x u(t ) x (0) x ′(0) 0 dt dt 2 2.4. Solve Prob. 2.1 using the MATLAB laplace command. 2.5. Solve Prob. 2.2 using the MATLAB ilaplace command. 2.6. Solve Prob. 2.3 using the MATLAB dsolve command, and then use ezplot to graph the solutions. 2.7. Use the MATLAB dsolve command to solve the differential equations that we developed for the mass and energy balances for the chemical mixing scenario, and then use ezplot to graph the solutions. Compare the results with those presented in the text. 2.8. Use the MATLAB ilaplace command to invert Eqs. (2.10) and (2.11), and then use ezplot to graph the solutions. Compare the results with those presented in the text. 2.9. Rework the chemical mixing scenario if at 3 P.M. the operator mistakenly increases the flow rate of stream 1 to 20 L/min while stream 2 and the heater input remain unchanged. cou9789x_ch02_009-031.indd 30 8/14/08 1:55:26 PM Confirming Pages CHAPTER 2 CAPSULE SUMMARY Definition of the Laplace transform: L{ f (t )} f (s) ∞ st ∫0 f (t )e dt Linearity: L{af1(t) bf2(t)} aL{f1(t)} bL{f2(t)} Transform of first derivative: L { } df (t ) sf (s) f (0) dt Transform of nth derivative: d n f L n s n f (s) s n 1 f (0) s n 2 f (1) (0) n sf ( n 2) (0) f ( n 1) (0) dt Transforms of some simple functions: See Table 2.1. The following procedure uses Laplace transforms for solving linear, ordinary, differential equations (ODEs) with constant coefficients: 1. 2. 3. Take the Laplace transform of both sides of the equation. The initial conditions are incorporated at this step in the transforms of the derivatives. Solve the resulting equation for the Laplace transform of the unknown function algebraically. Find the function of t that has the Laplace transform obtained in step 2. This function satisfies the differential equation and initial conditions and hence is the desired solution. This third step is frequently the most difficult or tedious step. We will make consistent use of partial fraction expansions to accomplish this (see Chap. 3). Useful MATLAB Commands: syms—declares variables to be symbolic laplace—takes the Laplace transform of a symbolic expression ilaplace—inverts a symbolic Laplace transform expression dsolve—solves a differential equation symbolically 31 cou9789x_ch02_009-031.indd 31 8/14/08 1:55:27 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS O ur study of the application of Laplace transforms to linear differential equations with constant coefficients has enabled us to rapidly establish the Laplace transform of the solution. We now wish to develop methods for inverting the transforms to obtain the solution in the time domain. In the first part of this chapter we give a series of examples that illustrate the partial fraction technique. After a generalization of these techniques, we proceed to a discussion of the qualitative information that can be obtained from the transform of the solution without inverting it. The equations to be solved are all of the general form an dnx d n ⫺1 x dx ⫹ an ⫺1 n ⫺1 ⫹ . . . ⫹ a1 ⫹ a0 x ⫽ r (t ) n dt dt dt The unknown function of time is x(t), and an, an 1, . . . , a1, a0 are constants. The given function r(t) is called the forcing function. In addition, for all problems of interest in control system analysis, the initial conditions are given. In other words, values of x, dx/dt, . . . , dn 1 x/dtn 1 are specified at time 0. The problem is to determine x(t) for all t 0. 3.1 PARTIAL FRACTIONS In the series of examples that follow, the technique of partial fraction inversion for solution of this class of differential equations is presented. 32 cou9789x_ch03_032-068.indd 32 8/14/08 2:32:38 PM Confirming Pages CHAPTER 3 Example 3.1. INVERSION BY PARTIAL FRACTIONS 33 Solve dx ⫹x ⫽1 dt x (0) ⫽ 0 Application of the Laplace transform yields sx (s) ⫹ x (s) ⫽ x (s) ⫽ or 1 s 1 s(s ⫹ 1) The theory of partial fractions enables us to write this as x (s ) ⫽ 1 A B ⫽ ⫹ s(s ⫹ 1) s s ⫹1 (3.1) where A and B are constants. Hence, from Table 2.1, it follows that x (t ) ⫽ A ⫹ Be⫺t (3.2) Therefore, if A and B were known, we would have the solution. The conditions on A and B are that they must be chosen to make Eq. (3.1) an identity in s. To determine A, multiply both sides of Eq. (3.1) by s. 1 Bs ⫽ A⫹ s ⫹1 s ⫹1 (3.3) Since this must hold for all s, it must hold for s 0. Putting s 0 in Eq. (3.3) yields A⫽1 To find B, multiply both sides of Eq. (3.1) by s 1. 1 A ⫽ (s ⫹ 1) ⫹ B s s (3.4) Since this must hold for all s, it must hold for s 1. This yields B ⫽ ⫺1 This procedure for determining the coefficients is called the Heaviside expansion. There is an easy way to visualize the Heaviside procedure and quickly determine the coefficients of the partial fraction expansion (A and B in this case). Considering Eq. (3.1), we can determine A, the numerator of the 1/s factor, by ignoring (or “covering up”) this term in the denominator of x(s) and letting all the remaining s’s equal the value of s that makes the “covered up” term equal to zero. The other coefficients are found in a similar manner. cou9789x_ch03_032-068.indd 33 8/14/08 2:32:39 PM Confirming Pages 34 PART 1 MODELING FOR PROCESS DYNAMICS For example, to solve for A, we “cover up” the s factor and let all the other s values equal 0. 1 ⫽1 s ( s ⫹ 1) A⫽ 0 Similarly for B, we cover up the s 1 term and let the other s values equal 1, so B ⫽ 1 s (s ⫹ 1) ⫽ 1 ⫽ ⫺1 ⫺1 ⫺1 Cross-multiplication (as well as the quick visualization method) works for distinct roots (non-repeated factors in the denominator) and in a limited way for repeated roots. We will discuss the case of repeated roots shortly. Now that we’ve found A and B, we have x (s) ⫽ 1 1 1 ⫽ ⫺ s(s ⫹ 1) s s ⫹1 (3.5) x (t ) ⫽ 1 ⫺ e⫺t (3.6) and therefore, Equation (3.5) may be checked by putting the right side over a common denominator, and Eq. (3.6) by substitution into the original differential equation and initial condition. Example 3.2. Chemical mixing scenario revisited. In Chap. 2 we solved the chemical mixing scenario problem to the point where we had obtained the transformed solution to the material and energy balances. The transformed solutions, Eqs. (2.10) and (2.11), are repeated in this example for convenience. 2 15 ⫹ s(5s ⫹ 1) 5s ⫹ 1 Ca ( s ) ⫽ (2.10) We can now invert this expression for the concentration in the tank to the time domain. Considering the first term on the left-hand side, we can separate it into partial fractions by using the same method that was employed in Example 3.1. 2 2 5 ⫽ s(5s ⫹ 1) s s⫹ ( cou9789x_ch03_032-068.indd 34 1 5 ) ⫽ A B ⫹ s s⫹ 1 5 ⫽ ⫺2 2 ⫹ s s ⫹ 15 8/14/08 2:32:39 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 35 Equation (2.10) may now be written as Ca ( s ) ⫽ 2 15 2 2 ⫹ ⫽ ⫺ s(5s ⫹ 1) s 5s ⫹ 1 s⫹ 1 5 ⫹ 3 s⫹ 1 5 ⫽ 2 1 ⫹ s s⫹ 1 5 We can now readily invert this expression to the time domain Ca (t ) ⫽ 2 ⫹ e⫺t / 5 This is the same solution that we previously obtained by separation and integration of the original mass balance differential equation in the time domain, which is plotted in Fig. 2–3. Similarly, we can obtain the time domain solution for the temperature in the mixing vessel by inverting Eq. (2.11). T (s ) ⫽ 70 /s ⫹ 5(80) 70 400 ⫽ ⫹ 5s ⫹ 1 s(5s ⫹ 1) 5s ⫹ 1 (2.11) Separating the right-hand side by using partial fractions, we get T (s ) ⫽ 70 400 70 ⫺70 80 70 10 ⫹ ⫽ ⫹ ⫹ ⫽ ⫹ 1 1 s(5s ⫹ 1) 5s ⫹ 1 s s s⫹ s⫹ s⫹1 5 ⫺t / 5 5 5 T (t ) ⫽ 70 ⫹ 10e This is the same solution that we previously obtained by separation and integration of the original energy balance differential equation in the time domain which is plotted in Fig. 2–7. Example 3.3. Solve d3x d2x dx ⫹ 2 ⫺ ⫺ 2 x ⫽ 4 ⫹ e2t 3 2 dt dt dt x (0) ⫽ 1 x ′(0) ⫽ 0 x ′′(0) ⫽ ⫺1 Taking the Laplace transform of both sides yields 4 1 s 3 x (s) ⫺ s 2 ⫹ 1 ⫹ 2 s 2 x (s) ⫺ s ⫺ [ sx (s) ⫺ 1] ⫺ 2 x (s) ⫽ ⫹ s s⫺2 Solving algebraically for x(s), we find x (s ) ⫽ cou9789x_ch03_032-068.indd 35 s 4 ⫺ 6s 2 ⫹ 9s ⫺ 8 s(s ⫺ 2)(s 3 ⫹ 2s 2 ⫺ s ⫺ 2) (3.7) 8/14/08 2:32:40 PM Confirming Pages 36 PART 1 MODELING FOR PROCESS DYNAMICS The cubic in the denominator may be factored, and x(s) is expanded in partial fractions. x (s ) ⫽ s 4 ⫺ 6s 2 ⫹ 9s ⫺ 8 A B C D E ⫽ ⫹ ⫹ ⫹ ⫹ s(s ⫺ 2)(s ⫹ 1)(s ⫹ 2)(s ⫺ 1) s s⫺2 s ⫹1 s⫹2 s ⫺1 (3.8) To find A, multiply both sides of Eq. (3.8) by s and then set s 0; the result is A = ⫺8 ⫽ ⫺2 (⫺2)(1)(2)(⫺1) The other constants are determined in the same way. The procedure and results are summarized in the following table. To determine Multiply Eq. (3.8) by and set s to Result B s2 2 B ⫽ 112 C s1 1 C ⫽ 11 3 D s2 2 D ⫽ ⫺17 12 E s1 1 E = 2 3 Accordingly, the solution to the problem is x (t ) ⫽ ⫺2 ⫹ 1 2t 12 e ⫹ 11 ⫺t 3 e ⫺ 17 ⫺2 t 12 e ⫹ 23 e t A comparison between this method and the classical method, as applied to Example 3.2, may be profitable. In the classical method for solution of differential equations, we first write down the characteristic function of the homogeneous equation: s 3 ⫹ 2s 2 ⫺ s ⫺ 2 ⫽ 0 This must be factored, as was also required in the Laplace transform method, to obtain the roots 1, 2, and 1. Thus, the complementary solution is xc (t ) ⫽ C1e⫺t ⫹ C2e⫺2 t ⫹ C3e t Furthermore, by inspection of the forcing function, we know that the particular solution has the form x p (t ) ⫽ A ⫹ Be 2 t The constants A and B are determined by substitution into the differential equation and, 1 , respectively. Then as expected, are found to be 2 and 12 x (t ) ⫽ ⫺2 ⫹ cou9789x_ch03_032-068.indd 36 1 2t 12 e ⫹ C1e⫺t ⫹ C2e⫺2 t ⫹ C3e t 8/14/08 2:32:41 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 37 Using MATLAB for Symbolic Processing—Partial Fractions Remember that we have previously declared some variables symbolic (a, k, x, y, z, t, and s). MATLAB does not have a built-in function for performing partial fractions. However, we can force MATLAB to do the work for us by taking advantage of two other MATLAB functions, diff and int. We have MATLAB integrate x(s), which it does internally by using partial fractions, and then immediately differentiate the resulting expression. The result will be the partial fraction expansion of x(s). Let’s have MATLAB find the partial fraction expansion represented by Eq. (3.8). x=(s^4–6*s^2+9*s–8)/s/(s–2)/(s+1)/(s+2)/(s–1) x= (s^4–6*s^2+9*s–8)/s/(s–2)/(s+1)/(s+2)/(s–1) diff(int(x)) ans= –2/s+1/12/(s–2)+11/3/(s+1)–17/12/(s+2)+2/3/(s–1) pretty(ans) -2 / s + 1 / 12 1 1 17 1 1 + 11 / 3 + 2/3 s-2 s + 1 12 s + 2 s-1 Thus, MATLAB arrives at the same result as we did by hand. and constants C1, C2, and C3 are determined by the three initial conditions. The Laplace transform method has systematized the evaluation of these constants, avoiding the solution of three simultaneous equations. Four points are worth noting: 1. 2. 3. 4. In both methods, one must find the roots of the characteristic equation. The roots give rise to terms in the solution whose form is independent of the forcing function. These terms make up the complementary solution. The forcing function gives rise to terms in the solution whose form depends on the form of the forcing function and is independent of the left side of the equation. These terms comprise the particular solution. The only interaction between these sets of terms, i.e., between the right side and left side of the differential equation, occurs in the evaluation of the constants involved. The only effect of the initial conditions is in the evaluation of the constants. This is so because the initial conditions affect only the numerator of x(s), as may be seen from the solution of this example. In the three examples we have discussed, the denominator of x(s) factored into real factors only. In the next example, we consider the complications that arise when the denominator of x(s) has complex factors. cou9789x_ch03_032-068.indd 37 8/14/08 2:32:42 PM Confirming Pages 38 PART 1 MODELING FOR PROCESS DYNAMICS Using MATLAB for Symbolic Processing—Solving ODEs MATLAB can symbolically solve ODEs. It uses the DSOLVE command for this purpose. We illustrate the use of this command with Example 3.2. The problem consisted of d3x d2x dx ⫹ 2 ⫺ ⫺ 2 x ⫽ 4 ⫹ e2t 3 2 dt dt dt x (0) ⫽ 1 x ′(0) ⫽ 0 x ′′(0) ⫽ ⫺1 The DSOLVE command is straighforward for solving this equation. dsolve('D3x+2*D2x–Dx–2*x=4+exp(2*t)','x(0)=1','Dx(0)=0','D2x(0)=–1') ans= –1/12*exp(2*t)*(24*exp(–2*t)–1)+2/3*exp(t)–17/12*exp(–2*t)+11/3*exp(–t) expand(ans) This command multiplies out the expression to make it easier to compare with our original answer. ans= –2+1/12*exp(t)^2+2/3*exp(t)–17/12/exp(t)^2+11/3/exp(t) ⫺t ⫺2 t 1 2t which is the same result we obtained by hand: x (t ) ⫽ ⫺2 ⫹ 12 e ⫹ 11 ⫺ 17 ⫹ 23 e t 3 e 12 e We can verify this result with MATLAB by inverting the partial fraction expansion we obtained with MATLAB previously. ilaplace(–2/s+1/12/(s–2)+11/3/(s+1)–17/12/(s+2)+2/3/(s–1)) ans= –2+1/12*exp(2*t)+11/3*exp(–t)–17/12*exp(–2*t)+2/3*exp(t) The result is the same! Example 3.4. Inversion of a transform that has complex roots in the denominator. Solve d2x dx ⫹ 2x ⫽ 2 2 ⫹ 2 dt dt x (0) ⫽ 0 x ′(0) ⫽ 0 Application of the Laplace transform yields x (s) ⫽ ( 2 s s 2 ⫹ 2s ⫹ 2 ) The quadratic term in the denominator may be factored by use of the quadratic formula. The roots are found to be 1 j and 1 j. If we use these complex cou9789x_ch03_032-068.indd 38 8/14/08 2:32:43 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 39 roots in the partial fraction expansion, the algebra can get quite tedious. We present a method to obtain the partial fraction expansion for the case of complex roots, without resorting to the use of complex algebra. Avoiding the use of complex algebra with a quadratic term. If we choose not to factor the quadratic term, we can use an alternate form of the partial fraction expansion. 2 A Bs ⫹ C ⫽ ⫹ 2 s s(s ⫹ 2s ⫹ 2) s ⫹ 2s ⫹ 2 x (s) ⫽ 2 Note that the second term of the expansion has the unfactored quadratic in the denominator. The numerator of each term in the expansion is a polynomial in s of one less degree than the denominator, hence the Bs C in the numerator (a first-order numerator with a second-order denominator). As before, we can determine A. A⫽ 2 ⫽1 0 ⫹ 2(0) ⫹ 2 So we now have x (s ) ⫽ 2 1 Bs ⫹ C ⫽ ⫹ 2 s s(s ⫹ 2s ⫹ 2) s ⫹ 2s ⫹ 2 2 Clearing the denominator on the left-hand side, we obtain 2 ⫽ s 2 ⫹ 2s ⫹ 2 ⫹ Bs 2 ⫹ Cs Collecting like terms, we get ( B ⫹ 1)s 2 ⫹ (2 ⫹ C )s ⫹ 2 ⫽ 2 We now match coefficients of like terms on the left and right sides of the equation. s2 : B ⫹ 1 ⫽ 0 s: 2 ⫹ C ⫽ 0 B ⫽ ⫺1 C ⫽ ⫺2 Thus, x (s) ⫽ 1 s⫹2 ⫺ 2 s s ⫹ 2s ⫹ 2 (3.9) To invert the second term, we complete the square in the denominator to get a familiar transform. Remember that for a perfect square, the quadratic must have cou9789x_ch03_032-068.indd 39 8/14/08 2:32:43 PM Confirming Pages 40 PART 1 MODELING FOR PROCESS DYNAMICS the form s2 a s (a /2)2 (s a /2)2, where the constant is one-half the middle coefficient squared. The second term on the right-hand side becomes s⫹2 s⫹2 s⫹2 ⫽ 2 ⫽ s ⫹ 2s ⫹ 2 (s ⫹ 2s ⫹ 1) ⫹ 2 ⫺ 1 (s ⫹ 1)2 ⫹ 1 2 where we’ve added and subtracted one-half the middle coefficent squared, (2/2)2 1, so the denominator remains unchanged. The transform of the solution is now x (s ) ⫽ 1 s⫹2 ⫺ s (s ⫹ 1)2 ⫹ 1 (3.10) One last modification of the second term is required before inversion. A term of this type will lead to a sine term and a cosine term in the solution. From Table 2.1, we see that { } (3.11a) { } (3.11b) k (s ⫹ a) 2 ⫹ k 2 s⫹a L e⫺at cos( kt ) ⫽ (s ⫹ a)2 ⫹ k 2 L e⫺at sin( kt ) ⫽ Note that everywhere s appears in these forms, it appears as the quantity s a. Thus, comparing these transforms with Eq. (3.10), we see that we need an s 1 term in the numerator, to go with the s 1 in the denominator. So we regroup as x (s ) ⫽ 1 (s ⫹ 1) ⫹ 1 1 s ⫹1 1 ⫺ ⫽ ⫺ ⫺ s s (s ⫹ 1)2 ⫹ 12 (s ⫹ 1)2 ⫹ 12 (s ⫹ 1)2 ⫹ 12 We can easily invert these terms to obtain the solution to the differential equation. x (t ) ⫽ 1 ⫺ e⫺t (cos t ⫹ sin t ) We now summarize the steps in this method for inverting quadratic terms with complex roots while avoiding the use of complex algebra. Step 1. Form the partial expansion term for the quadratic with a first-order term in s in the numerator. Step 2. Determine the numerators of the other terms in the expansion, using the Heaviside expansion. Step 3. Cross-multiply the equation for x(s) by the denominator of x(s), and equate coefficients of like terms to determine the constants in the numerator of the quadratic term. Step 4. Complete the square for the quadratic term. Step 5. Regroup the terms in the numerator, such that if the quadratic is now (s a)2, everywhere else that s appears, it appears as s a. Step 6. Invert the resulting two terms to a sine and a cosine term (probably multiplied by an exponential). cou9789x_ch03_032-068.indd 40 8/14/08 2:32:48 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 41 In the next example, an exceptional case is considered; the denominator of x(s) has repeated roots. The procedure in this case will vary slightly from that of the previous cases. Example 3.5. Inversion of a transform with repeated roots. Solve d3x 3 d2x 3 dx ⫹ ⫹x ⫽1 3 2 ⫹ dt dt dt x (0) ⫽ x ′(0) ⫽ x ′′(0) ⫽ 0 Application of the Laplace transform yields x (s) ⫽ ( 1 3 2 ) s s ⫹ 3s ⫹ 3s ⫹ 1 Factoring and expanding in partial fractions, we find x (s) ⫽ 1 A B C D ⫹ 3 ⫽ 3 ⫹ 2 ⫹ s s ⫹1 s(s ⫹ 1) (s ⫹ 1) (s ⫹ 1) (3.12) As in the previous cases, to determine A, multiply both sides by s and then set s to zero. This yields A⫽1 Multiplication of both sides of Eq. (3.12) by (s 1)3 results in 1 A(s ⫹ 1)3 ⫽ ⫹ B ⫹ C (s ⫹ 1) ⫹ D(s ⫹ 1)2 s s (3.13) Setting s 1 in Eq. (3.13) gives B ⫽ ⫺1 Having found A and B, we introduce these values into Eq. (3.12) and place the right side of the equation over a common denominator; the result is 1 (s ⫹ 1)3 ⫺ s ⫹ Cs(s ⫹ 1) ⫹ Ds(s ⫹ 1)2 3 ⫽ s(s ⫹ 1) s(s ⫹ 1)3 (3.14) Expanding the numerator of the right side gives 1 (1 ⫹ D)s 3 ⫹ (3 ⫹ C ⫹ 2 D)s 2 ⫹ (2 ⫹ C ⫹ D)s ⫹ 1 3 ⫽ s(s ⫹ 1) s(s ⫹ 1)3 (3.15) We now equate the numerators on each side to get 1 ⫽ (1 ⫹ D)s 3 ⫹ (3 ⫹ C ⫹ 2 D)s 2 ⫹ (2 ⫹ C ⫹ D)s ⫹ 1 cou9789x_ch03_032-068.indd 41 8/14/08 2:32:48 PM Confirming Pages 42 PART 1 MODELING FOR PROCESS DYNAMICS Equating the coefficients of like powers of s gives 1⫹ D ⫽ 0 3 ⫹ C ⫹ 2D ⫽ 0 2⫹C ⫹ D ⫽ 0 Solving these equations gives C 1 and D 1. The final result is then x (s) ⫽ 1 1 1 1 ⫺ 3 ⫺ 2 ⫺ s s ⫹1 (s ⫹ 1) (s ⫹ 1) (3.16) By referring to Table 2.1, this can be inverted to t2 x (t ) ⫽ 1 ⫺ e⫺t ⫹ t ⫹ 1 2 (3.17) The reader should verify that Eq. (3.16) placed over a common denominator results in the original form x (s) ⫽ 1 s(s ⫹ 1)3 and that Eq. (3.17) satisfies the differential equation and initial conditions. The result of Example 3.5 may be generalized. The appearance of the factor (s a)n in the denominator of x(s) leads to n terms in the partial fraction expansion C1 C2 Cn , ,..., n n ⫺1 s⫹a (s ⫹ a) (s ⫹ a) The constant C1 can be determined as usual by multiplying the expansion by (s a)n and setting s a. The other constants are determined by the method shown in Example 3.5. These terms, according to Table 2.1, lead to the following expression as the inverse transform: C2 C1 n ⫺1 ⫹ t n ⫺2 ⫹ . . . ⫹ Cn ⫺1t ⫹ Cn e⫺at (n ⫺ 1)! t ( n ⫺ 2 )! (3.18) It is interesting to recall that in the classical method for solving these equations, one treats repeated roots of the characteristic equation by postulating the form of Eq. (3.18) and selecting the constants to fit the initial conditions. cou9789x_ch03_032-068.indd 42 8/14/08 2:32:49 PM Confirming Pages CHAPTER 3 3.2 INVERSION BY PARTIAL FRACTIONS 43 QUALITATIVE NATURE OF SOLUTIONS If we are interested only in the form of the solution x(t), which is often the case in our work, this information may be obtained directly from the roots of the denominator of x(s). As an illustration of this “qualitative” approach to differential equations, consider Example 3.4 in which x (s) ⫽ ( 2 2 s s ⫹ 2s ⫹ 2 ) ⫽ A B C ⫹ ⫹ s s ⫹1⫹ j s ⫹1⫺ j is the transformed solution of d2x 2 ′ dx ⫹ ⫹ 2x ⫽ 2 2 dt dt x (0) ⫽ x ′(0) ⫽ 0 It is evident by inspection of the partial fraction expansion, without evaluation of the constants, that the s in the denominator of x(s) will give rise to a constant in x(t). Also, since the roots of the quadratic term are 1 j, it is known that x(t) must contain terms of the form et (C1 cos t C2 sin t) as we saw previously. This may be sufficient information for our purposes. Alternatively, we may be interested in the behavior of x(t) as t → . It is clear that the terms involving sin and cos vanish because of the factor et. Therefore, x(t) ultimately approaches the constant, which by inspection must be unity. The qualitative nature of the solution x(t) can be related to the location of the roots of the denominator of x(s) in the complex plane. These roots are the roots of the characteristic equation and the roots of the denominator of the transformed forcing function. Consider Fig. 3–1, a drawing of the complex plane, in which several typical Imaginary axis 1.5 1 0.5 0 –0.5 –1 0 0.5 1 1.5 2 2.5 S2 S3 (0, b3) 2 1.5 1 0 0.5 –1 –1.5 –2 0 0.1 0.20.30.40.50.60.70.80.9 1 (–a2, b2) S4 S1 (a4, b4) S6 (0, 0) (–a1, 0) S4* 2.5 1.5 S2* 1 (–a2, –b2) 0.5 0 1.5 1 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 (a4, –b4) 2 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 S3* (0, –b3) 12 10 8 6 4 2 0 –2 –4 –6 –8 0 0.1 0.20.30.40.50.60.70.80.9 1 S5 (a5, 0) 10 9 8 7 6 5 4 3 2 1 Real axis 0 0.1 0.20.30.40.50.60.70.80.9 1 2 1.5 1 0 0.5 –1.5 –2 0 0.1 0.20.30.40.50.60.70.80.9 1 FIGURE 3–1 Location of typical roots of characteristic equation. cou9789x_ch03_032-068.indd 43 8/14/08 2:32:50 PM Confirming Pages 44 PART 1 MODELING FOR PROCESS DYNAMICS TABLE 3.1 Nature of terms in the solution x(t) based on roots in the denominator of X(s) from Fig. 3–1 Roots in denominator of X(s) Terms in x(t) for t > 0 s1 C1e⫺a1 t s2 , s2* e⫺a2 t (C1 cos b2 t ⫹ C2 sin b2 t ) s3 , s3* C1 cos b3 t C2 sin b3 t s4 , s4* e a4 t (C1 cos b4 t ⫹ C2 sin b4 t ) s5 C1e a5 t s6 C1 roots are located and labeled with their coordinates. Table 3.1 gives the form of the terms in the equation for x(t), corresponding to these roots. Note that all constants a1, a2, . . . , b1, b2, . . . are taken as positive. The constants C1 and C2 are arbitrary and can be determined by the partial fraction expansion techniques. As discussed above, this determination is often not necessary for our work. If any of these roots are repeated, the term given in Table 3.1 is multiplied by a power series in t K1 ⫹ K 2 t ⫹ K 3 t 2 ⫹ . . . ⫹ K r t r ⫺1 where r is the number of repetitions of the root and the constants K1, K2, . . . , Kr can be evaluated by partial fraction expansion. It is thus evident that the imaginary axis divides the root locations into distinct areas, with regard to the behavior of the corresponding terms in x(t) as t becomes large. Terms corresponding to roots to the left of the imaginary axis vanish exponentially in time, while those corresponding to roots to the right of the imaginary axis increase exponentially in time. Terms corresponding to roots at the origin behave as power series in time, a constant being considered as a degenerate power series. Terms corresponding to roots located elsewhere on the imaginary axis oscillate with constant amplitude in time unless they are multiple roots, in which case the amplitude of oscillation increases as a power series in time. Much use will be made of this information in later sections of the text. SUMMARY The reader now has available the basic tools for the use of Laplace transforms to solve differential equations. In addition, it is now possible to obtain considerable information about the qualitative nature of the solution with a minimum of labor. It should be pointed out that it is always necessary to factor the denominator of x(s) to obtain any information about x(t). If this denominator is a polynomial of order 3 or more, this may be far from a trivial problem. cou9789x_ch03_032-068.indd 44 8/14/08 2:32:51 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 45 The appendix to this chapter is a grouping of several Laplace transform theorems that will find later application. In addition, a discussion of the impulse function d (t) is presented there. Unavoidably, this appendix is rather dry. It may be desirable for the reader to skip directly to Chap. 4, where our control studies begin. At each point where a theorem of App. 3A is applied, reference to the appropriate section of the appendix can be made. PROBLEMS 3.1. Solve the following by using Laplace transforms. (a) d 2x dx ⫹ ⫹x ⫽1 dt dt 2 (b) d2x dx ⫹2 ⫹x ⫽1 dt dt 2 x (0) ⫽ x ′(0) ⫽ 0 (c) d2x dx ⫹3 ⫹x ⫽1 dt dt 2 x (0) ⫽ x ′(0) ⫽ 0 x (0) ⫽ x ′(0) ⫽ 0 Sketch the behavior of these solutions on a single graph. What is the effect of the coefficient of dx/dt? 3.2. Solve the following differential equations by Laplace transforms. (a) d4x d 3x ⫽ cos t 4 ⫹ dt dt 3 (b) d 2q dq ⫹ ⫽ t 2 ⫹ 2t 2 dt dt x (0) ⫽ x ′(0) ⫽ x ′′′(0) ⫽ 0 q(0) ⫽ 4 x ′′(0) ⫽ 1 q′(0) ⫽ ⫺2 3.3. Invert the following transforms. 3s (a) s 2 ⫹ 1 s 2 ⫹ 4 ( )( 1 (b) s s 2 ⫺ 2s ⫹ 5 ( (c) ) ) 3s 3 ⫺ s 2 ⫺ 3s ⫹ 2 s 2 (s ⫺ 1)2 3.4. Expand the following functions by partial fraction expansion. Do not evaluate coefficients or invert expressions. 2 (a) X (s) ⫽ 2 2 (s ⫹ 1) s ⫹ 1 (s ⫹ 3) ( ) (b) X (s) ⫽ 1 s 3 (s ⫹ 1)(s ⫹ 2)(s ⫹ 3)3 (c) X (s) ⫽ 1 (s ⫹ 1)(s ⫹ 2)(s ⫹ 3)(s ⫹ 4) cou9789x_ch03_032-068.indd 45 8/14/08 2:32:52 PM Confirming Pages 46 PART 1 MODELING FOR PROCESS DYNAMICS 3.5. (a) Invert: x(s) 1/[s(s 1)(0.5s 1)] (b) Solve: dx/dt 2x 2 x(0) 0 3.6. Obtain y(t) for (a) y(s) ⫽ s ⫹1 s ⫹ 2s ⫹ 5 (b) y(s) ⫽ s 2 ⫹ 2s s4 (c) y(s) ⫽ 2s (s ⫺ 1)3 2 3.7. (a) Invert the following function. ( ) y(s) ⫽ 1/ s 2 ⫹ 1 2 (b) Plot y versus t from 0 to 3p. 3.8. Determine f (t) for f(s) 1/[s2 (s 1)]. 3.9. Solve the following differential equations. (a) d2x dx ⫹4 ⫹ 3 x ⫽ u(t ) dt dt 2 (b) d2x dx ⫹2 ⫹ x ⫽ u(t ) dt dt 2 (c) 2 d2x dx ⫹2 ⫹ x ⫽ u(t ) 2 dt dt x (0) ⫽ x ′(0) ⫽ 0 x (0) ⫽ x ′(0) ⫽ 1 x (0) ⫽ x ′(0) ⫽ 0 3.10. Use the trigonometric identities below to express the solution to Prob. 3.9c in terms of sine only. (Note: A sine and a cosine wave with the same frequency can be expressed as a single sine wave of the same frequency. The resulting sine wave will have a different amplitude and be phase-shifted from the original waves. This result will be important when we discuss frequency response in Chap. 15.) a1 cos b ⫹ a2 sin b ⫽ a3 sin(b ⫹ ⌽) sin( A ⫹ B) ⫽ sin A cos B ⫹ sin B cos A 3.11. Find f (t) if F(s) is cou9789x_ch03_032-068.indd 46 (a) 1 (s ⫹ 1)3 (s ⫹ 2) (b) s ⫹1 s 2 ⫹ 2s ⫹ 5 (c) s2 ⫺ s ⫺ 6 s 3 ⫺ 2s 2 ⫺ s ⫹ 2 (d) s ⫹1 s 2 (s ⫹ 2) 8/14/08 2:32:53 PM Confirming Pages CHAPTER 3 (e) 1 s( As ⫹ 1)( Bs ⫹ 1) (f) s ⫹1 s(2s ⫹ 1) (g) s ⫹1 s 2 ⫹ 3s ⫹ 1 (h) s ⫹1 s 2 (2s ⫹ 1) INVERSION BY PARTIAL FRACTIONS 47 3.12. Find the solution to the following set of equations. dx1 ⫽ 2 x1 ⫹ 3 x 2 ⫹ 1 dt dx 2 t ⫽ 2 x1 ⫹ x 2 ⫹ e dt x1 (0) ⫽ x 2 (0) ⫽ 0 Hint: Transformed equations can be manipulated algebraically to solve for each unknown (i.e., two equations in two unknowns) and then inverted separately. 3.13. Use MATLAB. (a) Find the partial fraction expansions for Prob. 3.11. (b) Invert the transforms in Prob. 3.11, using the ILAPLACE command. (c) Graph the solutions to Prob. 3.11 (skip Prob. 3.11e). 3.14. Use the MATLAB DSOLVE command to solve Prob. 3.12. 3.15. Use the MATLAB DSOLVE command to solve Prob. 3.9. 3.16. (a) Solve the differential equations in Prob. 2.3, using partial fractions. (b) Use the MATLAB DSOLVE command to solve the ODEs in Prob. 2.3. cou9789x_ch03_032-068.indd 47 8/14/08 2:32:55 PM Confirming Pages CHAPTER 3 CAPSULE SUMMARY Any proper fraction may be resolved into a number of partial fractions subject to the following rules. 1. 2. Factors such as (as b) in the denominator F(s)/… (as b) . . . will produce a term of type A/(as b), where A is a nonzero constant, in the expansion. If there are repeated factors in the denominator, such as F(s)/(as b)n, they will produce n terms in the partial fraction expansion. F (s ) ( as ⫹ b )n root repeated n times 3. ⫽ A1 A2 An ⫹ ⫹ .. . ⫹ 2 as ⫹ b ⫹ ⫹ b ( as b ) ( as )n there are n part ial fractions in expansion Quadratic or polynomial factors that you do not choose to factor yield F (s) ⫽ (as ⫹ bs ⫹ c) 2 As ⫹ B ⫹bs ⫹c as 2 ⫹ numerator is po o lynomial of one less degree than denominato or 4. If you have repeated quadratics, F (s ) ⫽ ( as ⫹ bs ⫹ c )n 2 yields → A2 s ⫹ B2 An s ⫹ Bn A1s ⫹ B1 ⫹ ⫹ ... ⫹ ⫹ .. . 2 2 2 n ⫹ bs ⫹ c ( as ⫹ bs ⫹ c ) ( as ⫹ b s ⫹ c ) as 2 n terms in expansion Useful MATLAB commands: dsolve—used to symbolically solve ODEs ilaplace—used to invert Laplace transforms diff(int(...))—used to “force” MATLAB to perform a partial fraction expansion 48 cou9789x_ch03_032-068.indd 48 8/14/08 2:32:55 PM Confirming Pages CHAPTER APPENDIX 3 FURTHER PROPERTIES OF TRANSFORMS AND PARTIAL FRACTIONS This appendix is a collection of theorems and results relative to the Laplace transformation. The theorems are selected because of their applicability to problems in control theory. Other theorems and properties of the Laplace transformation are available in standard texts on elementary differential equations. In later chapters, the theorems presented here will be used as needed. 3A.1 FINAL-VALUE THEOREM If f(s) is the Laplace transform of f (t), then lim [ f (t )] ⫽ lim[sf (s)] t→∞ s→0 provided that sf (s) does not become infinite for any value of s satisfying Re(s) 0. If this condition does not hold, f (t) does not approach a limit as t → . In the practical application of this theorem, the limit of f (t) that is found by use of the theorem is correct only if f (t) is bounded as t approaches infinity. Proof. From the Laplace transform of a derivative, we have ∞df ⫺st ∫0 dt e dt ⫽ sf (s) ⫺ f (0) 49 cou9789x_ch03_032-068.indd 49 8/14/08 2:32:56 PM Confirming Pages 50 PART 1 MODELING FOR PROCESS DYNAMICS Hence, ∞ df ∫ s→0 0 lim dt e⫺st dt ⫽ lim [sf (s)] ⫺ f (0) s→0 It can be shown that the order of the integration and limit operation on the left side of this equation can be interchanged if the conditions of the theorem hold. Doing this gives ∞df ∫0 dt dt ⫽ lim[sf (s)] ⫺ f (0) s→0 Evaluating the integral gives lim[ f (t )] ⫺ f (0) ⫽ lim[sf (s)] ⫺ f (0) t →∞ s→0 which immediately yields the desired result. Example 3A.1. transform is Find the final value of the function x(t) for which the Laplace x (s) ⫽ 1 ( ) s s 3 ⫹ 3s 2 ⫹ 3s ⫹ 1 Direct application of the final-value theorem yields lim[ x (t )] ⫽ lim s → 0 s3 t →∞ 1 ⫽1 ⫹ 3s 2 ⫹ 3s ⫹ 1 As a check, note that this transform was inverted in Example 3.6 to give t2 x (t ) ⫽ 1 ⫺ e⫺t ⫹ t ⫹ 1 2 which approaches unity as t approaches infinity. Note that since the denominator of sx(s) can be factored to (s 1)3, the conditions of the theorem are satisfied; that is, (s 1)3 0 unless s 1. Example 3A.2. transform is Find the final value of the function x(t) for which the Laplace x (s ) ⫽ s 4 ⫺ 6s 2 ⫹ 9s ⫺ 8 ( s(s ⫺ 2) s 3 ⫹ 2s 2 ⫺ s ⫺ 2 ) In this case, the function sx(s) can be written sx (s) ⫽ cou9789x_ch03_032-068.indd 50 s 4 ⫺ 6s 2 ⫹ 9s ⫺ 8 (s ⫹ 1)(s ⫹ 2)(s ⫺ 1)(s ⫺ 2) 8/14/08 2:32:56 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 51 Since this becomes infinite for s 1 and s 2, the conditions of the theorem are not satisfied. Note that we inverted this transform in Example 3.3, where it was found that x (t ) ⫽ ⫺2 ⫹ 1 2t 11 17 ⫺2 t e ⫹ e⫺t ⫺ e ⫹ 23 e t 12 3 12 This function continues to grow exponentially with t and, as expected, does not approach a limit. Mathematically we can understand why this is so, by referring to Fig. 3–1. A term in the denominator (the characteristic equation) such as s 1 yields a root in the right half-plane (to the right of the imaginary axis in Fig. 3–1). Referring to Table 3.1, we see that the solution will involve a term of the form C1et which will grow without bound and not approach a limit as t → . Thus, if any roots of the denominator have positive real parts, the final-value theorem does not apply. The proof of the next theorem closely parallels the proof of the last one and is left as an exercise for the reader. 3A.2 INITIAL-VALUE THEOREM lim [ f (t )] ⫽ lim [sf (s)] x→0 s→∞ The conditions on this theorem are not so stringent as those for the previous one because for functions of interest to us the order of integration and limiting process need not be interchanged to establish the result. Example 3A.3. Find the initial value x(0) of the function that has the transform x (s ) ⫽ s 4 ⫺ 6s 2 ⫹ 9s ⫺ 8 ( s(s ⫺ 2) s 3 ⫹ 2s 2 ⫺ s ⫺ 2 ) The function sx(s) is written in the form sx (s) ⫽ s 4 ⫺ 6s 2 ⫹ 9s ⫺ 8 s 4 ⫺ 5s 2 ⫹ 4 Dividing the numerator and denominator by s4, we get sx (s) ⫽ 1 ⫺ 6 /s 2 ⫹ 9 /s3 ⫺ 8 /s 4 1 ⫺ 5/s 2 ⫹ 4 /s 4 which clearly goes to unity as s becomes infinite. Hence x(0) ⫽ 1 which again checks Example 3.3. cou9789x_ch03_032-068.indd 51 8/14/08 2:32:57 PM Confirming Pages 52 PART 1 MODELING FOR PROCESS DYNAMICS 3A.3 TRANSLATION OF TRANSFORM If L{f (t)} f (s), then L{e⫺at f (t )} ⫽ f (s ⫹ a) In other words, the variable in the transform s is translated by a. We can transform f (t) just as if the eat were not present, and then replace all the s’s by s a. Proof L{e⫺at f (t )} ⫽ Example 3A.4. ∞ ⫺at ⫺st ∫0 f (t )e e dt ⫽ ∞ ⫺( s ⫹ a ) t ∫0 f (t )e d t ⫽ f (s ⫹ a) Find L{eat cos kt}. Since s s ⫹ k2 L{cos kt} ⫽ 2 then by the previous theorem, L{e⫺at cos kt} ⫽ s⫹a (s ⫹ a) 2 ⫹ k 2 which matches the result shown in Table 2.1. A primary use for this theorem is in the inversion of transforms. For example, by using this theorem the transform x (s ) ⫽ 1 (s ⫹ a) 2 can be immediately inverted to x (t ) ⫽ te⫺at In obtaining this result, we made use of the following transform pair from Table 2.1: L{t} ⫽ 3A.4 1 s2 TRANSLATION OF FUNCTION If L{f (t)} f (s), then L{ f (t ⫺ t0 )} ⫽ e⫺st0 f (s) provided that f (t ) ⫽ 0 for t ⬍ 0 (which will always be true for functions we use). cou9789x_ch03_032-068.indd 52 8/14/08 2:32:58 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 53 Before we prove this theorem, it may be desirable to clarify the relationship between f (t t0) and f (t). This is done for an arbitrary function f (t) in Fig. 3A–1, where it can be seen that f (t t0) is simply translated horizontally from f (t) through a distance t0. Proof. L{ f (t ⫺ t0 )} ⫽ ⬁ ∫0 f (t ⫺ t0 )e⫺st dt ⫽ e⫺st0 ∫ ⬁ ⫺t0 f (t ⫺ t0 )e⫺s ( t ⫺ t0 ) d (t ⫺ t0 ) But since f (t) 0 for t < 0, the lower limit of this integral may be replaced by zero. Since t t0 is now the dummy variable of integration, the integral may be recognized as the Laplace transform of f (t); thus, the theorem is proved. This result is also useful in inverting transforms. It follows that if f (t) is the inverse transform of f (s), then the inverse transform of e⫺st0 f (s) is the function g(t ) ⫽ {( 0 f t ⫺t 0 ) t ⬍ t0 t ⬎ t0 In other words, g(t) f (t t0)u(t t0). This property of transforms is very useful when pure time delays are present in a system. For example, f(t) consider a case where we have a stream exiting a chemical reactor, and the sensor that measures the f f (t–t0) exiting concentration lies downstream at a distance of 20 ft. If the exiting concentration changes at time t0 t 0, the sensor will not “see” the change until the O t fluid can travel the 20 ft down the pipeline to the FIGURE 3A–1 sensor. The concentration signal at the sensor is Illustration of f(t t0) as related to f (t). delayed, and the delay time is due to the transportation lag in the pipeline. If the signal at the exit of the reactor is C(t), then the signal at the sensor will be C(t t0)u(t t0), where t0 is the delay time. We will discuss this phenomenon further in Chap. 7. Example 3A.5. Find the Laplace transform of 0 1 f (t ) ⫽ h 0 cou9789x_ch03_032-068.indd 53 t ⬍ 0 0 ⬍ t ⬍ h t ⬎ h 8/14/08 2:33:00 PM Confirming Pages 54 PART 1 MODELING FOR PROCESS DYNAMICS This function is pictured in Fig. 3A–2. It is clear that f (t) may be represented by the difference of two functions 1 f (t ) ⫽ [u(t ) ⫺ u(t ⫺ h)] h 1 h f(t) O h t where u(t h) is the unit-step function translated h units to the right. We may now use the linearity of the transform and the previous theorem to write immediately FIGURE 3A–2 Pulse function of Example 3A.5. f (s ) ⫽ 1 1 ⫺ e⫺hs h s This result is of considerable value in establishing the transform of the unitimpulse function, as will be described in the next section. 3A.5 TRANSFORM OF THE UNIT-IMPULSE FUNCTION Consider again the function of Example 3A.5. If we allow h to shrink to zero, we obtain a new function which is zero everywhere except at the origin, where it is infinite. However, it is important to note that the area under this function always remains equal to unity. We call this new function d (t), and the fact that its area is unity means that ⬁ ∫⫺⬁ d (t ) dt ⫽1 The graph of d (t) appears as a line of infinite height at the origin, as indicated in Table 2.1. The function d (t) is called the unit-impulse function or, alternatively, the delta function. It is mentioned here that, in the strict mathematical sense of a limit, the function f (t) does not possess a limit as h goes to zero. Hence, the function d (t) does not fit the strict mathematical definition of a function. To assign a mathematically precise meaning to the unit-impulse function requires use of the theory of distributions, which is clearly beyond the scope of this text. However, for our work in automatic control, we will be able to obtain useful results by formal manipulation of the delta function, and hence we ignore these mathematical difficulties. We have derived in Example 3A.5 the Laplace transform of f (t) as L{ f (t )} ⫽ 1 ⫺ e⫺hs hs Formally, then, the Laplace transform of d (t) can be obtained by letting h go to 0 in L{f (t)}. Applying L’Hôpital’s rule, we find se⫺hs 1 ⫺ e⫺hs ⫽ lim ⫽1 h→0 h→0 hs s L{d (t )} ⫽ lim (3A.1) This “verifies” the entry in Table 2.1. cou9789x_ch03_032-068.indd 54 8/14/08 2:33:01 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 55 It is interesting to note that since we rewrote f (t) in Example 3A.5 as 1 [u(t ) ⫺ u(t ⫺ h)] h f (t ) ⫽ then d (t) can be written as u(t ) ⫺ u(t ⫺ h) h→0 h d (t ) ⫽ lim In this form, the delta function appears as the derivative of the unit-step function. The reader may find it interesting to ponder this statement in relation to the graphs of d (t) and u(t) and in relation to the integral of d (t) discussed previously. The unit-impulse function finds use as an idealized disturbance in control systems analysis and design. For example if one were to add water to a tank very quickly, the disturbance could be modeled as an impulse whose magnitude would be equal to the volume of the addition. If the contents of a 55-gal drum were quickly dumped into a storage tank, the addition could be modeled using an impulse function. 3A.6 TRANSFORM OF AN INTEGRAL If L{f (t)} f (s), then L {∫ f (t) dt} ⫽ f (ss) t 0 This important theorem is closely related to the theorem on differentiation. Since the operations of differentiation and integration are inverses of each other when applied to the time functions, i.e., d dt t ∫0 f (t ) dt ⫽ t df ∫0 dt dt ⫽ f (t ) (3A.2) it is to be expected that these operations when applied to the transforms will also be inverses. Thus assuming the theorem to be valid, Eq. (3A.2) in the transformed variable s becomes s f (s ) 1 ⫽ sf (s) ⫽ f (s) s s In other words, multiplication of f(s) by s corresponds to differentiation of f (t) with respect to t, and division of f (s) by s corresponds to integration of f (t) with respect to t. The proof follows from a straightforward integration by parts. f (s ) ⫽ cou9789x_ch03_032-068.indd 55 ∞ ∫0 f (t )e⫺st dt 8/14/08 2:33:02 PM Confirming Pages 56 PART 1 MODELING FOR PROCESS DYNAMICS Let u ⫽ e⫺st dv ⫽ f (t ) dt Then du ⫽ ⫺se⫺st dt v ⫽ t ∫0 f (t ) dt Hence, f (s) ⫽ e⫺st ∞ t ∫0 f (t ) dt 0 t ∞ ⫹ s ∫ ∫ f (t ) dt e⫺st dt 0 0 Since f (t) must satisfy the requirements for possession of a transform, it can be shown that the first term on the right, when evaluated at the upper limit of , vanishes because of the factor est. Furthermore, the lower limit clearly vanishes, and hence there is no t contribution from the first term. The second term may be recognized as sL{∫ f (t ) dt}, 0 and the theorem follows immediately. Example 3A.6. Solve the following equation for x(t). t dx ⫽ ∫ x (t ) dt ⫺ t 0 dt x (0) ⫽ 3 Taking the Laplace transform of both sides and making use of the previous theorem yield sx (s) ⫺ 3 ⫽ x (s ) 1 ⫺ 2 s s Solving for x(s) gives x (s) ⫽ 3s 2 ⫺ 1 ( ) s s2 ⫺ 1 ⫽ 3s 2 ⫺ 1 s(s ⫹ 1)(s ⫺ 1) This may be expanded into partial fractions according to the usual procedure to give x (s) ⫽ 1 1 1 ⫹ ⫹ s s ⫹1 s ⫺1 Hence, x (t ) ⫽ 1 ⫹ e⫺t ⫹ e t The reader should verify that this function satisfies the original equation. cou9789x_ch03_032-068.indd 56 8/14/08 2:33:03 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 57 Example 3A.7. We mentioned above that the unit impulse is the derivative of the unit-step function, and therefore the unit step must be the integral of the unitimpulse function. Use the theorem for the transform of an integral to determine the transform of the unit-step function if we know that L{d (t)} 1. ⬁ Since we know that u(t ) ⫽ ∫ d (t ) dt , then 0 ⬁ 1 1 L{u(t )} ⫽ L ∫ d (t ) dt ⫽ L{d (t )} ⫽ s s 0 Similarly, since we know that du(t)/dt d (t), then L{d (t )} ⫽ L 3A.7 { } du(t ) ⫽ sL{u(t )} ⫽ s dt 1 ⋅s ⫽1 CUSTOM INPUTS We can produce “custom” input signals by appropriately constructing them using standard input signals. These custom inputs are frequently useful when we analyze a process disturbance. 1.5 1 f 0.5 0 Example 3A.8. Determine f (t) and f (s) for the input signal in Fig. 3A–3. This signal represents the step change in temperature of a furnace at hour 1, and then after holding for 2 h at that temperature, the slow ramping down to the original temperature over an hour-long period. We can consider this input signal to be constructed from several individual “pieces,” as shown in Fig. 3A–4. Thus, –0.5 –1 –1 0 1 2 time 3 4 f (t ) ⫽ u(t ⫺ 1) ⫺ (t ⫺ 3)u(t ⫺ 3) ⫹ (t ⫺ 4)u(t ⫺ 4) 5 A physical example where a custom input would be useful occurs if we decide to change the temperature set point of a process up by 5F and then 10 min later we step the set FIGURE 3A–3 Custom input signal. 1.5 1 0.5 0.5 0.5 f 1 f 1.5 1 f 1.5 0 0 0 –0.5 –0.5 –0.5 –1 –1 0 1 2 time 3 4 Start with this piece . . . 5 –1 –1 0 1 2 time 3 4 5 Add this piece . . . –1 –1 0 1 2 time 3 4 5 Then add this piece. FIGURE 3A–4 Constructing the custom input signal by combining standard inputs. cou9789x_ch03_032-068.indd 57 8/14/08 2:33:04 PM Confirming Pages 58 PART 1 MODELING FOR PROCESS DYNAMICS point back down to its original value. We’ve introduced a square wave “pulse” that we can construct from two unit steps. 3A.8 GENERAL DISCUSSION OF PARTIAL FRACTIONS ON A QUADRATIC TERM In Chap. 3 we discussed how to express a quadratic term in the denominator of a fraction using partial fractions without resorting to complex algebra if we had complex roots. For completeness, we present the method using complex algebra. Consider the general expression involving a quadratic term x (s ) ⫽ F (s ) s2 ⫹ a s ⫹ b (3A.3) where F(s) is some function of s (say, 1/s). Expanding the terms on the right side gives x (s) ⫽ F1 (s) ⫹ Bs ⫹ C s ⫹ as ⫹ b 2 (3A.4) where F1(s) represents other terms in the partial fraction expansion. First solve for B and C algebraically by placing the right side over a common denominator and equating the coefficients of like powers of s. The next step is to express the quadratic term in the form s 2 ⫹ a s ⫹ b ⫽ (s ⫹ a) 2 ⫹ k 2 The terms a and k can be found by solving for the roots of s2 a s b 0 by the quadratic formula to give s1 a jk and s2 a jk. The quadratic term can now be written s 2 ⫹ a s ⫹ b ⫽ ( s ⫺ s1 ) ( s ⫺ s2 ) ⫽ (s ⫹ a ⫺ jk )(s ⫹ a ⫹ jk ) ⫽ (s ⫹ a)2 ⫹ k 2 Equation (3A.4) now becomes x (s) ⫽ F1 (s) ⫹ Bs ⫹ C (s ⫹ a)2 ⫹ k 2 (3A.5) The numerator of the quadratic term is written to correspond to the transform pairs given by Eqs. (3.11a) and (3.11b). C /B ⫺ a C ⫺ aB Bs ⫹ C ⫽ B s ⫹ a ⫹ k ⫽ B( s ⫹ a ) ⫹ k k k Equation (3A.5) becomes x (s) ⫽ F1 (s) ⫹ B cou9789x_ch03_032-068.indd 58 s⫹a C ⫺ aB k ⫹ k (s ⫹ a)2 ⫹ k 2 (s ⫹ a)2 ⫹ k 2 8/14/08 2:33:07 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 59 Applying the transform pairs of Eqs. (3.11a) and (3.11b) to the quadratic terms on the right gives x (t ) ⫽ F1 (t ) ⫹ Be⫺at cos kt ⫹ C ⫺ aB ⫺at e sin kt k (3A.6) where F1 (t) is the result of inverting F1 (s); B and C are coefficients of polynomial Bs C in numerator of quadratic term; and a and k correspond to the roots of the quadratic, roots a kj. Let’s use this generalized approach to solve a problem that we’re already familiar with, Example 3.4. Recall from Eq. (3.9) that x (s ) ⫽ A Bs ⫹ C ⫺s ⫺ 2 1 ⫹ 2 ⫽ ⫹ 2 s s s ⫹ 2s ⫹ 2 s ⫹ 2s ⫹ 2 The roots of the quadratic in the denominator are Roots ⫽ ⫺2 ⫾ 4 ⫺ 8 ⫽ ⫺1 ⫾ j 2 Summarizing the constants required for Eq. (3A.6), we have A 1 B 1 C 2 a 1 F1(s) 1/s k 1 Substituting these quantities into Eq. (3A.6), we obtain the solution x (t ) ⫽ 1 ⫹ (⫺1)e⫺t cos t ⫹ ⫺2 ⫺ (1)(⫺1) ⫺t e sin t ⫽ 1 ⫺ e⫺t (cos t ⫹ sin t ) 1 which is the same as our previous result. We now apply this method to another example. Example 3A.9. Solve x (s) ⫽ ( 1 2 s s ⫺ 2s ⫹ 5 ) ⫽ A Bs ⫹ C ⫹ 2 s s ⫺ 2s ⫹ 5 Applying the quadratic equation to the quadratic term gives Roots ⫽ 2 ⫾ 4 ⫺ 20 ⫽ 1 ⫾ 2j 2 Thus, we find that a 1 and k 2. Solving for A, B, and C gives A ⫽ 15 , B ⫽ ⫺ 15 , and C ⫽ 25 . Introducing these values into the expression for x(s) and applying Eq. (3A.6) give x (t ) ⫽ cou9789x_ch03_032-068.indd 59 1 5 ⫺ 15 e t cos 2t ⫹ 1 t e sin 10 2t 8/14/08 2:33:08 PM Confirming Pages 60 PART 1 MODELING FOR PROCESS DYNAMICS 3A.9 USING COMPLEX ALGEBRA FOR A QUADRATIC TERM Reworking Example 3.4 using the complex roots of the quadratic, we can arrive at the partial fraction expansion x (s) ⫽ 2 A B C ⫽ ⫹ ⫹ s(s ⫹ 1 ⫹ j)(s ⫹ 1 ⫺ j) s s ⫹1⫹ j s ⫹1⫺ j (3A.7) where A, B, and C are constants to be evaluated, so that this relation is an identity in s. The presence of complex factors does not alter the Heaviside procedure at all. However, the computations are more tedious. To obtain A, multiply Eq. (3A.7) by s and set s 0: A⫽ 2 ⫽1 (1 ⫹ j)(1 ⫺ j) To obtain B, multiply Eq. (3A.7) by s 1 j and set s 1 j: 2 ⫺1 ⫺ j ⫽ (⫺1 ⫺ j)(⫺2 j) 2 B ⫽ To obtain C, multiply Eq. (3A.7) by s 1 j and set s 1 j: C ⫽ 2 ⫺1 ⫹ j ⫽ (⫺1 ⫹ j)(2 j) 2 Therefore, x (s) ⫽ 1 ⫺1 ⫺ j 1 ⫺1 ⫹ j 1 ⫹ ⫹ s 2 s ⫹1⫹ j 2 s ⫹1⫺ j This is the desired result. To invert x(s), we may now use the fact that 1/(s a) is the transform of et. The fact that a is complex does not invalidate this result, as can be seen by returning to the derivation of the transform of eat. The result is x (t ) ⫽ 1 ⫹ ⫺1 ⫺ j ⫺(1⫹ j ) t ⫺1 ⫹ j ⫺(1⫺ j ) t e ⫹ e 2 2 By using the identity e( a ⫹ jb ) t ⫽ e at (cos bt ⫹ j sin bt ) this can be converted to x (t ) ⫽ 1 ⫺ e⫺t (cos t ⫹ sin t ) The details of this conversion are recommended as an exercise for the reader. A more general discussion of this case will promote understanding. It was seen in Example 3.4 that the complex conjugate roots of the denominator of x(s) gave rise to a pair of complex terms in the partial fraction expansion. The constants in these terms, B cou9789x_ch03_032-068.indd 60 8/14/08 2:33:09 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 61 and C, proved to be complex conjugates (1 j)/2 and (1 j)/2. When these terms were combined through a trigonometric identity, it was found that the complex terms canceled, leaving a real result for x(t). Of course, it is necessary that x(t) be real, since the original differential equation and initial conditions are real. This information may be utilized as follows: The general case of complex conjugate roots arises in the form x (s) ⫽ F (s ) ( s ⫹ k1 ⫹ jk2 )( s ⫹ k1 ⫺ jk2 ) (3A.8) where F(s) is some real function of s. For instance, in Example 3A.9 we had F (s ) ⫽ 2 s k1 ⫽ 1 k2 ⫽ 1 Expanding Eq. (3A.8) in partial fractions gives a1 ⫹ jb1 F (s ) a2 ⫹ jb2 ⫹ ⫽ F1 (s) ⫹ (3A.9) ⫹ s ⫹ k ⫹ jk s ⫹ k ⫺ jk s ⫹ k ⫹ j k s k1 ⫺ jk2 ( 1 2 )( 1 2) 2 1 where a1, a2, b1, and b2 are the constants to be evaluated in the partial fraction expansion and F1 (s) is a series of fractions arising from F(s). Again, in Example 3A.9, a1 ⫽ ⫺ 1 2 a2 ⫽ ⫺ 1 2 b1 ⫽ ⫺ 1 2 b2 ⫽ 1 2 F1 (s) ⫽ 1 s Now, since the left side of Eq. (3A.9) is real for all real s, the right side must also be real for all real s. Since two complex numbers will add to form a real number if they are complex conjugates, the right side will be real for all real s if and only if the two terms are complex conjugates. Since the denominators of the terms are conjugates, this means that the numerators must also be conjugates, or a2 ⫽ a1 b2 ⫽ ⫺b1 This is exactly the result obtained in the specific case of Example 3.4. With this information, Eq. (3A.9) becomes a1 ⫹ jb1 F (s ) a1 ⫺ jb1 ⫹ ⫽ F1 (s) ⫹ s ⫹ k1 ⫹ jk2 s ⫹ k1 ⫺ jk2 (3A.10) ( s ⫹ k1 ⫹ jk2 )( s ⫹ k1 ⫺ jk2 ) Hence, it has been established that terms in the inverse transform arising from the complex conjugate roots may be written in the form ( a1 ⫹ cou9789x_ch03_032-068.indd 61 jb1 ) e(⫺k1⫺ jk 2 )t ⫹ ( a1 ⫺ jb1 ) e(⫺k1⫹ jk 2 )t 8/14/08 2:33:10 PM Confirming Pages 62 PART 1 MODELING FOR PROCESS DYNAMICS Again, by using the identity e(C 1⫹ jC 2 )t ⫽ eC 1t ( cos C2 t ⫹ j sin C2 t ) this reduces to 2e⫺k1t ( a1 cos k2 t ⫹ b1 sin k2 t ) (3A.11) Let us now rework Example 3A.9, using Eq. (3A.11). We return to the point at which we arrived, by our usual techniques, with the conclusion that B ⫽ ⫺1 ⫺ j 2 Comparison of Eqs. (3A.7) and (3A.10) and the result for B show that we have two possible ways to assign a1, b1, k1, and k2 so that we match the form of Eq. (3A.10). They are a1 ⫽ ⫺ 12 a1 ⫽ ⫺ 12 b1 ⫽ ⫺ 12 b1 ⫽ 1 2 or k1 ⫽ 1 k2 ⫽ 1 k1 ⫽ 1 k2 ⫽ ⫺ 1 The first way corresponds to matching the term involving B with the first term of the conjugates of Eq. (3A.10), and the second to matching it with the second term. In either case, substitution of these constants into Eq. (3A.11) yields ⫺e⫺t (cos t ⫹ sin t ) which is, as we have discovered, the correct term in x(t). What this means is that one can proceed directly from the evaluation of one of the partial fraction constants, in this case B, to the complete term in the inverse transform, in this case et (cos t sin t). It is not necessary to perform all the algebra, since it has been done in the general case to arrive at Eq. (3A.11). Another example will serve to emphasize the application of this technique. Example 3A.10. Solve d2x ⫹ 4 x ⫽ 2e⫺1 dt 2 x (0) ⫽ x ′(0) ⫽ 0 The Laplace transform method yields x (s) ⫽ cou9789x_ch03_032-068.indd 62 (s 2 2 ) ⫹ 4 (s ⫹ 1) 8/14/08 2:33:11 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 63 Factoring and expanding into partial fractions give 2 A B C ⫽ ⫹ ⫹ (s ⫹ 1)(s ⫹ 2 j)(s ⫺ 2 j) s ⫹1 s ⫹ 2j s ⫺ 2j (3A.12) Multiplying Eq. (3A.12) by s 1 and setting s 1 yields 2 2 ⫽ (⫺1 ⫹ 2 j)(⫺1 ⫺ 2 j) 5 A⫽ Multiplying Eq. (3A.12) by s 2j and setting s 2j yields B ⫽ 2 ⫺2 ⫹ j ⫽ (⫺2 j ⫹ 1)(⫺4 j) 10 Matching the term (⫺2 ⫹ j) / 10 s ⫹ 2j with the first term of the conjugates of Eq. (3A.10) requires that 2 a1 ⫽ ⫺ 10 ⫽ ⫺ 15 b1 ⫽ 1 10 k1 ⫽ 0 k2 ⫽ 2 Substituting in Eq. (3A.11) results in ⫺ 25 cos 2t ⫹ 15 sin 2t Hence the complete answer is x (t ) ⫽ 2 ⫺t e 5 ⫺ 25 cos 2t ⫹ 15 sin 2t Readers should verify that this answer satisfies the differential equation and initial conditions. In addition, they should show that it can also be obtained by matching the term with the second term of the conjugates of Eq. (3A.10) or by determining C instead of B. SUMMARY In this appendix, we have presented and discussed several properties of Laplace transforms. As we continue our studies with first-order systems, these properties will prove quite useful in our understanding and analysis of the process dynamics. PROBLEMS 3A.1. If a forcing function f (t) has the Laplace transform f (s ) ⫽ e⫺s e⫺3s 1 ⫹ 2 ⫺ s s s graph the function f (t). cou9789x_ch03_032-068.indd 63 8/14/08 2:33:13 PM Confirming Pages 64 PART 1 MODELING FOR PROCESS DYNAMICS 3A.2. Solve the following equation for y(t). t ⫺2 ∫ y(t ) d t ⫽ 0 dy ⫹ 3y dt y( 0 ) ⫽ 1 3A.3. Express the function given in Fig. P3A–3 in the t domain and the s domain. f (t) 2 1 0 0 1 2 3 4 5 6 t FIGURE P3A–3 3A.4. Sketch the following functions. (a) f (t) u(t) 2u(t 1) u(t 3) (b) f (t) 3tu(t) 3u(t 1) u(t 2) 3A.5. The function f (t) has the Laplace transform f (s) ⫽ (1 ⫺ 2e⫺s ⫹ e⫺2 s ) / s 2 Obtain the function f (t) and graph f (t). 3A.6. Determine f (t) at t 1.5 and at t 3 for the function f (t ) ⫽ 0 . 5u(t ) ⫺ 0 . 5u(t ⫺ 1) ⫹ (t ⫺ 3)u(t ⫺ 2) 1 dx ⫹ x ⫽ f (t ) x (0) ⫽ 0 2 dt (a) Use f (t) from Prob. 3A.1 and determine x(t). Plot f (t) and x(t) on the same graph. (b) Use f (t) from Prob. 3A.3 and determine x(t). Plot f (t) and x(t) on the same graph. (c) Use f (t) from Prob. 3A.4a and determine x(t). Plot f (t) and x(t) on the same graph. (d) Use f (t) from Prob. 3A.4b and determine x(t). Plot f (t) and x(t) on the same graph. (e) Use f (t) from Prob. 3A.5 and determine x(t). Plot f (t) and x(t) on the same graph. (f) Use f (t) from Prob. 3A.6 and determine x(t). Plot f (t) and x(t) on the same graph. 3A.8. Use the final-value theorem to predict the final values for the ODEs in Prob. 2.3. Verify your answers, using the time domain solutions obtained in Prob. 3.16. 3A.7. For the following ODE: 3A.9. Find and sketch the solution to the following differential equations, using Laplace transforms. (a) y y d (t) y(0) 0 (b) y y d (t 1) y(0) 0 (c) y y u(t) y(0) 0 (d) y y u(t 1) y(0) 0 cou9789x_ch03_032-068.indd 64 8/14/08 2:33:14 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 65 3A.10. For the following transforms, find lim f (t ). t→⬁ (a) f (s) ⫽ 1 s(s ⫹ 1)2 (b) f (s) ⫽ 1 s(s ⫺ 1)2 . 3A.11. Use the formula for the Laplace transform of an integral to find L{sinh (kt)} if you are given that L{cosh (kt)} s/(s2 k2). 3A.12. Transform the function f (t) 5e2t t sin (3t). cou9789x_ch03_032-068.indd 65 8/14/08 2:33:16 PM Confirming Pages APPENDIX 3A CAPSULE SUMMARY Final-value theorem: lim[ f (t )] ⫽ lim [sf (s)] t →∞ s→0 [Be careful, this does not apply if the denominator of f(s) has roots with positive real parts, such as s 1.] Initial-value theorem: lim[ f (t )] ⫽ lim[sf (s)] t→0 s→∞ Translation of a transform: If L{f (t)} f(s), then L{eat f (t)} f(s a). If the function of interest is multiplied by eat, we can transform f (t) just as if the eat were not present, and then replace all the s’s by s a. Translation of a function (time delay): If L{f (t)} f(s), then L{ f (t ⫺ t0 )} ⫽ e⫺st 0 f (s), provided that f (t) 0 for t < 0 (which will always be true for functions we use). Alternatively, if g(s) ⫽ e⫺st 0 f (s), then g(t ) ⫽ {( t ⬍ t0 0 f t ⫺ t0 ) t ⬎ t0 or, more compactly, g(t) f (t t0)u(t t0). Transform of the unit-impulse function: 1 ⫺ e⫺hs se⫺hs ⫽ lim ⫽1 h→0 hs h→0 s L{d (t )} ⫽ lim Relationship between unit step and unit impulse: du(t ) and ⫽ d (t ) dt ⬁ ∫ d (t ) dt ⫽ u(t ) 0 Transform of an integral and a derivative: Multiplication of f (s) by s corresponds to differentiation of f (t) with respect to t, and division of f(s) by s corresponds to integration of f (t) with respect to t. L { } df (t ) ⫽ sf (s) dt provided f (0) ⫽ 0 t f (s ) L ∫ f (t ) dt ⫽ s 0 66 cou9789x_ch03_032-068.indd 66 8/14/08 2:33:17 PM Confirming Pages CHAPTER 3 INVERSION BY PARTIAL FRACTIONS 67 When inverting an expansion containing a quadratic factor x(s) F1(s) Bs C/(s2 a s b ), we obtain C ⫺ aB ⫺at x (t ) ⫽ F1 (t ) ⫹ Be⫺at cos kt ⫹ e sin kt k where F1(t) is the result of inverting F1(s), B and C are the coefficients of the polynomial Bs C, in the numerator of the quadratic term, and a and k correspond to the roots of the quadratic, roots a kj. “Custom” input signals can be constructed using standard input signals. “Custom” signal 1.5 1 f 0.5 0 –0.5 –1 –1 0 1 2 time 3 4 5 We can consider this input signal to be constructed from several individual “pieces” as follows: 1 0.5 0.5 0.5 f 1.5 1 f 1.5 1 f 1.5 0 0 0 –0.5 –0.5 –0.5 –1 –1 0 1 2 time 3 4 Start with this piece . . . step. cou9789x_ch03_032-068.indd 67 5 –1 –1 0 1 2 time 3 4 5 Add this piece . . . ramp. –1 –1 0 1 2 time 3 4 5 Then add this piece . . . ramp. 8/14/08 2:33:20 PM cou9789x_ch03_032-068.indd 68 8/14/08 2:33:21 PM Confirming Pages PART II LINEAR OPEN-LOOP SYSTEMS cou9789x_ch04_069-098.indd 69 8/14/08 3:01:50 PM cou9789x_ch04_069-098.indd 70 8/14/08 3:01:50 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS B efore we discuss a complete control system, it is necessary to become familiar with the responses of some of the simple, basic systems that often are the building blocks of a control system. This chapter and the three that follow describe in detail the behavior of several basic systems and show that a great variety of physical systems can be represented by a combination of these basic systems. Some of the terms and conventions that have become well established in the field of automatic control will also be introduced. By the end of this part of the book, systems for which a transient must be calculated will be of high order and require calculations that are time-consuming if done by hand. Several software packages exist for streamlining this effort. We will use MATLAB as a tool throughout the book to demonstrate the applications of such software. 4.1 TRANSFER FUNCTION MERCURY THERMOMETER. We develop the transfer function for a first-order sys- tem by considering the unsteady-state behavior of an ordinary mercury-in-glass thermometer. A cross-sectional view of the bulb is shown in Fig. 4–1a. Consider the thermometer to be located in a flowing stream of fluid for which the temperature x varies with time. Our problem is to calculate the response or the time variation of the thermometer reading y for a particular change in x. (In order that the result of the analysis of the thermometer be general and therefore applicable to other first-order systems, the symbols x and y have been selected to represent surrounding temperature and thermometer reading, respectively.) The following assumptions will be used in this analysis: 1. All the resistance to heat transfer resides in the film surrounding the bulb (i.e., the resistance offered by the glass and mercury is neglected). 71 cou9789x_ch04_069-098.indd 71 8/14/08 3:01:50 PM Confirming Pages 72 PART 2 LINEAR OPEN-LOOP SYSTEMS Fluid y = thermometer reading x = fluid temperature Glass wall Mercury FIGURE 4–1a Cross-sectional view of themometer. Glass wall resistance Glass wall x x Fluid Fluid y Mercury y Mercury Film resistances Film resistance Resistance to heat transfer distributed throughout the system All resistance to heat transfer lumped in the fluid Actual temperature profile Lumped temperature profile FIGURE 4–1b Temperature profiles in themometer. 2. All the thermal capacity is in the mercury. Furthermore, at any instant the mercury assumes a uniform temperature throughout. (Making these first two assumptions is often referred to as the lumping of parameters because all the resistance is “lumped” into one location and all the capacitance into another. As shown in the analysis, these assumptions make it possible to represent the dynamics of the system by an ordinary differential equation. If such assumptions were not made, the analysis would lead to a partial differential equation, and the representation would be referred to as a distributed-parameter system. In Chap. 20, distributedparameter systems will be considered in detail. See the difference between the actual temperature and lumped temperature profiles in Fig. 4–1b.) cou9789x_ch04_069-098.indd 72 8/14/08 3:01:51 PM Confirming Pages CHAPTER 4 3. RESPONSE OF FIRST-ORDER SYSTEMS 73 The glass wall containing the mercury does not expand or contract during the transient response. (In an actual thermometer, the expansion of the wall has an additional effect on the response of the thermometer reading. The glass initially expands and the cavity containing the mercury grows, resulting in a mercury reading that initially falls. Once the mercury warms and expands, the reading increases. This is an example of an inverse response. Inverse responses will be discussed in greater detail later.) It is assumed that the thermometer is initially at steady state. This means that, before time 0, there is no change in temperature with time. At time 0, the thermometer will be subjected to some change in the surrounding temperature x(t). By applying the unsteady-state energy balance ( Input rate ) ( Output rate ) ( Rate of accumulation ) we get the result hA( x y) 0 mC dy dt (4.1) where A ⫽ surface area of bulb for heat transfer, ft2 C ⫽ heat capacity of mercury, Btu/(lb m · °F) m ⫽ mass of mercury in bulb, lb m t ⫽ time, h h ⫽ film coefficient of heat transfer, Btu/(ft2 · h · °F) For illustrative purposes, typical engineering units have been used. Equation (4.1) states that the rate of flow of heat through the film resistance surrounding the bulb causes the internal energy of the mercury to increase at the same rate. The increase in internal energy is manifested by a change in temperature and a corresponding expansion of mercury, which causes the mercury column, or “reading” of the thermometer, to rise. The coefficient h will depend on the flow rate and properties of the surrounding fluid and the dimensions of the bulb. We will assume that h is constant for a particular installation of the thermometer. Our analysis has resulted in Eq. (4.1), which is a first-order differential equation. Before we solve this equation by means of the Laplace transform, deviation variables will be introduced into Eq. (4.1). The reason for these new variables will soon become apparent. Prior to the change in x, the thermometer is at steady state and the derivative dy/dt is zero. For the steady-state condition, Eq. (4.1) may be written hA ( xs ys ) 0 t 0 (4.2) The subscript s is used to indicate that the variable is the steady-state value. Equation (4.2) simply states that ys ⫽ xs, or the thermometer reads the true, bath temperature. Subtracting Eq. (4.2) from Eq. (4.1) gives hA ( x xs ) ( y ys ) mC cou9789x_ch04_069-098.indd 73 d ( y ys ) dt (4.3) 8/14/08 3:01:52 PM Confirming Pages 74 PART 2 LINEAR OPEN-LOOP SYSTEMS Notice that d(y ⫺ ys)/dt ⫽ dy/dt because ys is a constant. If we define the deviation variables to be the differences between the variables and their steady-state values X x xs Y y ys then Eq. (4.3) becomes hA( X Y ) mC dY dt (4.4) If we let mC/hA ⫽ t, Eq. (4.4) becomes X Y t dY dt (4.5) The parameter t is called the time constant of the system and has the units of time. From above, we have Btu lb m ⋅ F mC [] [] h t Btu hA 2 2 ft ft ⋅ h ⋅ F ( lbm ) ( ) Remember, in Eq. (4.5), X is the input to the system (the bath temperature) and Y is the output from the system (the indicated thermometer temperature). Taking the Laplace transform of Eq. (4.5) gives X (s) Y (s) t sY (s) Y (0) t sY (s) (4.6) The Laplace transform of the differential equation results in an equation that is free of initial conditions because the initial values of X and Y are zero. Since we start from steady state, Y(0) must be zero, Y (0) y(0) ys ys ys 0 And X(0) is zero for the same reason. In control system engineering, we are primarily concerned with the deviations of system variables from their steady-state values. The use of deviation variables is, therefore, natural as well as convenient. Rearranging Eq. (4.6) as a ratio of Y(s) to X(s) gives Y (s ) 1 output X (s ) ts 1 input (4.7) The expression on the right side of Eq. (4.7) is called the transfer function of the system. It is the ratio of the Laplace transform of the deviation in thermometer reading (output) to the Laplace transform of the deviation in the surrounding temperature (input). In examining other physical systems, we usually attempt to obtain a transfer function. Any physical system for which the relation between Laplace transforms of input and output deviation variables is of the form given by Eq. (4.7) is called a first-order cou9789x_ch04_069-098.indd 74 8/14/08 3:01:52 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 75 system. Synonyms for first-order systems are first-order lag and single exponential stage. The naming of all these terms is motivated by the fact that Eq. (4.7) results from a first-order, linear differential equation, Eq. (4.5). In Chap. 5 we discuss a number of other physical systems that are first-order. To summarize the procedure for determining the transfer function for a process: Step 1. Write the appropriate balance equations (usually mass or energy balances for a chemical process). Step 2. Linearize terms if necessary (details on this step are given in Chap. 5). Step 3. Place balance equations in deviation variable form. Step 4. Laplace-transform the linear balance equations. Step 5. Solve the resulting transformed equations for the transfer function, the output divided by the input. This procedure is a very useful summary for developing the transfer function for a process. Standard Form for First-Order Transfer Functions The general form for a first-order system is t dy y K p x (t ) dt (4.8) where y is the output variable and x(t) is the input forcing function. The initial conditions are y(0) ys K p x (0) K p xs Introducing deviation variables gives X x xs Y y ys Eq. (4.8) becomes t dY Y K p X (t ) dt Y (0) 0 (4.9) Transforming Eq. (4.9), we obtain t sY (s) Y (s) K p X (s) and rearranging, we obtain the standard first-order transfer function Kp Y (s ) X (s ) ts 1 cou9789x_ch04_069-098.indd 75 (4.10) 8/14/08 3:01:52 PM Confirming Pages 76 PART 2 LINEAR OPEN-LOOP SYSTEMS The important characteristics of the standard form are as follows: • The denominator must be of the form t s ⫹ 1. • The coefficient of the s term in the denominator is the system time constant t. • The numerator is the steady-state gain Kp. Example 4.1. Place the following transfer function in standard first-order form, and identify the time constant and the steady state gain. Y (s ) 2 X (s ) s 1 3 Rearranging to standard form, we get Y (s ) 6 X (s ) 3s 1 Thus, the time constant is 3, and the steady-state gain is 6. The physical significance of the steady-state gain becomes clear if we let X(s) ⫽ 1/s, the unit-step function. Then Y(s) is given by Y (s ) 6 s(3s 1) The ultimate value of Y(t) is 6 lim[sY (s)] lim 6 Kp s→0 s → 0 3s 1 Thus the steady-state gain Kp is the steady-state value that the system attains after being disturbed by a unit-step input. It can be obtained by setting s ⫽ 0 in the transfer function. PROPERTIES OF TRANSFER FUNCTIONS. In general, a transfer function relates two variables in a physical process; one of these is the cause (forcing function or input variable), and the other is the effect (response or output variable). In terms of the example of the mercury thermometer, the surrounding temperature is the cause or input, whereas the thermometer reading is the effect or output. We may write Transfer function G (s) Y (s ) X (s ) where G(s) ⫽ symbol for transfer function X(s) ⫽ transform of forcing function or input, in deviation form Y(s) ⫽ transform of response or output, in deviation form cou9789x_ch04_069-098.indd 76 8/14/08 3:01:52 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 77 The transfer function completely describes the dynamic characteristics of the system. If we select a particular input variation X(t) for which the transform is X(s), the response of the system is simply Y (s ) G (s ) X (s ) (4.11) By taking the inverse of Y(s), we get Y(t), the response of the system. The transfer function results from a linear differential equation; therefore, the principle of superposition is applicable. This means that the transformed response of a system with transfer function G(s) to a forcing function X (s) a1 X1 (s) a2 X 2 (s) where X1 and X2 are particular forcing functions and a1 and a2 are constants, is Y (s ) G (s ) X (s ) a1G (s) X1 (s) a2G (s) X 2 (s) a1Y1 (s) a2Y2 (s) where Y1 (s) and Y2 (s) are the responses to X1 and X2 alone, respectively. For example, the response of the mercury thermometer to a sudden change in surrounding temperature of 10°F is simply twice the response to a sudden change of 5°F in surrounding temperature. The functional relationship contained in a transfer function is often expressed by a block diagram representation, as shown in Fig. 4–2. The arrow entering the box is the forcing function or input variable, and the arrow leavX(s) Y(s) G(s) ing the box is the response or output variable. The transfer function is placed inside the box. We state that the transfer function G(s) in the Forcing box “operates” on the input function X(s) to proFunction Response Transfer duce an output function Y(s). The usefulness of Function Input Output the block diagram will be appreciated in Chap. 8, when a complete control system containing FIGURE 4–2 Block diagram. several blocks is analyzed. 4.2 TRANSIENT RESPONSE Now that the transfer function of a first-order system has been established, we can easily obtain its transient response to any forcing function. Since this type of system occurs so frequently in practice, it is worthwhile to study its response to several common forcing functions: step, impulse, ramp, and sinusoidal. These forcing functions have been found to be very useful in theoretical and experimental aspects of process control. They will be used extensively in our studies, so let’s explore each before we study the transient response of the first-order system to these forcing functions. cou9789x_ch04_069-098.indd 77 8/14/08 3:01:53 PM Confirming Pages 78 PART 2 LINEAR OPEN-LOOP SYSTEMS 4.3 FORCING FUNCTIONS STEP FUNCTION. Mathematically, the step function of magnitude A can be expres- sed as X(t) X (t ) Au(t ) X = 0; t < 0 X = A; t > 0 A X (s) = s A O 0 t FIGURE 4–3 Step input. where u(t) is the unit-step function defined in Chap. 2. A graphical representation is shown in Fig. 4–3. The transform of this function is X(s) ⫽ A/s. A step function can be approximated very closely in practice. For example, a step change in flow rate can be obtained by the sudden opening of a valve. IMPULSE FUNCTION. Mathematically, the impulse function of magnitude A is defined as X ( t ) Ad ( t ) where d (t) is the unit-impulse function defined and discussed in App. 3A. A graphical representation of this function, before the limit is taken, is A shown in Fig. 4–4. b The true impulse function, obtained by letting b → 0 in Fig. 4–4, has a Laplace transform of A. It is used more frequently as a mathematical aid O 0 b t than as an actual input to a physical system. For some systems it is difficult even to approximate FIGURE 4–4 an impulse forcing function. For this reason the Impulse function. representation of Fig. 4–4 is valuable, since this form can usually be approximated physically by application and removal of a step function. If the time duration b is sufficiently small, we will see in Chap. 5 that the forcing function of Fig. 4–4 gives a response that closely resembles the response to a true impulse. In this sense, we often justify the use of A as the Laplace transform of the physically realizable X forcing function of Fig. 4–4. X(t) X = 0; t < 0 A X= ;0<t<b b X = 0; t > b lim X(t) = Aδ(t) b→0 L{Aδ(t)} = A X = 0; t < 0 X = bt; t > 0 X(s) = b/s2 RAMP FUNCTION. This function increases linearly with time and is described by the equations Slope = b 0 0 FIGURE 4–5 Ramp function. cou9789x_ch04_069-098.indd 78 t X 0 X bt t 0 t 0 The ramp is shown graphically in Fig. 4–5. The transform of the ramp forcing function is X(s) ⫽ b/s2. We might, for example, desire to ramp up the temperature of an oven by 10°F/min. This would be an example of a ramp function. 8/14/08 3:01:58 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 79 SINUSOIDAL INPUT. This function is represented mathematically by the equations X 0 X Asin t X = 0; t < 0 X = A sin wt; t > 0 Aw X(s) = 2 s + w2 X(t) Period = 2p w A 0 0 t 0 t 0 t FIGURE 4–6 Sinusoidal input. where A is the amplitude and w is the radian frequency. The radian frequency w is related to the frequency f in cycles per unit time by w ⫽ 2p f. Figure 4–6 shows the graphical representation of this function. The transform is X(s) ⫽ Aw /(s2 ⫹ w 2). This forcing function forms the basis of an important branch of control theory known as frequency response. Historically, a large segment of the development of control theory was based on frequency-response methods, which will be presented in Chaps. 15 and 16. Physically, it is more difficult to obtain a sinusoidal forcing function in most process variables than to obtain a step function. This completes the discussion of some of the common forcing functions. We now devote our attention to the transient response of the first-order system to each of the forcing functions just discussed. 4.4 STEP RESPONSE If a step change of magnitude A is introduced into a first-order system, the transform of X(t) is A X (s ) (4.12) s The transfer function, which is given by Eq. (4.7), is Y (s ) 1 X (s ) ts 1 (4.7) Combining Eqs. (4.7) and (4.12) gives Y (s ) A 1 s ts 1 (4.13) This can be expanded by partial fractions to give Y (s ) cou9789x_ch04_069-098.indd 79 A /t C1 C2 s(s 1/t ) s s 1/t (4.14) 8/14/08 3:01:59 PM Confirming Pages 80 PART 2 LINEAR OPEN-LOOP SYSTEMS Solving for the constants C1 and C2 by the techniques covered in Chap. 3 gives C1 ⫽ A and C2 ⫽ ⫺A. Inserting these constants into Eq. (4.14) and taking the inverse transform give the time response for Y: Y (t ) 0 t 0 ( Y (t ) A 1 e t /t ) t 0 (4.15) Hereafter, for the sake of brevity, it will be understood that, as in Eq. (4.15), the response is zero before t ⫽ 0. Equation (4.15) is plotted in Fig. 4–7 in terms of the dimensionless quantities Y(t)/A and t/t. (Note that if we refer to the standard form for a first-order system, Eq. (4.10), Kp ⫽ A in this case.) Having obtained the step response, Eq. (4.15), from a purely mathematical approach, we should consider whether the result seems to be correct from physical principles. Immediately after the thermometer is placed in the new environment, the temperature difference between the mercury in the bulb and the bath temperature is at its maximum value. With our simple lumped-parameter model, we should expect the flow of heat to commence immediately, with the result that the mercury temperature rises, causing a corresponding rise in the column of mercury. As the mercury temperature rises, the driving force causing heat to flow into the mercury will diminish, with the result that the mercury temperature changes at a slower rate as time proceeds. We see that this description of the response based on physical grounds does agree with the response given by Eq. (4–15) and shown graphically in Fig. 4–7. Several features of this response are worth remembering: 1. Y(t) A 2. The value of Y(t) reaches 63.2 percent of its ultimate value when the time elapsed is equal to one time constant t. When the time elapsed is 2t, 3t, and 4t, the percent response is 86.5, 95, and 98, respectively. From these facts, one can consider the response essentially completed in three to four time constants. One can show from Eq. (4.15) that the slope of the response curve at the origin in Fig. 4–7 is 1. This means that if the initial rate of change of Y(t) were maintained, the response would be complete in one time constant. (See the dotted line in Fig. 4–7.) 3. A consequence of the principle of super1.0 position is that the response to a step input of any magnitude A may be obtained 0.8 directly from Fig. 4–7 by multiplying the ordinate by A. Figure 4–7 actually gives 0.6 the response to a unit-step function input, from which all other step responses are 0.4 derived by superposition. 0.2 0 0 1 2 3 4 5 t/t FIGURE 4–7 Response of a first-order system to a step input. cou9789x_ch04_069-098.indd 80 These results for the step response of a firstorder system will now be applied to the following example. Example 4.2. A thermometer having a time constant of 0.1 min is at a steady-state 8/14/08 3:01:59 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 81 temperature of 90°F. At time t ⫽ 0, the thermometer is placed in a temperature bath maintained at 100°F. Determine the time needed for the thermometer to read 98°F. (Note: The time constant given in this problem applies to the thermometer when it is located in the temperature bath. The time constant for the thermometer in air will be considerably different from that given because of the lower heattransfer coefficient in air.) In terms of symbols used in this chapter, we have xs 90 F t 0 . 1 min A 10 F The ultimate thermometer reading will, of course, be 100°F, and the ultimate value of the deviation variable Y(⬁) is 10°F. When the thermometer reads 98°F, Y(t) ⫽ 8°F. Substituting into Eq. (4.12) the appropriate values of Y, A, and t gives ( 8 10 1 et / 0.1 ) Solving this equation for t yields t 0 . 161 min The same result can also be obtained by referring to Fig. 4–7, where it is seen that Y/A ⫽ 0.8 at t/t ⫽ 1.6. Using MATLAB to Obtain the Response of a First-Order System to a Step Function The transform of the response is 10/s(0.1s ⫹ 1). We can simulate that in MATLAB by defining a system using the numerator and denominator of the response: num=[10]; den=[0.1 1]; We then use the step function in MATLAB to obtain the response (Fig. 4–8). step(num,den) To obtain numerical values for the plot, we use the tf (transfer function command). sys=tf(num,den) Transfer function: 10 0.1 s + 1 % Assigns the variables, temp and t, to the response. data=[t,temp] % Concatenates time and temperature into one matrix and displays them. [temp,t]=step(sys); cou9789x_ch04_069-098.indd 81 8/14/08 3:01:59 PM Confirming Pages 82 PART 2 LINEAR OPEN-LOOP SYSTEMS 10 9 8 7 T 6 5 4 3 2 1 0 0 0.1 0.2 0.3 0.4 Time (min) 0.5 0.6 FIGURE 4–8 Step response of thermometer in Example 4.1. data = 0 0.0055 0.0110 0.0166 0.0221 0.0276 0.0331 0.0387 0.0442 0.0497 ...... 0.1380 0.1436 0.1491 0.1546 0.1601 0.1656 0.1712 0.1767 ...... 0 0.5372 1.0455 1.5265 1.9817 2.4124 2.8200 3.2057 3.5707 3.9161 ...... 7.4851 7.6202 7.7481 7.8690 7.9835 8.0918 8.1943 8.2913 ...... 0.5908 0.5963 9.9728 9.9743 cou9789x_ch04_069-098.indd 82 temp=8.0 at approximately t = 0.16 min 8/14/08 3:02:00 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 83 Using Simulink to Obtain the Response of a First-Order System to a Step Function We constructed the block diagram for the system using Simulink (Fig. 4–9). The simulation was run for 0.6 min, and the Scope output is shown in Fig. 4–10. The data were also exported to the MATLAB workspace and graphed in Fig. 4–11 using the plot command. [plot (ScopeData.time, ScopeData.signals.values)] 1 0.1s + 1 Step Thermometer Scope FIGURE 4–9 Simulink block diagram for thermometer. FIGURE 4–10 Thermometer response to step input from Simulink scope. 10 9 8 7 T 6 5 4 3 2 1 0 0 0.1 0.2 0.3 Time 0.4 0.5 0.6 FIGURE 4–11 Thermometer response to step input using MATLAB plot command. cou9789x_ch04_069-098.indd 83 8/14/08 3:02:00 PM Confirming Pages 84 PART 2 LINEAR OPEN-LOOP SYSTEMS Note that the Simulink results are the same as we obtained previously by hand calculation and with MATLAB. The speed of the response of a first-order system is determined by the time constant for the system. Consider the following first-order system disturbed by a step input (Fig. 4–12). The response of a first-order system for several values of t is shown in Fig. 4–13. It can be seen that as t increases, it takes longer for the system to respond to the step disturbance. 1 tau. s + 1 Step Scope First-Order System tau = 2,4,6,8,10 min FIGURE 4–12 Simulink model for examining the effect of t on the step response. 1 0.9 0.8 increasing t slows the response 0.7 63.2% of the ultimate value Response 0.6 0.5 This curve is t = 8, note that it passes 63.2% of the ultimate response at t = t = 8 0.4 0.3 0.2 0.1 0 0 2 4 6 8 10 Time 12 14 16 18 20 FIGURE 4–13 Effect of t on the step response of a first-order system. cou9789x_ch04_069-098.indd 84 8/14/08 3:02:00 PM Confirming Pages CHAPTER 4 4.5 RESPONSE OF FIRST-ORDER SYSTEMS 85 IMPULSE RESPONSE The impulse response of a first-order system will now be developed. Anticipating the use of superposition, we consider a unit impulse for which the Laplace transform is X (s ) 1 (4.16) Combining this with the transfer function for a first-order system, which is given by Eq. (4.7), results in Y (s ) 1 ts 1 (4.17) Y (s ) 1/t s 1/t (4.18) This may be rearranged to The inverse of Y(s) can be found directly from the table of transforms and can be written in the form t Y (t ) et /t (4.19) A plot of this response is shown in Fig. 4–14 in terms of the variables t/t and t Y(t). The response to an impulse of magnitude A is obtained, as usual, by multiplying t Y(t) from Fig. 4–14 by A/t. Notice that the response rises immediately to 1.0 and then decays exponentially. Such an abrupt rise is, of course, physically impossible, but as we will see in Chap. 5, it is approached by the response to a finite pulse of narrow width, such as that of Fig. 4–4. 1.0 t Y(t) 0.8 0.6 0.4 0.2 0 0 1 2 3 4 5 t/t FIGURE 4–14 Unit-impulse response of a first-order system. cou9789x_ch04_069-098.indd 85 8/14/08 3:02:01 PM Confirming Pages 86 PART 2 LINEAR OPEN-LOOP SYSTEMS Using MATLAB to Generate the Impulse Response to a First-Order System num=[1]; den=[1 1]; sys=tf(num,den) Transfer function: 1 s + 1 [x,y]=impulse(sys); % Assigns the variables x and y to the response. data=[x,y] % Concatenates x and y into one matrix and displays them. data = 0 0.0552 0.1104 0.1656 0.2209 0.2761 0.3313 0.3865 0.4417 0.4969 ...... 5.0245 5.0797 5.1350 5.1902 5.2454 5.3006 5.3558 5.4110 5.4662 5.5215 5.5767 5.6319 5.6871 5.7423 5.7975 5.8527 5.9080 5.9632 cou9789x_ch04_069-098.indd 86 1.0000 0.9463 0.8954 0.8473 0.8018 0.7588 0.7180 0.6794 0.6429 0.6084 ...... 0.0066 0.0062 0.0059 0.0056 0.0053 0.0050 0.0047 0.0045 0.0042 0.0040 0.0038 0.0036 0.0034 0.0032 0.0030 0.0029 0.0027 0.0026 8/14/08 3:02:01 PM Confirming Pages CHAPTER 4 plot=[x,y] RESPONSE OF FIRST-ORDER SYSTEMS 87 % The result of this command is shown in Fig. 4–15. 1 0.9 0.8 0.7 Y 0.6 0.5 0.4 0.3 0.2 0.1 0 0 1 2 3 X 4 5 6 FIGURE 4–15 Impulse response of a first-order system using MATLAB. 4.6 RAMP RESPONSE For a ramp input of x(t) ⫽ bt, where X(s) ⫽ b/s2, the output is Y (s ) b s (t s 1) 2 Rearranging and using partial fractions yield. Y (s ) b b /t b bt bt 2 2 2 s s 2 (t s 1) s ( s 1/t ) s s 1/t ( Y (t ) bt b t 1 e − t / t ) b(t t ) b t e − t / t A plot of this response is shown in Fig. 4–16. 4.7 SINUSOIDAL RESPONSE To investigate the response of a first-order system to a sinusoidal forcing function, the example of the mercury thermometer will be considered again. Consider a thermometer to be in equilibrium with a temperature bath at temperature xs. At some time t ⫽ 0, the bath temperature begins to vary according to the relationship x xs A sin w t cou9789x_ch04_069-098.indd 87 t 0 (4.20) 8/14/08 3:02:01 PM Confirming Pages 88 PART 2 LINEAR OPEN-LOOP SYSTEMS 5 After an initial transient period, the response is parallel with input. 4.5 4 Steady-state difference between input and output (after transient) is b*t 3.5 Y/bKp 3 2.5 2 bt Output lags input by t 1.5 1 Input 0.5 Output 0 0 1 2 3 4 5 t /t FIGURE 4–16 Response of a first-order system to a ramp input. x ⫽ temperature of bath xs ⫽ temperature of bath before sinusoidal disturbance is applied A ⫽ amplitude of variation in temperature w ⫽ radian frequency, rad/time In anticipation of a simple result, we introduce a deviation variable X which is defined as where X x xs (4.21) Using this new variable in Eq. (4.20) gives X A sin w t (4.22) By referring to a table of transforms, the transform of Eq. (4.22) is X (s ) Aw s2 w 2 (4.23) Combining Eqs. (4.7) and (4.23) to eliminate X(s) yields Y (s ) Aw 1/t 2 s 1/ t s w 2 (4.24) This equation can be solved for Y(t) by means of a partial fraction expansion, as described in Chap. 3. The result is Y (t ) cou9789x_ch04_069-098.indd 88 Awt et /t Awt A 2 2 cos w t 2 2 sinw t 2 2 t w 1 t w 1 t w 1 (4.25) 8/14/08 3:02:01 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 89 Equation (4.25) can be written in another form by using the trigonometric identity p cos B q sin B r sin ( B q ) (4.26) where r p2 q 2 tan q p q Applying the identity of Eq. (4.26) to Eq. (4.25) gives Y (t ) Awt A sin(w t f ) et / t 2 2 t w 1 t w2 1 2 (4.27) where f tan1 (wt ) As t → ⬁, the first term on the right side of Eq. (4.27) vanishes and leaves only the ultimate periodic solution, which is sometimes called the steady-state solution Y (t ) s A t 2w 2 1 sin( wt f ) (4.28) By comparing Eq. (4.25) for the input forcing function with Eq. (4.28) for the ultimate periodic response, we see that 1. The output is a sine wave with a frequency w equal to that of the input signal. 2. The ratio of output amplitude to input amplitude is 1 t 2 2 1 . This ratio is always smaller than 1. We often state this by saying that the signal is attenuated. The output lags behind the input by an angle f. It is clear that lag occurs, for the sign of f is always negative.* 3. *By convention, the output sinusoid lags the input sinusoid if f in Eq. (4.28) is negative. In terms of a recording of input and output, this means that the input peak occurs before the output peak. If f is positive in Eq. (4.28), the system exhibits phase lead, or the output leads the input. In this book we always use the term phase angle (f) and interpret whether there is lag or lead by the convention f 0 phase lag f 0 phase lead For a particular system for which the time constant t is a fixed quantity, it is seen from Eq. (4.28) that the attenuation of amplitude and the phase angle f depend only on the frequency w. The attenuation and phase lag increase with frequency, but the phase lag can never exceed 90° and approaches this value asymptotically. cou9789x_ch04_069-098.indd 89 8/14/08 3:02:02 PM Confirming Pages 90 PART 2 LINEAR OPEN-LOOP SYSTEMS Sinusoidal Response of a First-Order System Using MATLAB For a first-order transfer function 1/(t s ⫹ 1), determine the response to an input function x ⫽ sin (4t). Plot the input and output on the same set of axes, and indicate the transient portion as well as the steady-state portion of the response. num=[1]; % Set up the transfer function as before… den=[1 1]; sys=tf(num,den) Transfer function: 1 s + 1 t=0:0.1:10; u=sin(4*t); z=lsim(sys,u,t); [plot(t,z,t,u)] [hold on] [w=0.2353*exp(-t)+0.2425;] [plot(t,w)] [q=0.2425;] [plot(t,q)] % Sets up a time vector to be used for the sine wave input. % Defines the sine wave input function. % Invokes the linear simulator within MATLAB and assigns the output to z. % Plots the input and output on the same axes. % Holds the axes for further graphs. % Transient envelope. % Plots the transient envelope. % Peak height for the ultimate periodic response. % Plots the steady-state peak height The resulting MATLAB graph is shown in Fig. 4–17. 1 Input sine wave 0.8 0.6 Transient envelope 0.4 Y 0.2 0 –0.2 Output sine wave –0.4 –0.6 –0.8 –1 0 1 2 3 4 5 Time 6 7 8 9 10 FIGURE 4–17 Response of a first-order system to a sine wave using MATLAB. cou9789x_ch04_069-098.indd 90 8/14/08 3:02:02 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 91 Note that the system response has pretty much settled down to the steady periodic output wave after approximately 4 min (or 4 time constants). By using Eq. (4.27), the analytic solution for the response is Y 0 . 2353et 0 . 2425 sin(4 t 1. 326 rad) Note that the phase angle is ⫺1.326 rad (⫺75.96°), and the response is nearly peaking as the input is zero and vice versa. The sinusoidal response is interpreted in terms of the mercury thermometer by the following example. Example 4.3. A mercury thermometer having a time constant of 0.1 min is placed in a temperature bath at 100°F and allowed to come to equilibrium with the bath. At time t ⫽ 0, the temperature of the bath begins to vary sinusoidally about its average temperature of 100°F with an amplitude of 2°F. If the frequency of oscillation is 10/p cycles/min, plot the ultimate response of the thermometer reading as a function of time. What is the phase lag? In terms of the symbols used in this chapter t 0 . 1 min xs 100 F A 2F 10 cycles/min p 10 w 2p f 2p 20 rad/min p f From Eq. (4.28), the amplitude of the response and the phase angle are calculated; thus A 2 2 t w 1 2 0 . 896 F 4 1 f tan1 2 63 . 5 1. 11 rad or Phase lag 63 . 5 The response of the thermometer is therefore Y (t ) 0 . 896 sin (20 t 1 . 11) or cou9789x_ch04_069-098.indd 91 y(t ) 100 0 . 896 sin (20 t 1 . 11) 8/14/08 3:02:02 PM Confirming Pages 92 PART 2 LINEAR OPEN-LOOP SYSTEMS To obtain the lag in terms of time rather than angle, we proceed as follows: A frequency of 10/p cycles/min means that a complete cycle (peak to peak) occurs in (10/p )⫺1 min. Since one cycle is equivalent to 360° and the lag is 63.5°, the time corresponding to this lag is Lag time 63 . 5 360 ( time for 1 cycle ) or 63 . 5 p Lag time 0 . 0555 min 360 10 thus, y(t ) 100 0 . 896 sin[20(t 0 . 0555 min)] In general, the lag in units of time is given by Lag time f 360 f when f is expressed in degrees. The response of the thermometer reading and the variation in bath temperature are shown in Fig. 4–18. Note that the response shown in this figure holds only after sufficient time has elapsed for the nonperiodic term of Eq. (4.27) to become negligible. For all practical purposes this term becomes negligible after a time equal to about 3t. If the response were desired beginning from the time the bath temperature begins to oscillate, it would be necessary to plot the complete response as given by Eq. (4.27). Period = 0.314 min lag = 0.056 min 102.0 100.9 100.0 99.1 98.0 Bath temperature Transient Thermometer temperature Ultimate periodic response 0 t (min) FIGURE 4–18 Response of a thermometer in Example 4.3. cou9789x_ch04_069-098.indd 92 8/14/08 3:02:03 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 93 SUMMARY In this chapter several basic concepts and definitions of control theory have been introduced. These include input variable, output variable, deviation variable, transfer function, response, time constant, first-order system, block diagram, attenuation, and phase lag. Each of these ideas arose naturally in the study of the dynamics of the first-order system, which was the basic subject matter of the chapter. As might be expected, the concepts will find frequent use in succeeding chapters. In addition to introducing new concepts, we have listed the response of the firstorder system to forcing functions of major interest. This information on the dynamic behavior of the first-order system will be of significant value in the remainder of our studies. PROBLEMS 4.1. A thermometer having a time constant of 0.2 min is placed in a temperature bath, and after the thermometer comes to equilibrium with the bath, the temperature of the bath is increased linearly with time at a rate of 1°/min. Find the difference between the indicated temperature and the bath temperature. (a) 0.1 min after the change in temperature begins (b) 1.0 min after the change in temperature begins (c) What is the maximum deviation between indicated temperature and bath temperature, and when does it occur? (d) Plot the forcing function and response on the same graph. After a long enough time, by how many minutes does the response lag the input? 4.2. A mercury thermometer bulb is 12 in long by 81 -in diameter. The glass envelope is very thin. Calculate the time constant in water flowing at 10 ft/s at a temperature of 100°F. In your solution, give a summary that includes (a) Assumptions used (b) Source of data (c) Results 4.3. Given: a system with the transfer function Y(s)/X(s) ⫽ (T1 s ⫹ 1)/(T2 s ⫹ 1). Find Y(t) if X(t) is a unit-step function. If T1/T2 ⫽ 5, sketch Y(t) versus t/T2. Show the numerical values of minimum, maximum, and ultimate values that may occur during the transient. Check these by using the initial-value and final-value theorems of App. 3A. 4.4. A thermometer having first-order dynamics with a time constant of 1 min is placed in a temperature bath at 100°F. After the thermometer reaches steady state, it is suddenly placed in a bath at 110°F at t ⫽ 0 and left there for 1 min, after which it is immediately returned to the bath at 100°F. (a) Draw a sketch showing the variation of the thermometer reading with time. (b) Calculate the thermometer reading at t ⫽ 0.5 min and at t ⫽ 2.0 min. 4.5. Repeat Prob. 4.4 if the thermometer is in the 110°F bath for only 10 s. 4.6. A mercury thermometer, which has been on a table for some time, is registering the room temperature, 75°F. Suddenly, it is placed in a 400°F oil bath. The following data are obtained for the response of the thermometer. cou9789x_ch04_069-098.indd 93 8/14/08 3:02:03 PM Confirming Pages 94 PART 2 LINEAR OPEN-LOOP SYSTEMS Give two independent estimates of the thermometer time constant. Time, s Thermometer reading, °F 0 1 2.5 5 8 10 15 30 75 107 140 205 244 282 328 385 4.7. Rewrite the sinusoidal response of a firstorder system [Eq. (4.27)] in terms of a cosine wave. Reexpress the forcing function [Eq. (4.22)] as a cosine wave, and compute the phase difference between input and output cosine waves. 4.8. The mercury thermometer of Prob. 4.6 is again allowed to come to equilibrium in the room air at 75°F. Then it is placed in the 400°F oil bath for a length of time less than 1 s and quickly removed from the bath and reexposed to the 75°F ambient conditions. It may be estimated that the heat-transfer coefficient to the thermometer in air is one-fifth that in the oil bath. If 10 s after the thermometer is removed from the bath it reads 98°F, estimate the length of time that the thermometer was in the bath. 4.9. A thermometer having a time constant of 1 min is initially at 50°C. It is immersed in a bath maintained at 100°C at t ⫽ 0. Determine the temperature reading at t ⫽ 1.2 min. 4.10. In Prob. 4.9, if at t ⫽ 1.5 min the thermometer is removed from the bath and put in a bath at 75°C, determine the maximum temperature indicated by the thermometer. What will be the indicated temperature at t ⫽ 20 min? 4.11. A process of unknown transfer function is subjected to a unit-impulse input. The output of the process is measured accurately and is found to be represented by the function y(t) ⫽ te⫺t. Determine the unit-step response of this process. 4.12. The temperature of an oven being heated using a pulsed resistance heater varies as T 120 5 cos ( 25t 30 ) where t is the time in seconds. The temperature of the oven is being measured with a thermocouple having a time constant of 5 s. (a) What are the maximum and minimum temperatures indicated by the thermocouple? (b) What is the maximum difference between the actual temperature and the indicated temperature? (c) What is the time lag between the actual temperature and the indicated temperature? 4.13. The temperature of a experimental heated enclosure is being ramped up from 80 to 450°F at the rate of 20°F/min. A thermocouple, embedded in a thermowell for protection, is being used to monitor the oven temperature. The thermocouple has a time constant of 6 s. (a) At t ⫽ 10 min, what is the difference between the actual temperature and the temperature indicated by the thermocouple? What is it at 60 min? (b) When the thermocouple indicates 450°F, the heater will begin to modulate and maintain the temperature at the desired 450°F. What is the actual oven temperature when the thermocouple first indicates 450°F? X(t) FIGURE P4–14 cou9789x_ch04_069-098.indd 94 2 4s + 1 Y(t) 4.14. For the transfer function in Fig. P4–14, the response Y(t) is sinusoidal. The amplitude of the output wave is 0.6 and it lags behind the input by 1.5 min. Find X(t). Note: the time constant in the transfer function is in minutes. 8/14/08 3:02:03 PM Confirming Pages CHAPTER 4 RESPONSE OF FIRST-ORDER SYSTEMS 95 4.15. The graph in Fig. P4–15 is the response of a suspected first-order process to an impulse function of magnitude 3. Determine the transfer function G(s) of the unknown process. 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 FIGURE P4–15 Time (min) Level (ft) 0 0.138 0.2761 0.4141 0.5521 0.6902 0.8282 0.9663 1.1043 1.2423 1.3804 1.5184 1.6564 1.7945 1.9325 2.0705 2.2086 2.3466 2.4847 2.6227 2.7607 ............ 14.3558 14.4938 14.6319 14.7699 4.8 5.3673 5.9041 6.412 6.8927 7.3475 7.7779 8.1852 8.5706 8.9354 9.2805 9.6071 9.9161 10.2085 10.4853 10.7471 10.9949 11.2294 11.4513 11.6612 11.8599 ............ 15.3261 15.328 15.3297 15.3313 Level (ft) 4.16. The level in a tank responds as a first-order system with changes in the inlet flow. Given the following level versus time data that were gathered (Fig. P4–16) after the inlet flow was 16 15 14 13 12 11 10 9 8 7 6 5 4 0 2 4 6 8 10 Time (min) 12 14 16 Inlet flow 1.5 gal/min → 4.8 gal/min at time = 0 LI Note: LI = level indicator FIGURE P4–16 cou9789x_ch04_069-098.indd 95 8/14/08 3:02:04 PM Confirming Pages 96 PART 2 LINEAR OPEN-LOOP SYSTEMS increased quickly from 1.5 to 4.8 gal/min, determine the transfer function that relates the height in the tank to the inlet flow. Be sure to use deviation variables and include units on the steady-state gain and the time constant. Salt Water @ 5% salt 4.17. A simple mixing process follows first-order behavior. A 200-gal mixing tank process, initially at steady state, is shown in Fig. P4–17. At time t ⫽ 0, the inlet flow is switched from 5% salt to freshwater. What does the inlet flow rate need to be to reduce the exit concentration to less than 0.5% in 30 min? Volume = 200 gal FIGURE P4–17 4.18. Joe, the maintenance man, dumps the contents of a 55-gal drum of water into the tank process shown below. 40 gal/min 5 ft Current volume = 200 gal of water 40 gal/min 3 ft FIGURE P4–18 (a) Will the tank overflow? (b) Plot the height as f(t), starting at t ⫽ 0, the time of the dump. (c) Plot the output flow as f(t), starting at t ⫽ 0, the time of the dump. NOTE: The output flow is proportional to the height of fluid in the tank. cou9789x_ch04_069-098.indd 96 8/14/08 3:02:05 PM Rev. Confirming Pages CHAPTER 4 CAPSULE SUMMARY Standard form for first-order system transfer function: G (s ) Kp Y (s ) X (s) ts 1 where Kp is the steady-state gain and is the time constant (having units of time). Note the 1 in the denominator when the transfer function is in standard form. Deviation variables: The difference between the process system variables and their steady-state values. When transfer functions are used, deviation values are always used. The convenience and utility of deviation variables lie in the fact that their initial values are most often zero. X x xs Y y ys Procedure for determining the transfer function for a process: Step 1. Write the appropriate balance equations (usually mass or energy balances for a chemical process). Step 2. Linearize terms if necessary (details on this step are given in Chap. 5). Step 3. Place balance equations in deviation variable form. Step 4. Laplace-transform the linear balance equations. Step 5. Solve the resulting transformed equations for the transfer function, the output divided by the input. Block diagrams: Graphically depict the relationship between the input variable, the transfer function, and the output variables Y(s) X(s)G(s). We always use transformed deviation variables with block diagrams. X(s) Y(s) G(s) Forcing function Input Transfer function Response Output Standard responses of first-order systems to common inputs: G (s ) Kp Y (s ) X (s) ts 1 97 cou9789x_ch04_069-098.indd 97 8/30/08 3:41:13 PM Confirming Pages 98 PART 2 LINEAR OPEN-LOOP SYSTEMS Input Output X(t) X(s) Step u(t) 1 s Kp s(t s 1) Impulse ␦ (t) 1 Kp ts 1 Ramp btu(t) Sinusoid Y(s) b s2 u(t) A sin (w t) Y(t) Kp (1 ⫺ e⫺t/) K p t / t e t Kp [bt ⫺ b (1 ⫺ e⫺t/t)] bK p 2 s (t s 1) Aw s2 w 2 ( Aw K p AK pwt s 2 w 2 (t s 1) 1 (wt )2 ) et / t AK p 1 (wt )2 sin w t tan1 (wt ) Key features of standard responses of first-order systems to common inputs Step Response of First-Order System Ultimate Value Impulse Response of a First-Order System 1 1 0.9 0.9 0.8 0.8 0.7 Response is 63.2% complete at t = t Y*t/Kp 0.6 Y/Kp 0.5 0.4 0.6 0.5 Response is 63.2% complete at t = t 0.4 0.3 (Initial “jump” has decayed to 36.8%) 0.3 0.2 Initial slope intersects ultimate value at t = t 0.2 0.1 0 Initial “jump” is to Kp /t 0.7 0.1 0 1 2 3 0 4 0 1 2 t/t t/t Sinusoidal Response of a First-Order System 4 Response of First-Order System to Ramp Input 1.5 5 phase lag = tan −1(wt) 1 After an initial transient period, the response is parallel with input. 4.5 Ratio =1/ 1+(wt )2 4 Steady-state difference between input and output (after transient) is b*t. 3.5 0.5 3 Y/bKp Y/AKp 3 0 2.5 2 −0.5 1 −1 −1.5 After an initial transient period, the response is periodic with the same frequency 5 10 15 Time/t cou9789x_ch04_069-098.indd 98 b*t 1.5 Output 0.5 20 0 Output lags input byt Input 0 1 2 3 4 5 Time/t 8/14/08 3:02:11 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS I n the first part of this chapter, we will consider several physical systems that can be represented by a first-order transfer function. In the second part, a method for approximating the dynamic response of a nonlinear system by a linear response will be presented. This approximation is called linearization. 5.1 EXAMPLES OF FIRST-ORDER SYSTEMS Liquid Level Consider the system shown in Fig. 5–1, which consists of a tank of uniform crosssectional area A to which is attached a flow resistance R such as a valve, a pipe, or a weir. Assume that qo, the volumetric flow rate (volume/time) through the resistance, is related to the head h by the linear relationship h (5.1) R A resistance that has this linear relationship between flow and head is referred to as a linear resistance. (A pipe is a linear resistance if the flow is in the laminar range. A specially contoured weir, called a Sutro weir, produces a linear head-flow relationship. Turbulent flow through pipes and valves is generally proportional to h. Flow through weirs having simple geometric shapes can be expressed as Khn, where K and n are positive constants. For example, the flow through a rectangular weir is proportional to h3/2.) A time-varying volumetric flow q of liquid of constant density r enters the tank. Determine the transfer function that relates head to flow. We can analyze this system by writing a transient mass balance around the tank: qo Rate of accumulation Rate of Rate of mass flow in − mass flow out = of mass in tank 99 cou9789x_ch05_099-122.indd 99 8/14/08 3:19:47 PM Confirming Pages 100 PART 2 LINEAR OPEN-LOOP SYSTEMS In terms of the variables used in this analysis, the mass balance becomes q (t) d ( rAh) dt dh q(t ) qo (t ) A dt r q(t ) r qo (t ) h (t) R qo(t) (5.2) FIGURE 5–1 Liquid-level system. Combining Eqs. (5.1) and (5.2) to eliminate qo(t) gives the following linear differential equation: q h dh A R dt (5.3) We will introduce deviation variables into the analysis before proceeding to the transfer function. Initially, the process is operating at steady state, which means that dh/dt 0 and we can write Eq. (5.3) as qs hs 0 R (5.4) where the subscript s has been used to indicate the steady-state value of the variable. Subtracting Eq. (5.4) from Eq. (5.3) gives q qs 1 d (h hs ) ( h hs ) A R dt (5.5) If we define the deviation variables as Q q qs H h hs then Eq. (5.5) can be written 1 dH HA R dt Q (5.6) Taking the transform of Eq. (5.6) gives Q (s ) 1 H (s) AsH (s) R (5.7) Notice that H(0) is zero, and therefore the transform of dH/dt is simply sH(s). Equation (5.7) can be rearranged into the standard form of the first-order lag to give H (s ) R Q (s ) ts 1 (5.8) where t AR. cou9789x_ch05_099-122.indd 100 8/14/08 3:19:47 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 101 In comparing the transfer function of the tank given by Eq. (5.8) with the transfer function for the thermometer given by Eq. (4.7), we see that Eq. (5.8) contains the factor R. The term R is simply the conversion factor that relates h(t) to q(t) when the system is at steady state. As we saw in Chap. 4, this value is the steady-state gain. We can again verify the physical significance of this value (as we did in Chap. 4) by applying the final-value theorem of App. 3A to the determination of the steady-state value of H when the flow rate Q(t) changes according to a unit-step change; thus Q (t ) u(t ) where u(t) is the symbol for the unit-step change. The transform of Q(t) is Q(s) 1 s Combining this forcing function with Eq. (5.8) gives H (s ) 1 R s ts 1 Applying the final-value theorem, proved in App. 3A, to H(s) gives H (t ) |tS lim CsH (s) D lim R R s→0 t s 1 s→0 This shows that the ultimate change in H(t) for a unit change in Q(t) is simply R. If the transfer function relating the inlet flow q(t) to the outlet flow is desired, note that we have from Eq. (5.1) qos hs R (5.9) Subtracting Eq. (5.9) from Eq. (5.1) and using the deviation variable Qo qo qos give Qo H R (5.10) H (s ) R (5.11) Taking the transform of Eq. (5.10) gives Qo (s) Combining Eqs. (5.11) and (5.8) to eliminate H(s) gives Qo (s) 1 Q(s ) ts 1 (5.12) Notice that the steady-state gain for this transfer function is dimensionless, which is to be expected because the input variable q(t) and the output variable qo(t) have the same units (volume/time). The possibility of approximating an impulse forcing function in the flow rate to the liquid-level system is quite real. Recall that the unit-impulse function is defined as a cou9789x_ch05_099-122.indd 101 8/14/08 3:19:49 PM Confirming Pages 102 PART 2 LINEAR OPEN-LOOP SYSTEMS pulse of unit area as the duration of the pulse approaches zero, and the impulse function can be approximated by suddenly increasing the flow to a large value for a very short time; that is, we may pour very quickly a volume of liquid into the tank. The nature of the impulse response for a liquid-level system will be described by the following example. A tank having a time constant of 1 min and a resistance of 19 ft/cfm is operating at steady state with an inlet flow of 10 ft3/min (or cfm). At time t 0, the flow is suddenly increased to 100 ft3/min for 0.1 min by adding an additional 9 ft3 of water to the tank uniformly over a period of 0.1 min. (See Fig. 5–2a for this input disturbance.) Plot the response in tank level and compare with the impulse response. Before proceeding with the details of the computation, we should observe that as the time interval over which the 9 ft3 of water is added to the tank is shortened, the input approaches an impulse function having a magnitude of 9. From the data given in this example, the transfer function of the process is Example 5.1. 1 H (s ) 9 Q (s ) s 1 The input may be expressed as the difference in step functions, as was done in Example 3A.5. Q(t ) 90[u(t ) u(t 0 . 1)] The transform of this is Q(s ) ( 90 1 e0.1s s ) Combining this and the transfer function of the process, we obtain 1 e0.1s H (s) 10 s(s 1) s(s 1) (5.13) The first term in Eq. (5.13) can be inverted as shown in Eq. (4.15) to give 10(1 et). The second term, which includes e0.1s, must be inverted by use of the theorem on translation of functions given in App. 3A. According to this theorem, the inverse of est0 f (s) is f(t t0)u(t t0) with u(t t0) 0 for t t0 < 0 or t < t0 . The inverse of the second term in Eq. (5.13) is thus e0.1s L−1 0 s(s 1) for t 0 . 1 ( 10 1 e( t 0.1) ) for t 0 . 1 or ( ) 10 1 e( t 0.1) u(t 0 . 1) cou9789x_ch05_099-122.indd 102 8/14/08 3:19:50 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 103 The complete solution to this problem, which is the inverse of Eq. (5.13), is ( ) ( ) H (t ) ⫽ 10 1 ⫺ e⫺t u(t ) ⫺ 10 1 ⫺ e⫺( t ⫺ 0.1) u(t ⫺ 0 . 1) (5.14) which is equivalent to ( H (t ) ⫽ 10 1 ⫺ e⫺t ( ) t ⬍ 0.1 ) ( ) H (t ) ⫽ 10 1 ⫺ e⫺t ⫺ 1 ⫺ e⫺( t ⫺ 0.1) t ⬎ 0.1 Simplifying this expression for H(t) for t > 0.1 gives H (t ) ⫽ 1 . 052e⫺t t ⬎ 0.1 From Eq. (4.19), the response of the system to an impulse of magnitude 9 is given by H (t ) impulse ⫽ (9) ( 19 e⫺t ) ⫽ e⫺t In Fig. 5–2, the pulse response of the liquid-level system and the ideal impulse response are shown for comparison. Notice that the level rises very rapidly during the 0.1 min that additional flow is entering the tank; the level then decays exponentially and follows very closely the ideal impulse response. 1.0 q (ft3/min) 100 10 0 0.1 0.2 t (min) (a) H(t) Pulse response Area = 9 ft3 Impulse response (ideal) 0 0 2 1 t (min) (b) FIGURE 5–2 Approximation of an impulse function in a liquid-level system (Example 5.1). (a) Pulse input; (b) response of tank level. The responses to step and sinusoidal forcing functions are the same for the liquidlevel system as for the mercury thermometer of Chap. 4. Hence, they need not be rederived. This is the advantage of characterizing all first-order systems by the same transfer function. Liquid-Level Process with Constant-Flow Outlet An example of a transfer function that often arises in control systems may be developed by considering the liquid-level system shown in Fig. 5–3. The resistance shown in cou9789x_ch05_099-122.indd 103 9/4/08 10:39:50 AM Confirming Pages 104 PART 2 LINEAR OPEN-LOOP SYSTEMS q (t ) h(t ) qo = constant FIGURE 5–3 Liquid-level system with constant outlet flow. Fig. 5–1 is replaced by a constant-flow pump. The same assumptions of constant crosssectional area and constant density that were used before also apply here. For this system, Eq. (5.2) still applies, but qo(t) is now a constant; thus q(t ) qo A dh dt (5.15) At steady state, Eq. (5.15) becomes qs qo 0 (5.16) Subtracting Eq. (5.16) from Eq. (5.15) and introducing the deviation variables Q q qs and H h hs give dH (5.17) Q A dt Taking the Laplace transform of each side of Eq. (5.17) and solving for H/Q give H (s ) 1 Q (s ) As (5.18) Notice that the transfer function 1/As in Eq. (5.18) is equivalent to integration. (Recall from App. 3A that multiplying the transform by s corresponds to differentiation of the function in the time domain, while dividing by s corresponds to integration in the time domain.) Therefore, the solution of Eq. (5.18) is h(t ) hs 1 t Q(t ) dt A ∫0 (5.19) Clearly, if we increase the inlet flow to the tank, the level will increase because the outlet flow remains constant. The excess volumetric flow rate into the tank accumulates, and the level rises. For instance, if a step change Q(t) u(t) were applied to the system shown in Fig. 5–3 the result would be h(t ) hs t /A (5.20) The step response given by Eq. (5.20) is a ramp function that grows without limit. Such a system that grows without limit for a sustained change in input is said to have cou9789x_ch05_099-122.indd 104 8/14/08 3:19:51 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 105 nonregulation. Systems that have a limited change in output for a sustained change in input are said to have regulation. An example of a system having regulation is the step response of a first-order system, such as that shown in Fig. 5–1. If the inlet flow to the process shown in Fig. 5–1 is increased, the level will rise until the outlet flow becomes equal to the inlet flow, and then the level stops changing. This process is said to be self-regulating. The transfer function for the liquid-level system with constant outlet flow given by Eq. (5.18) can be considered as a special case of Eq. (5.8) as R → . lim R→ ∞ R 1 ARs 1 As The next example of a first-order system is a mixing process. Mixing Process Consider the mixing process shown in Fig. 5–4 in which a stream of solution containing dissolved salt flows at a constant volumetric flow rate q into a tank of constant holdup volume V. The concentration of the salt in the entering x(t) stream x (mass of salt/volume) varies with time. It is q desired to determine the transfer function relating the y(t) q outlet concentration y to the inlet concentration x. If we assume the density of the solution to be V constant, the flow rate in must equal the flow rate y (t) out, since the holdup volume is fixed. We may analyze this system by writing a transient mass balance FIGURE 5–4 for the salt; thus Mixing process. Rate of accumulation Flow rate of Flow rate of = − of salt in tank salt in salt out Expressing this mass balance in terms of symbols gives d (Vy) dy (5.21) V dt dt We will again introduce deviation variables as we have in the previous examples. At steady state, Eq. (5.21) may be written qx qy qxs qys 0 (5.22) Subtracting Eq. (5.22) from Eq. (5.21) and introducing the deviation variables X x xs Y y ys give qX qY V cou9789x_ch05_099-122.indd 105 dY dt 8/14/08 3:19:52 PM Confirming Pages 106 PART 2 LINEAR OPEN-LOOP SYSTEMS Taking the Laplace transform of this expression and rearranging the result give Y (s ) 1 X (s ) ts 1 where t V/q. This mixing process is, therefore, another first-order process for which the dynamics are now well known. Our last example of a first-order system is a heating process. Steam or electricity q w, Ti Heating Process w, T FIGURE 5–5 Heating process. Rate of energy flow into tank (5.23) Rate of − energy flow out of tank Consider the heating process shown in Fig. 5–5. A stream at temperature Ti is fed to the tank. Heat is added to the tank by means of an electric heater. The tank is well mixed, and the temperature of the exiting stream is T. The flow rate to the tank is constant at w lb/h. A transient energy balance on the tank yields Rate o f + energy flow in from heater Rate o f = accumulation of energy in tank Converting this energy balance to symbols results in wC ( Ti Tref ) wC ( T Tref ) q rVC d ( T Tref ) dT rVC dt dt (5.24) where T ref is the reference temperature and C is the heat capacity of the fluid. At steady state, dT/dt is zero, and Eq. (5.24) can be written wC ( Tis Ts ) qs 0 (5.25) where the subscript s has been used to indicate steady state. Subtracting Eq. (5.25) from Eq. (5.24) gives wC ( Ti Tis ) wC ( T Ts ) q qs r VC d ( T Ts ) dt (5.26) If we assume that Ti is constant (and so Ti Tis) and introduce the deviation variables T ′ T Ts Q q qs cou9789x_ch05_099-122.indd 106 8/14/08 3:19:53 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 107 Eq. (5.26) becomes wCT ′ Q rVC dT ′ dt (5.27) Taking Laplace transforms of Eq. (5.27) gives wCT ′(s) Q(s) r VCsT ′(s) (5.28) Rearranging Eq. (5.28) produces the following first-order transfer function relating T(s) and Q(s): T ′(s ) 1/wC K Q(s) ( r V /w) s 1 t s 1 (5.29) Thus, this process exhibits first-order dynamics as the tank temperature T responds to changes in the heat input to the tank. Example 5.2. Consider the mixed tank heater shown in Fig. 5–6. Develop a transfer function relating the tank outlet temperature to changes in the inlet temperature. Determine the response of the outlet temperature of the tank to a step change in the inlet temperature from 60 to 70C. Before we proceed, intuitively what would we expect to happen? If the inlet temperature rises by 10C, we expect the outlet temperature to eventually rise by 10C if nothing else changes. Let’s see what modeling the process will tell us. From Eq. (5.26) we can write the following simplified balance, realizing that q qs: wC ( Ti Tis ) wC ( T Ts ) rVC d (T Ts ) dt In terms of deviation variables, this becomes wCTi′ wCT ′ rVC dT ′ dt Transforming, we get 200 L/min Ti = 60°C Water wCTi ′(s) wCT ′(s) rVCsT ′(s) V = 1,000 L T = 80°C Heat input q and finally, after rearranging, T ′(s ) 1 1 Ti ′ (s) ( rV /w ) s 1 t s 1 FIGURE 5–6 Mixed tank heater. cou9789x_ch05_099-122.indd 107 8/14/08 3:19:54 PM Confirming Pages 108 PART 2 LINEAR OPEN-LOOP SYSTEMS Substituting in numerical values for the variables, we obtain the actual transfer function for this mixed tank heater. rV V V tank volume 1, 000 L ⫽ ⫽ ⫽ ⫽ ⫽ 5 min w w /r u volumetric flow ratte 200 L /min t ⫽ T ′(s ) 1 ⫽ Ti′ (s) 5s ⫹ 1 If the inlet temperature is stepped from 60 to 70⬚C, Ti ′(t ) ⫽ 70 ⫺ 60 ⫽ 10 and Ti ′(s) ⫽ 10/s . Thus, T ′(s ) ⫽ 10 1 s 5s ⫹ 1 Inverting to the time domain, we obtain the expression for T⬘(t) ( T ′(t ) ⫽ 10 1 ⫺ e⫺t / 5 ) and finally, we obtain the expression for T(t), the actual tank outlet temperature. ( T (t ) ⫽ Ts ⫹ T ′(t ) ⫽ 80 ⫹ 10 1 ⫺ e⫺t / 5 ) A plot of the outlet temperature (in deviation variables) is shown in the Fig. 5–7a. The actual outlet temperature is shown in Fig. 5–7b. Note that for the uncontrolled mixing tank, a step change of 10⬚C in the inlet temperature 90 10 Outlet temperature (°C) deviation variables 9 8 88 7 87 6 86 5 85 4 84 3 83 2 82 1 81 0 0 5 10 15 Time (min) (a) Actual outlet temperature (°C) 89 20 25 80 0 5 10 15 Time (min) (b) 20 25 FIGURE 5–7 (a) Tank outlet temperature (deviation variable); (b) actual tank outlet temperature. cou9789x_ch05_099-122.indd 108 8/14/08 7:46:26 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 109 ultimately produces a 10⬚C change in the outlet temperature, just as we predicted intuitively before we began our modeling. This result is just what we expected. The three examples presented in this section are intended to show that the dynamic characteristics of many physical systems can be represented by a first-order transfer function. In the remainder of the book, more examples of first-order systems will appear as we discuss a variety of control systems. In summarizing the previous examples of first-order systems, the time constant for each has been expressed in terms of system parameters; thus mC hA t ⫽ AR V t ⫽ q rV t ⫽ w t ⫽ 5.2 for thermometer , Eq. (4. 5) for liq u id-level process, Eq. (5. 8) for mixing process, Eq. (5. 23) for heating proce ss, Eq. (5. 29) LINEARIZATION Thus far, all the examples of physical systems, including the liquid-level system of Fig. 5–1, have been linear. Actually, most physical systems of practical importance are nonlinear. Characterization of a dynamic system by a transfer function can be done only for linear systems (those described by linear differential equations). The convenience of using transfer functions for dynamic analysis, which we have already seen in applications, provides significant motivation for approximating nonlinear systems by linear ones. A very important technique for such approximation is illustrated by the following discussion of the liquid-level system of Fig. 5–1. We now assume that the flow out of the tank follows a square root relationship qo ⫽ Ch1/ 2 (5.30) where C is a constant. For a liquid of constant density and a tank of uniform cross-sectional area A, a material balance around the tank gives q(t ) ⫺ qo (t ) ⫽ A dh dt (5.31) Combining Eqs. (5.30) and (5.31) gives the nonlinear differential equation q ⫺ Ch1/ 2 ⫽ A dh dt (5.32) At this point, we cannot proceed as before and take the Laplace transform. This is due to the presence of the nonlinear term h1/2, for which there is no simple transform. This difficulty can be circumvented by linearizing the nonlinear term. cou9789x_ch05_099-122.indd 109 8/14/08 7:46:28 PM Confirming Pages 110 PART 2 LINEAR OPEN-LOOP SYSTEMS By means of a Taylor series expansion, the function qo(h) may be expanded around the steady-state value hs; thus qo qo ( hs ) qo′ ( hs ) ( h hs ) q ″o ( hs ) ( h hs )2 . . . 2 where qo′ ( hs ) is the first derivative of qo evaluated at hs , q ″o ( hs ) is the second derivative, etc. If we keep only the linear term, the result is qo 艑 qo ( hs ) qo′ ( hs ) ( h hs ) (5.33) Taking the derivative of qo with respect to h in Eq. (5.30) and evaluating the derivative at h hs give qo′ ( hs ) 1 / 2 1 Chs 2 Introducing this into Eq. (5.33) gives qo qos 1 / 2 where qos qo ( hs ) and 1/R1 12 Ch s 1 (h hs ) R1 (5.34) . Substituting Eq. (5.34) into Eq. (5.31) gives q qos h hs dh A R1 dt (5.35) At steady state the flow entering the tank equals the flow leaving the tank; thus qs qos (5.36) Introducing this last equation into Eq. (5.35) gives A dh h hs q qs dt R1 (5.37) Introducing deviation variables Q q qs and H h hs into Eq. (5.37) and transforming give H (s ) R1 Q(s) ts 1 (5.38) where R1 2hs1 / 2 C t R1 A We see that a transfer function is obtained that is identical in form with that of the linear system, Eq. (5.8). However, in this case, the resistance R1 depends on the steady-state conditions around which the process operates. Graphically, the resistance R1 is the reciprocal of the slope of the tangent line passing through the point ( qos, hs ) , as shown in cou9789x_ch05_099-122.indd 110 8/14/08 3:19:56 PM Confirming Pages CHAPTER 5 111 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS q (t) Tangent line dqo (hs) Slope = 1 = R1 dh qo h (t) Nonlinear resistance qo(t) qos qo = Ch1/2 0 0 hs h FIGURE 5–8 Liquid-level system with nonlinear resistance. Fig. 5–8. Furthermore, the linear approximation given by Eq. (5.35) is the equation of the tangent line itself. From the graphical representation, it should be clear that the linear approximation improves as the deviation in h becomes smaller. If one does not have an analytic expression such as h1/2 for the nonlinear function, but only a graph of the function, the technique can still be applied by representing the function by the tangent line passing through the point of operation. Whether or not the linearized result is a valid representation depends on the operation of the system. If the level is being maintained by a controller, at or close to a fixed level hs, then by the very nature of the control imposed on the system, deviations in level should be small (for good control) and the linearized equation is adequate. On the other hand, if the level should change over a wide range, the linear approximation may be very poor and the system may deviate significantly from the prediction of the linear transfer function. In such cases, it may be necessary to use the more difficult methods of nonlinear analysis, some of which are discussed in Chaps. 24 and 25. We shall extend the discussion of linearization to more complex systems in Chap. 20. Using MATLAB to Compare Nonlinear (Exact) Solutions and Linearized Solutions For the tank draining models of Eqs. (5.32) and (5.38) we have the following systems: Nonlinear model Linearized model q Ch1/2 A dh dt H (s ) R1 Q(s) ts 1 (5.32) (5.38) where 2hs1/ 2 C t R1 A R1 cou9789x_ch05_099-122.indd 111 8/14/08 3:19:57 PM Confirming Pages 112 PART 2 LINEAR OPEN-LOOP SYSTEMS Consider the case where A 3 ft2 and the steady-state height is 4 ft when the inlet flow is 16 cfm. Compare the linearized and nonlinear (exact) solutions for the transient response of the tank height to a step change in feed flow from 16 to 20 cfm. Solution: From Eq. (5.30), 1/ 2 qos Ch s 16 cfm C (4 ft )1/ 2 thus, C 8 cfm ft1/ 2 and ft 2hs1/ 2 2 4 ft 0.5 C cfm 8 cfm/ft1/ 2 ft 2 t R1 A 0 . 5 (3 ft ) 1 . 5 min cfm R1 Substituting the numerical values into the nonlinear model, Eq. (5.32), yields 20 8 h 3 dh dt h(0) 4 ft The MATLAB m-file necessary to simulate this equation is shown below. % FILENAME is level.m function hprime=level(t,h) hprime=(20—8*sqrt(h))/3; This file calculates the derivative dh/dt at any given t and h. We call the m-file using the numerical differential equation solver ODE45. >> [t,h] = ode45(@level Matrices that you want the answers returned into. MATLAB routine to numerically solve ODE , [0,10] , [4] m-file level.m contains the model ); time span... tinitial to tfinal initial condition h(0) must correspond to tinitial The linearized model, with numerical values substituted in, is H (s ) 0.5 Q (s ) 1 . 5s 1 cou9789x_ch05_099-122.indd 112 8/14/08 3:19:57 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 113 where H h4 H (s ) Q q 16 4 s deviation variables 2 0.5 1 . 5s 1 s (1 . 5s 1) step change Q by 4 cfm Inverting gives ( H (t ) 2 1 et / 1.5 ( ) h(t ) 4 2 1 et / 1.5 ) linearizzed solution Entering this equation into MATLAB yields >> hlin 4 2*(1 exp (t/1.5)); Plot the linearized and nonlinear solutions on the same axes using MATLAB (see Fig. 5–9). >>plot(t,h,t,hlin); >> xlabel(’Time’); >> ylabel(’h’); >> title(’Comparison of Nonlinear and Linearized Solutions’); Comparison of Nonlinear and Linearized Solutions 6.5 Nonlinear solution 6 Linear solution h 5.5 5 4.5 4 0 1 2 3 4 5 6 7 8 9 Time FIGURE 5–9 Comparison of nonlinear and linearized solutions for tank draining. cou9789x_ch05_099-122.indd 113 8/14/08 3:20:02 PM Confirming Pages 114 PART 2 LINEAR OPEN-LOOP SYSTEMS In general, the linearization of a nonlinear function is accomplished using a Taylor series expansion truncated to include only the linear terms. Thus for a singlevariable function f ( x ) f ( xs ) df dx ( x xs ) (higher-order terms)) (5.39) xs For functions of two variables, we have f ( x , y) f ( xs , ys ) ∂f ∂x ( xs , ys ) ( x xs ) ∂f ( y ys ) ∂y ( x s , ys ) (5.40) (higher-order terms) Consider the differential equation describing the dynamics of a system dy f ( y) x ( t ) dt nonlinear (5.41) term Linearizing the nonlinear term gives ∂f dy f ( ys ) ( y ys ) x (t ) ∂y ys dt (5.42) linearized approximattion Writing this equation again for the steady-state case gives dys ∂f f ( ys ) dt ∂y ( ys ys ) xs (5.43) ys Subtracting the steady-state case in Eq. (5.43) from Eq. (5.42), we can convert the original differential equation to deviation variables: d ( y ys ) ∂f dt ∂y ( y ys ) x xs ys dY ∂f dt ∂y Y X ys where X x xs and Y y ys. Note that the f (ys) term is eliminated in the process of forming deviation variables, and we are left with only linear terms in the equation which is now amenable to solution using Laplace transforms. cou9789x_ch05_099-122.indd 114 8/14/08 3:20:03 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 115 SUMMARY In this chapter, we demonstrated several physical examples of first-order systems. Transfer functions were developed for those physical systems and placed into the standard form for first-order systems. We will see more examples of first-order systems as we discuss control systems in later chapters. We have also characterized, in an approximate sense, a nonlinear system by a linear transfer function. In general, this technique may be applied to any nonlinearity that can be expressed in a Taylor series (or, equivalently, has a unique slope at the operating point). Since this includes most nonlinearities arising in process control, we have ample justification for studying linear systems in considerable detail. PROBLEMS 5.1. Derive the transfer function H(s)/Q(s) for the liquid-level system of Fig. P5–1 when (a) The tank level operates about the steady-state value of hs 1 ft (b) The tank level operates about the steady-state value of hs 3 ft q, ft3/min The pump removes water at a constant rate of 10 cfm (cubic feet per minute); this rate is independent of head. The cross-sectional area of the tank is 1.0 ft2, and the resistance R is 0.5 ft/cfm. R h(t) 5.2. A liquid-level system, such as the one shown in Fig. 5–1, has a cross-sectional area of 3.0 ft2. The valve characteristics are 2 ft FIGURE P5–1 q 8 h qo (ft3/min) Outlet flow 2.4 1.0 0.3 1.0 h(ft) FIGURE P5–3 q α h0 R B FIGURE P5–4 cou9789x_ch05_099-122.indd 115 where q flow rate, cfm, and h level above the valve, ft. Calculate the time constant for this system if the average operating level above the valve is (a) 3 ft (b) 9 ft 5.3. A tank having a cross-sectional area of 2 ft2 is operating at steady state with an inlet flow rate of 2.0 cfm. The flow-head characteristics are shown in Fig. P5–3. (a) Find the transfer function H(s)/Q(s). (b) If the flow to the tank increases from 2.0 to 2.2 cfm according to a step change, calculate the level h two minutes after the change occurs. 5.4. Develop a formula for finding the time constant of the liquid-level system shown in Fig. P5–4 when the average operating level is h0. The resistance R is linear. The tank has three vertical walls and one that slopes at an angle a from the vertical as shown. The distance separating the parallel walls is 1. 8/14/08 3:20:03 PM Confirming Pages 116 PART 2 LINEAR OPEN-LOOP SYSTEMS 5.5. Consider the stirred-tank reactor shown in Fig. P5–5. Ci, F The reaction occurring is A → B Volume V Co, F and it proceeds at a rate r kCo FIGURE P5–5 where r (moles A reacting)/(volume)(time) k reaction rate constant Co (t) concentration of A in reactor at any time t (mol A/volume) V volume of mixture in reactor Further, let F constant feed rate, volume/time Ci (t ) concentration of A in feed stream, moles/volume Assuming constant density and constant volume V, derive the transfer function relating the concentration in the reactor to the feed-stream concentration. Prepare a block diagram for the reactor. Sketch the response of the reactor to a unit-step change in Ci. 5.6. A thermocouple junction of area A, mass m, heat capacity C, and emissivity e is located in a furnace that normally is at TisC. At these temperatures convective and conductive heat transfer to the junction is negligible compared with radiative heat transfer. Determine the linearized transfer function between the furnace temperature Ti and the junction temperature T0. For the case m 0.1 g C 0 . 12 cal/ (g i C) e 0.7 A 0 . 1 cm 2 Tis 1100 C plot the response of the thermocouple to a 10C step change in furnace temperature. Compare this with the true response obtained by integration of the differential equation. 5.7. A liquid-level system has the following properties: Tank dimensions: 10 ft high by 5-ft diameter Steady-state operating characteristics: cou9789x_ch05_099-122.indd 116 Inflow, gal/h Steady-state level, ft 0 5,000 10,000 15,000 20,000 25,000 30,000 0 0.7 1.1 2.3 3.9 6.3 8.8 8/14/08 3:20:05 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 117 C (Ib/gal) (a) Plot the level response of the tank under the following circumstances: The inlet flow rate is held at 300 gal/min for 1 h and then suddenly raised to 400 gal/min. (b) How accurate is the steady-state level calculated from the dynamic response in part (a) when compared with the value given by the table above? (c) The tank is now connected in series with a second tank that has identical operating characteristics, but which has dimensions 8 ft high by 4-ft diameter. Plot the response of the original tank (which is upstream of the new tank) to the change described in part (a) when the connection is such that the tanks are (1) interacting and (2) noninteracting. (See Chap. 6.) 5.8. A mixing process may be described as follows: A stream with solute concentration Ci (pounds/volume) is fed to a perfectly stirred tank at a constant flow rate of q (volume/time). The perfectly mixed product is withdrawn from the tank, also at the flow rate q at the same concentration as the material in the tank C0. The total volume of solution in the tank is constant at V. Density may be considered to be inde2.1 pendent of concentration. A trace of the tank concentration versus time appears as shown in Fig. P5–8. 2.0 (a) Plot on this same figure your best guess of the quantitative behavior of the inlet 1.9 concentration versus time. Be sure to label the graph with quantitative information regarding times and magnitudes 1.8 and any other data that will demonstrate your understanding of the situation. (b) Write an equation for Ci as a function of 1.7 6:52 6:30 time. A.M. A.M. Data: tank dimensions: 8 ft high by 5-ft diameter Time Tank volume V: 700 gal FIGURE P5–8 Flow rate q: 100 gal/min Average density: 70 lb/ft3 10 cfm Disturbance 5.9. The liquid-level process shown in Fig. P5–9 is operating at steady state when the following disturbance occurs: At time t 0, 1 ft3 water is added suddenly (unit impulse) to the tank; at t 1 min, 2 ft3 of water is added suddenly to = 1 min h the tank. Sketch the response of the level in the tank versus time, and determine the level at t 0.5, 1, and 1.5 min. R = 0.5 q (cfm) FIGURE P5–9 2 1 0 0 1 3 2 t (min) 5.10. A tank having a cross-sectional area of 2 ft2 and a linear resistance of R 1 ft/cfm is operating at steady state with a flow rate of 1 cfm. At time 0, the flow varies as shown in Fig. P5–10. (a) Determine Q(t) and Q(s) by combining simple functions. Note that Q is the deviation in flow rate. (b) Obtain an expression for H(t) where H is the deviation in level. (c) Determine H(t) at t 2 and t . FIGURE P5–10 cou9789x_ch05_099-122.indd 117 8/14/08 3:20:05 PM Confirming Pages 118 PART 2 LINEAR OPEN-LOOP SYSTEMS q ft3/min A = 2 ft2 5.12. Derive the transfer function H/Q for the liquid-level system shown in Fig. P5–12. The resistances are linear; H and Q are deviation variables. Show clearly how you derived the transfer function. You are expected to give numerical values in the transfer function. h R1 = 2 ft/cfm R2 = 5 ft/cfm FIGURE P5–12 1 cfm Train of impulses A = 1 ft2 5.11. Determine y(t 5) if Y(s) e3s/s(7s 1). 5.13. The liquid-level system shown in Fig. P5–13 is initially at steady state with the inlet flow rate at 1 cfm. At time 0, 1 ft3 of water is suddenly added to the tank; at t 1, 1 ft3 is added; etc. In other words, a train of unit impulses is applied to the tank at intervals of 1 min. Ultimately the output wave train becomes periodic as shown in the sketch. Determine the maximum and minimum values of this output. Hmax H Hmin h R=1 0 n n +1 n +2 n +3 i FIGURE P5–13 5.14. The two-tank mixing process shown in Fig. P5–14 contains a recirculation loop that transfers solution from tank 2 to tank 1 at a flow rate of a qo. αq0 q0 = 1 cfm x(t) = feed concentration 1 ft3 1 ft3 c1 c2 q0 FIGURE P5–14 cou9789x_ch05_099-122.indd 118 8/14/08 3:20:06 PM Confirming Pages CHAPTER 5 PHYSICAL EXAMPLES OF FIRST-ORDER SYSTEMS 119 (a) Develop a transfer function that relates the concentration c2 in tank 2 to the concentration x in the feed, that is, C2 (s)/X(s) where C2 and X are deviation variables. For convenience, assume that the initial concentrations are x c1 c2 0. (b) If a unit-step change in x occurs, determine the time needed for c2 to reach 60 percent of its ultimate value for the cases where 0, 1, and . (c) Sketch the response for a . Assume that each tank has a constant holdup volume of 1 ft3. Neglect transportation lag in the line connecting the tanks and the recirculation line. Try to answer parts (b) and (c) by intuition. 5.15. Dye for our new line of blue jeans is being blended in a mixing tank. The desired color of blue is produced using a concentration of 1500 ppm blue dye, with a minimum acceptable concentration of 1400 ppm. At 9 A.M. today the dye injector plugged, and the dye flow was interrupted for 10 min, until we realized the problem and unclogged the nozzle. For how many minutes was the flow leaving the mixer off-specification (< 1400 ppm)? How many gallons of off-spec dye were made? See Fig. P5–15. Water 20 gal/min 20 gal/min aqueous dye for jeans (1500 ppm blue dye) Concentrated dye injector V = 100 gal FIGURE P5–15 5.16. For the reactor (CSTR) shown in Fig. P5–16, determine the transfer function that relates the exit concentration from the reactor to changes in the feed concentration. If we instantaneously double the feed concentration from 1 to 2 mol/L, what is the new exiting concentration 1 min later? What is the new steady-state reactor concentration? The rate constant is k 5 L/min CA0 = 1 mol/L 2 (mol/L )(min) Reaction : 2A B Rate law : -rA = kCA2 5 L/min CA0 = 0.2 mol/L Volume = 50 L FIGURE P5–16 cou9789x_ch05_099-122.indd 119 8/14/08 3:20:06 PM Confirming Pages 120 PART 2 LINEAR OPEN-LOOP SYSTEMS The reaction rate law is rA kCA2, where rA is the production rate of A in moles per liter per minute. 5.17. The Antoine equation for the vapor pressure of a liquid at a given temperature is given by P * e A B / (T C ) The constants for benzene are A 15 . 9008 B 2788 . 51 °C C 220 . 80 °C for the vapor pressure in millimeters of mercury (mmHg). Linearize the equation about a temperature of 40C. Compare the actual vapor pressure (from the Antoine equation) at 45 and 60C with the vapor pressure calculated from the linearized equation. What is the percent difference in each case? Comment on the suitability of the linearized equation. 5.18. Find the transfer function that relates the height in the vessel (Fig. P5–18) to changes in the inlet flow rate. qi (cfm) h θ = 30° q0 (cfm) R1 (valve resistance) FIGURE P5–18 cou9789x_ch05_099-122.indd 120 8/14/08 3:20:07 PM Confirming Pages CHAPTER 5 CAPSULE SUMMARY Here are some physical examples of first-order systems: System Transfer function q (t) A = area h (t) R Figure 5–1 Liquid level system H (s ) R Q (s ) ARs 1 Figure 5–4 Mixing process Y (s ) 1 X (s ) (V / q)s 1 Figure 5–5 Heating process T ′(s ) 1/ ( wC ) Q (s ) ( rV /w ) s 1 q0(t) x(t) q y(t) q V y (t) Steam or electricity w, Ti q w, T (continued) 121 cou9789x_ch05_099-122.indd 121 8/14/08 3:20:07 PM Confirming Pages 122 PART 2 LINEAR OPEN-LOOP SYSTEMS System Transfer function Fluid y = thermometer reading Y (s ) 1 X (s ) ( mc /hA) s 1 x = fluid temperature Figure 4–1a Thermometer Mercury Glass wall TAYLOR SERIES EXPANSIONS FOR LINEARIZING NONLINEAR TERMS Functions of a single variable: f ( x ) f ( xs ) df dx ( x xs ) xs Functions of two variables: f ( x , y) f ( xs , ys ) cou9789x_ch05_099-122.indd 122 ∂f ∂f ( x xs ) ( y ys ) ∂x ( xs , ys ) ∂y ( xs , ys ) 8/14/08 3:20:08 PM Confirming Pages CHAPTER 6 RESPONSE OF FIRST-ORDER SYSTEMS IN SERIES 6.1 INTRODUCTORY REMARKS Very often, a physical system can be represented by several first-order processes connected in series. To illustrate this type of system, consider the liquid-level systems shown in Fig. 6–1 in which two tanks are arranged so that the outlet flow from the first tank is the inlet flow to the second tank. Two possible piping arrangements are shown in Fig. 6–1. In Fig. 6–1a the outlet flow from tank 1 discharges directly into the atmosphere before spilling into tank 2, and the flow through R1 depends only on h1. The variation in h2 in tank 2 does not affect the transient response occurring in tank 1. This type of system is referred to as a noninteracting system. In contrast to this, the system shown in Fig. 6–1b is said to be interacting because the flow through R1 now depends on the difference between h1 and h2. We will consider first the noninteracting system of Fig. 6–1a. 6.2 NONINTERACTING SYSTEM As in the previous liquid-level example, we shall assume the liquid to be of constant density, the tanks to have uniform cross-sectional area, and the flow resistances to be linear. Our problem is to find a transfer function that relates h2 to q, that is, H2(s)/Q(s). The approach will be to obtain a transfer function for each tank, Q1(s)/Q(s) and H2(s)/ Q1(s), by writing a transient mass balance around each tank; these transfer functions will then be combined to eliminate the intermediate flow Q1(s) and produce the desired transfer function. A balance on tank 1 gives q − q1 ⫽ A1 dh1 dt (6.1) 123 cou9789x_ch06_123-136.indd 123 8/14/08 6:21:43 PM Confirming Pages 124 PART 2 LINEAR OPEN-LOOP SYSTEMS q (t) A1 q (t) A1 h1 R1 q1 A2 h2 A2 h1 R2 R1 h2 q1 q2 (a) R2 q2 (b) FIGURE 6–1 Two-tank liquid-level system: (a) Noninteracting; (b) interacting. A balance on tank 2 gives q1 ⫺ q2 ⫽ A2 dh2 dt (6.2) The flow-head relationships for the two linear resistances are given by the expressions q1 ⫽ h1 R1 (6.3) q2 ⫽ h2 R2 (6.4) Combining Eqs. (6.1) and (6.3) in exactly the same manner as was done in Chap. 5 and introducing deviation variables give the transfer function for tank 1 Q1 (s) 1 ⫽ Q(s ) t 1s ⫹ 1 (6.5) where Q1 ⫽ q1 ⫺ q1s , Q ⫽ q ⫺ qs , and t1 ⫽ R1A1. In the same manner, we can combine Eqs. (6.2) and (6.4) to obtain the transfer function for tank 2 H 2 (s ) R2 ⫽ Q1 (s) t 2s ⫹ 1 (6.6) where H 2 ⫽ h2 ⫺ h2s and t 2 ⫽ R2A2. Having the transfer function for each tank, we can obtain the overall transfer function H2(s)/Q(s) by multiplying Eqs. (6.5) and (6.6) to eliminate Q1(s): H 2 (s ) 1 R2 ⫽ Q (s ) t 1s ⫹ 1 t 2 s ⫹ 1 (6.7) Notice that the overall transfer function of Eq. (6.7) is the product of two first-order transfer functions, each of which is the transfer function of a single tank operating independently cou9789x_ch06_123-136.indd 124 8/14/08 6:21:44 PM Confirming Pages CHAPTER 6 RESPONSE OF FIRST-ORDER SYSTEMS IN SERIES 125 of the other. In the case of the interacting system of Fig. 6–1b, the overall transfer function cannot be found by simply multiplying the separate transfer functions; this will become apparent when the interacting system is analyzed later. Example 6.1. Two noninteracting tanks are connected in series as shown in Fig. 6–1a. The time constants are t 2 ⫽ 1 and t 1 ⫽ 0.5; R2 ⫽ 1. Sketch the response of the level in tank 2 if a unit-step change is made in the inlet flow rate to tank 1. The transfer function for this system is found directly from Eq. (6.7); thus H 2 (s ) R2 ⫽ (6.8) (t 1s ⫹ 1) (t 2 s ⫹ 1) Q (s ) For a unit-step change in Q, we obtain H 2 (s ) ⫽ R2 1 s (t 1s ⫹ 1)(t 2 s ⫹ 1) (6.9) Inversion by means of partial fraction expansion gives 1 t 1t 2 1 ⫺t /t 1 H 2 (t ) ⫽ R2 1 ⫺ ⫺ e⫺t /t 2 e ⫺ t t t t 1 2 2 1 (6.10) Substituting in the values of t1, t2, and R2 gives ( H 2 (t ) ⫽ 1 ⫺ 2e⫺t ⫺ e⫺2 t ) (6.11) A plot of this response is shown in Fig. 6–2. Notice that the response is S-shaped and the slope dH2/dt at the origin is zero. If the change in flow rate were introduced into the second tank, the response would be first-order and is shown for comparison in Fig. 6–2 by the dotted curve. 1.0 H2(t) One tank 0.5 0 Two tanks 0 1 2 3 t FIGURE 6–2 Transient response of liquid-level system (Example 6.1). cou9789x_ch06_123-136.indd 125 8/14/08 6:21:45 PM Confirming Pages 126 PART 2 LINEAR OPEN-LOOP SYSTEMS MATLAB/Simulink Simulation of the Transient Response in Fig. 6–2 It’s quite easy to verify the result in Fig. 6–2 using Simulink (Fig. 6–3). The model is shown below at the left, and the output from the model is shown in the graph. 1 Feed, Q 1 0.5s+1 1 s+1 Tank 1 Tank 2 One Tank 1 s+1 Tank 2 Feed, Q Two Tanks Response 0.5 0 Tank 2 only 0 1 2 3 4 FIGURE 6–3 Simulink simulation of one tank and two tanks in series. From Example 6.1, notice that the step response of a system consisting of two firstorder systems is S-shaped and that the response changes very slowly just after introduction of the step input. This sluggishness or delay is sometimes called transfer lag and is always present when two or more first-order systems are connected in series. For a single first-order system, there is no transfer lag; i.e., the response begins immediately after the step change is applied, and the rate of change of the response (slope of response curve) is maximal at t ⫽ 0. Generalization for Several Noninteracting Systems in Series We have observed that the overall transfer function for two noninteracting first-order systems connected in series is simply the product of the individual transfer functions. We may now generalize this concept by considering n noninteracting first-order systems as represented by the block diagram of Fig. 6–4. X0 k1 t1s⫹1 X1 k2 t2s⫹1 X2 Xi⫺1 ki kn tis⫹1 tns⫹1 Xn FIGURE 6–4 Noninteracting first-order systems. The block diagram is equivalent to the relationships X1 (s) k1 ⫽ t 1s ⫹ 1 X 0 (s) X 2 (s) k2 ⫽ t 2s ⫹ 1 X1 (s) etc. X n (s ) kn ⫽ X n ⫺1 (s) t ns ⫹ 1 cou9789x_ch06_123-136.indd 126 8/14/08 6:21:46 PM Confirming Pages CHAPTER 6 RESPONSE OF FIRST-ORDER SYSTEMS IN SERIES 127 To obtain the overall transfer function, we simply multiply the individual transfer functions; thus X n (s ) ⫽ X 0 (s ) n ∏ i ⫽1 ki t is ⫹ 1 (6.12) To show how the transfer lag is increased as the number of stages increases, Fig. 6–5 gives the unit-step response curves for several systems containing one or more firstorder stages in series. 1.0 n= 0.8 1 n = 2 0.6 Y(t) n = 3 n 0.4 0.2 0 = 4 1 u(t) 0 1 3 2 4 n Y(t) 5 FIGURE 6–5 Step response of noninteracting first-order systems in series. MATLAB/Simulink Simulation of Noninteracting First-Order Systems in Fig. 6–5 Let’s reproduce the response in Fig. 6–5 for four tanks in series using Simulink (Fig. 6–6). Note how simple the model is and that the result is identical to Fig. 6–5. 1 Input 1 s+1 Tank 1 1 s+1 Tank 2 1 s+1 Tank 3 1 s+1 Tank 4 0.8 0.6 0.4 Response 0.2 0 0 2 4 6 FIGURE 6–6 Simulink diagram for noninteracting first-order systems in series. cou9789x_ch06_123-136.indd 127 8/14/08 6:21:47 PM Confirming Pages 128 PART 2 LINEAR OPEN-LOOP SYSTEMS 6.3 INTERACTING SYSTEM To illustrate an interacting system, we will derive the transfer function for the system shown in Fig. 6–1b. The analysis is started by writing mass balances on the tanks as was done for the noninteracting case. The balances on tanks 1 and 2 are the same as before and are given by Eqs. (6.1) and (6.2). dh1 dt Tank 1 q ⫺ q1 ⫽ A1 Tank 2 q1 ⫺ q2 ⫽ A2 dh2 dt (6.1) (6.2) However, the flow-head relationship for R1 is now q1 ⫽ 1 ( h1 ⫺ h2 ) R1 (6.13) The flow-head relationship for R2 is the same as before [Eq. (6.4)]. q2 ⫽ h2 R2 (6.4) A simple way to combine Eqs. (6.1), (6.2), (6.4), and (6.13) is to first express them in terms of deviation variables, transform the resulting equations, and then combine the transformed equations to eliminate the unwanted variables. At steady state, Eqs. (6.1) and (6.2) can be written qs ⫺ q1s ⫽ 0 (6.14) q1s ⫺ q2s ⫽ 0 (6.15) Subtracting Eq. (6.14) from Eq. (6.1) and Eq. (6.15) from Eq. (6.2) and introducing deviation variables give dH1 dt Tank 1 Q ⫺ Q1 ⫽ A1 Tank 2 Q1 ⫺ Q2 ⫽ A2 dH 2 dt (6.16) (6.17) Expressing Eqs. (6.13) and (6.4) in terms of deviation variables gives cou9789x_ch06_123-136.indd 128 Valve 1 Q1 ⫽ H1 ⫺ H 2 R1 Valve 2 Q2 ⫽ H2 R2 (6.18) (6.19) 8/14/08 6:21:47 PM Confirming Pages CHAPTER 6 RESPONSE OF FIRST-ORDER SYSTEMS IN SERIES 129 Transforming Eqs. (6.16) through (6.19) gives Tank 1 Q(s) ⫺ Q1 (s) ⫽ A1sH1 (s) (6.20) Tank 2 Q1 (s) ⫺ Q2 (s) ⫽ A2 sH 2 (s) (6.21) Valve 1 R1Q1 (s) ⫽ H1 (s) ⫺ H 2 (s) (6.22) Valve 2 R2Q2 (s) ⫽ H 2 (s) (6.23) The analysis has produced four algebraic equations containing five unknowns: Q, Q1, Q2, H1, and H2. These equations may be combined to eliminate Q1, Q2, and H1 and to arrive at the desired transfer function: H 2 (s ) R2 ⫽ Q (s ) t 1t 2 s 2 ⫹ (t 1 ⫹ t 2 ⫹ A1 R2 ) s ⫹ 1 (6.24) Notice that the product of the transfer functions for the tanks operating separately, Eqs. (6.5) and (6.6), does not produce the correct result for the interacting system. The difference between the transfer function for the noninteracting system, Eq. (6.7), and that for the interacting system, Eq. (6.24), is the presence of the cross-product term A1R2 in the coefficient of s. The term interacting is often referred to as loading. The second tank of Fig. 6–1b is said to load the first tank. To understand the effect of interaction on the transient response of a system, consider a two-tank system for which the time constants are equal (t 1 ⫽ t 2 ⫽ t). If the tanks are noninteracting, the transfer function relating inlet flow to outlet flow is 2 Q2 (s) 1 ⫽ t s ⫹ 1 Q(s ) (6.25) The unit-step response for this transfer function can be obtained by the usual procedure to give Q2 (t ) ⫽ 1 ⫺ e⫺t /t ⫺ t ⫺t /t e t (6.26) If the tanks are interacting, the overall transfer function, according to Eq. (6.24), is (assuming further that A1 ⫽ A2) Q2 (s) 1 ⫽ 2 2 Q(s) t s ⫹ 3t s ⫹ 1 (6.27) By application of the quadratic formula, the denominator of this transfer function can be written as Q2 (s) 1 ⫽ (6.28) Q(s) (0 . 38t s ⫹ 1)(2 . 62t s ⫹ 1) cou9789x_ch06_123-136.indd 129 8/14/08 6:21:48 PM Confirming Pages 130 PART 2 LINEAR OPEN-LOOP SYSTEMS For this example, we see that the effect of interaction has been to change the effective time constants of the interacting system. One time constant has become considerably larger and the other smaller than the time constant t of either tank in the noninteracting system. The response of Q2(t) to a unit-step change in Q(t) for the interacting case [Eq. (6.28)] is Q2 (t ) ⫽ 1 ⫹ 0 . 17e⫺t / 0.38t ⫺ 1 . 17e⫺t / 2.62t (6.29) In Fig. 6–7, the unit-step responses [Eqs. (6.26) and (6.29)] for the two cases are plotted to show the effect of interaction. From this figure, it can be seen that interaction slows up the response. This result can be understood on physical grounds in the following way: If the same size step change is introduced into the two systems of Fig. 6–1, the flow from tank 1 (q1) for the noninteracting case will not be reduced by the increase in level in tank 2. However, for the interacting case, the flow q1 will be reduced by the buildup of level in tank 2. At any time t1 following the introduction of the step input, q1 for the interacting case will be less than for the noninteracting case with the result that h2 (or q2) will increase at a slower rate. 1.0 Q = u(t) 0.8 1 Q2 1 Q2/Q 0.6 Noninteracting Interacting 0.4 0 1 Q = u(t) 0.2 0 1 2 1 Q2 3 FIGURE 6–7 Effect of interaction on step response of two-tank system. In general, the effect of interaction on a system containing two first-order lags is to change the ratio of effective time constants in the interacting system. In terms of the transient response, this means that the interacting system is more sluggish than the noninteracting system. cou9789x_ch06_123-136.indd 130 8/14/08 6:21:49 PM Confirming Pages CHAPTER 6 RESPONSE OF FIRST-ORDER SYSTEMS IN SERIES 131 SUMMARY In this chapter we discussed the response of first-order systems in series. We observed that the nature of the response is dependent upon whether the first-order systems in series form a noninteracting or an interacting system. We used MATLAB to visualize and analyze the response of these two different types of systems and to study their behavior. This chapter concludes our specific discussion of first-order systems. We will make continued use of the material developed here in the succeeding chapters. PROBLEMS 6.1. Determine the transfer function H(s)/Q(s) for the liquid-level system shown in Fig. P6–1. Resistances R1 and R2 are linear. The flow rate from tank 3 is maintained constant at b by means of a pump; i.e., the flow rate from tank 3 is independent of head h. The tanks are noninteracting. q (t) A1 A2 R1 Tank 1 A3 R2 Tank 2 h qo = b Tank 3 FIGURE P6–1 6.2. The mercury thermometer in Chap. 4 was considered to have all its resistance in the convective film surrounding the bulb and all its capacitance in the mercury. A more detailed analysis would consider both the convective resistance surrounding the bulb and that between the bulb and the mercury. In addition, the capacitance of the glass bulb would be included. Let Ai ⫽ inside area of bulb, for heat transfer to mercury Ao ⫽ outside area of bulb, for heat transfer from surrounding ﬂuid m ⫽ mass of mercury in bulb mb ⫽ mass of glass bulb cou9789x_ch06_123-136.indd 131 8/14/08 6:21:50 PM Confirming Pages 132 PART 2 LINEAR OPEN-LOOP SYSTEMS C ⫽ heat capacity of mercury Cb ⫽ heat capacity of glass bulb hi ⫽ convective coefﬁcient between bulb and mercury ho ⫽ convective coefﬁcient between bulb and surrounding ﬂuid T ⫽ temperature of mercury Tb ⫽ temperature of glass bulb Tf ⫽ temperature of surrounding ﬂuid Determine the transfer function between Tf and T. What is the effect of the bulb resistance and capacitance on the thermometer response? Note that the inclusion of the bulb results in a pair of interacting systems, which give an overall transfer function somewhat different from that of Eq. (6.24). 6.3. There are N storage tanks of volume V arranged so that when water is fed into the first tank, an equal volume of liquid overflows from the first tank into the second tank, and so on. Each tank initially contains component A at some concentration Co and is equipped with a perfect stirrer. At time 0, a stream of zero concentration is fed into the first tank at a volumetric rate q. Find the resulting concentration in each tank as a function of time. 6.4. (a) Find the transfer functions H2/Q and H3/Q for the three-tank system shown in Fig. P6–4 where H2, H3, and Q are deviation variables. Tank 1 and tank 2 are interacting. (b) For a unit-step change in q (that is, Q ⫽ 1/s), determine H3(0) and H3(⬁), and sketch H3(t) versus t. q A1 = 1 A2 = 1 h2 R1 = 2 Tank 1 R2 = 2 Tank 2 A3 = 0.5 h3 R3 = 4 Tank 3 FIGURE P6–4 6.5. Three identical tanks are operated in series in a noninteracting fashion as shown in Fig. P6–5. For each tank, R ⫽ 1 and t ⫽ 1. The deviation in flow rate to the first tank is an impulse function of magnitude 2. cou9789x_ch06_123-136.indd 132 8/14/08 6:21:50 PM Confirming Pages CHAPTER 6 RESPONSE OF FIRST-ORDER SYSTEMS IN SERIES 133 (a) Determine an expression for H(s) where H is the deviation in level in the third tank. (b) Sketch the response H(t). (c) Obtain an expression for H(t). h FIGURE P6–5 6.6. In the two-tank mixing process shown in Fig. P6–6, x varies from 0 lb salt/ft3 to 1 lb salt/ft3 according to a step function. At what time does the salt concentration in tank 2 reach 0.6 lb salt/ft3? The holdup volume of each tank is 6 ft3. 3 ft3/min X C Tank 1 Tank 2 FIGURE P6–6 cou9789x_ch06_123-136.indd 133 8/14/08 6:21:51 PM Confirming Pages 134 PART 2 LINEAR OPEN-LOOP SYSTEMS 6.7. Starting from first principles, derive the transfer functions H1(s)/Q(s) and H2(s)/Q(s) for the liquid-level system shown in Fig. P6–7. The resistances are linear and R1 ⫽ R2 ⫽ 1. Note that two streams are flowing from tank 1, one of which flows into tank 2. You are expected to give numerical values of the parameters in the transfer functions and to show clearly how you derived the transfer functions. q(t) A2 = 2 ft2 h1 Ra = 2 R1 = 1 Tank 1 A2 = 1 ft2 h2 R2 = 1 Tank 2 FIGURE P6–7 cou9789x_ch06_123-136.indd 134 8/14/08 6:21:51 PM Rev. Confirming Pages CHAPTER 6 CAPSULE SUMMARY Noninteracting systems For two systems in series, if the output from system 1 is not affected by the output from system 2, the systems are said to be noninteracting. q (t) A1 X0 G1 X1 X2 G2 Xn⫺1 Gn Xn h1 R1 A2 q1 X n (s ) ⫽ X 0 (s ) R2 h2 n ∏ Gi i ⫽1 q2 H 2( s ) Q (s ) ⫽ 1 R2 t 1s ⫹ 1 t 2 s ⫹ 1 Interacting systems The output from system 1 is affected by the output from system 2. The overall transfer function for the process is not merely the product of the transfer functions in series. q (t) A1 A2 h1 R1 h2 q1 X n (s ) ⫽ X 0 (s ) n ∏ Gi i ⫽1 R2 q2 135 cou9789x_ch06_123-136.indd 135 8/30/08 3:43:38 PM Confirming Pages 136 PART 2 LINEAR OPEN-LOOP SYSTEMS For the interacting two-tank system, the transfer function is H 2 (s ) R2 ⫽ Q (s ) t 1t 2 s 2 ⫹ (t 1 ⫹ t 2 ⫹ A1 R2 )s ⫹ 1 Note the presence of the cross-product term in the denominator. This term has the effect of slowing down the response of the process. 1.0 Q = u(t) 0.8 1 Q2 1 Q2/Q 0.6 Noninteracting Interacting 0.4 0 cou9789x_ch06_123-136.indd 136 1 Q = u(t) 0.2 0 1 2 1 Q2 3 8/14/08 6:21:53 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 7.1 SECOND-ORDER SYSTEM Transfer Function This section introduces a basic system called a second-order system or a quadratic lag. Second-order systems are described by a second-order differential equation that relates the ouput variable y to the input variable x (the forcing function) with time as the independent variable. A d2y dy ⫹ Cy ⫽ x (t ) 2 ⫹ B dt dt (7.1) A second-order system can arise from two first-order systems in series, as we saw in Chap. 6. Some systems are inherently second-order, and they do not result from a series combination of two first-order systems. Inherently second-order systems are not extremely common in chemical engineering applications. Most second-order systems that we encounter will result from the addition of a controller to a first-order process. Let’s examine an inherently second-order system and develop some terminology that will be useful in our analysis of the control of chemical processes. Consider a simple manometer as shown in Fig. 7–1. The pressure on both legs of the manometer is initially the same. The length of the fluid column in the manometer is L. At time t 0, a pressure difference is imposed across the legs of the manometer. Assuming the resulting flow in the manometer to be laminar and the steady-state friction law for drag force in laminar flow to apply at each instant, we will determine the transfer function between the applied pressure difference P and the manometer reading h. If we perform a momentum balance on the fluid in the manometer, we arrive at the following terms: (Sum of forces causing fluid to move ) ⫽ (Rate of change of momentum of fluid) (7.2) 137 cou9789x_ch07_137-162.indd 137 8/14/08 5:58:48 PM Confirming Pages 138 PART 2 LINEAR OPEN-LOOP SYSTEMS P1 = 0 P2 = 0 D P1 Reference level t=0 P2 h h/2 h/2 L Before (Initial) After (Final) FIGURE 7–1 Manometer. where Sum of forces Frictional forces Un b alanced pressure forces ⫺ causing fluid to move ⫽ opposing motion causing motion p D2 p D2 Unbalanced pressure forces ⫽ ( P1 ⫺ P2 ) ⫺ rgh causing motion 4 4 Frictional forces Skin friction Shear stress Area in contact opposing motion ⫽ at wall ⫽ at wall ⫻ with wall 8 mV Frictional forces 8 m 1 dh opposing motion ⫽ t Wall (p DL) ⫽ D (p DL ) ⫽ D 2 dt (p DL ) The term for the skin friction at the wall is obtained from the Hagen-Poiseuille relationship for laminar flow (McCabe, Smith and Harriott, 2004). Note that V is the average velocity of the fluid in the tube, which is also the velocity of the interface, which is equal to 12 dh /dt (see Fig. 7–2). cou9789x_ch07_137-162.indd 138 8/14/08 5:58:50 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 139 P2 P1 V V h h/2 h/2 Reference level t=0 FIGURE 7–2 Average velocity of the fluid in the manometer. The rate of change of momentum of the fluid [the right side of Eq. (7.2)] may be expressed as d (Rate of change of momentum) ⫽ (mass ⫻ velocity ⫻ momentum correction factor ) dt p D2 ⫽ r 4 dV L ( b ) dt p D2 ⫽ r 4 1 d 2h L ( b ) 2 dt 2 The momentum correction factor b accounts for the fact that the fluid has a parabolic velocity profile in the tube, and the momentum must be expressed as b m V for laminar flow (see McCabe, Smith and Harriott, 2004). The value of b for laminar flow is 4/3. Substituting the appropriate terms into Eq. (7.2) produces the desired force balance equation for the manometer. p D2 r 4 4 1 d 2h p D2 p D 2 8 m 1 dh L ⫽ ( P ⫺ P ⫺ r gh ⫺ ) (p DL (7.3) 1 2 D 2 dt 4 4 3 2 dt 2 Rearranging Eq. (7.3), we obtain p D2 r 4 4 1 d 2 h 8 m 1 dh p D2 p D2 + L ( p DL ⫹ r gh ⫽ P ⫺ P ) ( ) 1 2 4 4 3 2 dt 2 D 2 dt and finally, dividing both sides by rg(p D2/4), we arrive at the standard form for a second-order system. cou9789x_ch07_137-162.indd 139 8/14/08 5:58:53 PM Confirming Pages 140 PART 2 LINEAR OPEN-LOOP SYSTEMS 2L d 2h 16 m L dh P1 ⫺ P2 ∆P ⫹ ⫹h ⫽ ⫽ 2 dt 3g dt 2 r g rg rD g (7.4) (A more detailed version of the analysis of the manometer can be found in Bird et al., 1960). Note that as with first-order systems, standard form has a coefficient of 1 on the dependent variable term, h in this case. Second-order systems are described by a secondorder differential equation. We may rewrite this Eq. (7.4) in general terms as t2 d 2Y dY ⫹ 2zt ⫹ Y ⫽ X (t ) dt dt 2 (7.5) where 2L 3g (7.6) 2zt ⫽ 16 m L rD 2 g (7.7) ∆P rg and t2 ⫽ X (t ) ⫽ Y ⫽ h (7.8) Solving for t and z from Eqs. (7.6) and (7.7) gives t ⫽ z ⫽ 8m rD 2 2L s 3g 3L dimensionless 2g (7.9) (7.10) By definition, both t and z must be positive. The reason for introducing t and z in the particular form shown in Eq. (7.5) will become clear when we discuss the solution of Eq. (7.5) for particular forcing functions X(t). Equation (7.5) is written in a standard form that is widely used in control theory. If the fluid column is motionless (dY/dt 0) and located at its rest position (Y 0) before the forcing function is applied, the Laplace transform of Eq. (7.4) becomes t 2 s 2Y (s) ⫹ 2zt sY (s) ⫹ Y (s) ⫽ X (s) (7.11) From this, the transfer function follows: Y (s) 1 ⫽ 2 2 X (s) t s ⫹ 2zt s ⫹ 1 (7.12) The transfer function given by Eq. (7.12) is written in standard form, and we will show later that other physical systems can be represented by a transfer function having the cou9789x_ch07_137-162.indd 140 8/14/08 5:58:54 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 141 denominator of t2s2 2zt s 1. All such systems are defined as second-order. Note that it requires two parameters, t and z, to characterize the dynamics of a second-order system in contrast to only one parameter for a first-order system. We now discuss the response of a second-order system to some of the common forcing functions, namely, step, impulse, and sinusoidal. Step Response If the forcing function is a unit-step function, we have 1 (7.13) s In terms of the manometer shown in Fig. 7–1, this is equivalent to suddenly applying a pressure difference [such that X(t) P/rg 1] across the legs of the manometer at time t 0. Superposition will enable us to determine easily the response to a step function of any other magnitude. Combining Eq. (7.13) with the transfer function of Eq. (7.12) gives X (s ) ⫽ Y (s ) ⫽ 1 1 s t 2 s 2 ⫹ 2zt s ⫹ 1 (7.14) The quadratic term in this equation may be factored into two linear terms that contain the roots sa ⫽ ⫺ z2 ⫺1 z ⫹ t t sb ⫽ ⫺ z2 ⫺1 z ⫹ t t (7.15) (7.16) Equation (7.14) can now be written 1/t 2 (7.17) s ( s ⫺ sa ) ( s ⫺ sb ) The response of the system Y(t) can be found by inverting Eq. (7.17). The roots sa and sb will be real or complex depending on value of the parameter z. The nature of the roots will, in turn, affect the form of Y(t). The problem may be divided into the three cases shown in Table 7.1. Each case will now be discussed. Y (s ) ⫽ TABLE 7–1 Step response of a second-order system Case I II III cou9789x_ch07_137-162.indd 141 y <1 1 >1 Nature of roots Description of response Complex Real and equal Real Underdamped or oscillatory Critically damped Overdamped or nonoscillatory 8/14/08 5:58:55 PM Confirming Pages 142 PART 2 LINEAR OPEN-LOOP SYSTEMS CASE I STEP RESPONSE FOR y < 1. For this case, the inversion of Eq. (7.17) yields the result Y (t ) ⫽ 1 ⫺ 1 1⫺z2 1⫺z2 t e⫺z t /t sin 1 ⫺ z 2 ⫹ tan⫺1 t z (7.18) To derive Eq. (7.18), use is made of the techniques of Chap. 3. Since z < 1, Eqs. (7.15) to (7.17) indicate a pair of complex conjugate roots in the left half-plane and a root at the origin. In terms of the symbols of Fig. 3–1, the complex roots correspond to s2 and s2* and the root at the origin to s6. The reader should realize that in Eq. (7.18), the argument of the sine function is in radians, as is the value of the inverse tangent term. By referring to Table 3.1, we see that Y(t) has the form t t Y (t ) ⫽ C1 ⫹ e⫺z t /t C2 cos 1 ⫺ z 2 ⫹ C3 sin 1 ⫺ z 2 t t (7.19) The constants C1, C2, and C3 are found by partial fractions. The resulting equation is then put in the form of Eq. (7.18) by applying the trigonometric identity used in Chap. 4, Eq. (4.26). The details are left as an exercise for the reader. It is evident from Eq. (7.18) that Y(t) → 1 as t → . The nature of the response can be understood most clearly by plotting the solution to Eq. (7.17) as shown in Fig. 7–3, where Y(t) is plotted against the dimensionless variable t/t for several values of z, including those above unity, which will be considered in the next section. Note that for z < 1 all the response curves are oscillatory in nature and become less oscillatory as z is increased. The slope at the origin in Fig. 7–3 is zero for all values of z. The response of a second-order system for z < 1 is said to be underdamped. What is the physical significance of an underdamped response? Using the manometer as an example, if we step-change the pressure difference across an underdamped manometer, the liquid levels in the two legs will oscillate before stabilizing. The oscillations are characteristic of an underdamped response. CASE II STEP RESPONSE FOR y ⴝ 1. For this case, the response is given by the expression t Y (t ) ⫽ 1 ⫺ 1 ⫹ e⫺t / t t (7.20) This is derived as follows: Equations (7.15) and (7.16) show that the roots s1 and s2 are real and equal. By referring to Fig. 3–1 and Table 3.1, it is seen that Eq. (7.20) is the correct form. The constants are obtained, as usual, by partial fractions. The response, which is plotted in Fig. 7–3, is nonoscillatory. This condition, z 1, is called critical damping and allows the most rapid approach of the response to Y 1 without oscillation. cou9789x_ch07_137-162.indd 142 8/14/08 5:58:56 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 143 1.6 1.4 1.2 Y(t) 1.0 ζ = 0.8 0.6 0.4 0.2 0.8 1.4 1.2 ζ = 1.0 0.6 0.4 0.2 0 0 2 4 6 8 10 t/ FIGURE 7–3 Response of a second-order system to a unit-step forcing function. CASE III STEP RESPONSE FOR z > 1. For this case, the inversion of Eq. (7.17) gives the result t z t sinh z 2 ⫺ 1 Y (t ) ⫽ 1 ⫺ e⫺z t /t cosh z 2 ⫺ 1 ⫹ t t z2 − 1 (7.21) where the hyperbolic functions are defined as sinh a ⫽ e a ⫺ e⫺a 2 e a ⫹ e⫺a 2 The procedure for obtaining Eq. (7.21) is parallel to that used in the previous cases. The response has been plotted in Fig. 7–3 for several values of z. Notice that the response is nonoscillatory and becomes more “sluggish” as z increases. This is known as an overdamped response. As in previous cases, all curves eventually approach the line Y 1. cosh a ⫽ cou9789x_ch07_137-162.indd 143 8/14/08 5:58:57 PM Confirming Pages 144 PART 2 LINEAR OPEN-LOOP SYSTEMS Actually, the response for z > 1 is not new. We saw it previously in the discussion of the step response of a system containing two first-order systems in series, for which the transfer function is Y (s ) 1 ⫽ X (s ) (t 1s ⫹ 1) (t 2 s ⫹ 1) (7.22) This is true for z > 1 because the roots s1 and s2 are real, and the denominator of Eq. (7.12) may be factored into two real linear factors. Therefore, Eq. (7.12) is equivalent to Eq. (7.22) in this case. By comparing the linear factors of the denominator of Eq. (7.12) with those of Eq. (7.22), it follows that ( ) (7.23) ( ) (7.24) t1 ⫽ z ⫹ z 2 ⫺ 1 t t2 ⫽ z ⫺ z 2 ⫺ 1 t Note that if t1 t 2, then t t 1 t 2 and z 1. The reader should verify these results. Using MATLAB/Simulink to Determine the Step Response of the Manometer Consider a manometer as illustrated in Fig. 7–1. The manometer is being used to determine the pressure difference between two instrument taps on an air line. The working fluid in the manometer is water. Determine the response of the manometer to a step change in pressure across the legs of the manometer. Data L 200 cm g 980 cm/s2 m ⫽ 1 cP ⫽ 0 . 01 g/ (cm⭈ s) for th e working fluid, water r ⫽ 1 . 0 g/cm 3 for t ⬍ 0 ∆P 0 ⫽ 10 cm for t ⱖ 0 rg D 0.11 cm, 0.21 cm, 0.31 cm (Three Cases) Solution. From Eq. (7.4), we have the governing differential equation for the manometer: 2L d 2h 16 mL dh P1 ⫺ P2 ∆P ⫹ ⫹h ⫽ ⫽ 2 dt 3g dt 2 r g rg rD g cou9789x_ch07_137-162.indd 144 8/14/08 5:58:58 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 145 In terms of transformed deviation variables, this becomes t 2 s 2Y (s) ⫹ 2zt sY (s) ⫹ Y (s) ⫽ X (s) Y ⫽ h ⫺ hs where t ⫽ ∆P ∆P X⫽ ⫺ rg s rg and 3L 2g z ⫽ and 8m rD 2 3L 2g Let’s calculate the time constant for the manometer. t ⫽ 2L ⫽ 3g 2(200 cm) ⫽ 0 . 369 s 3(980 cm 2 /s ) and the damping coefficient for the three different tube diameters z ⫽ 8m rD 2 3L 8[0 . 01 g/ (cm ⭈ s)] 3(200 cm) 0 . 0443 ⫽ ⫽ 3 2 2 2g (1 . 0 g/cm )( D ) 2(980 cm/s ) D2 Diameter (cm) 0.11 0.21 0.31 3.66 1.00 0.46 Clearly we have one underdamped system (z < 1), one critically damped system (z 1), and one overdamped system (z > 1). One method of obtaining the responses is to substitute the values of t and z into Eqs. (7.18), (7.20), and (7.21) and plot the resulting equations, realizing that the forcing function is 10 times a unit step. Another way to obtain the responses is to use MATLAB and Simulink to obtain the response of the transfer function Y/X to the forcing function input X. Y (s ) 1 ⫽ 22 X (s ) t s ⫹ 2zt s ⫹ 1 X ⫽ and 10 s The three necessary transfer functions are as follows: Diameter (cm) s z s2 2zs Transfer function 0.11 0.369 3.66 0.136 2.70 1 0 . 136s 2 ⫹ 2 . 70 s ⫹ 1 0.21 0.369 1.00 0.136 0.738 1 0 . 136s 2 ⫹ 0 . 738s ⫹ 1 0.31 0.369 0.46 0.136 0.340 1 0 . 136s 2 ⫹ 0 . 340 s ⫹ 1 (continued) cou9789x_ch07_137-162.indd 145 8/14/08 5:58:58 PM Confirming Pages 146 PART 2 LINEAR OPEN-LOOP SYSTEMS The Simulink model for simulating the transfer functions is shown in Fig. 7–4, and the response is shown in Fig. 7–5. 1 0.136s2 + 2.70s + 1 Overdamped manometer 1 0.136s2 + 0.738s + 1 Pressure forcing function 10/s Scope Critically damped manometer 1 0.136s2 + 0.340s + 1 Underdamped manometer FIGURE 7–4 Simulink diagram for manometer simulation. 12 Underdamped 10 Overdamped 8 Critically damped Height 6 4 2 0 0 1 2 3 4 5 6 7 8 9 10 Time FIGURE 7–5 Manometer response to step input. cou9789x_ch07_137-162.indd 146 8/14/08 5:58:59 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 147 Substituting the values for t and z into Eqs. (7.18), (7.20), and (7.21), we get 1 ⫺z 2 1 t Y (t ) ⫽ 10 1 ⫺ e⫺zt / t sin 1 ⫺ z 2 ⫹ tan⫺1 2 t z 1 ⫺ z underdamped manometer ⫺1 1.93 ⫽ 10 1 ⫺ 1.1 3 e⫺1.25t sin 2 . 41t ⫹ tan 1.09 rad t t ⫺t / 0.369 Y (t ) ⫽ 10 1 ⫺ 1 ⫹ e⫺t /t ⫽ 10 1 ⫺ 1 ⫹ e critically damped manometer 0 . 369 t Y (t ) ⫽ 10 1 ⫺ e⫺zt / t { t z t 2 sinh z 2 ⫺ 1 cosh z ⫺ 1 ⫹ t t z2 ⫺1 overdampee d manometer } ⫽ 10 1 ⫺ e⫺9.92 t [ cosh(9.54 t ) ⫹ 1.04 sin h(9.54 t ) ] Plotting these responses gives the same results as the Simulink model. On a practical note, notice that t increases with the total length of the fluid column and that z increases with the viscosity of the fluid. If the damping coefficient z is small (< < 1.0), the response of the manometer to a change in pressure can be very oscillatory, and it becomes difficult to obtain accurate readings of the pressure. To dampen the oscillations, it is common practice to place a restriction on the bend of the tube. This increases the drag force of the fluid and is equivalent to increasing m in the equation for z. Such a restriction (a partially open valve) is called a snubber. Terms Used to Describe an Underdamped System Of these three cases, the underdamped response occurs most frequently in control systems. Hence a number of terms are used to describe the underdamped response quantitatively. Equations for some of these terms are listed below for future reference. In general, the terms depend on z and/or t. All these equations can be derived from the time response as given by Eq. (7.18); however, the mathematical derivations are left to the reader as exercises. 1. Overshoot. Overshoot is a measure of how much the response exceeds the ultimate value following a step change and is expressed as the ratio A/B in Fig. 7–6. The overshoot for a unit step is related to z by the expression ⫺pz Overshoot ⫽ exp 1 ⫺ z 2 (7.25) This relation is plotted in Fig. 7–7. The overshoot increases for decreasing z. cou9789x_ch07_137-162.indd 147 8/14/08 5:59:01 PM Confirming Pages 148 PART 2 LINEAR OPEN-LOOP SYSTEMS Period T Response time limit C A Y(t) 1.0 T B 0 0 tr Rise time Response time t FIGURE 7–6 Terms used to describe an underdamped second-order response. 1.0 0.8 f fn 0.6 0.4 Overshoot 0.2 Decay ratio 0 0 0.2 0.4 0.6 0.8 1.0 ζ FIGURE 7–7 Characteristics of a step response of an underdamped second-order system. 2. Why are we concerned about overshoot? Perhaps the temperature in our chemical reactor cannot be allowed to exceed a specified temperature to protect the catalyst from deactivation, or if it’s a level control system, we don’t want the tank to overflow. If we know these physical limitations, we can determine allowable values of z and choose our control system parameters to be sure to stay within those limits. Decay ratio. The decay ratio is defined as the ratio of the sizes of successive peaks and is given by C/A in Fig. 7–6. The decay ratio is related to z by the expression ⫺2p z 2 Decay ratio ⫽ exp ⫽ (overshoot) 1 ⫺ z 2 (7.26) which is plotted in Fig. 7–7. Notice that larger z means greater damping, hence greater decay. cou9789x_ch07_137-162.indd 148 8/14/08 5:59:01 PM Confirming Pages CHAPTER 7 3. 4. 5. HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 149 Rise time. This is the time required for the response to first reach its ultimate value and is labeled tr in Fig. 7–6. The reader can verify from Fig. 7–3 that tr increases with increasing z. Response time. This is the time required for the response to come within 5 percent of its ultimate value and remain there. The response time is indicated in Fig. 7–6. The limits 5 percent are arbitrary, and other limits can be used for defining a response time. Period of oscillation. From Eq. (7.18), the radian frequency (radians/time) is the coefficient of t in the sine term; thus, radian frequency w ⫽ 1⫺z2 t (7.27) Since the radian frequency w is related to the cyclical frequency f by w 2pf, it follows that f ⫽ 6. 1⫺z2 t 1 1 ⫽ T 2p (7.28) where T is the period of oscillation (time/cycle). In terms of Fig. 7–6, T is the time elapsed between peaks. It is also the time elapsed between alternate crossings of the line Y 1. Natural period of oscillation. If the damping is eliminated [B 0 in Eq. (7.1), or z 0], the system oscillates continuously without attenuation in amplitude. Under these “natural” or undamped conditions, the radian frequency is 1/t, as shown by Eq. (7.27) when z 0. This frequency is referred to as the natural frequency w n: wn ⫽ 1 t (7.29) The corresponding natural cyclical frequency fn and period Tn are related by the expression 1 1 fn ⫽ ⫽ (7.30) Tn 2pt Thus, t has the significance of the undamped period. From Eqs. (7.28) and (7.30), the natural frequency is related to the actual frequency by the expression f ⫽ 1⫺z2 fn which is plotted in Fig. 7–7. Notice that for z < 0.5 the natural frequency is nearly the same as the actual frequency. In summary, it is evident that z is a measure of the degree of damping, or the oscillatory character, and t is a measure of the period, or speed, of the response of a second-order system. cou9789x_ch07_137-162.indd 149 8/14/08 5:59:02 PM Confirming Pages 150 PART 2 LINEAR OPEN-LOOP SYSTEMS Impulse Response If a unit impulse d (t) is applied to the second-order system, then from Eqs. (7.12) and (3A.1) the transform of the response is Y (s ) ⫽ 1 t s ⫹ 2zt s ⫹ 1 2 2 (7.31) As in the case of the step input, the nature of the response to a unit impulse will depend on whether the roots of the denominator of Eq. (7.31) are real or complex. The problem is again divided into the three cases shown in Table 7.1, and each is discussed below. CASE I IMPULSE RESPONSE FOR y < 1. The inversion of Eq. (7.31) for z < 1 yields the result Y (t ) ⫽ 1 t 1 e⫺z t / t sin 1 ⫺ z 2 2 t t 1⫺z (7.32) which is plotted in Fig. 7–8. The slope at the origin in Fig. 7–8 is 1.0 for all values of z. A simple way to obtain Eq. (7.32) from the step response of Eq. (7.18) is to take the derivative of Eq. (7.18) with respect to t (remember from App. 3A that the derivative of the unit-step function is the impulse function). Comparison of Eqs. (7.14) and (7.31) shows that (7.33) Y (s)impulse ⫽ sY (s)step 0.8 ζ = 0.2 0.6 0.4 0.6 0.8 0.4 τ Y(t) 1.0 = 1.2 1.4 0.2 0 −0.2 −0.4 0 2 4 6 8 10 t/ FIGURE 7–8 Response of a second-order system to a unit-impulse forcing function. cou9789x_ch07_137-162.indd 150 8/14/08 5:59:03 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 151 The presence of s on the right side of Eq. (7.33) implies differentiation with respect to t in the time response. In other words, the inverse transform of Eq. (7.31) is Y (t )impulse ⫽ ( d Y (t )step dt ) (7.34) Application of Eq. (7.34) to Eq. (7.18) yields Eq. (7.32). This principle also yields the results for the next two cases. CASE II IMPULSE RESPONSE FOR y 1. For the critically damped case, the response is given by Y (t ) ⫽ 1 ⫺t / t te t2 (7.35) which is plotted in Fig. 7–8. CASE III IMPULSE RESPONSE FOR y > 1. For the overdamped case, the response is given by Y (t ) ⫽ t 1 1 e⫺zt / t sinh z 2 ⫺ 1 t t z 2⫺1 (7.36) which is also plotted in Fig. 7–8. To summarize, the impulse–response curves of Fig. 7–8 show the same general behavior as the step response curves of Fig. 7–3. However, the impulse response always returns to zero. Terms such as decay ratio, period of oscillation, etc., may also be used to describe the impulse response. Many control systems exhibit transient responses such as those of Fig. 7–8. Sinusoidal Response If the forcing function applied to the second-order system is sinusoidal X (t ) ⫽ A sin w t then it follows from Eqs. (7.12) and (4.23) that Y (s ) ⫽ (s Aw 2 ⫹w 2 ) (t 2s2 ⫹ 2zt s ⫹ 1) (7.37) The inversion of Eq. (7.37) may be accomplished by first factoring the two quadratic terms to give Y (s ) ⫽ cou9789x_ch07_137-162.indd 151 Aw /t 2 ( s ⫺ jw ) ( s ⫹ jw ) ( s ⫺ sa ) ( s ⫺ sb ) (7.38) 8/14/08 5:59:05 PM Confirming Pages 152 PART 2 LINEAR OPEN-LOOP SYSTEMS Here sa and sb are the roots of the denominator of the transfer function and are given by Eqs. (7.15) and (7.16). For the case of an underdamped system (z < 1), the roots of the denominator of Eq. (7.38) are a pair of pure imaginary roots (1jw, jw) contributed by the forc- ( ing function and a pair of complex roots ⫺z /t ⫹ j 1 ⫺ z 2 / t , ⫺ z /t ⫺ j 1 ⫺ z 2 / t We may write the form of the response Y(t) by referring to Fig. 3–1 and Table 3.1; thus t t Y (t ) ⫽ C1 cos w t ⫹ C2 sin w t ⫹ e⫺z t / t C3 cos 1 ⫺ z 2 ⫹ C4 sin 1 ⫺ z 2 t t (7.39) The constants are evaluated by partial fractions. Notice in Eq. (7.39) that as t → , only the first two terms do not become zero. These remaining terms are the ultimate periodic solution; thus Y (t )t → ⬁ ⫽ C1 cos w t ⫹ C2 sin w t (7.40) The reader should verify that Eq. (7.40) is also true for z 1. From this little effort, we see already that the response of the second-order system to a sinusoidal driving function is ultimately sinusoidal and has the same frequency as the driving function. If the constants C1 and C2 are evaluated, we get from Eqs. (4.26) and (7.40) Y (t ) ⫽ A 1 ⫺ (w t )2 2 ⫹ ( 2zwt )2 f ⫽ ⫺tan⫺1 where sin (w t ⫹ f ) (7.41) 2zwt 1 ⫺ (wt )2 By comparing Eq. (7.41) with the forcing function X (t ) ⫽ A sin w t it is seen that 1. The ratio of the output amplitude to the input amplitude is Amplitude ratio ⫽ 2. output amplitude ⫽ input amplitude 2 1 ⫺ (wt ) ⫹ ( 2zwt ) It will be shown in Chap. 15 that this may be greater or less than 1, depending upon the values of z and w t. This is in direct contrast to the sinusoidal response of the first-order system, where the ratio of the output amplitude to the input amplitude is always less than 1. The output lags the input by phase angle |f |. f ⫽ phase angle ⫽ ⫺tan⫺1 cou9789x_ch07_137-162.indd 152 1 2 2 2 zwt 1 ⫺ (wt )2 8/14/08 5:59:06 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 153 It can be seen from Eq. (7.41), and will be shown in Chap. 15, that |f | approaches 180° asymptotically as w increases. The phase lag of the first-order system, on the other hand, can never exceed 90°. Discussion of other characteristics of the sinusoidal response will be deferred until Chap. 15. We now have at our disposal considerable information about the dynamic behavior of the second-order system. It happens that many control systems that are not truly second-order exhibit step responses very similar to those of Fig. 7–3. Such systems are often characterized by second-order equations for approximate mathematical analysis. Hence, the second-order system is quite important in control theory, and frequent use will be made of the material in this chapter. 7.2 TRANSPORTATION LAG A phenomenon that is often present in flow systems is the transportation lag. Synonyms for this term are dead time and distance velocity lag. As an example, consider the system shown in Fig. 7–9, in which a liquid flows through an insulated tube of uniform cross-sectional area A and length L at a constant volumetric flow rate q. The density r Cross-sectional area = A and the heat capacity C are constant. The tube y(t) x(t) wall has negligible heat capacity, and the velocq q ity profile is flat (plug flow). L The temperature x of the entering fluid FIGURE 7–9 varies with time, and it is desired to find the System with transportation lag. response of the outlet temperature y(t) in terms of a transfer function. As usual, it is assumed that the system is initially at steady state; for this system, it is obvious that the inlet temperature equals the outlet temperature; i.e., xs ⫽ ys (7.42) If a step change were made in x(t) at t 0, the change would not be detected at the end of the tube until t s later, where t is the time required for the entering fluid to pass through the tube. This simple step response is shown in Fig. 7–10. If the variation in x(t) were some arbitrary function, as shown in Fig. 7–10, the response y(t) at the end of the pipe would be identical with x(t) but again delayed by t x(t) x(t) τ y(t) y(t) τ 0 (a) t 0 τ (b) t FIGURE 7–10 Response of transportation lag to various inputs. cou9789x_ch07_137-162.indd 153 8/14/08 5:59:07 PM Confirming Pages 154 PART 2 LINEAR OPEN-LOOP SYSTEMS units of time. The transportation lag parameter t is simply the time needed for a particle of fluid to flow from the entrance of the tube to the exit, and it can be calculated from the expression t ⫽ volume of tube volumetric flow rate t ⫽ or AL q (7.43) It can be seen from Fig. 7–10 that the relationship between y(t) and x(t) is y( t ) ⫽ x ( t ⫺ t ) (7.44) Subtracting Eq. (7.42) from Eq. (7.44) and introducing the deviation variables X x xs and Y y ys give Y (t ) ⫽ X ( t ⫺ t ) (7.45) If the Laplace transform of X(t) is X(s), then the Laplace transform of X(t t) is est X(s). This result follows from the theorem on translation of a function, which was discussed in App. 3A. Equation (7.45) becomes Y (s) ⫽ e⫺st X (s) Y (s ) ⫽ e⫺st X (s ) or (7.46) Therefore, the transfer function of a transportation lag is est. The transportation lag is quite common in the chemical process industries where a fluid is transported through a pipe. We shall see in a later chapter that the presence of a transportation lag in a control system can make it much more difficult to control. In general, such lags should be avoided if possible by placing equipment close together. They can seldom be entirely eliminated. APPROXIMATION OF TRANSPORT LAG. The transport lag is quite different from the other transfer functions (first-order, second-order, etc.) that we have discussed in that it is not a rational function (i.e., a ratio of polynomials.) As shown in Chap. 13, a system containing a transport lag cannot be analyzed for stability by the Routh test. The transport lag can also be difficult to simulate by computer. For these reasons, several approximations of transport lag that are useful in control calculations are presented here. One approach to approximating the transport lag is to write et s as 1/et s and to express the denominator as a Taylor series; the result is e⫺t s ⫽ cou9789x_ch07_137-162.indd 154 1 1 ⫽ 2 2 ts 1 ⫹ t s ⫹ t s / 2 ⫹ t 3 s 3 / 3! ⫹ e 8/14/08 5:59:08 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 155 Keeping only the first two terms in the denominator gives 1 (7.47) 1 ⫹ ts This approximation, which is simply a first-order lag, is a crude approximation of a transport lag. An improvement can be made by expressing the transport lag as e⫺t s ⬵ e⫺t s / 2 et s / 2 Expanding numerator and denominator in a Taylor series and keeping only terms of first-order give e⫺t s ⫽ 1 ⫺ ts / 2 (7.48) first-order Padé 1 ⫹ ts / 2 This expression is also known as a first-order Padé approximation. Another well-known approximation for a transport lag is the second-order Padé approximation: e⫺t s ⬵ e⫺t s ⬵ 1 ⫺ t s / 2 ⫹ t 2 s 2 /12 1 ⫹ t s / 2 ⫹ t 2 s 2 /12 second-ordee r Padé (7.49) Equation (7.48) is not merely the ratio of two Taylor series; it has been optimized to give a better approximation. The step responses 1.0 of the three approximations of transport lag presented 0.6 (1) here are shown in Fig. 7–11. The step response of ets is (4) 0.2 also shown for comparison. Y 0 Notice that the response for the first-order Padé approxi−0.2 mation drops to 1 before ris(3) ing exponentially toward 1. −0.6 (2) The response for the secondorder Padé approximation −1.0 0 t/τ 2 1 jumps to 1 and then descends to below 0 before returning FIGURE 7–11 gradually back to 1. Step response to approximation of the transport lag ets: Although none of the 1 ; (2) first-order Padé; (3) second-order Padé ; (4) ets. (1) approximations for ets is ts ⫹ 1 very accurate, the approximation for et s is more useful when it is multiplied by several first-order or second-order transfer functions. In this case, the other transfer functions filter out the high-frequency content of the signals passing through the transport lag, with the result that the transport lag approximation, when combined with other transfer functions, provides a satisfactory result in many cases. The accuracy of a transport lag can be evaluated most clearly in terms of frequency response, a topic covered later in this book. cou9789x_ch07_137-162.indd 155 8/14/08 5:59:12 PM Confirming Pages 156 PART 2 LINEAR OPEN-LOOP SYSTEMS SUMMARY After studying the material in this chapter, we now have at our disposal considerable information about the dynamic behavior of the second-order systems and transportation lags. We noted that even though many control systems are not truly second-order, they frequently exhibit step responses very similar to what we have observed in this chapter. Such systems are often characterized by second-order equations for an approximate mathematical analysis. Therefore, the second-order system is quite important in control theory, and we will make use of this material often in future chapters. PROBLEMS 7.1. A step change of magnitude 4 is introduced into a system having the transfer function Y (s ) 10 ⫽ 2 X (s ) s ⫹ 1 . 6s ⫹ 4 Determine (a) Percent overshoot (b) Rise time (c) Maximum value of Y(t) (d) Ultimate value of Y(t) (e) Period of oscillation 20 ft3/min A1 10 ft3 h1 R1 A2 h2 R2 A1 ⫽ A2 ⫽ 10 ft 2 R1 ⫽ 0. 1 ft/cfm R2 ⫽ 0. 35 ft/cfm FIGURE P7–2 q A1 h1 R1 A2 h2 FIGURE P7–3 cou9789x_ch07_137-162.indd 156 7.2. The two-tank system shown in Fig. P7–2 is operating at steady state. At time t 0, 10 ft3 of water is quickly added to the first tank. Using appropriate figures and equations in the text, determine the maximum deviation in level (feet) in both tanks from the ultimate steady-state values and the time at which each maximum occurs. Data: R2 7.3. The two-tank liquid-level system shown in Fig. P7–3 is operating at steady state when a step change is made in the flow rate to tank 1. The transient response is critically damped, and it takes 1.0 min for the change in level of the second tank to reach 50 percent of the total change. If the ratio of the cross-sectional areas of the tanks is A1/A2 2, calculate the ratio R1/R2. Calculate the time constant for each tank. How long does it take for the change in level of the first tank to reach 90 percent of the total change? 8/14/08 5:59:14 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 157 7.4. Use Simulink to solve Prob. 7.3. 7.5. Design a mercury manometer that will measure pressures up to 2 atm absolute and will give responses that are slightly underdamped (that is, z 0.7). 7.6. Verify Eqs. (7.18), (7.20), and (7.21). 7.7. Verify Eqs. (7.25) and (7.26). 7.8. Verify Eq. (7.41). 7.9. If a second-order system is overdamped, it is more difficult to determine the parameters z and t experimentally. One method for determining the parameters from a step response has been suggested by R. C. Oldenbourg and H. Sartorius (The Dynamics of Automatic Controls. ASME, p. 78, 1948), as described below. (a) Show that the unit-step response for the overdamped case may be written in the form S (t ) ⫽ 1 ⫺ r1er2t ⫺ r2er1t r1 ⫺ r2 where r1 and r2 are the (real and negative) roots of t 2 s 2 ⫹ 2zt s ⫹ 1 ⫽ 0 (b) Show that S(t) has an inflection point at ti ⫽ ln ( r2 / r1 ) r1 / r2 (c) Show that the slope of the step response at the inflection point dS (t ) ⫽ S⬘( ti ) dt t ⫽ ti has the value S⬘( ti ) ⫽ ⫺r1er1ti ⫽ ⫺r2er2ti r ⫽ ⫺r1 2 r1 r1 / ( r1 ⫺r2 ) (d) Show that the value of the step response at the inflection point is S A ti B ⫽ 1 ⫹ and that hence r1 ⫹ r2 S⬘( ti ) r1r 2 1 ⫺ S ( ti ) 1 1 ⫽ ⫺ S⬘( ti ) r1 r2 (e) On a typical sketch of a unit-step response, show distances equal to 1 S⬘( ti ) and 1 ⫺ S ( ti ) S⬘( ti ) and hence present two simultaneous equations resulting from a graphical method for determination of r1 and r2. ( f ) Relate z and t to r1 and r2. cou9789x_ch07_137-162.indd 157 8/14/08 5:59:15 PM Confirming Pages 158 PART 2 LINEAR OPEN-LOOP SYSTEMS 7.10. Determine Y(0), Y(0.6), and Y() if Y (s ) ⫽ 1 25 ( s ⫹ 1) s s 2 ⫹ 2s ⫹ 25 7.11. In the liquid-level system shown in Fig. P7–11, the deviation in flow rate to the first tank is an impulse function of magnitude 5. The following data apply: A1 1 ft2, A2 A3 2 ft2, R1 1 ft /cfm, and R2 1.5 ft /cfm. Q(t) = 5δ (t) A1 h1 R1 Tank 1 A2 h2 R2 Tank 2 A3 h3 Constant flow Tank 3 FIGURE P7–11 (a) Determine expressions for H1(s), H2(s), and H3(s) where H1, H2, and H3 are deviations in tank level for tanks 1, 2, and 3. (b) Sketch the responses of H1 (t), H2 (t), and H3 (t). (You need show only the shape of the responses; do not plot.) (c) Determine H1 (3.46), H2 (3.46), and H3 (3.46). For H2 and H3, use graphs in Chap. 7 of this text after first finding values of t and z for an equivalent second-order system. 7.12. Sketch the response Y(t) if Y(s) e2s/[s2 1.2s 1]. Determine Y(t) for t 0, 1, 5, and . 7.13. The two tanks shown in Fig. P7–13 are connected in an interacting fashion. The system is initially at steady state with q 10 cfm. The following data apply to the tanks: A1 1 ft2, A2 1.25 ft2, R1 1 ft/cfm, and R2 0.8 ft/cfm. q A1 A2 h1 h2 R1 R2 FIGURE P7–13 cou9789x_ch07_137-162.indd 158 8/14/08 5:59:16 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 159 (a) If the flow changes from 10 to 11 cfm according to a step change, determine H2(s), i.e., the Laplace transform of H2(t), where H2 is the deviation in h2. (b) Determine H2(1), H2(4), and H2(). (c) Determine the initial levels (actual levels) h1(0) and h2(0) in the tanks. (d) Obtain an expression for H1(s) for the unit-step change described above. 7.14. From figures in this chapter, determine Y(4) for the system response expressed by y( s ) ⫽ 2 2s ⫹ 4 s 4 s 2 ⫹ 0 . 8s ⫹ 1 7.15. A step change of magnitude 3 is introduced into the transfer function Y (s ) 10 ⫽ 2 X (s) 2s ⫹ 0 . 3s ⫹ 0 . 5 Determine the overshoot and the frequency of oscillation. cou9789x_ch07_137-162.indd 159 8/14/08 5:59:17 PM Confirming Pages CHAPTER 7 CAPSULE SUMMARY STANDARD FORM FOR A SECOND-ORDER SYSTEM: 2 dY 2 d Y t dt 2 ⫹ 2zt dt ⫹ Y ⫽ X (t ) t ⫽ Time constant z ⫽ Damping coefficient (magn i tude of this parameter determines the natu u re of the response) Sample Second-Order System Response to a Unit Impulse Input Sample Second-Order System Response to a Unit-Step Input 0.7 1.4 Underdamped ζ<1 1.2 0.6 1 Output 0.8 0.6 Critically damped ζ=1 0.4 Critically damped ζ=1 Output Underdamped ζ<1 0.5 Input Step Overdamped ζ>1 0.3 0.2 Overdamped ζ>1 0.1 0.4 0 0.2 −0.1 0 0 1 2 3 4 5 6 Time 7 8 9 10 −0.2 0 2 4 6 8 10 12 Time 14 16 18 20 TERMS TO DESCRIBE AN UNDERDAMPED (OSCILLATORY) SECOND-ORDER RESPONSE ⫺pz A Overshoot ⫽ exp ⫽ 1 ⫺ z 2 B ⫺2pz C Decay ratio ⫽ exp ⫽ (overshoot)2 ⫽ 1 ⫺ z 2 A Radian frequency w ⫽ 1⫺z 2 t w n ⫽ natural frequenc y ⫽ 1 t 160 cou9789x_ch07_137-162.indd 160 8/14/08 5:59:18 PM Confirming Pages CHAPTER 7 HIGHER-ORDER SYSTEMS: SECOND-ORDER AND TRANSPORTATION LAG 161 Period T Response time limit C A Y(t) 1.0 T B 0 0 tr Rise time Response time t SINUSOIDAL RESPONSE OF A SECONDORDER SYSTEM A 2 A sin w t → 1⫺(w t)2 ⫹ ( 2zwt )2 1 → t 2 s 2 ⫹ 2zt s ⫹ 1 ultimate response after transsients disappear 1 output amplitude ⫽ input ampli t ude Amplitude ratio ⫽ 2zwt sin w t ⫹⫺tan⫺1 1⫺ (wt )2 2 2 1 ⫺ (wt ) ⫹ ( 2zwt ) f ⫽ phase angle ⫽ ⫺t a n⫺1 2 2zwt 1 ⫺ (wt )2 TRANSPORTATION LAG x(t) x(t) τ y(t) τ 0 (a) cou9789x_ch07_137-162.indd 161 t 0 y(t) τ (b) Y (s) ⫽ e⫺st X (s) t 8/14/08 5:59:22 PM cou9789x_ch07_137-162.indd 162 8/14/08 5:59:23 PM Confirming Pages PART III LINEAR CLOSED-LOOP SYSTEMS cou9789x_ch08_163-185.indd 163 8/13/08 4:10:17 PM cou9789x_ch08_163-185.indd 164 8/13/08 4:10:18 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 8.1 INTRODUCTION In previous chapters, the dynamic behavior of several basic systems was examined. With this background, we can extend the discussion to a complete control system and introduce the fundamental concept of feedback. To work with a familiar system, the treatment will be based on a stirred-tank heater. Figure 8–1 is a sketch of the apparatus. To orient the reader, the physical description of the process will be reviewed. A liquid stream at a temperature Ti enters an insulated, well-stirred tank at a constant flow rate w (mass/time). It is desired to maintain (or control) the temperature in the tank at TR by means of the controller. If the measured tank temperature Tm differs from the desired temperature TR, the controller senses the difference or error ε TR Tm and changes the heat input in such a way as to reduce the magnitude of e. If the controller changes the heat input to the tank by an amount that is proportional to e, we have proportional control. In Fig. 8–1, it is indicated that the source of heat input q may be electricity or steam. If an electrical source were used, the final control element might be a variable transformer that is used to adjust current to a resistance heating element; if steam were used, the final control element would be a control valve that adjusts the flow of steam. In either case, the output signal from the controller should adjust q in such a way as to maintain control of the temperature in the tank. 8.2 COMPONENTS OF A CONTROL SYSTEM The system shown in Fig. 8–1 may be divided into the following components: 1. 2. 3. 4. Process (stirred-tank heater). Measuring element (thermometer). Controller. Final control element (variable transformer or control valve). 165 cou9789x_ch08_163-185.indd 165 8/13/08 4:10:18 PM Confirming Pages 166 PART 3 LINEAR CLOSED-LOOP SYSTEMS Final control element Electrical power or steam Controller Tm Desired temperature (set point, TR) q w, Ti Temperature measuring element Process w, Ti FIGURE 8–1 Control system for a stirred-tank heater. Each of these components can be readily identified as a separate physical item in the process. In general, these four components will constitute most of the control systems that we consider in this text; however, the reader should realize that more complex control systems exist in which more components are used. For example, some processes require a cascade control system in which two controllers and two measuring elements are used. A cascade system is discussed in Chap. 17. 8.3 BLOCK DIAGRAM For computational purposes, it is convenient to represent the control system of Fig. 8–1 by means of the block diagram shown in Fig. 8–2. Such a diagram makes it much easier to visualize the relationships among the various signals. New terms, which appear in Fig. 8–2, are set point and load. The set point is a synonym for the desired value of the controlled variable. The load refers to a change in any variable that may cause the controlled variable of the process to change. In this example, the inlet temperature Ti is a load variable. Other possible loads for this system are changes in flow rate and heat loss from the tank. (These loads are not shown on the diagram.) Controller mechanism Comparator + TR Set point _ Controller Ti, load Final control element Error Tm Measured variable + + Process T Controlled variable Measuring element FIGURE 8–2 Block diagram of a simple control system. cou9789x_ch08_163-185.indd 166 8/13/08 4:10:19 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 167 The control system shown in Fig. 8–2 is called a closed-loop system or a feedback system because the measured value of the controlled variable is returned or “fed back” to a device called the comparator. In the comparator, the controlled variable is compared with the desired value or set point. If there is any difference between the measured variable and the set point, an error is generated. This error enters a controller, which in turn adjusts the final control element to return the controlled variable to the set point. Negative Feedback Versus Positive Feedback Several terms have been used that may need further clarification. The feedback principle, which is illustrated by Fig. 8–2, involves the use of the controlled variable T to maintain itself at a desired value TR. The arrangement of the apparatus of Fig. 8–2 is often described as negative feedback to contrast with another arrangement called positive feedback. Negative feedback ensures that the difference between TR and Tm is used to adjust the control element so that the tendency is to reduce the error. For example, assume that the system is at steady state and that T Tm TR. If the load Ti should increase, T and Tm would start to increase, which would cause the error e to become negative. With proportional control, the decrease in error would cause the controller and final control element to decrease the flow of heat to the system, with the result that the flow of heat would eventually be reduced to a value such that T approaches TR. A verbal description of the operation of a feedback control system, such as the one just given, is admittedly inadequate, for this description necessarily is given as a sequence of events. Actually all the components operate simultaneously, and the only adequate description of what is occurring is a set of simultaneous differential equations. This more accurate description is the primary subject matter of the present and succeeding chapters. If the signal to the comparator were obtained by adding TR and Tm, we would have a positive feedback system, which is inherently unstable. To see that this is true, again assume that the system is at steady state and that T Tm TR. If Ti were to increase, T and Tm would increase, which would cause the signal from the comparator (e in Fig. 8–2) to increase, with the result that the heat to the system would increase. However, this action, which is just the opposite of that needed, would cause T to increase further. It should be clear that this situation would cause T to “run away” and control would not be achieved. For this reason, positive feedback would never be used intentionally in the system of Fig. 8–2. However, in more complex systems it may arise naturally. An example of this is discussed in Chap. 20. Servo Problem Versus Regulator Problem The control system of Fig. 8–2 can be considered from the point of view of its ability to handle either of two types of situations. In the first situation, which is called the servomechanism-type (or servo) problem, we assume that there is no change in load Ti and that we are interested in changing the bath temperature according to some prescribed function of time. For this problem, the set point TR would be changed in accordance with the desired variation in bath temperature. If the variation is sufficiently cou9789x_ch08_163-185.indd 167 8/13/08 4:10:20 PM Confirming Pages 168 PART 3 LINEAR CLOSED-LOOP SYSTEMS slow, the bath temperature may be expected to follow the variation in TR very closely. There are occasions when a control system in the chemical industry will be operated in this manner. For example, one may be interested in varying the temperature of a reactor according to a prescribed time-temperature pattern. However, the majority of problems that may be described as the servo type come from fields other than the chemical industry. The tracking of missiles and aircraft and the automatic machining of intricate parts from a master pattern are well-known examples of the servo-type problem. The servo problem can be viewed as trying to follow a moving target (i.e., the changing set point). The other situation will be referred to as the regulator problem. In this case, the desired value TR is to remain fixed, and the purpose of the control system is to maintain the controlled variable at TR in spite of changes in load Ti. This problem is very common in the chemical industry, and a complicated industrial process will often have many selfcontained control systems, each of which maintains a particular process variable at a desired value. These control systems are of the regulator type. In considering control systems in the following chapters, we will frequently discuss the response of a linear control system to a change in set point (servo problem) separately from the response to a change in load (regulator problem). However, it should be realized that this is done only for convenience. The basic approach to obtaining the response of either type is essentially the same, and the two responses may be superimposed to obtain the response to any linear combination of set point and load changes. 8.4 DEVELOPMENT OF BLOCK DIAGRAM Each block in Fig. 8–2 represents the functional relationship existing between the input and output of a particular component. In previous chapters, such input-output relations were developed in the form of transfer functions. In block diagram representations of control systems, the variables selected are deviation variables, and inside each block is placed the transfer function relating the input-output pair of variables. Finally, the blocks are combined to give the overall block diagram. This is the procedure to be followed in developing Fig. 8–2. Process Consider first the block for the process. This block will be seen to differ somewhat from those presented in previous chapters in that two input variables are present; however, the procedure for developing the transfer function remains the same. An unsteady-state energy balance around the tank gives q wC ( Ti To ) wC ( T To ) rCV dT dt (8.1) where To is the reference temperature. [Note: In this analysis, it is assumed that the flow rate of heat q is instantaneously available and independent of the temperature in the tank. In some stirred-tank heaters, such as a jacketed kettle, q depends on both the temperature of the fluid in the jacket and the temperature of the fluid in the kettle. In this introductory chapter, systems (electrically heated tank or direct steam-heated tank) are selected for which this complication can be ignored. In Chap. 20, the analysis of a steam-jacketed kettle is given in which the effect of kettle temperature on q is taken into account.] cou9789x_ch08_163-185.indd 168 8/13/08 4:10:20 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 169 At steady state, dT/dt is zero, and Eq. (8.1) can be written qs wC ( Tis To ) wC ( Ts To ) 0 (8.2) where the subscript s has been used to indicate steady state. Subtracting Eq. (8.2) from Eq. (8.1) gives q qs wC [( Ti Tis ) ( T Ts )] rCV d ( T Ts ) dt (8.3) Notice that the reference temperature To cancels in the subtraction. If we introduce the deviation variables Ti Ti Tis (8.4) Q q qs (8.5) T T Ts (8.6) Eq. (8.3) becomes Q wC ( Ti T ) rCV dT dt (8.7) Taking the Laplace transform of Eq. (8.7) gives Q ( s ) wC [ Ti ( s ) T ( s ) ] rCVsT ( s ) (8.8) Q (s) rV T ( s ) s 1 Ti ( s ) w wC (8.9) or This last expression can be written Stirred heater transfer function T ( s ) where t 1 1/wC Q (s) Ti ( s ) ts 1 ts 1 (8.10) rV kg [] [] min w kg / min The gain for Q(t) is 1 1 C [] kg [] wC kJ /min kJ/ ( kg C ) ] [ min Note that if we multiply this gain by a heat input with units of kilojoules per minute, we obtain a quantity with units of degrees Celsius, as we would expect. From Eq. (8.10) we can see that two independent quantities, the heater input and the inlet temperature, can cou9789x_ch08_163-185.indd 169 8/13/08 4:10:20 PM Confirming Pages 170 PART 3 LINEAR CLOSED-LOOP SYSTEMS cause changes in the outlet temperature. The effects of these two quantities are additive (a consequence of linear equations and superposition). If there is a change in Q(t) only, then Ti ( t ) 0 and the transfer function relating T to Q is T ( s ) 1/wC Q (s) ts 1 (8.11) If there is a change in Ti ( t ) only, then Q(t) 0 and the transfer function relating T to Ti is T( s ) 1 Ti ( s ) ts 1 (8.12) Equation (8.10) is represented by the block diagram shown in Fig. 8–3a. This diagram is simply an alternate way to express Eq. (8.10) in terms of the transfer functions of Eqs. (8.11) and (8.12). Superposition makes this representation possible. Notice that in Fig. 8–3 we have indicated summation of signals by the symbol shown in Fig. 8–4, which is called a summing junction. Subtraction can also be indicated with this symbol by placing a minus sign at the appropriate input. The summing junction was used previously as the symbol for the comparator of the controller (see Fig. 8–2). This symbol, which is standard in the control literature, may have several inputs but only one output. A block diagram that is equivalent to Fig. 8–3a is shown in Fig. 8–3b. That this diagram is correct can be seen by rearranging Eq. (8.10); thus 1 1 T ( s ) Q ( s ) Ti ( s ) wC ts 1 (8.13) 1 s+1 T′i (s) T ′i (s) 1 C s+ 1 + Q(s) + T ′(s) Q(s) (a) 1 C + + 1 s+1 T ′(s) (b) FIGURE 8–3 Block diagram for process. Input + Input + FIGURE 8–4 Summing junction. cou9789x_ch08_163-185.indd 170 Output In Fig. 8–3b, the input variables (1/wC) Q(s) and Ti ( s ) are summed before being operated on by the transfer function 1/(ts 1). Note that in all cases the units in the block diagram must be consistent. The quantities being combined at a summing junction must have the same units. The readers should convince themselves that this is indeed the 8/13/08 4:10:21 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 171 case. We will see this again and again as we develop more complex block diagrams in w, Ti future chapters. Note that the rearrangement in Eq. (8.13) is merely for mathematical convenience, and not what actually hapT w+f≅w pens. Physically the two inputs (heat and inlet temperature) have independent effects FIGURE 8–5 on the outlet temperature. Supplying heat by steam. The physical situation that exists for the control system (Fig. 8–1) if steam heating is used requires more careful analysis to show that Fig. 8–3 is an equivalent block diagram. Assume that a supply of steam at constant conditions is available for heating the tank. One method for introducing heat to the system is to let the steam flow through a control valve and discharge directly into the water in the tank, where it will condense completely and become part of the stream leaving the tank (see Fig. 8–5). If the flow of steam, f (lb/time) is small compared with the inlet flow w, the total outlet flow is approximately equal to w. When the system is at steady state, the heat balance may be written f Steam at constant conditions wC ( Tis To ) wC ( Ts To ) fs ( H g Hls ) 0 where (8.14) To reference temperature used to evaluate enthalpy of all streams entering and leaving tank Hg specific enthalpy of steam supplied, a constant Hls specific enthalpy of condensed steam flowing out at Ts, as part of total stream The term Hls may be expressed in terms of heat capacity and temperature (assuming no phase change occurs between Ts and To). Hls C ( Ts To ) (8.15) From this, we see that if the steady-state temperature changes, Hls changes. In Eq. (8.14), fs ( H g Hls ) is equivalent to the steady-state input qs used previously, as can be seen by comparing Eq. (8.2) with Eq. (8.14). Now consider an unsteady-state operation in which f is much less than w and the temperature T of the bath does not deviate significantly from the steady-state temperature Ts. For these conditions, we may write the unsteady-state balance approximately; thus wC ( Ti To ) wC ( T To ) f ( H g Hls ) rCV cou9789x_ch08_163-185.indd 171 dT dt (8.16) 8/13/08 4:10:23 PM Confirming Pages 172 PART 3 LINEAR CLOSED-LOOP SYSTEMS In a practical situation for steam, Hg will be about 1000 Btu/lbm. If the temperature of the bath T never deviates from Ts by more than 10, the error in using the term f ( H g Hls ) instead of f(Hg Hl) will be no more than 1 percent. Under these conditions, Eq. (8.16) represents the system closely, and by comparing Eq. (8.16) with Eq. (8.1), it is clear that q f ( H g Hls ) (8.17) Therefore, q is proportional to the flow of steam f, which may be varied by means of a control valve. It should be emphasized that the analysis presented here is only approximate. Both f and the deviation in T must be small. The smaller they become, the more closely Eq. (8.16) represents the actual physical system. An exact analysis of the problem leads to a differential equation with time-varying coefficients, and the transferfunction approach does not apply. The problem becomes considerably more difficult. A better approximation will be discussed in Chap. 20, where linearization techniques are used. Example 8.1. Stirred-tank heater model. Let’s revisit the stirred-tank heater introduced in Example 5.2 (Fig. 8–6). (a) Determine the response of the outlet temperature of the tank to a step change in the inlet temperature from 60 to 70C. (b) Determine the response of the outlet temperature of the tank to a step increase in the heat input of 42 kW. (c) Determine the response of the outlet temperature of the tank to a simultaneous step change in the inlet temperature from 60 to 70C and a step increase in the heat input of 42 kW. The energy balance for the stirred-tank heater is Eq. (8.10): T( s ) 200 L/min Water 1/wC 1 Q (s) Ti ( s ) ts 1 ts 1 (8.10) Ti = 60°C V = 1,000 L T = 80°C Heat input q FIGURE 8–6 Stirred-tank heater revisited. cou9789x_ch08_163-185.indd 172 8/13/08 4:10:24 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 173 Substituting in numerical values for the variables, we obtain the actual transfer function for this stirred-tank heater. t rV V V tank volume 1, 000 L 5 min w w /r u volumetric flow ratte 200 L/min 1 C C 1 1 14 kJ/s 14 kW wC . 1 min 200 L 1 kg 4 184 kJ min L kg C 60 s T (s) 1 1 C Ti (s) Q (s ) 5s 1 5s 1 14 kW The block diagram for the tank is shown in Fig. 8–7. Remember that T , Ti , and Q(s) are deviation variables: T ′i Q 1 14 + + 1 5s + 1 T′ FIGURE 8–7 Block diagram for stirred-tank heater (tank and heater only). T T 80 Ti Ti 60 Q q qs The steady-state heat input qs may be found from the steady-state energy balance, Eq. (8.2). qs wC ( Tis To ) wC ( Ts To ) 0 (8.2) kJ kgg 1 min qs wC ( Ts Tis ) 200 (80 C 60 C ) 280 kW 4 . 184 min 60 s kg C Thus, Q q qs q 280 kW Solution (a) If the inlet temperature is stepped from 60 to 70C, then Ti(t ) 70 60 10 and Ti (s) 10/ s. Note that Q 0. Thus, T (s) 10 1 s 5s 1 Inverting to the time domain, we obtain the expression for T(t). T (t ) 10(1 et / 5 ) cou9789x_ch08_163-185.indd 173 8/13/08 4:10:25 PM Confirming Pages 174 PART 3 LINEAR CLOSED-LOOP SYSTEMS 10 Outlet Temperature (°C) 9 Deviation 8 Variable 7 6 5 4 3 2 1 0 0 5 10 15 Time (min) 20 90 Ac tual Outlet 89 Temperature (°C) 88 87 86 85 84 83 82 81 80 0 5 10 15 Time (min) 25 (a) 20 25 (b) FIGURE 8–8 Outlet temperature from stirred-tank heater. (a) Deviation variable; (b) actual temperature. and finally, we obtain the expression for T(t), the actual tank outlet temperature. T (t ) Ts T (t ) 80 10(1 et / 5 ) A plot of the outlet temperature (in deviation variables) is shown in Fig. 8–8a. The actual outlet temperature is shown in Fig. 8–8b. Note that for the uncontrolled mixing tank, a step change of 10C in the inlet temperature ultimately produces a 10C change in the outlet temperature. This result is just what we would expect after considering the physics of the situation. (b) For a step increase in the heat input of 42 kW, Q 42/s. Note that Ti 0 for this case. The expression for T is given by T (s) 42 1 1 s 14 5s 1 Inverting to the time domain, we obtain the expression for T(t). T (t ) 3(1 et / 5 ) and finally, we obtain the expression for T(t), the actual tank outlet temperature. T (t ) Ts T (t ) 80 3(1 et / 5 ) Qualitatively, this response is the same as for part (a); however, the ultimate temperature change due to the increased heat input is 3C. (c) If both changes occur simultaneously, the expression for T(t) is given by T (s) 42 1 1 s s 1 14 5 ∆T due to change in heat input cou9789x_ch08_163-185.indd 174 10 1 1 s 5 s ∆T due to change in inlet temperature 8/13/08 4:10:28 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 175 Inverting, we obtain T (t ) 13(1 et / 5 ) T (t ) Ts T (t ) 80 13(1 et / 5 ) As we can see, because the system is linear, the principle of superposition applies and the effects are additive. Measuring Element The temperature measuring element, which senses the bath temperature T and transmits a signal Tm to the controller, may exhibit some dynamic lag. From the discussion of the mercury thermometer in Chap. 4, we observed this lag to be first-order. In this example, we will assume that the temperature measuring element is a first-order system, for which the transfer function is Measuring element transfer function Tm ( s ) 1 T( s ) t ms 1 (8.18) where the input-output variables T and Tm are deviation variables, defined as T T Ts Tm Tm Tms Note that when the control system is at steady state, Tms Ts , which means that the temperature measuring element reads the true bath temperature. The transfer function for the measuring FIGURE 8–9 element may be represented by the block diaBlock diagram of measuring element. gram shown in Fig. 8–9. Thermocouples, commonly used temperature sensing devices in industry, have time constants on the order of 6 to 20 seconds (Riggs, 2007). The size of the time constant depends on the mass (size) of the thermocouple. Sometimes thermocouples are installed inside a protective sleeve called a thermowell that extends into the pipe or tank. The thermowell provides a protective barrier for the thermocouple against corrosion and abrasion, but can increase the effective time constant for the sensor. 1 ms + 1 T′(s) T ′m(s) Example 8.2. The temperature sensing element for the stirred-tank heater in Example 8.1 is a thermocouple. The manufacturer’s specifications state that the thermocouple has a response time of 45 s (with the response time defined by the manufacturer as the time required for the thermocouple’s reading to be 90 percent complete after a step change). Assuming that the thermocouple behaves as a first-order system, determine the transfer function for the temperature measuring element. cou9789x_ch08_163-185.indd 175 8/13/08 4:10:29 PM Confirming Pages 176 PART 3 LINEAR CLOSED-LOOP SYSTEMS The model for the sensor is first-order. If T(s) 1/s, a unit-step change, then the sensor response is Tm (s) 1 1 s t m s 1 Inverting gives Tm (t ) (1 e t / t m ) Since the ultimate value of Tm is 1, we know from the manufacturer’s specifications that we can expect the response to be 90 percent complete at t 45 s, which enables us to determine tm. 0 . 9 (1 e 45 s /t m ) 45 s 2 . 303 tm t m 19 . 5 s 0 . 33 min Therefore, the transfer function relating the actual temperature in the tank T to the measured or indicated temperature Tm is Tm ( s ) 1 T( s ) 0 . 33s 1 The block diagram for the stirred-tank heater, including the thermocouple, is shown in Fig. 8–10. Example 8.3. For a step change in the inlet temperature to the stirred tank of 10C (no change in heat input Q 0), plot the actual tank temperature and the temperature indicated by the thermocouple in Example 8.2 as a function of time. We can solve the equations by hand to obtain: T (t ) 10(1 et /5 ) (from Example 8. 1) and Tm (s) 10 0 . 71 10 . 71 10 s ( 0 . 33s 1)( 5s 1) s 3 . 03 s 0.2 s Inverting to get Tm (t ) yields Tm (t ) 10 0 . 71e3.03t 10 . 71e0.2 t Plotting T and Tm , we obtain Fig. 8–11. From the graphs, it is clear that Tm , lags behind T. cou9789x_ch08_163-185.indd 176 8/13/08 4:10:30 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 177 T i′ Q 1 14 + + T′ 1 5s + 1 1 0.33s + 1 T ′m FIGURE 8–10 Block diagram for stirred-tank heater and measuring element. 10 Actual 9 temperature T′ 8 7 6 5 Indicated temperature T ′m 4 3 7 Temperature change (°C) Temperature change (°C) 2 1 0 0 Expanded view of short times 6 Actual temperature T ′ 5 4 3 2 Indicated temperature T ′m 1 5 10 15 Time (min) (a) 20 25 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Time (min) (b) FIGURE 8–11 Comparison of temperature response for measured and actual temperatures in the stirredtank heater. (a) Complete response; (b) expanded time scale for short times. 1 Inlet temp. T′i Step Actual temp. T′ 5s + 1 Tank model 1 0.33s + 1 Thermocouple Indicated temp. T ′m Scope Temps To workspace FIGURE 8–12 Simulink block diagram for Example 8.3. cou9789x_ch08_163-185.indd 177 8/13/08 4:10:31 PM Confirming Pages 178 PART 3 LINEAR CLOSED-LOOP SYSTEMS We can also obtain these plots easily using Simulink (see Figs. 8–12 and 8–13). If we choose not to use the Scope, we can plot from the MATLAB workspace and obtain the same graph that we did previously (Fig. 8–14). plot(temps.time,temps.signals.values) FIGURE 8–13 Scope output from Simulink simulation of Example 8.3. 10 Actual temperature T ′ 9 Temperature Change (°C) 8 7 6 Indicated temperature T′m 5 4 3 2 1 0 0 5 10 15 Time (min) 20 25 FIGURE 8–14 MATLAB plot of output for stirred-tank heater reponse. cou9789x_ch08_163-185.indd 178 8/13/08 4:10:35 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 179 Controller and Final Control Element For convenience, the blocks representing the controller and the final control element are combined into one block. In this way, we need be concerned only with the overall response between the error in the temperature and the heat input to the tank. Also, it is assumed that the controller is a proportional controller. (In Chap. 9, the response of other controllers, which are commonly used in control systems, will be described.) The relationship for a proportional controller is q K ce A where (8.19) e TR TM TR set point temperature Kc proportional sensitivity or controller gain A heat input when e 0, also called the bias value (shortly we will show that A qs, the steady-state heat input) At steady state, it is assumed that the set point, the process temperature, and the measured temperature are all equal to one another; thus TRs Ts Tms (8.20) Let e be the deviation variable for error; thus e e es (8.21) where e s TRs Tms . Since TRs Tms , e s 0 and Eq. (8.21) becomes e e 0 e (8.22) This result shows that e is itself a deviation variable. Since at steady state there is no error, and A is the heat input for zero error, we see from Eq. (8.19) that qs K ce s A 0 A A Note: The steady-state heat input qs is the heat required to raise the steady-state inlet temperature from Tis to TRs , the desired steady-state set point temperature. The steadystate output from the controller/heater is termed the bias value. It is the output from the controller when the error is zero (i.e., steady state). It is simply calculated in this case from qs wC (TRs Tis ), which is easily deduced from Eq. (8.2). Equation (8.19) may now be written in terms of qs; thus q K ce qs or Q K ce (8.23) where Q q qs. cou9789x_ch08_163-185.indd 179 8/13/08 4:10:35 PM Confirming Pages 180 PART 3 LINEAR CLOSED-LOOP SYSTEMS The transform of Eq. (8.23) is simply Proportional controller transfer function Q ( s ) K ce ( s ) (8.24) Note that e, which is also equal to e , may be expressed as e TR TRs ( Tm Tms ) (8.25) e TR Tm (8.26) or Equation (8.25) follows from the definition of e and the fact that TRs Tms . Taking the transform of Eq. (8.26) gives Error definition for stirred heater e ( s ) TR ( s ) Tm ( s ) (8.27) The transfer function for the proportional controller given by Eq. (8.24) and the definition of the error given by Eq. (8.27) may be expressed by the block diagram shown in Fig. 8–15. We have now completed the development of the separate blocks. If these are combined according to Fig. 8–2, we obtain the block diagram for the complete control system shown in Fig. 8–16. The reader should verify this figure. Note that when we construct a block diagram, it is indicating the flow of information around the control loop. If a line in the block diagram splits, the information in that line is merely being sent to two different places–there is not a process “stream” that is being divided in any way. + (s) T′R(s) Kc Q(s) _ T ′m(s) FIGURE 8–15 Block diagram of proportional controller. T ′i(s) + T′R(s) Q Kc 1 C + 1 s +1 + T′(s) Tm ′ (s) 1 ms + 1 FIGURE 8–16 Block diagram of control system. cou9789x_ch08_163-185.indd 180 8/13/08 4:10:37 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 181 Example 8.4 Using the numerical values given in Examples 8.1 and 8.2 and a value of 20 for the proportional gain (Kc 20), plot the response of the tank temperature to a change in the set point (or desired temperature) TR of 5C. At this point, we have not yet learned how to handle the block diagram algebra required to obtain this response by a hand calculation (this will be the topic of Chaps. 11 and 12). However, Simulink provides us with a very easy means of determining the response. We merely construct the block diagram in Simulink exactly as written (see Fig. 8–17). The plot of the response of the tank temperature to a change in the set point is shown in Fig. 8–18. Set point T r′ +_ Error Q′(kW) 20 1 5s+1 Tank 0.0714 Proportional gain Heater gain Kc Step change Measured temp, T ′m Actual tank temp. T' Scope temps 1 0.33s+1 Thermocouple To workspace FIGURE 8–17 Simulink model of stirred tank control system. 6 Set point T ′r Temperature change (°C) 5 4 Actual tank temperature T ′ 3 2 1 0 0 5 10 15 20 25 30 Time (min) FIGURE 8–18 Response of stirred tank heater control system for proportional control only to a step change in the set point of 5C (Kc 20). cou9789x_ch08_163-185.indd 181 8/13/08 4:10:37 PM Confirming Pages 182 PART 3 LINEAR CLOSED-LOOP SYSTEMS SUMMARY It has been shown that a control system can be translated to a block diagram that includes the transfer functions of the various components. It should be emphasized that a block diagram is simply a systematic way of writing the simultaneous differential and algebraic equations that describe the dynamic behavior of the components. In the present case, these were Eqs. (8.10), (8.18), (8.24), and (8.27). The block diagram clarifies the relationships among the variables of these simultaneous equations. Another advantage of the block diagram representation is that it clearly shows the feedback relationship between measured variable and desired variable and how the difference in these two signals (the error e ) is used to maintain control. A set of equations generally does not clearly indicate the relationships shown by the block diagram. In the next several chapters, tools will be developed that will enable us to reduce a block diagram such as the one in Fig. 8.16 to a single block that relates T(s) to Ti or TR . We will then obtain the transient response of the control system shown in Fig. 8.16 to some specific changes in Ti and TR . However, first we pause in Chap. 9 to look more carefully at the controller and control element blocks, which have been skimmed over in this chapter. PROBLEMS 8.1. The two-tank heating process shown in Fig. P8–1 consists of two identical, well-stirred tanks in series. A flow of heat can enter tank 2. At time t 0, the flow rate of heat to tank 2 suddenly increases according to a step function to 1,000 Btu/min, and the temperature of the inlet water Ti drops from 60 to 52F according to a step function. These changes in heat flow and inlet water temperature occur simultaneously. (a) Develop a block diagram that relates the outlet temperature of tank 2 to the inlet temperature to tank 1 and the flow of heat to tank 2. (b) Obtain an expression for T2 ( s ) where T2 is the deviation in the temperature of tank 2. This expression should contain numerical values of the parameters. (c) Determine T2(2) and T2 (). (d) Sketch the response T2 ( t ) versus t. Initially, Ti T1 T2 60F and q 0. The following data apply: w 250 lb/min Holdup volume of each tank 5 ft3 Density of fluid 50 lb/ft3 Heat capacity of fluid 1 Btu/(lb · F) q w w T1 T2 Ti Tank 1 Tank 2 FIGURE P8–1 cou9789x_ch08_163-185.indd 182 9/5/08 9:52:49 AM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 183 8.2. The two-tank heating process shown in Fig. P8–2 consists of two identical, well-stirred tanks in series. At steady state, Ta Tb 60F. At time t 0, the temperature of each stream entering the tanks changes according to a step function, that is, Ta 10u ( t ) and Tb 20u ( t ) where Ta and Tb are deviation variables. (a) Develop the block diagram that relates T2 , the deviation in temperature in tank 2, to Ta and Tb. (b) Obtain an expression for T2 ( s ) . (c) Determine T2(2). The following data apply: w1 w2 250 lb/min Holdup volume of each tank 10 ft3 Density of fluid 50 lb/ft3 Heat capacity of fluid 1 Btu/(lb · F) Tb w2 Ta T2 T1 w1 w3 = w1 + w2 w1 Tank 1 Tank 2 FIGURE P8 –2 8.3. The heat transfer equipment shown in Fig. P8–3 consists of two tanks, one nested inside the other. Heat is transferred by convection through the wall of the inner tank. The contents of each tank are well mixed. The following data and information apply: 1. The holdup volume of the inner tank is 1 ft3. The holdup of the outer tank is 1 ft3. 2. The cross-sectional area for heat transfer between the tanks is 1 ft2. 3. The overall heat-transfer coefficient for the flow of heat between the tanks is 10 Btu/ (h · ft2 · F). 4. The heat capacity of fluid in each tank is 1 Btu/(lb · F). The density of each fluid is 50 lb/ft3. Initially the temperatures of the feed stream to the outer tank and the contents of the outer tank are equal to 100F. The conQ tents of the inner tank are initially Inner tank at 100F. At time zero, the flow of 10 lb/h heat to the inner tank Q is changed according to a step change from 0 to 500 Btu/h. T (a) Obtain an expression for the Outer tank Laplace transform of the temperature of the inner tank T(s). (b) Invert T(s) and obtain T for FIGURE P8 –3 time 0, 5 h, 10 h, and . cou9789x_ch08_163-185.indd 183 8/13/08 4:10:38 PM Confirming Pages CHAPTER 8 CAPSULE SUMMARY Controller mechanism Comparator + TR Set point _ Controller Ti, load Final control element Error Tm Measured variable + + Process T Controlled variable Measuring element FIGURE 8–2 Block diagram of a simple control system. Closed loop: The measured value of the controlled variable is fed back to the controller. Controller: A device that outputs a signal to the process or final control element based on the magnitude of the error signal. A proportional controller outputs a signal proportional to the error. Deviation variable: The difference between the actual value of a variable and its steady-state value. Block diagrams are always constructed using deviation variables. Error: The difference between the value of the set point and that of the measured variable. Final control element: A device that provides a modulated input to the process in response to a signal from the controller. For example, this may be a heater, a control valve, or a variety of other devices. Load: The change in any process variable that can cause the controlled variable to change. Measuring element: A sensor used to determine the value of the controlled variable and to send it to the comparator/controller. Examples include a thermocouple (temperature), a strain gage (pressure), a gas chromatograph (composition), and a pH electrode (acidity). These sensors typically have some dynamic behavior associated with them and can affect the design of the control system. 184 cou9789x_ch08_163-185.indd 184 8/13/08 4:10:39 PM Confirming Pages CHAPTER 8 THE CONTROL SYSTEM 185 Negative feedback: The error is the difference between the set point and the measured variable (this is usually the desired configuration). Positive feedback: The measured variable is added to the set point. (This is usually an undesirable situation, and frequently it leads to instability.) Regulator problem: The goal of a control system for this type of problem is to enable the system to compensate for load changes and maintain the controlled variable at the set point. Servo problem: The goal of a control system for this type of problem is to force the system to “track” the requested set point changes. Set point: The desired value of the controlled variable. cou9789x_ch08_163-185.indd 185 8/13/08 4:10:39 PM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS I n Chap. 8, the block diagram representation of a simple control system (Fig. 8–2) was developed. This chapter will focus attention on the controller and final control element and will discuss the dynamic characteristics of some of these components that are in common use. As shown in Fig. 8–2, the input signal to the controller is the error, and the output signal of the controller is fed to the final control element. In many process control systems, this output signal is an air pressure, and the final control element is a pneumatic valve that opens and closes as the air pressure on the diaphragm changes. For the mathematical analysis of control systems, it is sufficient to regard the controller as a simple computer. For example, a proportional controller may be thought of as a device that receives the error signal and puts out a signal proportional to it. Similarly, the final control element may be regarded as a device that produces corrective action on the process. The corrective action is regarded as mathematically related to the output signal from the controller. However, it is desirable to have some appreciation of the actual physical mechanisms used to accomplish this. For this reason, we begin this chapter with a physical description of a pneumatic control valve and a simplified description of a proportional controller. Up to about 1960, most controllers were pneumatic. Although pneumatic controllers are still in use and function quite well in many installations, the controllers being installed today are electronic or computer-based instruments. For this reason, the proportional controller to be discussed in this chapter will be electronic or computer-based. The transfer functions that are presented in this chapter apply to either type of controller, and 186 cou9789x_ch09_186-204.indd 186 8/14/08 8:07:37 AM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS 187 the discussion is in no way restrictive. Other pneumatic devices, such as control valves, are found throughout chemical processing plants and are a very important part of chemical process control systems. After the introductory discussion, transfer functions will be presented for simplified or idealized versions of the control valve and the conventional controllers. These transfer functions, for practical purposes, will adequately represent the dynamic behavior of control valves and controllers. Hence, they will be used in subsequent chapters for mathematical analysis and design of control systems. 9.1 MECHANISMS Control Valve The control valve shown in Fig. 9–1 contains a pneumatic device (valve motor) that moves the valve stem as the pressure on a spring-loaded diaphragm changes. The stem positions a plug in the orifice of the valve body. In the air-to-close valve, as the air pressure increases, the plug moves downward and restricts the flow of fluid through the valve. In the air-to-open valve, the valve opens and allows greater flow as the valve-top air pressure increases. The choice between air-to-open and air-to-close is usually made based on safety considerations. If the instrument air pressure fails, we would like the valve to fail in a safe position for the process. For example, if the control valve were on the cooling water inlet to a cooling jacket for an exothermic chemical reactor, we would want the valve to fail open so that we do not lose cooling water flow to the reactor. In such a situation, we would choose an air-to-close valve. Valve motors are often constructed so that the valve stem position is proportional to the valve-top pressure. Most commercial valves move from fully open to fully closed as the valve-top pressure changes from 3 to 15 psig. In general, the flow rate of fluid through the valve depends upon the upstream and downstream fluid pressures and the size of the opening through the valve. The plug and seat (or orifice) can be shaped so that various relationships between stem position and size of opening (hence, flow rate) are obtained. In our example, we assume for simplicity that at steady state the flow (for fixed upstream and downstream fluid pressures) is Air p Stem Air Motor Valve p Stem Valve Plug Air-to-close (a) Motor Plug Air-to-open (b) FIGURE 9–1 Pneumatic control valves. (a) Air to close; (b) air to open. cou9789x_ch09_186-204.indd 187 8/14/08 8:07:37 AM Confirming Pages 188 PART 3 LINEAR CLOSED-LOOP SYSTEMS proportional to the valve-top pneumatic pressure. A valve having this relation is called a linear valve. A more complete discussion of control valves is presented in Chap. 19. Controller The control hardware required to control the temperature of a stream leaving a heat exchanger is shown in Fig. 9–2. This hardware consists of the following components listed here along with their respective conversions: Transducer (temperature-to-current) Computer/Controller (current-to-current) Converter (current-to-pressure) Control valve (pressure-to-flow rate) Figure 9–2 shows that a thermocouple is used to measure the temperature; the signal from the thermocouple is sent to a transducer, which produces a current output in the range of 4 to 20 mA, which is a linear function of the input. The output of the transducer enters the controller where it is compared to the set point to produce an error signal. The computer/controller converts the error to an output signal in the range of 4 to 20 mA in accordance with the computer control algorithm. The only control algorithm we have considered so far has been proportional. Later in this chapter other control algorithms will be described. The output of the computer/controller enters the converter, which produces an output in the range of 3 to 15 psig, as a linear function of the input. Finally, the air pressure output of the converter is sent to the top of the control valve, which adjusts the flow of steam to the heat exchanger. We assume that the valve is linear and 120 V Transducer 4–20 mA Computer / Controller 20 psig air 120 V Converter 4–20 mA Set point temp. Gain 3–15 psig Valve Set point Hot process stream 120 V Heat exchanger Cold process stream Temperature measuring unit (thermocouple) FIGURE 9–2 Schematic diagram of control system. cou9789x_ch09_186-204.indd 188 8/14/08 8:07:38 AM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS 189 is the air-to-open type. The external power (120 V) needed for each component is also shown in Fig. 9–2. Electricity is needed for the transducer, computer/controller, and converter. A source of 20 psig air is also needed for the converter. To see how the components interact with one another, consider the process to be operating at steady state with the outlet temperature equal to the set point. If the temperature of the cold process stream decreases, the following events occur: After some delay the thermocouple detects a decrease in the outlet temperature and produces a proportional change in the signal to the controller. As soon as the controller detects the drop in temperature, relative to the set point, the controller output increases according to proportional action. The increase in signal to the converter causes the output pressure from the converter to increase and to open the valve wider to admit a greater flow of the hot process stream. The increased flow of hot stream will eventually increase the output temperature and move it toward the set point. From this qualitative description, we see that the flow of signals from one component to the next is such that the outlet temperature of the heat exchanger should return toward the set point. An equivalent P&ID (piping and instrumentation diagram) for this control system is shown in Fig. 9–3 (for other P&ID symbols, see App. 9A). In a well-tuned control system, the response of the Temperature recording controller TRC TT Temperature transmitter TE Temperature element Cold process stream Heat exchanger Air-to-open control valve I P Current-to-pressure converter Hot process stream FIGURE 9–3 Piping and instrumentation diagram for control system of Fig. 9–2. cou9789x_ch09_186-204.indd 189 8/14/08 8:07:38 AM Confirming Pages 190 PART 3 LINEAR CLOSED-LOOP SYSTEMS temperature will oscillate around the set point before coming to steady state. We will give considerable attention to the transient response of a control system in the remainder of this book. Further discussion will also be given on control valves in Chap. 19. For convenience in describing various control laws (or algorithms) in the next part of this chapter, the transducer, controller, and converter will be lumped into one block, as shown in Fig. 9–4. This concludes our brief introduction to valves and controllers. We now present transfer functions for such devices. These transfer functions, especially for controllers, are based on ideal devices that can be only approximated in practice. The degree of approximation is sufficiently good to warrant use of these transfer functions to describe the dynamic behavior of controller mechanisms for ordinary design purposes. Transducer mA x Measured variable Converter Controller psig mA p (a) ''Controller'' p x (b) FIGURE 9–4 Equivalent block for transducer, controller, and converter. 9.2 IDEAL TRANSFER FUNCTIONS Control Valve A pneumatic valve always has some dynamic lag, which means that the stem position does not respond instantaneously to a change in the applied pressure from the controller. From experiments conducted on pneumatic valves, it has been found that the relationship between flow and valve-top pressure for a linear valve can often be represented by a first-order transfer function; thus Control valve first-order transfer function Q(s) Kv t vs 1 P (s) (9.1) where Kv is the steady-state gain, i.e., the constant of proportionality between the steadystate flow rate and the valve-top pressure, and tv is the time constant of the valve. In many practical systems, the time constant of the valve is very small when compared with the time constants of other components of the control system, and the transfer function of the valve can be approximated by a constant. Control valve (fast dynamics) transfer functio n cou9789x_ch09_186-204.indd 190 Q(s) Kv P (s) (9.2) 8/14/08 8:07:38 AM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS 191 Under these conditions, the valve is said to contribute negligible dynamic lag. To justify the approximation of a fast valve by a transfer function, which is simply Kv, consider a first-order valve and a first-order process connected in series, as shown in Fig. 9–5. P Value Process Kv Kp vs+1 Ps+1 Y FIGURE 9–5 Block diagram for first-order valve and a first-order process. According to the discussion of Chap. 6, if we assume no interaction (which is generally valid for this case), the relationship between the air pressure to the valve and the output from the process (perhaps a reactor temperature) is Y (s) Kv KP P (s) (t v s 1) (t P s 1) For a unit-step change in the valve-top pressure P, Y 1 K v KP s (t v s 1) (t P s 1) the inverse of which is t vt P 1 t / t v 1 t / t P Y ( t ) K v K P 1 e e t t t t v P P v If tv V tP, this equation is approximately ( Y ( t ) K v K P 1 et /t P ) The last expression is the unit-step response of the transfer function Y (s) KP Kv t Ps 1 P (s) so that the combination of process and valve is essentially first-order. This clearly demonstrates that when the time constant of the valve is much smaller than that of the process, the valve transfer function can be taken as Kv. A typical pneumatic valve has a time constant of the order of 1 s. Many industrial processes behave as first-order systems or as a series of first-order systems having time constants that may range from a minute to an hour. For these systems we have shown that the lag of the valve is negligible, and we will make frequent use of this approximation. cou9789x_ch09_186-204.indd 191 8/14/08 8:07:39 AM Confirming Pages 192 PART 3 LINEAR CLOSED-LOOP SYSTEMS Controllers In this section, we present the transfer functions for the controllers frequently used in industrial processes. Because the transducer and the converter will be lumped together with the controller for simplicity, the result is that the input will be the measured variable x (e.g., temperature and fluid level) and the output will be a pneumatic signal p. (See Fig. 9–4.) Actually this form (x as input and p as output) applies to a pneumatic controller. For convenience, we will refer to the lumped components as the controller in the following discussion, even though the actual electronic controller is but one of the components. PROPORTIONAL CONTROL. The simplest type of controller is the proportional con- troller. (The ON/OFF control is really the simplest, but it is a special case of the proportional controller as we’ll see shortly.) Our goal is to reduce the error between the process output and the set point. The proportional controller, as we will see, can reduce the error, but cannot eliminate it. If we can accept some residual error, proportional control may be the proper choice for the situation. The proportional controller has only one adjustable parameter, the controller gain. The proportional controller produces an output signal (pressure in the case of a pneumatic controller, current, or voltage for an electronic controller) that is proportional to the error e. This action may be expressed as Proportional controller where p K ce ps (9.3) p output signal from controller, psig or mA Kc proportional gain, or sensitivity e error (set point) (measured variable) ps a constant, the steady-state output from the controller [the bias value, see Eqs. (8.19) and (8.23)] The error e, which is the difference between the set point and the signal from the measuring element, may be in any suitable units. However, the units of the set point and the measured variable must be the same, since the error is the difference between these quantities. In a controller having adjustable gain, the value of the gain Kc can be varied by entering it into the controller, usually by means of a keypad (or a knob on older equipment). The value of ps is the value of the output signal when e is zero, and in most controllers ps can be adjusted to obtain the required output signal when the control system is at steady state and e 0. To obtain the transfer function of Eq. (9.3), we first introduce the deviation variable P p ps into Eq. (9.3). At time t 0, we assume the error es to be zero. Then e is already a deviation variable. Equation (9.3) becomes cou9789x_ch09_186-204.indd 192 8/14/08 8:07:39 AM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS P ( t ) K ce ( t ) 193 (9.4) Taking the transform of Eq. (9.4) gives the transfer function of an ideal proportional controller. Proportional controller transfer function P (s) Kc e (s) (9.5) The actual behavior of a proportional controller is depicted in Fig. 9–6. The controller output will saturate (level out) at pmax 15 psig or 20 mA at the upper end and at pmin 3 psig or 4 mA at the lower end of the output. The ideal transfer function Eq. (9.5) does not predict this saturation phenomenon. Proportional controller—Ideal behavior p slope = Kc Proportional controller—actual behavior p Saturation pmax ps = bias value 0 ε 0 ps = bias value Saturation 0 pmin 0 ε (b) (a) FIGURE 9–6 Proportional controller output as a function of error input to the controller. (a) Ideal behavior; (b) actual behavior. The next example will help to clarify the concept of controller gain. Example 9.1. A pneumatic proportional controller is used in the process shown in Fig. 9–2 to control the cold stream outlet temperature within the range of 60 to 120F. The controller gain is adjusted so that the output pressure goes from 3 psig (valve fully closed) to 15 psig (valve fully open) as the measured temperature goes from 71 to 75F with the set point held constant. Find the controller gain Kc. Gain ∆p 15 psig 3 psig 3 psi/ F ∆e 75 F 71 F Now assume that the gain of the controller is changed to 0.4 psi/F. Find the error in temperature that will cause the control valve to go from fully closed to fully open. ∆T cou9789x_ch09_186-204.indd 193 ∆p 12 psi 30 F gain 0 . 4 psi/ F 8/14/08 8:07:40 AM Confirming Pages 194 PART 3 LINEAR CLOSED-LOOP SYSTEMS At this level of gain, the valve will be fully open if the error signal reaches 30F. The gain Kc has the units of psi per unit of measured variable. [Regarding the units on controller gain, if the actual controller of Fig. 9–4 is considered, both the input and the output units are in milliamperes. In this case the gain will be dimensionless (i.e., mA/mA).] ON/OFF CONTROL. A special case of proportional control is on/off control. If the gain Kc is made very high, the valve will move from one extreme position to the other if the process deviates only slightly from the set point. This very sensitive action is called on/ off action because the valve is either fully open (on) or fully closed (off); i.e., the valve acts as a switch. This is a very simple controller and is exemplified by the thermostat used in a home-heating system. In practice, a dead band is inserted into the controller. With a dead band, the error reaches some finite positive value before the controller “turns on.” Conversely, the error must fall to some finite negative value before the controller “turns off.” This behavior is shown in Fig. 9–7. Expanding the width of the dead band makes the controller less sensitive to noise and prevents the phenomenon of chattering, where the controller will rapidly cycle on and off as the error fluctuates about zero. Chattering is detrimental to equipment performance. On/off controller with dead band Ideal on/off controller p p pmax pmax Dead band pmin Slope = Kc = infinite pmin 0 0 0 (a) ε 0 ε (b) FIGURE 9–7 Output from on/off controller as a function of error input to the controller. (a) Ideal on/off controller; (b) on/off controller with dead band. For various reasons, it is often desirable to add other modes of control to the basic proportional action. These modes, integral and derivative action, are discussed below with the objective of obtaining the ideal transfer functions of the expanded controllers. The reasons for introducing these modes will be discussed briefly at the end of this chapter and in greater detail in later chapters. PROPORTIONAL-INTEGRAL (PI) CONTROL. If we cannot tolerate any residual error, we will have to introduce an additional control mode: integral control. If we add integral control to our proportional controller, we have what is termed PI, or proportional-integral cou9789x_ch09_186-204.indd 194 8/14/08 8:07:41 AM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS 195 control. The integral mode ultimately drives the error to zero. This controller has two adjustable parameters for which we select values, the gain and the integral time. Thus it is a bit more complicated than a proportional controller, but in exchange for the additional complexity, we reap the advantage of no error at steady state. PI control is described by the relationship Proportional-integral controller where p K ce Kc tI t ∫0 e dt + ps (9.6) Kc proportional gain tI integral time, min ps constant (the bias value) In this case, we have added to the proportional action term Kce another term that is proportional to the integral of the error. The values of Kc and tI are both adjustable. To visualize the response of this controller, consider the response to a unit-step change in error, as shown in Fig. 9–8. This unit-step response is most directly obtained by inserting e 1 into Eq. (9.6), which yields Kc t ps tI p ( t ) Kc p (9.7) Kc τI } Kc ps 1 0 0 t FIGURE 9–8 Response of a PI controller to a unit-step change in error. Notice that p changes suddenly by an amount Kc and then changes linearly with time at a rate Kc/tI. To obtain the transfer function of Eq. (9.6), we again introduce the deviation variable P p ps into Eq. (9.6) and then take the transform to obtain Proportional-integral controller transfer funcction cou9789x_ch09_186-204.indd 195 1 P (s) Kc 1 e (s) t Is (9.8) 8/14/08 8:07:41 AM Confirming Pages 196 PART 3 LINEAR CLOSED-LOOP SYSTEMS Some manufacturers prefer to use the term reset rate, which is defined as the reciprocal of tI. The integral adjustment on a controller may be denoted by integral time or reset rate (carefully check the specific controller to be sure which value to enter). The calibration of the proportional and integral action is often checked by observing the jump and slope of a step response, as shown in Fig. 9–8. PROPORTIONAL-DERIVATIVE (PD) CONTROL. Derivative control is another mode that can be added to our proportional or proportional-integral controllers. It acts upon the derivative of the error, so it is most active when the error is changing rapidly. It serves to reduce process oscillations. This mode of control may be represented by Proportional-derivative controller p K ce K ct D de ps dt (9.9) Kc proportional gain tD derivative time, min ps constant (bias value) where In this case, we have added to the proportional term another term Kc tD de/dt, which is proportional to the derivative of the error. The values of Kc and tD are both adjustable. Other terms that are used to describe the derivative action are rate control and anticipatory control. The action of this controller can be visualized by considering the response to a linear change in error as shown in Fig. 9–9. AKc p AKc ps D } Derivative alone 1 Proportional alone A 1 0 0 t FIGURE 9–9 Response of a PD controller to a ramp input in error. This response is obtained by introducing the linear function e (t) At into Eq. (9.9) to obtain p ( t ) AK c t AK ct D ps Notice that p changes suddenly by an amount AKc tD as a result of the derivative action and then changes linearly at a rate AKc. The effect of derivative action in this case is to cou9789x_ch09_186-204.indd 196 8/14/08 8:07:42 AM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS 197 anticipate the linear change in error by adding output AKc tD to the proportional action. The controller is taking preemptive action to counter the anticipated change in the error that it predicted from the slope of the error versus time curve. To obtain the transfer function from Eq. (9.9), we introduce the deviation variable P p ps and then take the transform to obtain Proportional-derivative controller transfer fuu nction P (s) K c (1 t D s ) e (s) (9.10) PROPORTIONAL-INTEGRAL-DERIVATIVE (PID) CONTROL. This mode of control is a combination of the previous modes and is given by the expression Proportional-integral-derivative controller p K ce K ct D de Kc dt tI t ∫0 e dt ps (9.11) In this case, all three values Kc, tD, and tI can be adjusted in the controller. The transfer function for this controller can be obtained from the Laplace transform of Eq. (9.11); thus Proportional-integral-derivative controller trransfer function P (s) 1 Kc 1 t D s e (s) t Is (9.12) Derivative action is based on how rapidly the error is changing, not the magnitude of the error or how long the error has persisted. It is based on the slope of the error versus time curve at any instant in time. Therefore, a rapidly changing error signal will induce a large derivative response. “Noisy” error signals cause significant problems for derivative action because of the rapidly changing slope of the error caused by noise. Derivative control should be avoided in these situations unless the error signal can be filtered to remove the noise. Motivation for Addition of Integral and Derivative Control Modes Having introduced ideal transfer functions for integral and derivative modes of control, we now wish to indicate the practical motivation for use of these modes. The curves of Fig. 9–10 show the behavior of a typical feedback control system using different kinds of control when it is subjected to a permanent disturbance. This may be visualized in terms of the stirred-tank temperature control system of Chap. 8 after a step change in Ti. The value of the controlled variable is seen to rise at time zero owing to the disturbance. With no control, this variable continues to rise to a new steady-state value. With control, after some time the control system begins to take action to try to maintain the controlled variable close to the value that existed before the disturbance occurred. cou9789x_ch09_186-204.indd 197 8/14/08 8:07:42 AM Confirming Pages PART 3 LINEAR CLOSED-LOOP SYSTEMS Controlled variable, deviation from initial value 198 1 1 2 3 4 Control action None Proportional Proportional-integral Proportional-integral-derivative 2 3 4 0 Offset 4 8 Time, min 12 16 FIGURE 9–10 Response of a typical control system showing the effects of various modes of control. With proportional action only, the control system is able to arrest the rise of the controlled variable and ultimately bring it to rest at a new steady-state value. The difference between this new steady-state value and the original value (the set point, in this case) is called offset. For the particular system shown, the offset is seen to be only about 20 percent of the ultimate change that would have been realized for this disturbance in the absence of control. As shown by the PI curve, the addition of integral action eliminates the offset; the controlled variable ultimately returns to the original value. This advantage of integral action is balanced by the disadvantage of a more oscillatory behavior. The addition of derivative action to the PI action gives a definite improvement in the response. The rise of the controlled variable is arrested more quickly, and it is returned rapidly to the original value with little or no oscillation. Discussion of the PD mode is deferred to a later chapter. The selection among the control systems whose responses are shown in Fig. 9–10 depends on the particular application. If an offset of about 20 percent is tolerable, proportional action would likely be selected. If no offset were tolerable, integral action would be added. If excessive oscillations had to be eliminated, derivative action might be added. The addition of each mode means, as we will see in later chapters, more difficult controller adjustment. Our goal in forthcoming chapters will be to present the material that will enable the reader to develop curves such as those of Fig. 9–10 and thereby to design efficient, economic control systems. SUMMARY In this chapter we have presented a brief discussion of control valves and controllers. In addition, we presented ideal transfer functions to represent their dynamic behavior and some typical results of using these controllers. cou9789x_ch09_186-204.indd 198 8/14/08 8:07:43 AM Confirming Pages CHAPTER 9 CONTROLLERS AND FINAL CONTROL ELEMENTS 199 The ideal transfer functions actually describe the action of many types of controllers, including pneumatic, electronic, computer-based, hydraulic, mechanical, and electrical systems. Hence, the mathematical analyses of control systems to be presented in later chapters, which are based upon first- and second-order systems, transportation lags, and ideal controllers, generalize to many branches of the control field. After studying this text on process control, the reader should be able to apply the knowledge to, e.g., problems in mechanical control systems. All that is required is a preliminary study of the physical nature of the systems involved. PROBLEMS 9.1. A pneumatic PI temperature controller has an output pressure of 10 psig when the set point and process temperature coincide. The set point is suddenly increased by 10F (i.e., a step change in error is introduced), and the following data are obtained: Time, s psig 0 0 20 60 90 10 8 7 5 3.5 Determine the actual gain (psig per degree Fahrenheit) and the integral time. 9.2. A unit-step change in error is introduced into a PID controller. If Kc 10, tI 1, and tD 0.5, plot the response of the controller P(t). 9.3. An ideal PD controller has the transfer function P K c (t D s 1) e An actual PD controller had the transfer function t Ds 1 P Kc e (t D /b ) s 1 where b is a large constant in an industrial controller. If a unit-step change in error is introduced into a controller having the second transfer function, show that ( P ( t ) K c 1 Aebt / t D ) where A is a function of b which you are to determine. For b 5 and Kc 0.5, plot P(t) versus t/tD. As b → , show that the unit-step response approaches that for the ideal controller. 9.4. A PID temperature controller is at steady state with an output pressure of 9 psig. The set point and process temperature are initially the same. At time t 0, the set point is increased at the rate of 0.5F/min. The motion of the set point is in the direction of lower temperatures. If the current settings are cou9789x_ch09_186-204.indd 199 8/14/08 8:07:43 AM Confirming Pages 200 PART 3 LINEAR CLOSED-LOOP SYSTEMS K c 2 psig/ F t i 1 . 25 min t D 0 . 4 min plot the output pressure versus time. 9.5. The input e to a PI controller is shown in Fig. P9–5. Plot the output of the controller if Kc 2 and tI 0.50 min. 1 0.5 ε 0 2 0 3 1 4 t, min −0.5 −1 FIGURE P9–5 9.6. A PI controller has the transfer function Gc 5s 10 s Determine the values of Kc and tI. 9.7. Dye for our new line of blue jeans is being blended in a mixing tank. The desired color of blue is produced using a concentration of 1500 ppm blue dye, with a minimum acceptable concentration of 1400 ppm. At 9 A.M. today the dye injector plugged, and the dye flow was interrupted for 10 min, until we realized the problem and unclogged the nozzle. see Fig. P9–7. Water 20 gal/min 20 gal/min aqueous dye for jeans (1500 ppm blue dye) Concentrated dye injector V = 100 gal Color analyzer PI FIGURE P9–7 Plot the controller ouput from 9 A.M. to 9:10 A.M. The steady-state controller output (the bias value) is 8 psig. Does the controller output saturate (output range is 3 to 15 psig)? If so, at what time does it occur? The controller is a PI controller with Kc 0.001 psig/ppm and tI 1 min. cou9789x_ch09_186-204.indd 200 8/14/08 8:07:43 AM Confirming Pages CHAPTER 9 CAPSULE SUMMARY CONTROL VALVES Air Two basic types of control valves are air-to-close and air-to-open. The air pressure (pneumatic) signal is usually 3 to 15 psig. The dynamics of the valves are adequately modeled as first-order systems. The time constant is on the order of 1 s. Control valuve first-order transfer function Air Motor p Stem Valve Motor p Stem Valve Plug Q (s) Kv P (s) t vs 1 Plug Air-to-close Air-to-open (a) (b) Control valve (fast dynamics) transfer functi o n Q (s) Kv P (s) CONTROLLERS Controller type Time domain model Proportional (P) p Kc e ps Proportional-integral (PI) p Kce Transfer function P( s ) Kc e (s) ∫0 e dt ps P(s) 1 Kc 1 e (s) t I s Proportional-derivative (PD) p K c e K ct D de ps dt P(s) K c (1 t D s ) e (s) Proportional-integralderivative (PID) de Kc dt tI Kc tI t p K c e K ct D t ∫0 e dt ps P(s) 1 Kc 1 t D s e (s) t I s 201 cou9789x_ch09_186-204.indd 201 8/14/08 8:07:44 AM Confirming Pages PART 3 LINEAR CLOSED-LOOP SYSTEMS Controlled variable, deviation from initial value 202 0 1 1 2 3 4 Control action None Proportional Proportional-integral Proportional-integral-derivative 2 3 4 Offset 4 8 Time (min) 12 16 (Figure 9.10 Response of a typical control system showing the effects of various modes of control.) cou9789x_ch09_186-204.indd 202 8/14/08 8:07:45 AM Confirming Pages CHAPTER 9 APPENDIX PIPING AND INSTRUMENTATION DIAGRAM SYMBOLS Piping and instrumentation diagrams (P&IDs) are used by the chemical process industry to document the control systems for their processes. The Instrument Society of America (ISA) produces a standards document [ISA-5.1-1984-(R1992)] “Instrument Symbols and Identification” that establishes a uniform means of designating instruments, control systems, and sensors used for measurement and control in a process. The standard is suitable for use in the chemical, petroleum, power generation, air conditioning, metal refining, and numerous other process industries. Table 9A.1 shows some common conventions used for identifying process instrumentation and control on process drawings. Figure 9A–1 shows an example of a level control loop for a process tank. The operation is as follows. The tank level is measured using a sensor, perhaps a differential pressure cell, indicated by the bubble containing LE on the diagram. The sensor is connected to a level transmitter, LT, that sends an electrical signal (4 to 20 mA) to a level indicating controller LIC. Level alarms high and low, LAH and LAL, “monitor” the signal from the level transmitter and indicate an alarm situation if necessary. Notice from the symbols (the line through the bubble) that LIC, LAH, and LAL are all located in the control Tank T-101 LE LCV LT LAH LY LAL LIC FIGURE 9A–1 Example of a level control loop using P&ID symbols. 203 cou9789x_ch09_186-204.indd 203 8/14/08 8:07:45 AM Confirming Pages 204 PART 3 LINEAR CLOSED-LOOP SYSTEMS room, while LE and LT are mounted in the field. The level indicating controller determines the necessary signal to send to the valve, based on the current level in the tank, the set point, and the selected control algorithm being used (P or PID, for example). LY computes the necessary control air pressure signal (3 to 15 psig) to send to the level control valve LCV to properly respond to the controller output signal from LIC. TABLE 9A.1 Common symbols used on P&IDs [Instrument Society of America Standard, ISA-5.1-1984 (R1992)] Instrument Line Symbols Instrument supply or piping connection to process (capillary) Pneumatic ............................... Electrical Software signal Instrument Location and Identification ABB Instrument located in plant ABB Instrument located on front of panel in control room ABB Instrument located on back of panel in control room. ABB Instrument accessible as part of a distributed control system Common Identification Letters Used in Instrument Symbols First Letter (A) A B C D E F G H I J K L M O P Q R S T V W Y Z cou9789x_ch09_186-204.indd 204 Analysis Burner, combustion User choice (sometimes conductivity) User choice (sometimes density or specific gravity) Voltage Flow rate User choice Hand Current (electrical) Power Time or time schedule Level User choice (sometimes moisture or humidity) Pressure, vacuum Quantity Radioactivity or ratio Speed or frequency Temperature Viscosity or vibration Weight or force Event Position Second or Third Letter (B) Alarm Control Sensor element Glass, viewing device High Indicate Control station Light or low Middle or intermediate Orifice, restriction Point Record Switch Transmit Valve, damper or louver Well Relay, compute, convert Drive 8/14/08 8:07:45 AM Confirming Pages CHAPTER 10 BLOCK DIAGRAM OF A CHEMICAL-REACTOR CONTROL SYSTEM T o tie together the principles developed thus far and to illustrate further the procedure for reduction of a physical control system to a block diagram, we consider in this chapter the two-tank chemical-reactor control system of Fig. 10–1. This entire chapter serves as an example and may be omitted by the reader with no loss in continuity. Set point composition Pure A Controller m F + rm ≅ F A F c0 Product stream V, T1, c1, k1 Composition measuring element V, T2, c2, k2 Sample stream Heating coil FIGURE 10–1 Control of a stirred-tank chemical reactor. 205 cou9789x_ch10_205-217.indd 205 8/14/08 8:58:25 AM Confirming Pages 206 PART 3 LINEAR CLOSED-LOOP SYSTEMS 10.1 DESCRIPTION OF SYSTEM A liquid stream enters tank 1 at a volumetric flow rate F cfm and contains reactant A at a concentration of c0 mol A/ft3. Reactant A decomposes in the tanks according to the irreversible chemical reaction A → B The reaction is first-order and proceeds at a rate rA kc where rA rate of formation of A, (mol A)/(ft3 · time) c concentration of A, mol A/ft3 k reaction rate constant (a function of temperature), time1 (k1 ≡ tank 1, k2 ≡ tank 2) The reaction is to be carried out in a series of two continuous stirred-tank reactors. The tanks are maintained at different temperatures. The temperature in tank 2 is to be greater than the temperature in tank 1, with the result that k2, the reaction rate constant in tank 2, is greater than that in tank 1, k1. We will neglect any changes in physical properties due to chemical reaction. The purpose of the control system is to maintain c2, the concentration of A leaving tank 2, at some desired value in spite of variations in the inlet concentration c0. This will be accomplished by adding a stream of pure A to tank 1 through a control valve. We wish to produce a block diagram for the process so that we can simulate its response to changes in inlet concentration. 10.2 REACTOR TRANSFER FUNCTIONS We begin the analysis by making a material balance on A around tank 1; thus dc1 V dt accumulation of A in tank 1 where m Fc m F 0 1Vc1 c1 K r A reaction rate flow of A into tank 1 flow of A out of tank 1 (10.1) o f A in tank 1 m molar flow rate of pure A through valve, (lb · mol/min) rA density of pure A (lb · mol/ft3) V holdup volume of tank, a constant (ft3) It is assumed that the volumetric flow of A through the valve m/rA is much less than the inlet flow rate F, with the result that Eq. (10.1) can be written V cou9789x_ch10_205-217.indd 206 dc1 ( F k1V ) c1 Fc0 m dt (10.2) 8/14/08 8:58:26 AM Confirming Pages CHAPTER 10 BLOCK DIAGRAM OF A CHEMICAL-REACTOR CONTROL SYSTEM 207 This last equation may be written in the form V dc1 F 1 c1 c0 m F k1V dt F k1V F k1V 1/F dc1 1 m c0 t1 c1 1 k1t 1 k1t dt where t residence time for each tank (10.3) V , (time) F t1 effective time constant for tank 1 V t , (time) F k1V 1 k1t At steady state, dc1/dt 0, and Eq. (10.3) becomes c1s 1 1F c0s ms 1 k1t 1 k1t (10.4) where s refers to steady state. Subtracting Eq. (10.4) from Eq. (10.3) and introducing the deviation variables C1 c1 c1s C0 c0 c0s M m ms give t1 dC1 1 1F C1 C0 M dt 1 k1t 1 k1t (10.5) Taking the transform of Eq. (10.5) yields the transfer function of the first reactor: Transfer function for tank 1 C1 ( s ) 1 1/ (1 k1t ) C0 ( s ) M ( s ) t 1s 1 F (10.6) A material balance on A around tank 2 gives dc2 V dt accumulation of A in tank 2 Fc 1 flow of A into tank 2 Fc2 flow of A out tank 2 k 2Vc2 reaction rate of A in tank 2 (10.7) Rearranging gives dc2 ( F k2 V ) c2 Fc1 dt V dc2 F c2 c1 F k2V dt F k2 V V cou9789x_ch10_205-217.indd 207 (10.7a) 8/14/08 8:58:26 AM Confirming Pages 208 PART 3 LINEAR CLOSED-LOOP SYSTEMS As with tank 1, Eq. (10.7a) can be written in terms of deviation variables and arranged to give dC2 1 t2 C2 C1 (10.8) dt 1 k2t where C2 is the deviation variable for tank 2, C2 c2 c2s , and t2 is the effective time constant for tank 2, t2 t V V F F k2 V 1 k2 ( V F ) 1 k2t Taking the transform of Eq. (10.8) gives the transfer function for the second reactor: Transfer function for tank 2 C2 ( s ) 1/ (1 k2t ) C1 ( s ) t 2s 1 (10.9) To obtain some numerical results, we will assume the following data to apply to the system: MWA 100 lb/lb · mol A (molecular weight of A) rA 0.8 lb · mol/ft3 cos 0 . 1 lb mol A /ft 3 F 100 cfm ms 1 . 0 lb mol/min 1 . 25 cfm ms 1 . 0 lb mol/min Note: rA 0. 8 lb mol/ft 3 k1 k1 1 6 2 3 min1 k2 k1 because T2 T1 min1 V 300 ft3 t V 300 ft 3 3 min F 100 cfm From Eq. (10.4), we can calculate the steady-state concentration of A in tank 1. 1 c1s 1 1 lb mol 1F c0s ms (0.1) 100 1 (1.0) 0.0733 1 1 k1t 1 k1t 1 2 ft 3 1 6 ( 3) () 1 2 We can calculate the steady-state concentration of A in tank 2 by using Eq. (10.7a) written at steady state (when dc2/dt 0). lb moo l 1 1 c2s c1 (0 . 0733) 0 . 0244 2 1 k2t s ft 3 1 3 ( 3) () cou9789x_ch10_205-217.indd 208 8/14/08 8:58:27 AM Confirming Pages CHAPTER 10 10.3 BLOCK DIAGRAM OF A CHEMICAL-REACTOR CONTROL SYSTEM 209 CONTROL VALVE The air-to-open control valve selected for the process has the following characteristics. The flow of A through the valve varies linearly from 0 to 2 cfm as the valve-top pressure varies from 3 to 15 psig. The time constant tv of the valve is so small compared with the other time constants in the system that its dynamics can be neglected. From Eq. (10.2) the transfer function for a fast valve is Q(s ) Kv P (s ) (10.2) From this relationship, and remembering that Q and P are deviation variables, we can compute the valve gain using the given data Kv ∆Flow (cfm) 20 1 cfm ∆Pressure (psi) 15 3 6 p si Since ms /rA 1.25 cfm, the normal operating pressure on the valve is ps 3 1 . 25 (15 3) 10 . 5 psig 2 (10.10) Since our mole balances for the tanks are written in terms of molar flow rates instead of volumetric flow rates (cfm), we would like the valve equation to be written in terms of molar flow rates as well. Realizing that the volumetric flow rate (cfm) can be written as m/rA, we can write the equation for the control valve as (see Fig. 10–2) m /rA 1 . 25 Kv p 10 . 5 and finally, m [1 . 25 K v ( p 10 . 5) ] rA (10.11) In terms of deviation variables, this can be written M K v rA P where (10.12) M m 1 . 25 rA m 1 . 0 P p 10 . 5 Taking the transform of Eq. (10.12) gives 1 ft 3 /min 0 . 8 lb mol lb mol/min M (s) K v rA 0 . 133 3 psi P (s) ft 6 psi (10.13) as the valve transfer function. cou9789x_ch10_205-217.indd 209 8/14/08 8:58:28 AM Confirming Pages 210 PART 3 LINEAR CLOSED-LOOP SYSTEMS Control Valve Sensitivity 2 1.75 Slope = Kv 1.5 Steady-state flow = 1.25 cfm Flow (cfm) 1.25 1 0.75 (p, m/ρ) 0.5 0.25 0 10.5 psig 0 3 6 9 Pressure (psig) 12 15 FIGURE 10–2 Control valve sensitivity. 10.4 MEASURING ELEMENT For illustration, assume that the measuring element converts the concentration of A to an electronic signal. Specifically, the output of the measuring element varies from 4 to 20 mA as the concentration of A varies from 0.01 to 0.05 lb · mol A/ft3. We will assume that the concentration measuring device is linear and has negligible lag. The sensitivity (or gain) of the measuring device is therefore Km 20 4 mA 400 0 . 05 0 . 01 lb mol/ft 3 Since c2s is 0.0244 lb · mol/ft3, the normal signal from the measuring device is 0 . 0244 0 . 01 ( 20 4 ) 4 . 0 5 . 76 4 . 0 9 . 7 6 mA 0 . 05 0 . 01 The equation for the measuring device is therefore b 9.76 K m ( c2 0.0244 ) 9.76 mA 400 ( mA c2 0.0244 lb mol/ft 3 lb mol/ft 3 ) (10.14) where b is the output current (milliamperes) from the measuring device. In terms of deviation variables, Eq. (10.14) becomes cou9789x_ch10_205-217.indd 210 8/14/08 8:58:29 AM Confirming Pages CHAPTER 10 BLOCK DIAGRAM OF A CHEMICAL-REACTOR CONTROL SYSTEM B K m C2 211 (10.15) where B b 9.76 and C2 c2 c2s . The transfer function for the measuring device is therefore B(s) Km C2 ( s ) (10.16) A measuring device that changes the type of signal between its input and output is called a transducer; in the present case, the concentration signal is changed to a current signal. A thermocouple is another example of a transducer. It changes an input temperature signal to an output electrical signal (millivolts). 10.5 CONTROLLER For convenience, we will assume the controller to have proportional action and produce a current output signal. The relation between the controller output signal (milliamperes) and the error (milliamperes) is p ps K c ( cR b ) ps K ce (10.17) where cR desired current signal (or set point), mA Kc controller gain, mA/mA e error cR b, mA In terms of deviation variables, Eq. (10.17) becomes P K ce (10.18) The transform of this equation gives the transfer function of the controller P (s) Kc e (s) (10.19) Assuming the set point and the signal from the measuring device to be the same when the system is at steady state under normal conditions, we have for the reference value of the set point cRs bs 9 . 76 mA The corresponding deviation variable for the set point is C R cR cRs cou9789x_ch10_205-217.indd 211 8/14/08 8:58:30 AM Confirming Pages 212 PART 3 LINEAR CLOSED-LOOP SYSTEMS 10.6 CONTROLLER TRANSDUCER The output from the controller is an electronic current signal ranging from 4 to 20 mA. The signal that the control valve requires is a pneumatic signal of 3 to 15 psig. A transducer is required to convert the current signal from the controller to the pneumatic signal required by the control valve. We will assume that the transducer is linear and has negligible lag. The transfer function for the transducer is merely a gain that is given by Eq. (10.20). P T (s ) (15 3) psig psig KT 0 . 75 P (s ) (20 4) mA mA 10.7 (10.20) TRANSPORTATION LAG A portion of the liquid leaving tank 2 is continuously withdrawn through a sample line, containing a concentration measuring element, at a rate of 0.1 cfm. The measuring element must be remotely located from the process, because rigid ambient conditions must be maintained for accurate concentration measurements. The sample line has a length of 50 ft, and the cross-sectional area of the line is 0.001 ft2. The sample line can be represented by a transportation lag with parameter td volume (50)(0 . 001) 0 . 5 min flow rate 0.1 The transfer function for the sample line is, therefore, et d s e0.5s 10.8 BLOCK DIAGRAM We have now completed the analysis of each component of the control system and have obtained a transfer function for each. These transfer functions can now be combined so that the overall system is represented by the block diagram in Fig. 10–3. C0 CR Km C'R + mA Kc P mA KT PT psig − B mA KvρA M lb.moles min 1 F M + F + Km 1 1 + k1τ τ1s + 1 C1 1 1 + k2τ τ2s + 1 C2 e−τd s− FIGURE 10–3 Block diagram for a chemical-reactor control system. cou9789x_ch10_205-217.indd 212 8/14/08 8:58:30 AM Confirming Pages CHAPTER 10 BLOCK DIAGRAM OF A CHEMICAL-REACTOR CONTROL SYSTEM 213 An equivalent diagram is shown in Fig. 10–4 in which some of the blocks have been combined. K1 = C0 CR + KKc − 1 (1 + k1τ)(1 + k2τ) + + 1 (τ1s + 1)(τ2s + 1) C2 e−τds τ1 = 2, τ2 = 1, τd = 0.5, K1 = 1 4.5 KmKTKvρA Open-loop gain = KKc = K = 0.09 Kc F(1 + k1τ)(1 + k2τ) c FIGURE 10–4 Equivalent block diagram for a chemical-reactor control system (CR is now in concentration units). Numerical quantities for the parameters in the transfer functions are given in Fig. 10–4. It should be emphasized that the block diagram is written for deviation variables. The true steady-state values, which are not given by the diagram, must be obtained from the analysis of the problem. The example analyzed in this chapter will be used later in discussion of control system design. The design problem will be to select a value of Kc that gives satisfactory control of the composition C2 despite the rather long transportation lag involved in getting information to the controller. In addition, we will want to consider possible use of other modes of control for the system. SUMMARY We have now learned how to analyze a physical system and develop mathematical models for the various components of the system and then use those models to construct a block diagram of the model. We will make extensive use of these skills in upcoming chapters as we study control systems in greater depth. PROBLEMS 10.1. In the process shown in Fig. P10–1, the concentration of salt leaving the second tank is controlled using a proportional controller by adding concentrated solution through a control valve. The following data apply: (a) The controlled concentration is to be 0.1 lb salt/ft3 solution. The inlet concentration ci is always less than 0.1 lb/ft3. (b) The concentration of concentrated salt solution is 30 lb salt/ft3 solution. cou9789x_ch10_205-217.indd 213 8/14/08 8:58:31 AM Confirming Pages 214 PART 3 LINEAR CLOSED-LOOP SYSTEMS (c) Transducer: The output of the transducer varies linearly from 3 to 15 psig as the concentration varies from 0.05 to 0.15 lb/ft3. (d) Controller: The controller is a pneumatic, direct-acting, proportional controller. (e) Control valve: As valve-top pressure varies from 3 to 15 psig, the flow through the control valve varies linearly from 0 to 0.005 cfm. ( f ) It takes 30 s for the solution leaving the second tank to reach the transducer at the end of the pipe. Draw a block diagram of the control system. Place in each block the appropriate transfer function. Calculate all the constants and give the units. Concentrated solution Controller Set point Zero length Salt solution 1 ft3/min ci c1 c2 V1 = 3 ft3 V2 = 4 ft3 Transducer FIGURE P10–1 Use the process shown in Figs. 10–3 and 10–4 for Prob. 10.2 to 10.5. 10.2. Verify the values of t1 and t2. 10.3. Determine the steady-state value of the controller output ps in milliamperes. 10.4. Use Simulink to simulate the open-loop response of the two chemical reactors to a step change in the feed concentration C0 from 0.1 to 0.25 lb · mol A/ft3. 10.5. The open-loop process has an upset such that the flow rate to the process instantaneously rises to 120 cfm (from the original 100 cfm). How does the open-loop block diagram change? Plot the outlet concentration of A in both reactors as a function of time. 10.6. Two isothermal stirred-tank reactors (Fig. P10–6) are connected by a long pipe that acts as a pure time delay between the two tanks (no reaction takes place in the pipe). CSTR 1 is at a higher temperature than CSTR 2, but both temperatures remain constant. Assume constant throughputs and holdups (volumes) and a first-order, irreversible reaction taking place in each CSTR (A → B). The flow rate through the system is 4 ft3/min, and the delay time in the pipe is 30 s. The inlet concentration to CSTR 1 is initially at steady state at 1 lb · mol/ft3 and is increased at time 0 through a step change to 2 lb · mol/ft3. (a) Draw the block diagram for the process, and be sure to include all necessary constants. (b) Use Simulink to plot the exit concentration of A from each of the reactors. (c) Use Simulink to plot the exit concentration of B from each of the reactors. cou9789x_ch10_205-217.indd 214 8/14/08 8:58:31 AM Confirming Pages CHAPTER 10 BLOCK DIAGRAM OF A CHEMICAL-REACTOR CONTROL SYSTEM 215 DATA CSTR 1 Rate constant (min Volume (ft3) 1 ) Reactor 1 0.3 25 CSTR 2 0.15 15 Reactor 2 Dead time = 30 s FIGURE P10–6 cou9789x_ch10_205-217.indd 215 8/14/08 8:58:32 AM Confirming Pages CHAPTER 10 CAPSULE SUMMARY A model for a two CSTRs in series was developed. The reaction that occurs in each vessel is A → B. The rate of formation of A is given by rA ⫽ ⫺kc. A process schematic is shown below. Set point composition Pure A Controller F + rm ≅ F m A F c0 Composition measuring element Product stream V, T2, c2, k2 V, T1, c1, k1 Sample stream Heating coil FIGURE 10–1 Control of a stirred-tank chemical reactor. A block diagram for the process is as follows: C0 CR Km C'R + mA Kc P mA KT PT psig − KvρA M lb.moles min 1 F B mA M + F + Km 1 1 + k1τ τ1s + 1 C1 1 1 + k2τ τ2s + 1 C2 e−τd s FIGURE 10–3 Block diagram for a chemical-reactor control system. 216 cou9789x_ch10_205-217.indd 216 9/5/08 11:23:17 AM Confirming Pages CHAPTER 10 BLOCK DIAGRAM OF A CHEMICAL-REACTOR CONTROL SYSTEM 217 Numerical values for the process are given below. MWA 100 lb/lb · mol A rA 0.8 lb · mol/ft3 cos 0 . 1 lb mol A /ft 3 F 100 cfm ms 1.0 lb · mol/min ms 1 . 0 lb mol/min 1 . 25 cfm rA 0 . 8 lb mol/ft 3 k1 k2 1 6 2 3 min1 k2 k1 because T2 T1 min1 V 300 ft3 t Kv V 300 ft 3 3 min, t 1 2 min, t 2 1 min F 100 cfm 1 cfm 6 psig psig mA mA 400 lb mol/ft 3 K T 0 . 75 Km td 0.5 min c1s 0 . 01 lb mol/ft 3 c2s 0 . 0244 lb mol/ft 3 pTs 10 . 5 psig ms 1.0 lb · mol/min bs 9.76 mA ps 14 mA cou9789x_ch10_205-217.indd 217 8/14/08 8:58:32 AM Confirming Pages CHAPTER 11 CLOSED-LOOP TRANSFER FUNCTIONS 11.1 STANDARD BLOCK-DIAGRAM SYMBOLS In Chap. 8, a block diagram was developed for the control of a stirred-tank heater (Fig. 8–2). In Fig. 11–1, the block diagram has been redrawn and incorporates some standard symbols for the variables and transfer functions, which are widely used in the control literature. These symbols are defined as follows: R ⫽ set point or desired value C ⫽ controlled variable e ⫽ error B ⫽ variable produced by measuring element M ⫽ manipulated variable U ⫽ load variable or disturbance Gc ⫽ transfer function of controller G1 ⫽ transfer function of final control element G2 ⫽ transfer function of process H ⫽ transfer function of measuring element U + R + Gc + G1 M _ B G2 C H FIGURE 11–1 Standard control system nomenclature. 218 cou9789x_ch11_218-227.indd 218 8/14/08 8:05:28 AM Confirming Pages CHAPTER 11 CLOSED-LOOP TRANSFER FUNCTIONS 219 In some cases, the blocks labeled Gc and G1 will be lumped together into a single block. The series of blocks between the comparator and the controlled variable, which consist of Gc, G1, and G2, is referred to as the forward path. The block H between the controlled variable and the comparator is called the feedback path. The use of G for a transfer function in the forward path and H for one in the feedback path is a common convention. The product GH, which is the product of all transfer functions (GcG1G2H) in the loop, is called the open-loop transfer function. We call GH the open-loop transfer function because it relates the measured variable B to the set point R if the feedback loop (of Fig. 11–1) is disconnected (i.e., opened) from the comparator. The subject of this chapter is the closed-loop transfer function, which relates two variables when the loop of Fig. 11–1 is closed. In more complex systems, the block diagram may contain several feedback paths and several loads. An example of a multiloop system, which is shown in Fig. 11–2, is cascade control. Several multiloop systems of industrial importance are presented in Chap. 17. U1 Inner loop U2 + R + Gc1 _ + Gc2 G1 + + G2 + G3 C _ B2 B1 H2 H1 FIGURE 11–2 Block diagram for a multiloop, multiload system. 11.2 OVERALL TRANSFER FUNCTION FOR SINGLE-LOOP SYSTEMS Once a control system has been described by a block diagram, such as the one shown in Fig. 11–1, the next step is to determine the transfer function relating C to R or C to U. We refer to these transfer functions as overall transfer functions because they apply to the entire system. These overall transfer functions are used to obtain considerable information about the control system, as will be demonstrated in the succeeding chapters. For the present it is sufficient to note that they are useful in determining the response of C to any change in R and U. The response to a change in set point R, obtained by setting U ⫽ 0, represents the solution to the servo problem. The response to a change in load variable U, obtained by setting R ⫽ 0, is the solution to the regulator problem. A systematic approach for obtaining the overall transfer function for set point change and load change will now be presented. cou9789x_ch11_218-227.indd 219 8/14/08 8:05:28 AM Confirming Pages 220 PART 3 LINEAR CLOSED-LOOP SYSTEMS Overall Transfer Function for Change in Set Point For this case, U ⫽ 0 and Fig. 11–1 may be simplified or reduced as shown in Fig. 11–3. In this reduction, we have made use of a simple rule of block diagram reduction which states that a block diagram consisting of several transfer functions in series can be simplified to a single block containing a transfer function that is the product of the individual transfer functions. R + Gc _ B G1 G2 C R + G C R _ B H (a) H G 1 + GH C G = GcG1G2 (b) (c) FIGURE 11–3 Block diagram reduction to obtain overall transfer function. This rule can be proved by considering two noninteracting blocks in series as shown in Fig. 11–4. This block diagram is equivalent to the equations Y ⫽ GA X Z ⫽ GB Y Multiplying these equations gives Y Z ⫽ G AGB XY which simplifies to Z ⫽ G AGB X X GA Y GB Z FIGURE 11–4 Two noninteracting blocks in series. Thus, the intermediate variable Y has been eliminated, and we have shown the overall transfer function Z/X to be the product of the transfer functions GAGB. This proof for two blocks can be easily extended to any number of blocks to give the rule for the general case. This rule was developed in Chap. 7 for the specific case of several noninteracting, first-order systems in series. With this simplification the following equations can be written directly from Fig. 11–3b. cou9789x_ch11_218-227.indd 220 C ⫽ Ge (11.1) B ⫽ HC (11.2) 8/14/08 8:05:29 AM Confirming Pages CHAPTER 11 CLOSED-LOOP TRANSFER FUNCTIONS e ⫽ R⫺B 221 (11.3) Since there are four variables and three equations, we can solve the equations simultaneously for C in terms of R as follows: C ⫽ G( R ⫺ B) C ⫽ G ( R ⫺ HC ) C ⫽ GR ⫺ GHC or finally C G ⫽ (11.4) R 1 ⫹ GH This is the overall transfer function relating C to R and may be represented by an equivalent block diagram as shown in Fig. 11–3c. Overall Transfer Function for Change in Load In this case R ⫽ 0, and Fig. 11–1 is drawn as shown in Fig. 11–5a. From the diagram we can write the following equations: C ⫽ G2(U ⫹ M ) (11.5) M ⫽ Gc G1e (11.6) e ⫽ ⫺B (11.7) B ⫽ HC (11.8) U R=0 + Gc G1 M G2 C U _ B G2 1 +GH C H (b) (a) FIGURE 11–5 Block diagram for change in load. Again the number of variables (C, U, M, B, e) exceeds by 1 the number of equations, and we can solve for C in terms of U as follows: C ⫽ G2 (U ⫹ Gc G1e ) C ⫽ G2 [U ⫹ Gc G1 (⫺HC ) ] or finally C G2 ⫽ 1 ⫹ GH U cou9789x_ch11_218-227.indd 221 (11.9) 8/14/08 8:05:29 AM Confirming Pages 222 PART 3 LINEAR CLOSED-LOOP SYSTEMS where G ⫽ GcG1G2. Notice that the transfer functions for load change or set point change have denominators that are identical, 1 ⫹ GH. Another approach to finding the closed-loop transfer functions from the block diagram is a “brute-force” technique that involves “breaking the loop” and working your way across the block diagram. + U + Gc R M G1 G2 C _ ≈ ≈ CH H FIGURE 11–6 Modified block diagram for determining closed-loop transfer function. Starting at the left edge of the modified block diagram in Fig. 11–6, we can work our way across the diagram and develop the following equation. [( R ⫺ CH ) GcG1 ⫹ U ] G2 ⫽C Solving this equation for C, we obtain (after a bit of algebra) C ⫽ Gc G1G2 G2 R⫹ U ⫽ 1 ⫹ Gc G1G2 H 1 ⫹ Gc G1G2 H G2 G R⫹ U 1 ⫹ GH 1 ⫹ GH closed loop transfer function relating C and R closed loop transfer function relaa ting C and U This result is the same as the individual results for C/R and C/U that we found previously. The following simple rule serves to generalize the results for the single-loop feedback system shown in Fig. 11–1: The transfer function relating any pair of variables X, Y is obtained by the relationship Y p forward ⫽ 1 ⫹ p loop X negative feedback (11.10) where pforward ⫽ product of transfer functions in forward path between locations of X and Y ploop ⫽ product of all transfer functions in loop (i.e., in Fig. 11–1, ploop ⫽ GcG1G2H ) If this rule is applied to finding C/R in Fig. 11–1, we obtain cou9789x_ch11_218-227.indd 222 8/14/08 8:05:30 AM Confirming Pages CHAPTER 11 CLOSED-LOOP TRANSFER FUNCTIONS 223 C Gc G1G2 G ⫽ ⫽ R 1 ⫹ Gc G1G2 H 1 ⫹ GH which is the same as before. For positive feedback, the reader should show that the following result is obtained: Y p forward ⫽ 1 ⫺ p loop X positive feedback (11.11) Example 11.1. Determine the transfer functions C/R, C/U1, and B/U2 for the system shown in Fig. 11–7. Also determine an expression for C in terms of R and U1 for the situation when both set point change and load change occur simultaneously. U2 U1 R + + + Gc G1 G2 + + G3 C _ H2 B H1 FIGURE 11–7 Block diagram for Example 11.1. Using the rule given by Eq. (11.10), we obtain by inspection the results C Gc G1G2G3 ⫽ R 1⫹G (11.12) C G2G3 ⫽ U1 1⫹G (11.13) B G3 H1 H 2 ⫽ U2 1⫹G (11.14) where G ⫽ GcG1G2G3H1H2. The reader should check one or more of these results by the direct method of solution of simultaneous equations and the method described in Fig. 11–6. For separate changes in R and U1, we may obtain the response C from Eqs. (11.12) and (11.13); thus C ⫽ Gc G1G2G3 R 1⫹G (11.15) G2G3 U1 1⫹G (11.16) and C ⫽ cou9789x_ch11_218-227.indd 223 8/14/08 8:05:31 AM Confirming Pages 224 PART 3 LINEAR CLOSED-LOOP SYSTEMS If both R and U1 occur simultaneously, the principle of superposition requires that the overall response be the sum of the individual responses; thus Gc G1G2G3 G2G3 R⫹ U1 1⫹G 1⫹G C ⫽ 11.3 (11.17) OVERALL TRANSFER FUNCTION FOR MULTILOOP CONTROL SYSTEMS To illustrate how one obtains the overall transfer function for a multiloop system, consider the next example in which the method used is to reduce the block diagram to a single-loop diagram by application of the rules summarized by Eqs. (11.10) and (11.11). Example 11.2. Determine the transfer function C/R for the system shown in Fig. 11–8. This block diagram represents a cascade control system, which will be discussed later. R + _ Gc1 + _ + Gc2 U2 U1 Inner loop G1 + G2 + + G3 C H2 H1 (a) R + _ Gc1 Ga = Gc2G1 1 + Gc2G1H2 Gb = G2G3 C R Gc GaGb 1 1 + Gc1GaGbH1 C H1 (b) (c) FIGURE 11–8 Block diagram reduction. (a) Original diagram; (b) first reduction; (c) final block diagram. Obtaining the overall transfer function C/R for the system represented by Fig. 11–8a is straightforward if we first reduce the inner loop (or minor loop) involving Gc2 , G1 , and H2 to a single block, as we have just done in the case of Fig. 11–1. For convenience, we may also combine G2 and G3 into a single block. These reductions are shown in Fig. 11–8b. Figure 11–8b is a single-loop block diagram that can be reduced to one block, as shown in Fig. 11–8c. It should be clear without much detail that to find any other transfer function such as C/U1 in Fig. 11–8a, we proceed in the same manner, i.e., first reduce the inner loop to a single-block equivalent. cou9789x_ch11_218-227.indd 224 8/14/08 8:05:31 AM Confirming Pages CHAPTER 11 CLOSED-LOOP TRANSFER FUNCTIONS 225 SUMMARY In this chapter, we have illustrated the procedure for reducing the block diagram of a control system to a single block that relates one input to one output variable. This procedure consists of writing, directly from the block diagram, a sufficient number of linear algebraic equations and solving them simultaneously for the transfer function of the desired pair of variables. For single-loop control systems, a simple rule was developed for finding the transfer function between any desired pair of input-output variables. This rule is also useful in reducing a multiloop system to a single-loop system. It should be emphasized that regardless of the pair of variables selected, the denominator of the closed-loop transfer function will always contain the same term, 1 ⫹ G, where G is the open-loop transfer function of the single-loop control system. In the succeeding chapters, frequent use will be made of the material in this chapter to determine the overall response of control systems. PROBLEMS 11.1. Determine the transfer function Y(s)/X(s) for the block diagrams shown in Fig. P11–1. Express the results in terms of Ga, Gb, and Gc. _ Ga X + Gb + Gc Y + (a) _ X + + Gc Gb Ga Y + Gc Ga (b) FIGURE P11–1 11.2. Find the transfer function Y(s)/X(s) of the system shown in Fig. P11–2. X _ 0.5 1 + + + 2 Y +1 FIGURE P11–2 cou9789x_ch11_218-227.indd 225 8/14/08 8:05:32 AM Confirming Pages 226 PART 3 LINEAR CLOSED-LOOP SYSTEMS 11.3. R For the control system shown in Fig. P11–3 determine the transfer function C(s)/R(s). + + 2 _ _ 1 s + 2 C _ FIGURE P11–3 11.4. Derive the transfer function Y/X for the control system shown in Fig. P11–4. 2 + X + + 1 s 1 s Y _ 25 FIGURE P11–4 cou9789x_ch11_218-227.indd 226 11.5. Derive the transfer function T⬘/T⬘R for the temperature control system shown in Fig. 8–16. 11.6. Derive the transfer functions C2/C0 and C2/CR for the reactor control system shown in Fig. P10–3. 8/14/08 8:05:32 AM Confirming Pages CHAPTER 11 CAPSULE SUMMARY R ⫽ set point or desired value U C ⫽ controlled variable + + Gc G1 + G2 R e ⫽ error M _ B ⫽ variable produced by measuring element M ⫽ manipulated variable B H U ⫽ load variable or disturbance Gc ⫽ transfer function of controller FIGURE 11.1 G1 ⫽ transfer function of final control element Standard control system nomenclature. G2 ⫽ transfer function of process H ⫽ transfer function of measuring element G ⫽ Gc G1G2 (the product of the transfer functions in the forward path between R and C) C Closed-loop process Process in which the feedback loop is connected to the comparator. Closed-loop transfer function Transfer functions relating two variables in the process when the feedback loop is connected to the comparator. Feedback path The path that connects the controlled variable and the comparator. Forward path The transfer functions that lie between two signals in the block diagram moving left to right as drawn in the block diagram above. The complete forward path consists of G ⫽ Gc G1G2. The forward path between U and C is G2 only. Open-loop process Process in which the feedback loop is disconnected from the comparator. Open-loop transfer function Product of all transfer functions in the loop relating B and R when the feedback loop is disconnected from the comparator. Closed-loop formula (negative feedback) Y p forward ⫽ X 1 ⫹ p loop Closed-loop formula (positive feedback) Y p forward ⫽ X 1 ⫺ p loop pforward ⫽ product of transfer functions in forward path between locations of X and Y ploop ⫽ product of all transfer functions in loop (i.e., from figure above, ploop ⫽ GcG1G2H) 227 cou9789x_ch11_218-227.indd 227 8/14/08 8:05:32 AM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS I n this chapter the results of all the previous chapters will be applied to determining the transient response of a simple control system to changes in set point and load. Considerable use will be made of the results of Chaps. 4 through 7 (Part 2) because the overall transfer functions for the examples presented here reduce to first- and secondorder systems. Consider the control system for the heated, stirred tank that we previously discussed and is represented by Fig. 12–1. The reader may want to refer to Chap. 8 for a description of this control system. Steam or electricity w,Ti Power controller Electrical power Steam q w, T (a) T'R + − Gc G1 (b) 1 wC + + (c) A = 1/wC T'i 1 s+1 1 (d) T' T'R + Gc − + A + T'i s 1 +1 T' (e) FIGURE 12–1 Block diagram of temperature control system. 228 cou9789x_ch12_228-251.indd 228 8/14/08 4:33:44 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 229 In Fig. 12–1a, the sketch of the apparatus is drawn in such a way that the source of heat (electricity or steam) is not specified. To make this problem more realistic, we have shown in Fig. 12–1b that the source of heat is steam that is discharged directly into the water, and in Fig. 12–1c the source of heat is electrical. In the latter drawing, a device known as a power controller provides electric power to a resistance heater proportional to the signal from the controller. The block diagram is shown in Fig. 12–1d. The block representing the process is taken directly from Fig. 8–3. To reduce the number of symbols, 1/wC has been replaced by A in Fig. 12–1e. Throughout this chapter, we will assume that the valve does not have any dynamic lag, for which case the transfer function of the valve (G1 in Fig. 12–1) will be taken as a constant Kv. This assumption was shown to be reasonable in Chap. 9. To simplify the discussion further, Kv has been taken as 1. (If Kv were other than 1, we may simply replace Gc by GcKv in the ensuing discussion.) In the first part of the chapter, we will also assume that there is no dynamic lag in the measuring element (tm 0), so that it may be represented by a transfer function that is simply the constant 1. A bare thermocouple will have a response that is so fast that for all practical purposes it can be assumed to follow the slowly changing tank temperature without lag. When the feedback transfer function is unity, the system is called a unity-feedback system. Introducing these assumptions leads to the simplified block diagram of Fig. 12–1e, for which we will obtain overall transfer functions for changes in set point and load when proportional control and proportional-integral control are used. 12.1 PROPORTIONAL CONTROL FOR SET POINT CHANGE (SERVO PROBLEM— SET POINT TRACKING) The goal of the control system for this case is to force the system to “track” the desired set point changes. For proportional control, Gc Kc. By using the methods developed in Chap. 11, the overall transfer function in Fig. 12–1 e is T⬘ K c A / (ts ⫹ 1) Kc A ⫽ ⫽ TR⬘ 1 ⫹ K c A / (ts ⫹ 1) ts ⫹ (1 ⫹ K c A ) (12.1) This may be rearranged in the form of a first-order lag to give T⬘ K c A / (1 ⫹ K c A ) A1 ⫽ ⫽ TR⬘ t / ( 1 ⫹ K A s ⫹ 1 t s ) [ ] c 1 ⫹1 where t 1 ⫽ A1 ⫽ cou9789x_ch12_228-251.indd 229 (12.2) t 1 ⫹ Kc A Kc A 1 ⫽ 1 ⫹ Kc A 1 ⫹ (1/K c A ) 8/14/08 4:33:45 PM Confirming Pages 230 PART 3 LINEAR CLOSED-LOOP SYSTEMS According to this result, the response of the tank temperature to a change in set point is first-order. The time constant for the closed-loop control system t1 is less than that of the stirred tank itself t. This means that one of the effects of feedback control is to speed up the response. We may use the results of Chap. 4 to find the response to a variety of inputs. The response of the system to a unitstep change in set point TR is shown in T'R 1 Fig. 12–2. (We have selected a unit change in A1 set point for convenience; responses to steps Offset T' of other magnitudes are obtained by superposition.) Remember, the goal of a control 0 0 system is to force the system to track the set t point. Thus, the desired ultimate value of T, FIGURE 12–2 which is T (), is of course 1. For the case of Unit-step response for set point change a unit-step change in set point, T approaches (P control). A1 KcA/(1 KcA), a fraction of unity. The desired change is, of course, 1. Thus, the ultimate value of the temperature T () does not match the desired change. This discrepancy is called offset and is defined as Offset ⫽ TR⬘ ( ⬁ ) ⫺ T ⬘ ( ⬁ ) (12.3) The offset is actually the steady-state value of the error (for the case of unity feedback). In terms of the particular control system parameters Offset ⫽ 1 ⫺ Kc A 1 ⫽ 1 ⫹ Kc A 1 ⫹ Kc A (12.4) This difference or discrepancy between set point and tank temperature at steady state is characteristic of proportional control. In some cases offset cannot be tolerated. However, notice from Eq. (12.4) that the offset decreases as Kc increases, and in theory the offset could be made as small as desired by increasing Kc to a sufficiently large value. Proportional control : As K c ↑ Offset ↓ To give a full answer to the problem of eliminating offset by high controller gain requires a discussion of stability and the response of the system when other lags, which have been neglected, are included in the system. Both these subjects are to be covered later. For the present we simply say that whether or not proportional control is satisfactory depends on the amount of offset that can be tolerated, the speed of response of the system, and the amount of gain that can be provided by the controller without causing the system to go unstable. cou9789x_ch12_228-251.indd 230 8/14/08 4:33:45 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 200 L/min Water Ti = 60°C Example 12.1. Proportional control of a stirred tank heater for set point tracking. Consider the stirred-tank heater that we examined in Chap. 8 (Examples 8.1 to 8.3), shown in Fig. 12–3. Determine the response of a closed-loop proportional control system for Kc 5, 10, 20, and 100 for a set point change of 5C. Determine the offset for each value of Kc. V = 1,000 L T Heat input q TC 231 TR = 80°C FIGURE 12–3 Stirred-tank heater control system. As before, we define TR⬘ ⫽ TR ⫺ 80 deviati on variables q⬘ ⫽ q ⫺ 280 kW Ti⬘ ⫽ Ti ⫺ 60 t ⫽ 1, 000 L ⫽ 5 min 200 L/min A⫽ 1⬚C 0 . 0714 ⬚ C 1 ⫽ ⫽ 14 kW kW wC If we assume that the thermocouple that senses the tank exiting temperature has fast dynamics compared to the tank, a Simulink block diagram for the process is shown in Fig. 12–4. 1 Out1 Heater gain Set point Tr +− Set point Error Q(kW) kc 1 ++ 0.0714 Actual tank temp, T 5s+1 Proportional gain Kc Tank Scope Feed temp, Ti Measured temp, Tm 1 1 temps To workspace Thermocouple 2 Out2 FIGURE 12–4 Simulink model for Example 12.1. cou9789x_ch12_228-251.indd 231 8/14/08 4:33:46 PM Confirming Pages 232 PART 3 LINEAR CLOSED-LOOP SYSTEMS We can program a MATLAB m-file to run this Simulink model (named: 'example12_1') for each value of Kc and then to plot the results on the same graph. The m-file that calls the model is shown below: % M-file that calls Simulink Model 'Example12_1' for i=1:4 z=[5,10,20,100]; kc=z(1,i); % the variable y in the 'sim' statement is taken from the outport in the model [t,x,y]=sim('example12_1',15); plot(t,y(:,1) hold on % the last row in y contains the final values of the setpoint and the tank temp [norow,nocol]=size(y); offset(i,1)=y(norow,2)-y(norow,1); kcplot(i,1)=kc; end grid title('Temp vs time for Kc=5,10,20,100'); hold off figure; plot(kcplot,offset); title('Offset vs Kc'); The results are shown in Figs. 12–5 and 12–6. A hand calculation verifies the offset results obtained by MATLAB for the specified values of Kc. Kc 5 10 20 100 Offset ⫽ 5 1 ⫹ Kc A ⫽ 5 1 ⫹ 0 . 0714 K c 3.69 2.92 2.06 0.614 Note that as the proportional gain Kc increases, the process more closely approaches the desired set point (5C in deviation variables, actually 85C ). For a gain of 100, the outlet temperature is 4.386C (deviation variable; the actual outlet temperature is 84.386C) with an offset of 0.614C. cou9789x_ch12_228-251.indd 232 8/14/08 4:33:50 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 233 Deviation in Tank Temperature vs Time for Kc = 5,10,20,100 5 Kc = 100 Deviation in Tank Temperature t 4.5 4 3.5 Kc = 20 3 2.5 Kc = 10 2 Kc = 5 1.5 1 0.5 0 0 5 10 15 Time FIGURE 12–5 Response of stirred-tank temperature for proportional control—set point tracking. Offset vs Kc for Set Point Tracking 5 4.5 4 3.5 Offset 3 2.5 2 1.5 1 0.5 0 0 10 20 30 40 50 Kc 60 70 80 90 100 FIGURE 12–6 Effect of controller gain on offset for stirred-tank proportional control—set point tracking. cou9789x_ch12_228-251.indd 233 8/14/08 4:33:50 PM Confirming Pages 234 PART 3 LINEAR CLOSED-LOOP SYSTEMS 12.2 PROPORTIONAL CONTROL FOR LOAD CHANGE (REGULATOR PROBLEM— DISTURBANCE REJECTION) The same control system shown in Fig. 12–1e is to be considered. This time the set point remains fixed; that is, TR 0. We are interested in the response of the system to a change in the inlet stream temperature, i.e., to a load change. Remember that the goal of the control system in this case is to reject the effect of disturbances (changes in the inlet temperature Ti for this process) and maintain the controlled variable, the tank temperature, at the set point. Since the set point has not changed from its steady-state value for this case (TR 0), we want T () 0. By using the methods of Chap. 11, the overall transfer function becomes T⬘ 1/ (ts ⫹ 1) 1 ⫽ ⫽ 1 ⫹ K c A / (ts ⫹ 1) ts ⫹ 1 ⫹ K c A Ti⬘ (12.5) This may be arranged in the form of the first-order lag; thus T⬘ A2 ⫽ Ti⬘ t 1s ⫹ 1 where A2 ⫽ 1 1 ⫹ Kc A t1 ⫽ t 1 ⫹ Kc A (12.6) As for the case of set point change, we have an overall response that is first-order. The overall time constant t1 is the same as for set point changes. For a step change in inlet temperature, let us examine the response of the controlled process and how effectively the disturbance is rejected. Thus for Ti 1/s, 1 1/ (1 ⫹ K c A ) T⬘ ⫽ s [t / (1 ⫹ K c A )] s ⫹ 1 and the ultimate (steady-state) value of T is T ⬘(⬁) lim sT ⬘(s) ⫽ s→0 1 1 ⫹ Kc A 1 T' Without control (Kc = 0) 1 1+KcA 0 0 With control (KcA = 2) Offset t FIGURE 12–7 Unit-step response for load change (P control). cou9789x_ch12_228-251.indd 234 The response of the system to a unit-step change in inlet temperature Ti is shown in Fig. 12–7. It may be seen that T approaches 1/(1 Kc A). To demonstrate the benefit of control, we have shown the response of the tank temperature (open-loop response) to a unitstep change in inlet temperature if no control were present, that is, Kc 0. In this case, the 8/14/08 4:33:51 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 235 major advantage of control lies in the reduction of offset. From Eq. (12.3), the offset becomes Offset ⫽ TR⬘ ( ⬁ ) ⫺ T ⬘( ⬁ ) ⫽ 0 ⫺ Offset ⫽ ⫺ 1 1 ⫹ Kc A 1 1 ⫹ Kc A (12.7) As for the case of a step change in set point, the absolute value of the offset is reduced as controller gain Kc is increased. Example 12.2. Proportional control of a stirred-tank heater for disturbance rejection. Determine the response of a closed-loop proportional control system in Example 12.1 for Kc 5, 10, 20, and 100 for a disturbance of 5C (i.e., the feed temperature increases by 5C from 60 to 65C). Determine the offset for each value of Kc. We can use the same Simulink model for this analysis. We just have the step change enter the process as the disturbance and leave the set point set to zero. The results are as follows. Figure 12–8 shows the tank temperature (plotted as a deviation variable) versus time, and Fig. 12–9 shows the offset plotted as a function of Kc. Tank Deviation Temperature vs Time for Kc = 5, 10, 20, 100 4 Kc = 5 Tank Deviation Temperature T' 3.5 Kc = 10 3 2.5 Kc = 20 2 1.5 1 Kc = 100 0.5 0 0 5 10 15 Time FIGURE 12–8 Response of stirred-tank temperature for proportional control—disturbance rejection. cou9789x_ch12_228-251.indd 235 8/14/08 4:33:52 PM Rev. Confirming Pages 236 PART 3 LINEAR CLOSED-LOOP SYSTEMS Offset vs Kc for Disturbance Rejection 0 −0.5 −1 Offset −1.5 −2 −2.5 −3 −3.5 −4 −4.5 0 10 20 30 40 50 Kc 60 70 80 90 100 FIGURE 12–9 Effect of controller gain on offset for stirred-tank proportional control—disturbance rejection. Note the results are qualitatively the same as for set point tracking. As Kc increases for proportional control, the system does a better job at rejecting the disturbance and bringing the tank temperature back to the original set point (0⬚C in deviation variables, or 80⬚C). The final temperature is closer to zero, and the offset also approaches zero. 12.3 PROPORTIONAL-INTEGRAL CONTROL FOR LOAD CHANGE Again, for the case of a load change, the desired goal is the rejection of the disturbance, and the desired final value is zero (since the set point remains unchanged). For PI control, we replace Gc in Fig. 12–1e by Kc (1 ⫹ 1/tI s). The overall transfer function for load change is therefore T⬘ ⫽ Ti⬘ 1/ (ts ⫹ 1) K c (1 ⫹ 1/t I s ) A 1⫹ ts ⫹ 1 (12.8) Rearranging this gives T⬘ tIs ⫽ ( t s ⫹ 1 ( t s ⫹ K c A (t I s ⫹ 1) ⬘ ) ) Ti I or T⬘ tIs ⫽ 2 Ti⬘ tt I s ⫹ ( K c At I ⫹ t I ) s ⫹ K c A cou9789x_ch12_228-251.indd 236 9/4/08 10:42:17 AM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 237 Since the denominator contains a quadratic expression, the transfer function may be written in the standard form of a second-order system to give T⬘ (t I /K c A) s ⫽ 2 Ti⬘ (tt I /K c A) s ⫹ t I (1 ⫹ 1/K c A) s ⫹ 1 or T⬘ A1s ⫽ 2 2 Ti⬘ t 1 s ⫹ 2zt 1s ⫹ 1 (12.9) where A1 ⫽ t1 ⫽ z ⫽ tI Kc A tt I Kc A 1 t I 1 ⫹ Kc A 2 t Kc A For a unit-step change in load, Ti 1/s. Combining this with Eq. (12.9) gives T⬘ ⫽ t 12 s 2 A1 ⫹ 2zt 1s ⫹ 1 (12.10) Equation (12.10) shows that the response of the tank temperature to a step change in Ti is equivalent to the response of a second-order system to an impulse function of magnitude A1. Since we have studied the impulse response of a second-order system in Chap. 7, the solution to the present problem is already known. This justifies in part some of our previous work on transients. By using Eq. (7.32), the impulse response for this system may be written for z 1 as 1 t 1 T ⬘ ⫽ A1 e⫺zt /t 1 sin 1 ⫺ z 2 t 1 1 ⫺ z 2 t 1 (12.11) Although the response of the system can be determined from Eq. (12.11) or Fig. 7–8, the effect of varying Kc and tI on the system response can be seen more clearly by plotting response curves, such as those shown in Fig. 12–10. From Fig. 12–10a, we see that an increase in Kc, for a fixed value of tI, improves the response by decreasing the maximum deviation and by making the response less oscillatory. The formula for z in Eq. (12.9) shows that z increases with Kc, which indicates that the response is less oscillatory. Figure 12–10b shows that, for a fixed value of Kc, a decrease in tI decreases cou9789x_ch12_228-251.indd 237 8/14/08 4:33:53 PM Rev. Confirming Pages PART 3 LINEAR CLOSED-LOOP SYSTEMS 0.2 Kc = 2.0 T' Kc = 3.5 =1 A=1 I = 0.25 Kc = 8.2 0 0 1 2 =1 A=1 Kc = 3.5 0.2 I = 0.5 I= T' 238 0.25 I= 0 3 0 0.125 1 t (a) 2 3 t (b) FIGURE 12–10 Unit-step response for load change—disturbance rejection (PI control). the maximum deviation and period. However, a decrease in tI causes the response to become more oscillatory, which means that z decreases. This effect of tI on the oscillatory nature of the response is also given by the formula for z in Eq. (12.9). Summary for Proportional-Integral Control—Response to Step Change in Load For fixed value of t I: As Kc ↑, z ↑ Max. deviation ↓ Oscillations ↓ For fixed value of Kc: As t I ↓, z ↓ Max. deviation ↓ Oscillations ↑ For this case, the offset as defined by Eq. (12.3) is zero; thus Offset TR ( ) T ( ) 00 0 One of the most important advantages of PI control is the elimination of offset. Key Point Proportional-Integral Control ã No offset Example 12.3. Proportional-integral control of a stirred tank heater for disturbance rejection. (a) Determine the response of a closed-loop proportional control system in Example 12.1 for Kc 5, 10, 20, and 100 for a disturbance of 5C (i.e., the feed temperature increases by 5C from 60 to 65C). Use a value of 2 min for tI. (b) Repeat part (a), using a constant value of 20 for Kc and values of 1, 2, 5, and 10 min for tI. We can use nearly the same Simulink model for this analysis. We just change the gain block that we used for the proportional controller to a PID block (see Fig. 12–11). Within the PID block, the controller algorithm is given as P I/s Ds. Comparing this with our standard PID equation, we conclude that P Kc, I Kc/tI, and D KctD. So that we can vary the model parameters with a MATLAB m-file, we insert the variable kc for P and kc/taui for I. cou9789x_ch12_228-251.indd 238 1/1/70 3:35:30 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 239 FIGURE 12–11 Simulink PID block parameters for Example 12.3. For a disturbance change only, the step change enters the process as the disturbance, and we leave the set point input set to zero. The m-file that we use to drive the model is shown below. clear % the variable h can be used to plot each line in a different color h(1,:)='g- '; h(2,:)='r-.'; h(3,:)='b: '; h(4,:)='k--'; h(5,:)='k- '; taui=2 for i=1:4 z=[5,10,20,100]; kc=z(1,i); % the variable y in the 'sim' statement is taken from the outport in the model [t,x,y]=sim('example12_3',50); plot(t,y(:,1),h(i,:)) hold on end grid hold off The result for a constant tI of 2 min is shown in Fig. 12–12. cou9789x_ch12_228-251.indd 239 8/14/08 4:33:55 PM Confirming Pages 240 PART 3 LINEAR CLOSED-LOOP SYSTEMS Stirred-Tank Disturbance Rejection Using Pl Control, for I = 2 min 2.5 Kc = 5 Tank Temperature T' 2 Kc = 10 1.5 Kc = 20 1 0.5 Kc = 100 0 −0.5 5 0 10 15 20 25 30 Time (min) 35 40 45 50 FIGURE 12–12 Stirred-tank disturbance rejection using PI control, for tI 2 min. The result for a constant value of Kc 20 is shown in Fig. 12–13. Stirred -Tank Disturbance Rejection Using Pl Control, for Kc = 20 2 1.5 Tank Temperature T' 1 = 10 1 1 =2 1 =5 0.5 1 =1 0 −0.5 0 5 10 15 20 25 30 Time (min) 35 40 45 50 FIGURE 12–13 Stirred-tank disturbance rejection using PI control, for Kc 20. cou9789x_ch12_228-251.indd 240 8/14/08 4:33:55 PM Rev. Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 12.4 241 PROPORTIONAL-INTEGRAL CONTROL FOR SET POINT CHANGE Again, the controller transfer function is Kc(1 1/tIs), and we obtain from Fig. 12–1e the transfer function AK c (1 1/t I s ) [1/ (ts 1)] T TR 1 AK c (1 1/t I s ) [1/ (ts 1)] (12.12) This equation may be reduced to the standard quadratic form to give T tIs 1 2 2 TR t 1 s 2zt 1s 1 (12.13) where t1 and z are the same functions of the parameters as in Eq. (12.9). Introducing a unit-step change (TR 1/s) into Eq. (12.13) gives T 1 tIs 1 2 2 s t 1 s 2zt 1s 1 (12.14) To obtain the response of T in the time domain, Eq. (12.14) is expanded into two terms: 1 1 tI T 2 2 (12.15) 2 2 s t 1 s 2zt 1s 1 t 1 s 2zt 1s 1 The first term on the right is equivalent to the response of a second-order system to an impulse function of magnitude tI. The second term is the unit-step response of a second-order system. It is convenient to use Figs. 8–3 and 8–6 to obtain the response for Eq. (12.15). For z 1, an analytic expression for T is t 1 T ez t /t 1 sin 1 z 2 1 2 t1 t1 1 z (12.16) 1 z 1 t ez t /t 1 sin 1 z 2 tan1 2 t1 z 1 z 2 The last expression was obtained by combining Eqs. (7.18) and (7.32). A typical response for T is shown in Fig. 12–14. The offset as defined by Eq. (12.3) is zero; thus Offset TR ( ) T ( ) 11 0 Again notice that the integral action in the controller has eliminated the offset. cou9789x_ch12_228-251.indd 241 1/1/70 3:35:31 PM Confirming Pages 242 PART 3 LINEAR CLOSED-LOOP SYSTEMS T' 1 Kc = 1 I = 0.2 A=1 =1 0 0 2 4 6 t FIGURE 12–14 Unit-step response for set point change (PI control). Proportional-integral control of a stirred-tank heater for set Example 12.4. point tracking (a) Determine the response of a closed-loop proportional control system in Example 12.1 for Kc 5, 10, 20, and 100 for a set point change of 5C. Use a value of 2 min for tI. (b) Repeat part (a) using a constant value of 20 for Kc and values of 1, 2, 5, and 10 min for tI. By proceeding as before using the Simulink model driven with a MATLAB mfile, these results are obtained, shown in Figs. 12–15 and 12–16. Stirred Tank set Point Tracking with Pl Control, for I = 2 6 Kc = 20 Kc = 100 5 Kc = 10 Tank Temperature T' Kc = 5 4 3 2 1 0 0 5 10 15 20 25 30 Time (min) 35 40 45 50 FIGURE 12–15 Stirred-tank set point tracking with PI control (tI 2 min). cou9789x_ch12_228-251.indd 242 8/14/08 4:33:56 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 243 Stirred Tank Set Point Tracking with Pl Control for Kc = 20 7 1 =1 6 1 =2 1 =5 Tank Temperature T' 5 1 4 = 10 3 2 1 0 5 0 10 15 20 25 30 Time (min) 35 40 45 50 FIGURE 12–16 Stirred-tank set point tracking with PI control (Kc 20). Note that there is no offset with PI control. Eventually each system levels out at the desired set point of 5C. 12.5 PROPORTIONAL CONTROL OF SYSTEM WITH MEASUREMENT LAG In the previous examples the lag in the measuring element was assumed to be negligible (the feedback transfer function was taken as 1). We now consider the same control system, the stirred-tank heater of Fig. 12–1, with a first-order measuring element having a transfer function 1/(tms 1). The block diagram for the modified system is now shown in Fig. 12–17. T'i + T'R Kc − 1 wC + 1 s+1 + T' 1 ms + 1 FIGURE 12–17 Control system with measurement lag. cou9789x_ch12_228-251.indd 243 8/14/08 4:33:57 PM Confirming Pages 244 PART 3 LINEAR CLOSED-LOOP SYSTEMS By the usual procedure, the transfer function for set point changes may be written T⬘ A1 (t m s ⫹ 1) ⫽ 2 2 t 2 s ⫹ 2z 2t 2 s ⫹ 1 TR⬘ (12.17) where A1 ⫽ Kc A 1 ⫹ Kc A t2 ⫽ z2 ⫽ tt m 1 ⫹ Kc A t ⫹ tm 1 2 tt m 1 ⫹ K c A We will not obtain an expression for the transient response for this case, for it will be of the same form as Eq. (12.16). Adding the first-order measuring lag to the control system of Fig. 12–1 produces a second-order system even for proportional control. This means there will be an oscillatory response for an appropriate choice of the parameters t, tm, Kc, and A. To understand the effect of gain Kc and measuring lag tm on the behavior of the system, response curves are shown in Fig. 12–18 for various combinations of Kc and tm for a fixed value of t 1. In general, the response becomes more oscillatory, or less stable, as Kc or tm increases. =1 =1 T'R = 1s Kc = 8 T' T' Kc = 4 m =1 m = 0.5 =1 Kc = 8 T'R = 1s 1 1 0 =2 m m Kc = 2 0 1 2 3 4 (a) T'R 5 t + Kc 0 m 0 1 s+ 1 − =0 1 2 3 4 (b) 5 t T' A=1 1 ms + 1 (c) FIGURE 12–18 Effect of controller gain and measurement lag on system response for unit-step change in set point. For a fixed value of tm 1, Fig. 12–18a shows that the offset is reduced as Kc increases; however, this improvement in steady-state performance is obtained at the cou9789x_ch12_228-251.indd 244 8/14/08 4:33:57 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 245 expense of a poorer transient response. As Kc increases, the overshoot becomes excessive, and the response becomes more oscillatory. In general, we will find that a control system having proportional control will require a value of Kc that is based on a compromise between low offset and satisfactory transient response. For a fixed value of controller gain (Kc 8), Fig. 12–18b shows that an increase in measurement lag produces a poorer transient response in that the overshoot becomes greater and the response more oscillatory as tm increases. This behavior illustrates a general rule that the measuring element in a control system should respond quickly if satisfactory response is to be achieved. Example 12.5 Effect of measurement lag on the proportional-integral control of a stirred-tank heater for set point tracking. Let’s examine the effect of measurement lag on the PI control of the stirred-tank heater. We just concluded that as measurement lag increases, a proportional control only system becomes more oscillatory. What happens with PI control? The Simulink model is modified to include measurement lag (see Fig. 12–19). We’ll fix Kc at 20 and tI 2, which gave us reasonably good results in Example 12.4. 1 Out1 Heater Gain Set point Tr +− Set point Error Q(kW) PID 0.0714 1 ++ PID Controller Actual tank temp. T 5s+1 Tank Scope Feed Temp. Ti Measured Temp. Tm 1 taum s + 1 temps To workspace Thermocouple 2 Out2 FIGURE 12–19 Simulink model for Example 12.5. For values of tm 0.33, 1.0, 2.0, and 5.0, determine the nature of the response to a set point change of 5C. Figure 12–20 shows the results. It is clear that as the measurement lag increases, the response becomes more oscillatory and less stable. cou9789x_ch12_228-251.indd 245 8/14/08 4:33:58 PM Confirming Pages 246 PART 3 LINEAR CLOSED-LOOP SYSTEMS Effect of Measurement Lag on Pl Control of Stirred-Tank Heater 12 m Kc = 20 1=2 =5 Tank Temperature T' 10 8 m =2 m =1 6 m = 0.33 4 2 0 0 5 10 15 20 25 30 Time (min) 35 40 45 50 FIGURE 12–20 Effect of measurement lag on PI control of stirred-tank heater. SUMMARY In this chapter, we confined our attention to the response of simple control systems that were either first-order or second-order. This means that the transient response can be found by referring to Chaps. 4 and 7. However, if integral action were added to the controller in the system of Fig. 12–17, the overall transfer function would have a third-order polynomial in the denominator. Incorrect selection of controller parameters can lead to a response with increasing amplitude. These unstable responses can occur in all systems with third- or higher-order polynomials in the denominator of the overall transfer function. Inversion would require factoring a cubic, which is generally a difficult task by hand, but a routine task using a computer. Actually, systems with denominator polynomials of order greater than 2 are the rule rather than the exception. Hence, we will develop in forthcoming chapters convenient techniques for studying the response of higher-order control systems. These techniques will be of direct use in control system design. Unstable responses can occur in all systems with third- or higher-order polynomials in the denominator of the overall transfer function. In Chap.13 we will present a concrete definition of stability and begin the development of methods for determining stability in control systems. PROBLEMS 12.1. The set point of the control system shown in Fig. P12–1 is given a step change of 0.1 unit. Determine (a) The maximum value of C and the time at which it occurs (b) The offset (c) The period of oscillation cou9789x_ch12_228-251.indd 246 8/14/08 4:33:58 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 247 Draw a sketch of C(t) as a function of time. + R 5 (s + 1)(2s + 1) K = 1.6 − C FIGURE P12–1 12.2. The control system shown in Fig. P12–2 contains a PID controller. (a) For the closed loop, develop formulas for the natural period of oscillation t and the damping factor z in terms of the parameters K, tD, tI, and t1. For the following parts, tD tI 1 and t1 2, (b) Calculate z when K is 0.5 and when K is 2. (c) Do z and t approach limiting values as K increases, and if so, what are these values? (d ) Determine the offset for a unit-step change in load if K is 2. (e) Sketch the response curve (C versus t) for a unit-step change in load when K is 0.5 and when K is 2. ( f ) In both cases of part (e) determine the maximum value of C and the time at which it occurs. U + R K1+ Ds − + 1 τ Is + + 1s 1 +1 C FIGURE P12–2 12.3. The location of a load change in a control loop may affect the system response. In the block diagram shown in Fig. P12–3, a unit-step change in load enters at either location 1 or location 2. (a) What is the frequency of the transient response when the load enters at location 1 and when the load enters at location 2? (b) What is the offset when the load enters at location 1 and when it enters at location 2? (c) Sketch the transient response to a step change in U1 and to a step change in U2. U1 R + − Kc = 5 + + U2 2 2s + 1 + + 1 2s + 1 C FIGURE P12–3 cou9789x_ch12_228-251.indd 247 8/14/08 4:33:59 PM Confirming Pages 248 PART 3 LINEAR CLOSED-LOOP SYSTEMS 12.4. Consider the liquid-level control system shown in Fig. P12–4. The tanks are noninteracting. The following information is known: • The resistances on the tanks are linear. These resistances were tested separately, and it was found that if the steady-state flow rate q cfm is plotted against steady-state tank level h ft, the slope of the line dq/dh is 2 ft2/min. • The cross-sectional area of each tank is 2 ft2. • The control valve was tested separately, and it was found that a change of 1 psi in pressure to the valve produced a change in flow of 0.1 cfm. • There is no dynamic lag in the valve or the measuring element. (a) Draw a block diagram of this control system, and in each block give the transfer function, with numerical values of the parameters. (b) Determine the controller gain Kc for a critically damped response. (c) If the tanks were connected so that they were interacting, what is the value of Kc needed for critical damping? (d) Using 1.5 times the value of Kc determined in part (c), determine the response of the level in tank 2 to a step change in set point of 1 in of level. Proportional controller Water supply q R1 R2 FIGURE P12–4 12.5. A PD controller is used in a control system having a first-order process and a measurement lag as shown in Fig. P12–5. (a) Find expressions for z and t for the closed-loop response. (b) If t1 1 min and tm 10 s, find Kc so that z 0.7 for the two cases (1) tD 0 s and (2) tD 3 s. (c) Compare the offset and period realized for both cases, and comment on the advantage of adding the derivative mode. U R + + − Kc (1 + Ds) 1 +1 + 1s C 1 ms + 1 FIGURE P12–5 cou9789x_ch12_228-251.indd 248 8/14/08 4:33:59 PM Confirming Pages CHAPTER 12 TRANSIENT RESPONSE OF SIMPLE CONTROL SYSTEMS 249 12.6. The thermal system shown in Fig. P12–6 is controlled by a PD controller. These data are given: w ⫽ 250 lb/min r ⫽ 62.5 lb/ft 3 V1 ⫽ 4 ft 3 V2 ⫽ 5 ft 3 V3 ⫽ 6 ft 3 C ⫽ 1 Btu/( lb⭈⬚ F ) A change of 1 psi from the controller changes the flow rate of heat q by 500 Btu/min. The temperature of the inlet stream may vary. There is no lag in the measuring element. (a) Draw a block diagram of the control system with the appropriate transfer function in each block. Each transfer function should contain numerical values of the parameters. (b) From the block diagram, determine the overall transfer function relating the temperature in tank 3 to a change in set point. (c) Find the offset for a unit-step change in inlet temperature if the controller gain Kc is 3 psi/F of temperature error and the derivative time is 0.5 min. PD controller Final control element psi °F q w V1 V2 V3 FIGURE P12–6 12.7. (a) For the control system shown in Fig. P12–7, obtain the closed-loop transfer function C/U. (b) Find the value of Kc for which the closed-loop response has a z of 2.3. (c) Find the offset for a unit-step change in U if Kc 4. U + R + Kc − s+1 0.25s + 1 + 1 s C FIGURE P12–7 12.8. For the control system shown in Fig. P12–8, determine (a) C(s)/R(s) (b) C() (c) Offset (d) C(0.5) (e) Whether the closed-loop response is oscillatory cou9789x_ch12_228-251.indd 249 8/14/08 4:34:00 PM Confirming Pages 250 PART 3 LINEAR CLOSED-LOOP SYSTEMS U=0 + R= + 2 s 2 + 2 s(s + 1) C − FIGURE P12–8 12.9. For the control system shown in Fig. P12–9, determine an expression for C(t) if a unit-step change occurs in R. Sketch the response C(t) and compute C(2). R + 1+ 1 s 1 C − FIGURE P12–9 12.10. Compare the responses to a unit-step change in set point for the system shown in Fig. P12–10 for both negative feedback and positive feedback. Do this for Kc of 0.5 and 1.0. Compare these responses by sketching C(t). R + Kc 1 s+1 C + or − FIGURE P12–10 cou9789x_ch12_228-251.indd 250 8/14/08 4:34:01 PM Confirming Pages CHAPTER 12 CAPSULE SUMMARY Offset: The difference (at steady state) between the desired value of the controlled variable (set point) and the actual value of the controlled variable. Measurement lag: The response time of the sensor used to measure the value of the controlled variable and send it to the controller. It usually appears in the feedback loop of the control system. TRANSIENT RESPONSE OF A PROPORTIONAL-ONLY CONTROL SYSTEM Key Point - With proportional-only control there will always be some offset and As Kc ↑ Offset ↓ TRANSIENT RESPONSE OF A PROPORTIONAL-INTEGRAL CONTROL SYSTEM Key Point - With proportional-integral control there is no offset. However, the addition of the integral action in the controller can cause the system response to become more oscillatory and increase the maximum deviation (or overshoot) of the response. PI Control ⴝ No offset, but may be oscillatory TRANSIENT RESPONSE OF A SYSTEM WITH MEASUREMENT LAG The larger the measurement lag in a control system, the poorer the control performance. As measurement lag increases, the response of the system will degrade and will become more oscillatory. 251 cou9789x_ch12_228-251.indd 251 8/14/08 4:34:02 PM Confirming Pages CHAPTER 13 STABILITY 13.1 CONCEPT OF STABILITY In Chap. 12, the overall response of the control system was no higher than second-order. For these systems, the step response must resemble those of Fig. 4–7 or of Fig. 7–3. Hence, the system is inherently stable. In this chapter we consider the problem of stability in a control system (Fig. 13–1) only slightly more complicated than any studied previously. This system might represent proportional control of two stirred-tank heaters with measuring lag. In this discussion, only set point changes are to be considered. From the methods developed in Chap. 11 for determining the overall transfer function, we have from Fig. 13–1 C G1G2 (13.1) R 1 G1G2 H U + + R G1 = Kc _ G2 = + H= 1 ( 1s + 1) ( 2s + 1) C 1 ( 3s + 1) FIGURE 13–1 Third-order control system. 252 cou9789x_ch13_252-268.indd 252 8/14/08 8:04:12 AM Confirming Pages CHAPTER 13 STABILITY 253 In terms of the particular transfer functions shown in Fig. 13–1, C/R becomes, after some rearrangement, C K c (t 3 s 1) R (t 1s 1) (t 2 s 1) (t 3s 1) K c (13.2) The denominator of Eq. (13.2) is a third-order polynomial. For a unit-step change in R, the transform of the response is C 1 K c (t 3 s 1) s (t 1s 1) (t 2 s 1) (t 3 s 1) K c (13.3) To obtain the transient response C(t), it is necessary to find the inverse of Eq. (13.3). This requires obtaining the roots of the denominator of Eq. (13.2), which is third-order. We can no longer find these roots as easily as we did for the second-order systems by use of the quadratic formula. However, in principle they can always be obtained by algebraic methods or through the use of computer software such as MATLAB. It is apparent that the roots of the denominator depend upon the particular values of the time constants and Kc. These roots determine the nature of the transient response, according to the rules presented in Fig. 3–1 and Table 3.1. It is of interest to examine the nature of the response for the control system of Fig. 13–1 as Kc is varied, assuming the time constants t1,t2, and t3 to be fixed. To be specific, consider the step response for t 1 1, t 2 12 , and t 3 13 for several values of Kc. Without going into the detailed calculations at this time, the results of inversion of Eq. (13.3) are shown as response curves in Fig. 13–2. From these response curves, it is seen that as Kc increases, the system response becomes more oscillatory. In fact, beyond a certain value of Kc, the successive amplitudes of the response grow rather than decay; this type of response is called unstable. Evidently, for some values of Kc, there is a pair of roots corresponding to s4 and s4* of Fig. 3–1. As control system designers, we are clearly interested in being able to determine quickly the values of Kc that give unstable responses, such as that corresponding to Kc 12 in Fig. 13–2. C (t) Kc = 12 Kc = 9 Kc = 6 Unit step 1 Kc = 3 0 0 2 4 6 8 10 t FIGURE 13–2 Response of control system of Fig. 13–1 for a unit-step change in set point. cou9789x_ch13_252-268.indd 253 8/14/08 8:04:13 AM Confirming Pages 254 PART 3 LINEAR CLOSED-LOOP SYSTEMS If the order of Eq. (13.2) had been higher than 3, the hand calculations necessary to obtain Fig. 13–2 would have been even more difficult. In this chapter, the focus is on developing a clearer understanding of the concept of stability. In addition, we develop a quick test for detecting roots having positive real parts, such as s4 and s 4* in Fig. 3–1. 13.2 DEFINITION OF STABILITY (LINEAR SYSTEMS) For our purposes, a stable system will be defined as one for which the output response is bounded for all bounded inputs. A system exhibiting an unbounded response to a bounded input is unstable. This definition, although somewhat loose, is adequate for most of the linear systems and simple inputs that we shall study. STABLE SYSTEM → a bounded input produces a bounded output (BIBO) A bounded input function is a function of time that always falls within certain bounds during the course of time. For example, the step function and sinusoidal function are bounded inputs. The function f(t) t is obviously unbounded. Although the definition of an unstable system states that the output becomes unbounded, this is true only in the mathematical sense. An actual physical system always exhibits bounds or constraints. A linear mathematical model (set of linear differential equations describing the system) from which stability information is obtained is meaningful only over a certain range of variables. For example, a linear control valve gives a linear relation between flow and valve-top pressure only over the range of pressure (or flow) corresponding to values between which the valve is shut tight or wide open. When the valve is wide open, for example, further change in pressure to the diaphragm will not increase the flow. We often describe such a limitation by the term saturation. A physical system, when unstable, may not follow the response of its linear mathematical model beyond certain physical bounds but rather may saturate. However, the prediction of stability by the linear model is of utmost importance in a real control system since operation with the valve shut tight or wide open is clearly unsatisfactory control. 13.3 STABILITY CRITERION The purpose of this section is to translate the stability definition into a simpler criterion, one that can be used to ascertain the stability of control systems of the form shown in Fig. 13–3. U R + G1 − B + + G2 C H FIGURE 13–3 Basic single-loop control system. cou9789x_ch13_252-268.indd 254 8/14/08 8:04:13 AM Confirming Pages CHAPTER 13 STABILITY 255 From the block diagram of the control system (Fig. 13–3), we obtain by the methods of Chap. 12 CHARACTERISTIC EQUATION. C G1G2 G2 R U 1 G1G2 H 1 G1G2 H (13.4) To simplify the nomenclature, let G G1G2H. We call G the open-loop transfer function because it relates the measured variable B to the set point R if the feedback loop of Fig. 13–3 is disconnected from the comparator (i.e., if the loop is opened). In terms of the open-loop transfer function G, Eq. (13.4) becomes C G1G2 G2 R U 1G 1G (13.5) In principle, for given forcing functions R(s) and U(s), Eq. (13.5) may be inverted to give the control system response. To determine under what conditions the system represented by Eq. (13.5) is stable, it is necessary to test the response to a bounded input. Suppose a unit-step change in set point is applied. Then G1G2 F ( s ) G1G2 1 s ( s r1 )( s r2 )( s rn ) 1G s C (s) (13.6) where r1,r2, . . . , rn are the n roots of the equation 1 G (s) 0 (13.7) and F(s) is a function that arises in the rearrangement to the right-hand form of Eq. (13.6). Equation (13.7) is called the characteristic equation for the control system of Fig. 13–3. For example, for the control system of Fig. 13–1 the step response is C (s) G1G2 s (1 G ) K c / [(t 1s 1) (t 2 s 1)] s [1 K c / (t 1s 1) (t 2 s 1) (t 3 s 1)] which may be rearranged to C (s) K c (t 3 s 1) s t 1t 2t 3 s (t 1t 2 t 1t 3 t 2t 3 ) s 2 (t 1 t 2 t 3 ) s (1 K c ) 3 This is equivalent to C (s) K c (t 3 s 1) /t 1t 2t 3 s ( s r1) ( s r2 ) ( s r3 ) where r1, r2, and r3 are the roots of the characteristic equation cou9789x_ch13_252-268.indd 255 8/14/08 8:04:14 AM Confirming Pages 256 PART 3 LINEAR CLOSED-LOOP SYSTEMS t 1t 2t 3 s 3 (t 1t 2 t 1t 3 t 2t 3 ) s 2 (t 1 t 2 t 3 ) s (1 K c ) 0 (13.8) Evidently, for this case the function F(s) in Eq. (13.6) is F (s) (t 1s 1) (t 2 s 1) (t 3s 1) t 1t 2t 3 In Chap. 3, the qualitative nature of the inverse transforms of equations such as Eq. (13.6) was discussed. It was shown that (see Fig. 3–1 and Table 3.1) if there are any of the roots r1, r2, . . . , rn in the right half of the complex plane, the response C(t) will contain a term that grows exponentially in time and the system is unstable. If there are one or more roots of the characteristic equation at the origin, there is an sm in the denominator of Eq. (13.6) (where m 2) and the response is again unbounded, growing as a polynomial in time. This condition specifies m as greater than or equal to 2, not 1, because one of the s terms in the denominator is accounted for by the fact that the input is a unit step (1/s) in Eq. (13.6). (Note that a 1/s term will invert to a constant, while a 1/s2 term will invert to a term of the form C1t, which is unbounded.) Additionally, if there is a pair of conjugate roots of the characteristic equation on the imaginary axis, the contribution to the overall step response is a pure sinusoid, which is bounded. However, if the bounded input is taken as sin wt, where w is the imaginary part of the conjugate roots, the contribution to the overall response is a sinusoid with an amplitude that increases as a polynomial in time [the response will have a term of the form C1t sin(wt f)]. Thus, if a root lies on the imaginary axis, there is the potential for repeating the root of a bounded input (such as a step input or a sinusoid input), and the response will be unstable. Therefore, the right-half plane, including the imaginary axis, is the unstable region for location of roots of the characteristic equation. It is evident from Eq. (13.5) that precisely the same considerations apply to a change in the load U, as we just discussed for set point changes, since they have the same characteristic equation. Therefore, the definition of stability for linear systems may be translated to the following criterion: A linear control system is unstable if any roots of its characteristic equation are on, or to the right of, the imaginary axis. Otherwise the system is stable. It is important to note that the characteristic equation of a control system, which determines its stability, is the same for set point or load changes. The stability depends only upon G(s), the open-loop transfer function. Furthermore, although the rules derived above were based on a step input, they are applicable to any input. This is true, first, by the definition of stability and, second, because if there is a root of the characteristic equation in the right half-plane, it contributes an unbounded term in the response to any input. This follows from Eq. (13.5) after it is rearranged to the form of Eq. (13.6) for the particular input. cou9789x_ch13_252-268.indd 256 8/14/08 8:04:14 AM Confirming Pages CHAPTER 13 Imaginary axis Stable Region Left Half-Plane (LHP) 1.5 1 0.5 0 –0.5 –1 0 0.5 1 1.5 2 s3 (0, b3) s2 2.5 2 1.5 1 0 0.5 –1 –1.5 –2 0 0.1 0.20.30.40.50.60.70.80.9 1 s4 1.5 (–a2, –b2) 0.5 0 (a4, –b4) 2 s2* 1 0 0.1 0.20.30.40.50.60.70.80.9 1 s4* 2.5 1.5 12 10 8 6 4 2 0 –2 –4 –6 –8 (a4, b4) s6 (0, 0) (–a1, 0) 1 0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 s3* 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 257 Unstable Region Right Half-Plane (RHP) (–a2, b2) s1 STABILITY (0, –b3) s5 (a5, 0) 10 9 8 7 6 5 4 3 2 1 Real axis 0 0.1 0.20.30.40.50.60.70.80.9 1 2 1.5 1 0 0.5 –1.5 –2 0 0.1 0.20.30.40.50.60.70.80.9 1 FIGURE 13–4 Stability of typical roots of the characteristic equation. Therefore, the stability of a control system of the type shown in Fig. 13–3 is determined solely by its open-loop transfer function through the roots of the characteristic equation. This behavior is shown graphically in Fig. 13–4, which relates the nature of the response to the location of the roots of the characteristic equation. Example 13.1. In terms of Fig. 13–3, a control system has the transfer functions 0 . 5s 1 ( PI controller) s 1 G2 (stirred tank) 2s 1 H 1 (measuring element without lag) G1 10 We have suggested a physical system by the components placed in parentheses. Find the characteristic equation and its roots, and determine whether the system is stable. The first step is to write the open-loop transfer function G G1G2 H cou9789x_ch13_252-268.indd 257 10 ( 0 . 5s 1) s ( 2s 1) 8/14/08 8:04:15 AM Confirming Pages 258 PART 3 LINEAR CLOSED-LOOP SYSTEMS The characteristic equation is therefore 1 10 ( 0 . 5s 1) 0 s ( 2s 1) which is equivalent to s 2 3s 5 0 Solving by the quadratic formula gives s 3 2 9 20 2 or s1 3 11 j 2 2 s2 3 11 j 2 2 Since the real part of s1 and s2 is negative (3/2), the system is stable. 13.4 ROUTH TEST FOR STABILITY The Routh test is a purely algebraic method for determining how many roots of the characteristic equation have positive real parts; from this it can also be determined whether the system is stable, for if there are no roots with positive real parts, the system is stable. The test is limited to systems that have polynomial characteristic equations. This means that it cannot be used to test the stability of a control system containing a transportation lag. The procedure for application of the Routh test is presented without proof. The proof is available elsewhere (Routh, 1905) and is mathematically beyond the scope of this text. The procedure for examining the roots is to write the characteristic equation in the form (13.9) a0 s n a1s n 1 a2 s n 2 . . . an 0 where a0 is positive. (If a0 is originally negative, both sides are multiplied by 1.) In this form, it is necessary that all the coefficients a0 , a1 , a2 , ... , an 1 , an be positive if all the roots are to lie in the left half-plane. If any coefficient is negative, the system is definitely unstable, and the Routh test is not needed to answer the question of stability. (However, in this case, the Routh test will tell us the number of roots in the right half-plane.) If all the coefficients are positive, the system may be stable or unstable. It is then necessary to apply the following procedure to determine stability. cou9789x_ch13_252-268.indd 258 8/14/08 8:04:15 AM Confirming Pages CHAPTER 13 STABILITY 259 Routh Array Arrange the coefficients of Eq. (13.9) into the first two rows of the Routh array as follows: Row 1 2 3 4 5 6 7 n1 a0 a1 b1 c1 d1 e1 f1 g1 a2 a3 b2 c2 d2 e2 a4 a5 b3 c3 a6 a7 The array has been filled in for n 7 to simplify the discussion. For any other value of n, the array is prepared in the same manner. In general, there are n 1 rows. For n even, the first row has one more element than the second row. The elements in the remaining rows are found from the formulas a1a2 a0 a3 a1 b1a3 a1b2 c1 b1 a1a4 a0 a5 a1 b1a5 a1b3 c2 b1 b1 b2 ... ... The elements for the other rows are found from formulas that correspond to those just given. The elements in any row are always derived from the elements of the two preceding rows. During the computation of the Routh array, any row can be divided by a positive constant without changing the results of the test. (The application of this rule often simplifies the arithmetic.) Having obtained the Routh array, we can apply the following theorems to determine stability. THEOREMS OF THE ROUTH TEST Theorem 13.1. The necessary and sufficient condition for all the roots of the characteristic equation [Eq. (13.9)] to have negative real parts (stable system) is that all elements of the first column of the Routh array (a0, a1, b1, c1, etc.) be positive and nonzero. Theorem 13.2. If some of the elements in the first column are negative, the number of roots with a positive real part (in the right half-plane) is equal to the number of sign changes in the first column. Theorem 13.3. If one pair of roots is on the imaginary axis, equidistant from the origin, and all other roots are in the left half-plane, then all the elements of the nth row will vanish and none of the elements of the preceding row will vanish. The location of the pair of imaginary roots can be found by solving the equation Cs 2 D 0 (13.10) where the coefficients C and D are the elements of the array in the (n – 1)st row as read from left to right, respectively. cou9789x_ch13_252-268.indd 259 8/14/08 8:04:16 AM Confirming Pages 260 PART 3 LINEAR CLOSED-LOOP SYSTEMS The algebraic method for determining stability is limited in its usefulness in that all we can learn from it is whether a system is stable. It does not give us any idea of the degree of stability or the roots of the characteristic equation. Example 13.2. Given the characteristic equation s 4 3s 2 5s 2 4 s 2 0 determine the stability by the Routh criterion. Since all the coefficients are positive, the system may be stable. To test this, form the following Routh array: Row 1 2 3 4 5 1 3 11/3 26/11 2 5 4 6/3 0 2 The elements in the array are found by applying the formulas presented in the rules; for example, b1, which is the element in the first column, third row, is obtained by b1 a1a2 a0 a3 a1 or in terms of numerical values, b1 (3)(5) (1)(4) 15 4 11 3 3 3 3 Since there is no change in sign in the first column, there are no roots having positive real parts, and the system is stable. Example 13.3. (a) Using t 1 1, t 2 12 , and t 3 13 , determine the values of Kc for which the control system in Fig. 13–1 is stable. (b) For the value of Kc for which the system is on the threshold of instability, determine the roots of the characteristic equation with the help of Theorem 13.3. Solution (a) The characteristic equation 1 G(s) 0 becomes 1 cou9789x_ch13_252-268.indd 260 Kc ( s 1) ( s / 2 1) ( s / 3 1) 0 8/14/08 8:04:16 AM Confirming Pages CHAPTER 13 STABILITY 261 Using MATLAB to Determine the Roots of a Polynomial The MATLAB command roots(C) computes the roots of the polynomial whose coefficients are the elements of the vector C. If C has n 1 elements, the polynomial is C1 x n C2 x n 1 C3 x n 2 . . . Cn 1 x 2 Cn x Cn 1 For the characteristic equation of Example 13.2, s4 3s2 5s2 4s 2 0, the vector C would be C [1, 3, 5, 4, 2] The MATLAB command and the resulting solution are shown below. roots([1,3,5,4,2]) ans = -1.0000 -1.0000 -0.5000 -0.5000 1.0000i 1.0000i 0.8660i 0.8660i The real parts of all four roots are negative. Thus, the roots all lie in the left half-plane (LHP), and the system is stable. This is the same result that we obtained using the Routh test. Rearrangement of this equation for use in the Routh test gives s 3 6s 2 11s 6 (1 K c ) 0 (13.11) The Routh array is Row 1 2 3 4 1 6 10 Kc 6(1 Kc) 11 6(1 Kc) Since the proportional sensitivity of the controller Kc is a positive quantity, we see that the fourth entry in the first column, 6(1 Kc), is positive. According to Theorem 13.1, all the elements of the first column must be positive for stability; hence 10 K c 0 K c 10 It is concluded that the system will be stable only if Kc 10, which agrees with Fig. 13–2. cou9789x_ch13_252-268.indd 261 8/14/08 8:04:17 AM Confirming Pages 262 PART 3 LINEAR CLOSED-LOOP SYSTEMS (b) At Kc 10, the system is on the verge of instability, and the element in the nth (third) row of the array is zero. According to Theorem 13.3, the location of the imaginary roots is obtained by solving Cs 2 D 0 where C and D are the elements in the (n 1)st row. For this problem, with Kc 10, we obtain 6s 2 66 0 s j 11 Therefore, two of the roots on the imaginary axis are located at 11 and 11. The third root can be found by expressing Eq. (13.11) in factored form ( s s1 ) ( s s2 ) ( s s3 ) 0 (13.12) where s1, s2, and s3 are the roots. Introducing the two imaginary roots (s1 j 11 and s2 j 11 ) into Eq. (13.12) and multiplying out the terms give s 3 s3 s 2 11s 11s3 0 Comparing this equation with Eq. (13.11), we see that s3 6. The roots of the characteristic equation are therefore s1 j 11, s2 j 11, and s3 6. Using MATLAB and s3 6s2 11s 6(1 Kc) 0 with Kc 10, we can use the roots command as follows. roots([1,6,11,66]) ans = -6.0000 0.0000 + 3.3166i 0.0000 – 3.3166i The result is the same, 11 3 . 3166 . Example 13.4. Determine the stability of the system shown in Fig. 13–1 for which a PI controller is used. Use t 1 1, t 2 12 , t 3 13 , K c 5, and τI 0.25. Solution Characteristic equation is 1 ( K c / t 1t 2t 3 ) (t I s 1) 0 t I s [ s (1 / t 1 )][ s (1 / t 2 )][ s (1 / t 3 )] Using the parameters given above in this equation leads to s 4 6s 3 11s 2 36s 120 0 cou9789x_ch13_252-268.indd 262 8/14/08 8:04:17 AM Confirming Pages CHAPTER 13 STABILITY 263 Notice that the order of the characteristic equation has increased from three to four as a result of adding integral action to the controller. The Routh array becomes Row 1 2 3 4 5 1 6 5 108 120 11 36 120 120 Because there are two sign changes in the first column, we know from Theorem 13.2 of the Routh test that two roots have positive real parts. From Example 13.3 we know that for Kc 5 the system is stable with proportional control. With integral action present, however, the system is unstable for Kc 5. Employing MATLAB to help us again yields roots([1,6,11,36,120]) ans = -3.7953 -3.7953 0.7953 0.7953 + + - 0.8664i 0.8664i 2.6992i 2.6992i Note the two roots in the RHP (with the positive real parts of 0.7953). The system is unstable. This is the same result we obtained with the Routh test. SUMMARY A definition of stability for a control system has been presented and discussed. This definition was translated into a simple mathematical criterion relating stability to the location of roots of the characteristic equation. Briefly, it was found that a control system is stable if all the roots of its characteristic equation lie in the left half of the complex plane. The Routh criterion, a simple algebraic test for detecting roots of a polynomial lying in the right half of the complex plane, was presented and applied to control system stability analysis. This criterion suffers from two limitations: (1) It is applicable only to systems with polynomial characteristic equations, and (2) it gives no information about the actual location of the roots and, in particular, their proximity to the imaginary axis. This latter point is quite important, as can be seen from Fig. 13–2 and the results of Example 13.3. The Routh criterion tells us only that for Kc 10 the system is stable. However, from Fig. 13–2 it is clear that the value Kc 9 produces a response that is undesirable because it has a response time that is too long. In other words, the controlled variable oscillates too long before returning to steady state. It will be shown later that this happens because for Kc 9 there is a pair of roots close to the imaginary axis. cou9789x_ch13_252-268.indd 263 8/14/08 8:04:18 AM Confirming Pages 264 PART 3 LINEAR CLOSED-LOOP SYSTEMS In Chap. 14 tools will be developed for obtaining more information about the actual location of the roots of the characteristic equation. This will enable us to predict the form of the curves of Fig. 13–2 for various values of Kc. The advantage of these tools is that they are graphical and are easy to apply compared with standard algebraic solutions of the characteristic equation. There are two distinct approaches to this problem: root locus methods and frequency-response methods. The former are discussed in Chap. 14 and the latter in Chaps. 15 and 16. These groups of chapters are written in parallel, and the reader may study one or both groups in either order. As a guide to making this decision, here are some general comments concerning the two approaches. Root locus methods allow rapid determination of the location of the roots of the characteristic equation as functions of parameters such as Kc of Fig. 13–1. However, they are difficult to apply to systems containing transportation lags. Also, they require a reasonably accurate knowledge of the theoretical process transfer function. Frequency-response methods are an indirect solution to the location of the roots. They utilize the sinusoidal response of the open-loop transfer function to determine values of parameters such as Kc that keep these roots a “safe distance” from the right half-plane. The actual transient response for a given value of Kc can be only crudely approximated. However, frequency-response methods are easily applied to systems containing transportation lags and may be used with only experimental knowledge of the unsteady-state process behavior. A mastery of control theory requires knowledge of both methods because they are complementary. However, the reader may choose to study only frequency response and still be adequately prepared for most of the material in the remainder of this book. The choice of studying only root locus will be more restrictive in terms of preparation for subsequent chapters. In addition, much of the literature on process dynamics relies heavily on frequency-response methods. PROBLEMS 13.1. Write the characteristic equation and construct the Routh array for the control system shown in Fig. P13–1. Is the system stable for (a) Kc 9.5, (b) Kc 11, and (c) Kc 12? R + − 1 (s + 1)(0.5s + 1) Kc C 3 s+3 FIGURE P13–1 13.2. By means of the Routh test, determine the stability of the system shown in Fig. P13–2 when Kc 2. cou9789x_ch13_252-268.indd 264 8/14/08 8:04:19 AM Confirming Pages CHAPTER 13 R + − 3) Kc(1 + − s Kv = 1 2 STABILITY 265 C 1 0.2s2 + 0.4s +1 FIGURE P13–2 13.3. In the control system of Prob. 12.6, determine the value of gain (psi/ F) that just causes the system to be unstable if (a) tD 0.25 min and (b) tD 0.5 min. 13.4. Prove that if one or more of the coefficients a0, a1, . . . , an of the characteristic equation [Eq. (13.9)] is negative or zero, then there is necessarily an unstable root. Hint: First show that a1/a0 is minus the sum of all the roots, a2/a0 is plus the sum of all possible products of two roots, aj /a0 is (1) j times the sum of all possible products of j roots, etc. 13.5. Prove that the converse statement of Prob. 13.4—that an unstable root implies that one or more of the coefficients will be negative or zero—is untrue for all n > 2. Hint: To prove that a statement is untrue, it is only necessary to demonstrate a single counterexample. 13.6. Deduce an extension of the Routh criterion that will detect the presence of roots with real parts greater than σ for any specified s 0. 13.7. Show that any complex number s satisfying |s| 1 yields a value of z 1s 1s that satisfies Re ( z ) 0 (Hint: Let s x jy; z u jv. Rationalize the fraction, and equate real and imaginary parts of z and the rationalized fraction. Now consider what happens to the circle x2 y2 1. To show that the inside of the circle goes over to the right half-plane, consider a convenient point inside the circle.) On the basis of this transformation, deduce an extension of the Routh criterion that will determine whether the system has roots inside the unit circle. Why might this information be of interest? How can the transformation be modified to consider circles of other radii? 1 4s R + _ 1 1s + 1 ( 1s + 1) ( + 3s U +1 2s + 1) C +1 FIGURE P13–8 cou9789x_ch13_252-268.indd 265 8/14/08 8:04:19 AM Confirming Pages 266 PART 3 LINEAR CLOSED-LOOP SYSTEMS 13.8. Given the control diagram shown in Fig. P13–8, deduce by means of the Routh criterion those values of τI for which the output C is stable for all inputs R and U. 13.9. In the control system shown in Fig. P13–9, find the value of Kc for which the system is on the verge of instability. The controller is replaced by a PD controller, for which the transfer function is Kc (tDs 1). If Kc 10, determine the range of tD for which the system is stable. R + 1 (s + 1)3 Kc − C FIGURE P13–9 13.10. (a) Write the characteristic equation for the control system shown in Fig. P13–10. (b) Use the Routh test to determine if the system is stable for Kc 4. (c) Determine the ultimate value of Kc above which the system is unstable. U + R + ( + ) Kc 1 + −2s 1 2s + 1 C − 1 s+1 FIGURE P13–10 13.11. For the control system in Fig. P13–11, the characteristic equation is s 4 4 s 3 6s 2 4 s (1 K ) 0 (a) Determine the value of K above which the system is unstable. (b) Determine the value of K for which two of the roots are on the imaginary axis, and determine the values of these imaginary roots and the remaining two roots. R + 1 K (s + 1)3 C − 1 s+1 FIGURE P13–11 cou9789x_ch13_252-268.indd 266 8/14/08 8:04:19 AM Confirming Pages CHAPTER 13 CAPSULE SUMMARY STABLE SYSTEM → A bounded input produces a bounded output (BIBO). Stable Region Left Half-Plane (LHP) 1.5 1 0.5 0 –0.5 –1 0 0.5 1 1.5 2 2.5 s2 Imaginary axis Unstable Region Right Half-Plane (RHP) 2 1.5 1 0 0.5 –1 –1.5 –2 s3 0 0.1 0.20.30.40.50.60.70.80.9 1 (0, b3) (–a2, b2) Characteristic equation: 1 G(s) 0, where 1 G(s) is the denominator of the closed-loop transfer function for the process. Routh stability test: A test to determine the stability of a linear control system. To apply the Routh test, rewrite the characteristic equation as s4 (a4, b4) s1 s6 (–a1, 0) (0, 0) s4* 2.5 1.5 s2* 1 (–a2, –b2) 0.5 0 2 1.5 1 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 (a4, –b4) s3* 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 (0, –b3) 12 10 8 6 4 2 0 –2 –4 –6 –8 0 0.1 0.20.30.40.50.60.70.80.9 1 s5 (a5, 0) 10 9 8 7 6 5 4 3 2 1 Real axis 0 0.1 0.20.30.40.50.60.70.80.9 1 2 1.5 1 0 0.5 –1.5 –2 a0 s n a1s n 1 a2 s n 2 . . . an 0 (13.9) 0 0.1 0.20.30.40.50.60.70.80.9 1 FIGURE 13–4 Stability of typical roots of the characteristic equation. where a0 is positive. Form the following array by using the coefficients: Row 1 a0 a2 a4 a6 2 a1 a3 a5 a7 3 b1 b2 b3 c3 4 c1 c2 5 d1 d2 6 e1 e2 7 f1 n1 g1 The other elements are found from the formulas a1a2 a0 a3 a1 b1a3 a1b2 c1 b1 b1 a1a4 a0 a5 a1 b1a5 a1b3 c2 b1 b2 ... ... 267 cou9789x_ch13_252-268.indd 267 8/14/08 8:04:20 AM Confirming Pages 268 PART 3 LINEAR CLOSED-LOOP SYSTEMS THEOREMS OF THE ROUTH TEST Theorem 13.1. The necessary and sufficient condition for all the roots of the characteristic equation [Eq. (13.9)] to have negative real parts (stable system) is that all elements of the first column of the Routh array (a0, a1, b1, c1, etc.) be positive and nonzero. Theorem 13.2. If some of the elements in the first column are negative, the number of roots with a positive real part (in the right half-plane) is equal to the number of sign changes in the first column. Theorem 13.3. If one pair of roots is on the imaginary axis, equidistant from the origin, and all other roots are in the left half-plane, all the elements of the nth row will vanish and none of the elements of the preceding row will vanish. The location of the pair of imaginary roots can be found by solving the equation Cs 2 D 0 (13.10) where the coefficients C and D are the elements of the array in the (n – 1)st row as read from left to right, respectively. cou9789x_ch13_252-268.indd 268 8/14/08 8:04:20 AM Confirming Pages CHAPTER 14 ROOT LOCUS I n Chap. 13 on stability, Routh’s criterion was introduced to provide an algebraic method for determining the stability of a simple feedback control system (Fig. 13–3) from the characteristic equation of the system [Eq. (13.7)]. This criterion also yields the number of roots of the characteristic equation that are located in the right half of the complex plane. In this chapter, we develop a graphical method for finding the actual values of the roots of the characteristic equation, from which we can obtain the transient response of the system to an arbitrary forcing function. 14.1 CONCEPT OF ROOT LOCUS In Chap. 13, the response of the simple feedback control system, shown again in Fig. 14–1, was given by the expression C ⫽ G1G2 G2 R⫹ U 1⫹G 1⫹G (14.1) where G ⫽ G1 G2 H. The factor in the denominator, 1 ⫹ G, when set equal to zero, is called the characteristic equation of the closed-loop system. The roots of the characteristic equation determine the form (or character) of the response C(t) to any particular forcing function R(t) or U(t). The root locus method is a graphical procedure for finding the roots of 1 ⫹ G ⫽ 0, as one of the parameters of G varies continuously. In our work, the parameter that will be varied is the gain (or sensitivity) Kc of the controller. We can illustrate the concept 269 cou9789x_ch14_269-284.indd 269 8/14/08 7:53:03 AM Confirming Pages 270 PART 3 LINEAR CLOSED-LOOP SYSTEMS R + − + G1 U G2 + C H FIGURE 14–1 Simple feedback control system. of a root locus diagram by considering the example presented in Fig. 13–1, which is represented by the block diagram of Fig. 14–1 with G1 ⫽ K c 1 G2 ⫽ (t 1s ⫹ 1) (t 2 s ⫹ 1) H ⫽ 1 t 3s ⫹ 1 For this case, the open-loop transfer function is G ⫽ Kc (t 1s ⫹ 1) (t 2 s ⫹ 1) (t 3s ⫹ 1) which may be written in the alternate form G (s) ⫽ where K ⫽ (s ⫺ K p1 ) ( s ⫺ p2 ) ( s ⫺ p3 ) (14.2) Kc t 1t 2t 3 p1 ⫽ ⫺ 1 t1 p2 ⫽ ⫺ 1 t2 p3 ⫽ ⫺ 1 t3 The terms p1, p2, and p3 are called the poles of the open-loop transfer function. A pole of G(s) is any value of s for which G(s) approaches infinity. For example, it is clear from Eq. (14.2) that if s ⫽ p1, the denominator of Eq. (14.2) is zero and therefore G(s) approaches infinity. Hence p1 ⫽ ᎐1/t1 is a pole of G(s). The characteristic equation for the closed-loop system is 1⫹ K ⫽ 0 ( s ⫺ p1) ( s ⫺ p2 ) ( s ⫺ p3) This expression may be written (s ⫺ cou9789x_ch14_269-284.indd 270 p1) ( s ⫺ p2 ) ( s ⫺ p3 ) ⫹ K ⫽ 0 (14.3) 8/14/08 7:53:03 AM Confirming Pages CHAPTER 14 ROOT LOCUS 271 Using the same numerical values for the poles that were used at the beginning of Chap. 13 (᎐1, ᎐2, ᎐3) gives ( s ⫹ 1) ( s ⫹ 2 ) ( s ⫹ 3) ⫹ K ⫽ 0 (14.4) K ⫽ 6Kc where Expanding the product of this equation gives s 3 ⫹ 6s 2 ⫹ 11s ⫹ K ⫹ 6 ⫽ 0 (14.5) which is third-order. For any particular value of controller gain Kc, we can obtain the roots of the characteristic equation [Eq. (14.5)]. For example, if Kc ⫽ 4.41 (K ⫽ 26.5), Eq. (14.5) becomes s 3 ⫹ 6s 2 ⫹ 11s ⫹ 32 . 5 ⫽ 0 Solving this equation for the three roots gives r1 ⫽ ⫺5 . 10 r 2 ⫽ ⫺0 . 45 ⫺ j 2 . 5 r 3 ⫽ ⫺0 . 45 ⫹ j 2 . 5 Note: MATLAB can easily be used to find the roots of a polynomial: >> c⫽[1,6,11,32.5]; %placing the coefficients of the equation into a vector >> roots(c); %finding the roots ans = -5.0931 -0.4534 + 2.4851i By selecting other values of K, other sets of roots are obtained, as shown in Table 14.1. If the roots are all real, the response will be nonoscillatory. TABLE 14.1 Roots of the characteristic equation (s ⫹ 1)(s ⫹ 2)(s ⫹ 3) ⫹ K ⫽ 0 K ⫽ 6 Kc 0 0.23 0.39 1.58 6.6 26.5 60.0 100.0 cou9789x_ch14_269-284.indd 271 r1 ⫺3 ⫺3.10 ⫺3.16 ⫺3.45 ⫺4.11 ⫺5.10 ⫺6.00 ⫺6.72 r2 r3 ⫺2 ⫺1.75 ⫺1.42 ⫺1.28 ⫺ j0.75 ⫺0.95 ⫺ j1.5 ⫺0.45 ⫺ j2.5 0.0 ⫺ j3.32 0.35 ⫺ j4 ⫺1 ⫺1.15 ⫺1.42 ⫺1.28 ⫹ j0.75 ⫺0.95 ⫹ j1.5 ⫺0.45 ⫹ j2.5 0.0 ⫹ j3.32 0.35 ⫹ j4 8/14/08 7:53:04 AM Confirming Pages 272 PART 3 LINEAR CLOSED-LOOP SYSTEMS For convenience, we may plot the roots r1, r2, and r3 on the complex plane as K changes continuously. Such a plot is called a root locus diagram and is shown in Fig. 14–2. Notice that there are three loci or branches corresponding to the three roots and that they “emerge” or begin (for K ⫽ 0) at the poles of the open-loop transfer function (⫺1, ⫺2, ⫺3). The direction of increasing K (remember, K is our controller gain, which we can adjust) is indicated on the diagram by an arrow. Also the values of K are marked on each locus. The root locus diagram for this system and others to follow is symmetric with respect to the real axis, and only the portion of the diagram in the upper half-plane need be drawn. This follows from the fact that the characteristic equation for a physical system contains coefficients that are real, and therefore complex roots of such an equation must appear in conjugate pairs. The root locus diagram has the distinct advantage of giving at a glance the character of the response as the gain of the controller is continuously changed. The diagram of Fig. 14–2 reveals two critical values of K; one is at K2 where two of the roots become equal, and the other is at K3 where two of the roots are pure imaginary. It should be clear from the discussion in Chap. 13 that the nature of the response C(t) will depend only on the roots r1, r2, r3. Thus, if the roots are all real, which occurs for K < 0.39 in Fig. 14–2, the response will be nonoscillatory. If two of the roots are complex and have negative real parts (K2 < K < K3), the response will include damped sinusoidal terms, which will produce an oscillatory response. If we adjust the controller gain such that K > K3, two of the roots are complex and have positive real parts, and the response is a growing sinusoid. Some of these types of responses were shown in Fig. 13–2. j K3 = 60 100 26.5 3 2 K3 = 60 1.58 26.5 6.6 100 −8 6.6 −6 −4 0.39 0.23 1.58 0.23 0.23 K2 = 0.39 1 −1 1 2 −2 −3 K3 = 60 FIGURE 14–2 Root locus diagram for (s ⫹ 1)(s ⫹ 2)(s ⫹ 3) ⫹ K⫽ 0. cou9789x_ch14_269-284.indd 272 8/14/08 7:53:05 AM Confirming Pages CHAPTER 14 273 ROOT LOCUS MATLAB can also be used to quickly and easily construct a root locus diagram for the system. We enter the open-loop transfer function in the form of Eq. (14.2) with K ⫽ 1, and MATLAB generates the root locus diagram: >> num ⫽ [1]; >> den ⫽ [1 6 11 6]; >> sys ⫽ tf(num,den) Transfer function: 1 sˆ3 + 6 sˆ2 + 11 s + 6 >> rlocus(sys) Root Locus 5 5 4 4 3 3 2 2 1 1 Imag Axis Imag Axis Root Locus 0 −1 System: sys Gain: 0.385 Pole: -1.41 Damping: 1 Over shoot (%): 0 Frequency (rad/s): 1.41 0 −1 −2 −2 −3 −3 −4 −4 −5 System: sys Gain: 60 Pole: 0.00117 + 3.32i Damping: -0.000354 Over shoot (%): 100 Frequency (rad/s): 3.32 System: sys Gain: 60 Pole: 0.00117-3.32i Damping: -0.000354 Over shoot (%): 100 Frequency (rad/s): 3.32 −5 −8 −7 −6 −5 − 4 −3 Real Axis (a) −2 −1 0 1 −8 −7 −6 −5 − 4 −3 − 2 Real Axis (b) −1 0 1 FIGURE 14–3 Root locus diagram for (s ⫹ 1)(s ⫹ 2)(s ⫹ 3) ⫹ K ⫽ 0 generated using MATLAB. By clicking on one of the branches of the diagram in Fig. 14–3a with the mouse, MATLAB displays information regarding that particular root location, as shown in Fig. 14–3b. cou9789x_ch14_269-284.indd 273 8/14/08 7:53:05 AM Confirming Pages 274 PART 3 LINEAR CLOSED-LOOP SYSTEMS Additionally MATLAB has a graphical tool (SISO tool) that can be used to plot and manipulate root locus diagrams. The command for this case is sisotool(sys) The root locus diagram pops up in an interactive window (see Fig. 14–4). The root locations (the heavy blocks) can be dragged along the branches, and the value of K is displayed as the Current Compensator in the upper left-hand corner of the window. FIGURE 14–4 Root locus diagram generated using MATLAB SISO tool. cou9789x_ch14_269-284.indd 274 8/14/08 7:53:05 AM Confirming Pages CHAPTER 14 ROOT LOCUS 275 As another example of a root locus diagram, let the proportional controller be replaced with a PI controller, for which case G1 in Fig. 14–1 is 1 G1 ⫽ K c 1 ⫹ t Is For this case, the open-loop transfer function is K c (t I s ⫹ 1) t I s (t 1s ⫹ 1) (t 2 s ⫹ 1) (t 3 s ⫹ 1) G (s) ⫽ which may be written in an alternate form G (s) ⫽ where K ⫽ K ( s ⫺ z1 ) s ( s ⫺ p1) ( s ⫺ p2 ) ( s ⫺ p3 ) (14.6) Kc t 1t 2t 3 z1 ⫽ ⫺ p1 ⫽ ⫺ 1 tI 1 , t1 p2 ⫽ ⫺ 1 , t2 p1 ⫽ ⫺ 1 t3 The term z1 is called a zero of the open-loop transfer function. A zero of G(s) is any value of s for which G(s) approaches zero. By comparing Eq. (14.6) with Eq. (14.2), we see that the addition of integral action contributes to the open-loop transfer function one zero at z1 and one additional pole at the origin. The characteristic equation corresponding to Eq. (14.6) is 1⫹ K ( s ⫺ z1) ⫽ 0 s ( s ⫺ p1) ( s ⫺ p2 ) ( s ⫺ p3 ) (14.7) This expression may be written s ( s ⫺ p1 ) ( s ⫺ p2 ) ( s ⫺ p3 ) ⫹ K ( s ⫺ z1 ) ⫽ 0 (14.8) As a specific example of the root locus diagram corresponding to Eq. (14.8), let t 1 ⫽ 1, t 2 ⫽ 12 , t 3 ⫽ 13 , and t I ⫽ 14 . These parameters are the same as those used in Example 13.4. The root locus diagram is shown in Fig. 14–5. cou9789x_ch14_269-284.indd 275 8/14/08 7:53:05 AM Confirming Pages 276 PART 3 LINEAR CLOSED-LOOP SYSTEMS The MATLAB commands for constructing the root locus diagram for this system are >> num1 = [1 4]; >> den1 = [1 6 11 6 0]; >> sys1 = tf(num1,den1) Transfer function: s + 4 sˆ4 + 6 sˆ3 + 11 sˆ2 + 6 s >> rlocus(sys1) Root Locus 8 6 4 Image Axis 2 0 −2 −4 −6 −8 −8 −6 −4 −2 Real Axis 0 2 4 FIGURE 14–6 Reproduction of root locus diagram of Fig. 14–5 using MATLAB. cou9789x_ch14_269-284.indd 276 1/1/70 6:51:00 PM Confirming Pages CHAPTER 14 ROOT LOCUS 277 j 3 K = 3.84 K = 3.84 K = 0.74 K = 114 −6 −5 −4 −3 −2 2 1 −1 K = 3.384 K = 0.28 K = 3.84 1 2 −1 −2 −3 FIGURE 14–5 Root locus diagram for s(s ⫹ 1)(s ⫹ 2)(s ⫹ 3) ⫹ K(s ⫹ 4) ⫽ 0; K ⫽ 6Kc. Notice that for this case there are four loci corresponding to the four roots and that they emerge (at K ⫽ 0) from the open-loop poles (0, ⫺1, ⫺2, ⫺3). One of the loci moves toward the open-loop zero at ⫺4 as K approaches infinity. The MATLAB box shows the commands for reproducing the root locus diagram in Fig. 14–5. The resulting MATLAB graph is shown in Fig. 14–6. The diagram in Fig. 14–5 should be compared with the one in Fig. 14–2 to see the effect of adding integral action to the control system. Notice that the value of K ⫽ 3.84, above which the roots move into the right half-plane, is lower than the corresponding value of K ⫽ 60 for proportional control. The effect of adding integral action has been to destabilize the system in terms of the amount of proportional action that can be used before instability occurs. We can verify the points at which the loci cross the imaginary axis by using the Routh test (Theorem 13.3) of Chap. 13. The characteristic equation for proportional control from Eq. (14.4) is ( s ⫹ 1) ( s ⫹ 2 ) ( s ⫹ 3) ⫹ K ⫽ 0 or s 3 ⫹ 6s 2 ⫹ 11s ⫹ K ⫹ 6 ⫽ 0 from which we can write the Routh array: Row 1 2 3 1 6 b1 11 K⫹6 The theorem states that if one pair of roots is on the imaginary axis and all others are in the left half-plane, then all the elements of the nth row must be zero. From this we obtain for the element b1 cou9789x_ch14_269-284.indd 277 8/14/08 7:53:07 AM Confirming Pages 278 PART 3 LINEAR CLOSED-LOOP SYSTEMS b1 ⫽ (6)(11) ⫺ ( K ⫹ 6) ⫽ 0 6 Solving for K gives K ⫽ 60 as we found previously ( Fig. 14–2). A root on the imaginary axis is expressed as simply ja. Substituting s ⫽ ja and K ⫽ 60 into the polynomial gives ⫺ ja3 ⫺ 6a 2 ⫹ 11aj ⫹ 66 ⫽ 0 (66 ⫺ 6a2 ) ⫹ (11a ⫺ a3 ) j ⫽ 0 Equating the real part or the imaginary part to zero gives a ⫽ ⫾ 11 ⫽ ⫾3 . 32 Therefore the loci intersect the imaginary axis at ⫹3.32j and ⫺3.32j (which we also found previously). Example 14.1 Consider the block diagram for the control system shown in Fig. 14–7. This system may represent a two-tank, liquid-level system having a PID controller and a first-order measuring lag. The open-loop transfer function is G ⫽ Kc 1 ⫹ 2s / 3 ⫹ 1/ 3s ( 20 s ⫹ 1) (10 s ⫹ 1) ( 0 . 5s ⫹ 1) Rearranging this into the standard form gives G ⫽ where + R − ( K ( s ⫺ z1) ( s ⫺ z2 ) s ( s ⫺ p1) ( s ⫺ p2 ) ( s ⫺ p3 ) K ⫽ Kc/150 z1 ⫽ ⫺0.5 z2 ⫽ ⫺1 p1 ⫽ ⫺0.05 p2 ⫽ ⫺0.1 p3 ⫽ ⫺2 2 Kc 1 + − s + 1− 3 3s ) 1 (20s + 1) (10s + 1) C 1 0.5s + 1 FIGURE 14–7 Block diagram for Example 14.1. cou9789x_ch14_269-284.indd 278 8/14/08 7:53:07 AM Confirming Pages CHAPTER 14 ROOT LOCUS 279 In this case, there are four poles at 0, ⫺0.05, ⫺0.1, and ⫺2, and two zeros at ⫺0.5 and ⫺1. These are plotted in Fig. 14–8. Note that the PID controller contributes the pole at the origin and the zeros at ⫺0.5 and ⫺1. j j 0.10 2 −0.10 −0.05 −0.10 −2 1 −1 −0.5 0.5 Asymptote −1 FIGURE 14–8 Root locus diagram for Example 14.1. The two values of gain K which give a pair of roots of the characteristic equation that lie on the imaginary axis (and the corresponding roots) are K ⫽ 0 . 004 or K c ⫽ 0 . 6 s ⫽ ⫾ j 0 . 1 or K c ⫽ 360 s ⫽ ⫾ j1 . 1 K ⫽ 2.4 From these results, we conclude that the system will oscillate with constant amplitude with a frequency w ⫽ 0.1 rad/time when Kc ⫽ 0.6. It will also oscillate at constant amplitude with w ⫽ 1.1 when Kc ⫽ 360. The system is unstable for 0.6 < Kc < 360. The system is stable for Kc < 0.6 and for Kc > 360. The complete root locus diagram is sketched in Fig. 14–8. SUMMARY In this chapter, root locus diagrams have been presented and applied to a control system. MATLAB can be used to generate the complete root locus diagrams quickly and easily. It should be emphasized that the basic advantage of this method is the speed and ease with which the loci can be obtained. Once the roots are available, the response of the system to any forcing function can be obtained by the usual procedures of partial fractions and inversion given in Chap. 3. cou9789x_ch14_269-284.indd 279 8/14/08 7:53:08 AM Confirming Pages 280 PART 3 LINEAR CLOSED-LOOP SYSTEMS PROBLEMS 14.1. Draw the root locus diagram for the system shown in Fig. P14–1 where Gc ⫽Kc (1 ⫹ 0.5s ⫹ 1/s). R + 1 (s + 1) (2s + 1) Gc − C FIGURE P14–1 14.2. Draw the root locus diagram for the system shown in Fig. P13–4 for (a) tI ⫽ 0.4 min and (b) tI ⫽ 0.2 min. (The proportional controller is replaced by a PI controller.) Determine the controller gain that just causes the system to become unstable. The values of parameters of the system are K v ⫽ valve constant 0. 070 cfm/psi K m ⫽ transduce r constant 6. 74/ ( ft of tank level ) R2 ⫽ 0. 55 ft level/cfm t 1 ⫽ time constant of tank 1 ⫽ 2. 0 min t 2 ⫽ time constant of tank 2 ⫽ 0. 5 min The controller gain Kc has the units of pounds per square inch (psi). 14.3. Construct the root locus diagram for the system shown in Fig. P13–2. If the system is unstable at higher values of Kc, find the roots on the imaginary axis and the corresponding value of Kc. 14.4. Construct the root loci for the following equations. (a) 1 ⫹ K ( s ⫹ 1) ( 2s ⫹ 1) ⫽ 0 (b) 1 ⫹ K ⫽ 0 s ( s ⫹ 1) ( 2s ⫹ 1) (c) 1 ⫹ K ( 4 s ⫹ 1) ⫽ 0 s ( s ⫹ 1) ( 2s ⫹ 1) (d) 1 ⫹ K (1 . 5s ⫹ 1) ⫽ 0 s ( s ⫹ 1) ( 2s ⫹ 1) (e) 1 ⫹ K ( 0 . 5s ⫹ 1) ⫽ 0 s ( s ⫹ 1) ( 2s ⫹ 1) On your diagrams you should locate quantitatively all poles and zeros. In addition show the parameter that is being varied along the locus and the direction in which the loci travel as this parameter is increased. cou9789x_ch14_269-284.indd 280 8/14/08 7:53:08 AM Confirming Pages CHAPTER 14 14.5. The control system is shown in Fig. P14–5. There are two cases: case 1: t D ⫽ 1 case 2: t D ⫽ 9 . + R Kc(1 + − 1 (s + 1)(s/2 + 1) D s) ROOT LOCUS 2 3 281 and C 1 s/3 + 1 FIGURE P14–5 (a) Sketch the root locus diagram in each case. (b) If the system can go unstable, find the value of Kc that just causes instability. (c) Using Theorem 13.3 of the Routh test, find the locations (if any) at which the loci cross into the unstable region. 14.6. Draw the root locus diagram for the control system shown in Fig. P14–6. (a) Determine the value of Kc needed to obtain a root of the characteristic equation of the closed-loop response which has an imaginary part 0.75. (b) Using the value of Kc found in part (a), determine all the other roots of the characteristic equation from the root locus diagram. (c) If a unit impulse is introduced into the set point, determine the response of the system C(t). + R 2 0.5s 2 + s + 1 2 0.5s + 1 Kc(1 + 3/s) − C FIGURE P14–6 14.7. Plot the root locus diagram for the system shown in Fig. P14–7. We may consider this system to consist of a process having negligible lag; an underdamped, second-order measuring element; and a PD controller. This system may approximate the control of flow rate, in which case the block labeled Kp would represent a valve having no dynamic lag. The feedback element would represent a flow measuring device, such as a mercury manometer placed across an orifice plate. Mercury manometers are known to have underdamped, second-order dynamics. Plot the diagram for t D ⫽ 13 . R + Kc(1 + D s) Kp = 0.2 C − 1 0.2s 2 + 0.8s + 1 FIGURE P14–7 cou9789x_ch14_269-284.indd 281 8/14/08 7:53:08 AM Confirming Pages 282 PART 3 LINEAR CLOSED-LOOP SYSTEMS 14.8. Draw the root locus diagram for the proportional control of a plant having the transfer function 2/(s ⫹ 1)3. Determine the roots on the imaginary axis and the corresponding value of Kc. 14.9. (a) Show how you would adopt the usual root locus method for variation in controller gain to the problem of obtaining the root locus diagram for variation in tD for the control system shown in Fig. P14–9 for Kc ⫽ 2. (b) Plot the root locus diagram for variation in tD with Kc ⫽ 2. (c) Determine the response of the system C(t) for a unit-step change in R for tD ⫽ 0.5 and Kc ⫽ 2. Sketch the response. What is the ultimate value of C(t)? R + − Kc(1 + D s) 1 (s + 1) (2s + 1) C FIGURE P14–9 Hint: Rearrange the open-loop transfer function to be in the form G (s) ⫽ t Ds s ⫹ 1 . 5s ⫹ 1 . 5 2 Then apply the usual root locus rules with tD taking the place of Kc. cou9789x_ch14_269-284.indd 282 8/14/08 7:53:09 AM Confirming Pages CHAPTER 14 CAPSULE SUMMARY R + + − G1 U G2 + C H FIGURE 14–1 Simple feedback control system. For this case, the open-loop transfer function is G ⫽ G1G2 H ⫽ K (t 1s ⫹ 1)(t 2 s ⫹ 1)(t 3s ⫹ 1) or, in the alternate form, G (s) ⫽ where K⬘ ⫽ (s ⫺ K⬘ p1 ) ( s ⫺ p2 ) ( s ⫺ p3 ) K t 1t 2t 3 1 p1 ⫽ ⫺ t1 1 p2 ⫽ ⫺ t2 1 p3 ⫽ ⫺ t3 Here p1, p2, and p3 are called the poles of the open-loop transfer function. A pole of G(s) is any value of s for which G(s) approaches infinity (i.e., a root of the denominator). The characteristic equation for the closed-loop system is 1⫹ (s K' ⫽ 0 ⫺ p1 ) ( s ⫺ p2 ) ( s ⫺ p3 ) or (s ⫺ p1 ) ( s ⫺ p2 ) ( s ⫺ p3 ) ⫹ K ' ⫽ 0 283 cou9789x_ch14_269-284.indd 283 8/14/08 7:53:09 AM Confirming Pages 284 PART 3 LINEAR CLOSED-LOOP SYSTEMS The root locus diagram is a plot of the location of the roots of this equation as K ' is varied. MATLAB commands for generating the root locus diagram are % set up the open loop transfer function with K' = 1 >> >> >> >> num = [1]; %coefficients of numerator polynomial den = [1 6 11 6]; %coeffs. of denom. polynomial sys = tf(num,den); rlocus(sys) Root Locus 5 4 3 Imag Axis 2 1 0 −1 −2 −3 −4 −5 −8 cou9789x_ch14_269-284.indd 284 −7 −6 −5 − 4 −3 Real Axis (a) −2 −1 0 1 8/14/08 7:53:10 AM Confirming Pages PART IV FREQUENCY RESPONSE cou9789x_ch15_285-322.indd 285 8/14/08 4:38:28 PM cou9789x_ch15_285-322.indd 286 8/14/08 4:38:28 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE C hapters 4 and 7 discussed briefly the response of first- and second-order systems to sinusoidal forcing functions. These frequency responses were derived by using the standard Laplace transform technique. In this chapter, a convenient graphical technique will be established for obtaining the frequency response of linear systems. The motivation for doing so will become apparent in Chap. 16, where it will be found that frequency response is a valuable tool in the analysis and design of control systems. Many of the calculations in this chapter make use of complex numbers. The reader should review the two forms of complex numbers (rectangular and polar) and the basic operations used on complex numbers. 15.1 SUBSTITUTION RULE A Fortunate Circumstance Consider a simple first-order system with transfer function G (s) ⫽ 1 ts ⫹ 1 (15.1) Substituting the quantity jw for s in Eq. (15.1) gives G ( jw ) ⫽ 1 jwt ⫹ 1 287 cou9789x_ch15_285-322.indd 287 8/14/08 4:38:28 PM Confirming Pages 288 PART 4 FREQUENCY RESPONSE We may convert this expression to polar form by multiplying numerator and denominator by the conjugate of jwt ⫹ 1; the result is ⫺ jwt ⫹ 1 1 wt ⫽ ⫺ j 2 2 ( jwt ⫹ 1) (⫺ jwt ⫹ 1) 1 ⫹ w t 1 ⫹ w 2t 2 G ( jw ) ⫽ (15.2) To convert a complex number in rectangular form (a ⫹ jb) to polar form Re jf, where R ⫽ magnitude and φ ⫽ angle, one uses the relationships R ⫽ and f ⫽ tan⫺1 a2 ⫹ b2 b a For visualization of the polar form, see Fig. 15–1. Imaginary axis b R a Real axis FIGURE 15–1 Complex number representations. By fixing R and φ, we can define the complex number. Applying these relationships to Eq. (15.2) gives 1 tan⫺1 (⫺wt ) G ( jw ) ⫽ 2 2 1 w t ⫹ (15.3) f R The quantities on the right side of Eq. (15.3) are familiar. Recall Eqs. (4.27) and (4.28) Y (t ) ⫽ Awt A e⫺t /t ⫹ sin (w t ⫹ f ) 2 2 t w ⫹1 t w2 ⫹ 1 2 (4.27) where f ⫽ tan⫺1 (⫺wt ) As t → ∞, the first term on the right side of Eq. (4.27) vanishes and leaves only the ultimate periodic solution, which is sometimes called the steady-state solution and is shown in Fig. 15–2. Y (t ) s ⫽ cou9789x_ch15_285-322.indd 288 A 2 t w2 ⫹ 1 sin (w t ⫹ f ) (4.28) 8/14/08 4:38:28 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 289 So, after sufficient time elapses, the response of a first-order system to a sinusoidal input of frequency w is also a sinusoid of frequency w. Furthermore, from Eq. (4.28) 2 2 the ratio of the amplitude of the response to that of the input is 1 / w t ⫹ 1 , and the ⫺1 phase difference between output and input is tan (⫺wt). Hence, we have shown here that for the frequency response of a first-order system, Amplitude ratio ( AR ) ⫽ output amplitude ⫽ G ( jw ) ⫽ R input amplitude Phase angle ⫽ f ⫽ G ( jw ) That is, to obtain the amplitude ratio AR and phase angle, one merely substitutes jw for s in the transfer function and then finds the magnitude and argument (or angle) of the resulting complex number, respectively. Output Input Process A sin ( t) Transients + A · AR · sin ( t + ) Input Phase amplitude angle Output amplitude 2 1.5 Input Output 1 AR Signal/A 0.5 Time for transients to decay 0 −0.5 −1 −1.5 −2 0 1 2 3 4 Time 5 Phase lag 6 7 8 FIGURE 15–2 Characteristics of a steady-state sinusoidal response. cou9789x_ch15_285-322.indd 289 8/14/08 4:38:30 PM Confirming Pages 290 PART 4 FREQUENCY RESPONSE KEY FEATURES TO NOTE ABOUT THE FREQUENCY RESPONSE OF THE PROCESS • • • • • After transients die out, the output is a sine wave. Input frequency ⫽ output frequency ⫽ w. In general, the output is attenuated, that is, AR < 1. The output is shifted in time (it lags the input by the phase angle φ). Amplitude ratio and phase angle are both functions of frequency. Example 15.1 Rework Example 4.2. The pertinent transfer function is G (s) ⫽ 1 0 . 1s ⫹ 1 The frequency of the bath-temperature variation is given as 10/p cycles/min which is equivalent to w ⫽ (10 cycles/p min)(2p rad/cycle) ⫽ 20 rad/min. Hence, let s ⫽ w j ⫽ 20 j to obtain G ( 20 j ) ⫽ 1 2j ⫹ 1 In polar form, this is G ( 20 j ) ⫽ 1 1 ⫺ 1. 11 rad ⫽ ⫺ 63 . 5 ⬚ 5 5 which agrees with the previous result. MATLAB Solution of Example 15.1 Using Simulink The Simulink model of the process is shown in Fig. 15–3a. Once this Simulink model is run, we can use MATLAB to plot the process output and input as follows. >> plot(temperature.time,temperature.signals.values) The resulting output of this MATLAB command is shown in Fig. 15–3b. The tabular output can be examined to verify the actual peak locations so that we can check our hand-calculated results from Example 15.1. cou9789x_ch15_285-322.indd 290 8/14/08 4:38:31 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 291 temperature To Workspace 1 0.1s + 1 Sine wave Transfer fcn Scope (a) Simulink model of first-order process for Example 15.1 3 lag time = 0.055 minutes transients decay steady state oscillations Temperature(F) 2 Output Temperature 1 0 −1 −2 −3 Input Temperature 0 0.2 0.4 0.6 Time (min) 0.8 1 1.2 (b) MATLAB generated response for first-order system of Example 15.1. FIGURE 15–3 Using MATLAB and Simulink to solve Example 15.1. >> x=[temperature.time,temperature.signals.values] x = [time] [output] [input] 0 0 0 ... ... ... 1.0200 1.0250 1.0300 1.0350 1.0400 0.3837 0.4624 0.5365 0.6053 0.6680 allow transients to decay...pick up data at t=1 min, (from graph) 1.9996← input peak 1.9937 1.9678 1.9223 1.8576 (continues) cou9789x_ch15_285-322.indd 291 8/14/08 4:38:32 PM Confirming Pages 292 PART 4 1.0450 1.0500 1.0550 1.0600 1.0650 1.0700 1.0750 1.0800 1.0850 1.0900 1.0950 1.1000 FREQUENCY RESPONSE 0.7241 1.7743 0.7729 1.6733 0.8140 1.5556 0.8469 1.4223 0.8714 1.2748 0.8872 1.1146 0.8941 0.9433← output peak 0.8921 0.7625 0.8812 0.5741 0.8614 0.3800 0.8331 0.1820 0.7964 -0.0177 Using the MATLAB output to check our previous results, we get Amplitude ratio ⫽ 0 . 8941 1 ⫽ 0 . 447 compare to ⫽ 0 . 447 (check ) 1 . 9996 5 rad Phase lag ⫽ 20 ( 0.055 min ) ⫽ 1.11 rad ⫽ 63.6 ⬚ (check) min The procedure we have just discussed is general and can be used for other transfer functions with the important restriction that it applies only to systems whose transfer functions yield stable responses. A stable response is necessary so that the transients die out and we will be left with an oscillating, sinusoidal response. We prove the general result in the appendix at the end of this chapter for the reader who is interested. Example 15.2 Find the frequency response of the system with the general second-order transfer function, and compare the results with those of Chap. 7. The transfer function is 1 t s ⫹ 2ζts ⫹ 1 2 2 Putting s ⫽ jw yields 1 1 ⫺ t w ⫹ j 2zwt 2 2 which may be converted to polar form 1 (1 ⫺ w 2t 2 )2 ⫹ ( 2zwt ) 2 ⫺2zwt tan⫺1 1 ⫺ w 2t 2 Hence, cou9789x_ch15_285-322.indd 292 8/14/08 4:38:33 PM Confirming Pages CHAPTER 15 Amplitude ratio AR ⫽ INTRODUCTION TO FREQUENCY RESPONSE 293 1 2 1 ⫺ (wt )2 ⫹ ( 2zwt )2 Phase angle ⫽ f ⫽ ⫺tan⫺1 (15.4) 2 zwt 1 ⫺ (wt )2 which agree with Eq. (7.41). Example 15.3. Consider a second-order transfer function, with t ⫽ 1 and z ⫽ 0.8, being disturbed with a sine wave input of 3 sin (0.5t) (Fig. 15–4). Determine the form of the response after the transients have decayed and steady-state oscillations are established. 3 sin (0.5t) 1 s2 + 1.6s + 1 FIGURE 15–4 Block diagram for Example 15.3. The steady-state oscillations will have the form 3(AR) sin(0.5t ⫹ f), which is an attenuated (smaller-amplitude) sine wave of the same frequency as the input and shifted by a phase angle f. Thus, all we need to determine is AR and f. We can calculate both from Eq. (15.4). AR ⫽ 1 {1 ⫺ [ 0.5(1)] } 2 2 ⫹ [ (2)(0 . 8)(0 . 5)(1) ]2 ⫽ 1 ⫽ 0 . 91 0 . 5 6 25 ⫹ 0 . 64 (2)(0 . 8)(0 . 5)(1) 0.8 ⫽ ⫺tan⫺1 ⫽ ⫺0. 818 8 rad ⫽ ⫺46 . 8 ⬚ f ⫽ ⫺tan⫺1 2 0 . 75 1 ⫺ [ 0 . 5(1) ] The form of the steady-state oscillations is therefore 3 ( AR ) sin (0 . 5t ⫹ f ) ⫽ 2. 73 sin (0 . 5t ⫺ 0. 818 rad) ⫽ 2. 7 3 sin [ 0 . 5 ( t ⫺ 1 . 636 )] ⫽ 2. 73 sin (0 . 5t ⫺ 46 . 8⬚ ) We can easily verify this response by using MATLAB and Simulink. The Simulink model for this problem is shown in Fig. 15–5a. cou9789x_ch15_285-322.indd 293 8/14/08 4:38:34 PM Confirming Pages 294 PART 4 FREQUENCY RESPONSE Y To workspace 1 s2 + 1.6s + 1 Sine wave Transfer function Scope (a) Simulink model of first-order process for Example 15.3. 4 input 3 output 2 1 0 −1 −2 −3 −4 0 5 10 15 20 25 30 Time (min) 35 40 45 50 (b) MATLAB generated response for first-order system of Example 15.3. FIGURE 15–5 Using MATLAB and Simulink to solve Example 15.3. Once this Simulink model is run, we can use MATLAB to plot the process output and input as follows: >> plot(Y.time,Y.signals.values) The resulting output of this MATLAB command is shown in Fig. 15–5b. >> x = [Y.time,Y.signals.values] x = 0 0 0 0.1495 0.0008 0.2241 0.2991 0.0059 0.4469 ... ... ... allow transients to decay 26.6008 -0.2285 2.0096 27.3739 0.8162 2.7011 28.1273 1.7195 2.9919 ← input peak cou9789x_ch15_285-322.indd 294 8/14/08 4:38:35 PM Confirming Pages CHAPTER 15 28.7632 29.3446 29.9062 30.4614 31.0193 31.5891 32.1830 ... 47.4619 48.0523 48.6172 49.1731 49.7295 50.0000 2.2986 2.6274 2.7358 2.6324 2.3254 1.8266 1.1507 ... -2.1802 -2.5667 -2.7289 -2.6774 -2.4201 -2.2260 INTRODUCTION TO FREQUENCY RESPONSE 295 2.9108 2.5806 2.0556 ← output peak 1.3781 0.5911 -0.2594 -1.1226 ... -2.9573 -2.6825 -2.2019 -1.5583 -0.7944 -0.3971 Transportation Lag The response of a transportation lag is not described by a standard nth-order differential equation (that yields standard transfer functions). Rather, a transportation lag is described by the relation Y (t ) ⫽ X (t ⫺ t ) (15.5) which states that the output Y lags the input X by an interval of time t. If X is sinusoidal X ⫽ A sin w t then from Eq. (15.5) Y ⫽ A sin w ( t ⫺ t ) ⫽ A sin (w t ⫺ wt ) It is apparent that the AR is unity and the phase angle is (⫺wt). To check the substitution rule of the previous section, recall that the transfer function for a transportation lag is given by G (s) ⫽ Y (s) ⫽ e⫺ts X (s) Putting s ⫽ jw gives ( G ( jw ) ⫽ e⫺ jwt ⫽ form of: Re jf , R ⫽ 1, f ⫽ ⫺wt ) Then AR ⫽ e⫺ jwt ⫽ 1 (15.6) Phase angle ⫽ e⫺ jwt ⫽ ⫺wt and the validity of the rule is verified. cou9789x_ch15_285-322.indd 295 8/14/08 4:38:35 PM Confirming Pages 296 PART 4 FREQUENCY RESPONSE Example 15.4. Consider a stirred-tank heater with a capacity of 15 gal. Water is entering and leaving the tank at the constant rate of 600 lb/min. The heated water that leaves the tank enters a well-insulated section of 6-in-ID pipe. Two feet from the tank, a thermocouple is placed in this line for recording the tank temperature, as shown in Fig. 15–6. The electrical heat input is held constant at 1000 kW. Ti = 75 + 5 sin 46t 600 lb/min Tm 1000 kW T 600 lb/min 2' 6" ID pipe FIGURE 15–6 Tank temperature system for Example 15.4. If the inlet temperature is varied according to the relation Ti ⫽ 75 ⫹ 5 sin 46t where Ti is in degrees Fahrenheit and t is in minutes, find the eventual behavior of the thermocouple reading Tm. Compare this with the behavior of the tank temperature T. It may be assumed that the thermocouple has a very small time constant and effectively measures the true fluid temperature at all times. The problem is to find the frequency response of Tm to Ti. Deviation variables must be used. Define the deviation variable T⬘i as Ti⬘ ⫽ Ti ⫺ 75 ⫽ 5 sin 46t To define a deviation variable for Tm, note that if Ti were held at 75˚F, Tm would come to the steady state satisfying qs ⫽ wC ( Tms ⫺ Tis ) This may be solved for Tms : Tms W Btu /min (1000 kW ) 1000 0 . 0569 qs W kW ⫽ ⫹ Tis ⫽ ⫹ 75⬚ F ⫽ 170 ⬚ F wC lb Btu 1.0 600 min lb ⭈ ⬚ F Hence, define a deviation variable T⬘m as Tm⬘ ⫽ Tm ⫺ 170 cou9789x_ch15_285-322.indd 296 8/14/08 4:38:38 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 297 Now, the overall system between T⬘i and T⬘m is made up of two components in series: the tank and the 2-ft section of pipe. The transfer function for the tank is G1 ( s ) ⫽ 1 t 1s ⫹ 1 where, as we have seen before, t1 is given by lb 60 . 8 3 (15 gal ) rV ft t1 ⫽ ⫽ ⫽ 0 . 203 min lb gal w 600 7 . 48 3 min ft The transfer function of the 2-ft section of pipe, which corresponds to a transportation lag, is G2 ( s ) ⫽ e⫺t 2 s where t2 is the length of time required for the fluid to traverse the length of pipe. This is t2 ⫽ L ⫽ v 2 ft 600 lb/min ⫽ 0 . 0397 min (60.8 lb/ft3 ) ( 0.196 ft2 ) The factor 0.196 is the cross-sectional area of the pipe in square feet. Since the two systems are in series, the overall transfer function between T⬘i and Tm⬘ is Tm⬘ e⫺t 2 s e⫺0.0397 s ⫽ ⫽ Ti⬘ t 1s ⫹ 1 0 . 203s ⫹ 1 Notice that we must use the same time units throughout, minutes in this case. We are now ready to construct the block diagram for the process, which is shown in Fig. 15–7. T'i = 5 sin 46t 1 0.203s + 1 Tank T' e−0.0397s T'm Pipe FIGURE 15–7 Block diagram for Example 15.4. To find the AR and phase lag, we merely substitute s ⫽ jw ⫽ 46j and find the magnitude and argument of the resulting complex number. However, note that we have previously derived the individual frequency responses for the first-order system and transportation lag. The overall transfer function is the product of the individual transfer functions; hence, its magnitude will be the product of the magnitudes and its argument the sum of the arguments of the individual transfer cou9789x_ch15_285-322.indd 297 8/14/08 4:38:39 PM Confirming Pages 298 PART 4 FREQUENCY RESPONSE functions. In general, if we have several transfer functions in series, the overall transfer function is G ( s ) ⫽ G1 ( s ) G2 ( s ) Gn ( s ) and substituting s ⫽ jw, we obtain G ( jw ) ⫽ G1 ( jw ) G2 ( jw ) Gn ( jw ) or, in polar form, Re jf ⫽ R1e jf1 . R2e jf 2 Rn e jf n ⫽ R R2 Rn 1 j (f1⫹f 2 ⫹⫹f n ) e product of individual su m of individual phase angles magnitudes then G ( jw ) ⫽ G1 ( jw ) G2 ( jw ) Gn ( jw ) G ( jw ) ⫽ G1 ( jw ) ⫹ G2 ( jw ) ⫹ ⫹ Gn ( jw ) This rule makes it very convenient to find the frequency response of a number of systems in series. Using Eq. (15.3) for the tank, AR ⫽ 1 (46 ⫻ 0 . 203)2 ⫹ 1 ⫽ 1 ⫽ 0 . 107 9 . 39 Phase angle ⫽ tan⫺1 [(⫺46 ) ( 0 . 203)] ⫽ ⫺84 ⬚ For the section of pipe, AR is unity, so that the overall AR is just 0.107. The phase lag due to the pipe may be obtained from Eq. (15.6) as Phase angle ⫽ ⫺wt 2 ⫽ ⫺( 46 ) ( 0 . 0397 ) ⫽ ⫺1. 82 rad ⫽ ⫺1 0 4⬚ The overall phase lag from T⬘i to T⬘m is the sum of the individual lags, Tm⬘ Ti⬘ ⫽ ⫺84 ⫺ 104 ⫽ ⫺188 ⬚ Hence Tm ⫽ 170 ⫹ 5 ( 0 . 107 ) sin ( 46t ⫺ 188 ⬚ ) For comparison, a plot of T⬘i, T⬘m, and T⬘ is given in Fig. 15–8, where T⬘ ⫽ tank temperature ⫺ 170 ⬚ F cou9789x_ch15_285-322.indd 298 8/14/08 4:38:40 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 299 Overall lag, 188° Inlet temperature (T'i ) Tank lag, Transport lag, 84° 104° U Tank temp. (2T' ) Zero line Thermocouple temp. (2T'm) B Time FIGURE 15–8 Temperature variation in Example 15.4. These temperatures are shown on a block diagram in Fig. 15–9. It should be emphasized that this plot applies only after sufficient time has elapsed for the complementary solution to become negligible (i.e., the transients have decayed). This restriction applies to all the forthcoming work on frequency response. Also, note that, for convenience of scale (so that we easily can see them on the same axes), the tank and thermocouple temperatures have been plotted as 2T⬘ and 2T⬘m, respectively. T'i = 5 sin 46t 1 T' = 0.535 sin(46t − 84°) 0.203s +1 AR = 0.107 = −84° T'm = 0.535 sin (46t − 188°) e−0.0397s AR = 1.0 = −104° 3.28 rad = 0.535 sin[46(t − 0.07133)] 3.28/46 FIGURE 15–9 Response summary for Example 15.4. A Control Problem An interesting conclusion may be reached from a study of Fig. 15–8. Suppose that we are trying to control the tank temperature, using the deviation between the thermocouple reading and the set point as the error. A block diagram for proportional control might appear as in Fig. 15–10, where T⬘i is replaced by U, T⬘ by C, and T⬘m by B to conform with our standard block diagram nomenclature. The variable R denotes the deviation of the set point from 170˚F and is the desired value of the deviation C. The value of R is assumed to be zero in the following analysis (control at 170˚F). The following arguments, while not rigorous, serve to give some insight regarding the application of frequency response to control system analysis. cou9789x_ch15_285-322.indd 299 8/14/08 4:38:42 PM Confirming Pages 300 PART 4 FREQUENCY RESPONSE U = T'i R + Kc − Q + 1 (wC) + 1 ( 1s + 1) C = T' B = T'm e− 2s FIGURE 15–10 Proportional control of heated, stirred tank. If R ⫽ 0, then Error e ⫽ 0 ⫺ B ⫽ ⫺B Heat input Q ⫽ K ce ⫽ ⫺K c B Thus, the heat being added to the tank is given in deviation variables as ⫺KcB. With reference to Fig. 15–8, which shows the response of the uncontrolled tank to a sinusoidal variation in U, it can be seen that the peaks of U (which is T⬘i) and B (which is T⬘m) are almost exactly opposite because the phase difference is 188˚. This means that if the loop were closed, the control system would have a tendency to add more heat when the inlet temperature Ti was at its high peak, because B is then negative and ⫺KcB becomes positive. (Recall that the set point R is held constant at zero.) Conversely, when the inlet temperature is at a low point, the tendency will be for the control system to add less heat because B is positive. This is precisely opposite to the way the heat input should be controlled. Figure 15–11 clarifies this physical situation. Therefore, the possibility of an unstable control system exists for this particular sinusoidal variation in frequency. Indeed, we shall demonstrate in Chap. 16 that if Kc is taken too large, the tank temperature will oscillate with increasing amplitude for all variations in U, and hence we have an unstable control system. The fact that such information may be obtained by study of the frequency response (i.e., the particular solution for a sinusoidal forcing function) justifies further study of this subject. 15.2 BODE DIAGRAMS Thus far, it has been necessary to calculate AR and phase lag by direct substitution of s ⫽ jw into the transfer function for the particular frequency of interest. It can be seen from Eqs. (15.3), (15.4), and (15.6) that the AR and phase lag are functions of frequency. There is a convenient graphical representation of their dependence on the frequency that largely eliminates direct calculation. This is called a Bode diagram and consists of two graphs: logarithm of AR versus logarithm of frequency, and phase angle versus logarithm of frequency (see Fig. 15–12). The Bode diagram will be shown in Chap. 16 to be a convenient tool for analyzing control problems such as the one discussed in the preceding section. The remainder of this chapter is devoted to developing this tool and presenting Bode diagrams for the basic components of control loops. cou9789x_ch15_285-322.indd 300 8/14/08 4:38:43 PM Confirming Pages CHAPTER 15 Inlet Temp. too High INTRODUCTION TO FREQUENCY RESPONSE 301 Inlet Temp. too Low Inlet temperature (T'i) Zero line Thermocouple temp. (2T'm) Adding More Heat Than Req'd for Set-point Set Point Adding Less Heat Than Req'd for Set-point Heater Response (Q' = −KcT'm) Zero line Time FIGURE 15–11 Illustration of possible unstable response of tank temperature system. log AR log log FIGURE 15–12 Bode diagram format. cou9789x_ch15_285-322.indd 301 8/14/08 4:38:43 PM Confirming Pages 302 PART 4 FREQUENCY RESPONSE First-Order System As we saw earlier [(Eq. (15.3)], the AR and phase angle for the sinusoidal response of a first-order system are AR ⫽ 1 2 (15.7) 2 t w ⫹1 Phase angle ⫽ tan⫺1 (⫺wt ) (15.8) It is convenient to regard these as functions of wt for the purpose of generality. From Eq. (15.7) log AR ⫽ ⫺ 12 log (wt )2 ⫹ 1 (15.9) The first part of the Bode diagram is a plot of Eq. (15.9). The true curve is shown as the solid line on the upper part of Fig. 15–13. Phase angle 1 20 Corner frequency Low-frequency asymptote High-frequency asymptote True curve 0 0.1 −20 0.01 0 −40 Decibels Amplitude ratio 10 −45 −90 0.01 0.1 1 10 100 FIGURE 15–13 Bode diagram for first-order system. Some asymptotic considerations can simplify the construction of this plot. As wt → 0, Eq. (15.7) shows that AR → 1. This is indicated by the low-frequency asymptote on Fig. 15–13. As wt → ∞, Eq. (15.9) becomes asymptotic to log AR ⫽ ⫺log (wt ) which is a line of slope ⫺1, passing through the point wt ⫽ 1 AR ⫽ 1 This line is indicated as the high-frequency asymptote in Fig. 15–13. The frequency wc ⫽ 1/t, where the two asymptotes intersect, is known as the corner frequency; it may cou9789x_ch15_285-322.indd 302 8/14/08 4:38:44 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 303 be shown that the deviation of the true AR curve from the asymptotes is a maximum at the corner frequency. Using wc ⫽ 1/t in Eq. (15.7) gives AR ⫽ 1 2 ⫽ 0 . 707 as the true value, whereas the intersection of the asymptotes occurs at AR ⫽ 1. Since this is the maximum deviation and is an error of less than 30 percent, for engineering purposes it is often sufficient to represent the curve entirely by the asymptotes. Alternately, the asymptotes and the value of 0.707 may be used to sketch the curve if greater accuracy is required. Computer packages (MATLAB, for example, see below) can also be used to generate the diagrams fairly easily. In the lower half of Fig. 15–13, we have shown the phase curve as given by Eq. (15.8). Since f ⫽ tan⫺1 (⫺wt ) ⫽ ⫺tan⫺1 (wt ) it is evident that f approaches 0˚ at low frequencies and ⫺90˚ at high frequencies. This verifies the low- and high-frequency portions of the phase curve. At the corner frequency, wc ⫽ 1/t, fc ⫽ ⫺tan⫺1 (w ct ) ⫽ ⫺tan⫺1 1 ⫽ ⫺45 ⬚ There are asymptotic approximations available for the phase curve, but they are not so accurate or so widely used as those for the AR. Instead, it is convenient to note that the curve is symmetric about ⫺45˚. It should be stated that, in a great deal of the literature on control theory, amplitude ratios (or gains) are reported in decibels. The decibel (dB) is defined by Decibels ⫽ 20 log ( AR ) Thus, an AR of unity corresponds to 0 dB, and an amplitude ratio of 0.1 corresponds to ⫺20 dB. The value of the AR in decibels is given on the right-hand ordinate of Fig. 15–13. Example 15.5. Using MATLAB to Generate a Bode Diagram for a First-Order System To generate a Bode diagram, we must first form the transfer function in MATLAB. >> num= [1]; >> den= [10 1]; >> sys= tf (num,den);%this statement generates a transfer function assigned to sys Transfer function: 1 _______ 10 s + 1 >> bode(sys);%the “bode” command generates the Bode diagram shown in Fig. 15–14 cou9789x_ch15_285-322.indd 303 8/14/08 4:38:45 PM Confirming Pages 304 PART 4 FREQUENCY RESPONSE 0 Asymptote Magnitude (dB) −5 −10 −15 −20 Phase (deg) −25 0 −45 −90 10−2 10−1 100 Frequency (rad/s) FIGURE 15–14 Bode diagram for first-order system in Example 15.5. A drawback of this built-in Bode command is that the magnitude plot is always generated in decibels. An m-file that will generate a Bode plot without converting to decibels is shown below as function m-file named mybode.m. mybode.m file function mybode(sys); [mag,phase,w]=bode(sys); figure subplot(2,1,1) loglog(w,squeeze(mag)) grid ylabel('amplitude') xlabel('frequency(rad/time)') subplot(2,1,2) semilogx(w,squeeze(phase)) grid ylabel('phase(deg)') xlabel('frequency (rad/time)') The result of using mybode on the current system is shown in Fig. 15–15. Note that in both Figs. 15–14 and 15–15 the asymptotes were added after the magnitude graph was generated to illustrate the slopes of 0 and ⫺1 for the lowand high-frequency asymptotes, respectively. cou9789x_ch15_285-322.indd 304 8/14/08 4:38:46 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 305 Amplitude 100 10−1 10−2 10−1 100 Frequency (rad/time) 0 Phase (deg) −20 −40 −60 −80 −100 10−2 10−1 Frequency (rad/time) 100 FIGURE 15–15 Bode diagram for Example 15.5 generated using mybode m-file and MATLAB. First-Order Systems in Series The advantages of the Bode plot become evident when we wish to plot the frequency response of systems in series. As shown in Example 15.4, the rules for multiplication of complex numbers indicate that the AR for two first-order systems in series is the product of the individual ARs: AR ⫽ 1 w 2 t 12 ⫹ 1 w 2t 2 2 ⫹ 1 (15.10) Similarly, the phase angle is the sum of the individual phase angles f ⫽ tan⫺1 (⫺wt 1 ) ⫹ tan⫺1 (⫺wt 2 ) (15.11) Since the AR is plotted on a logarithmic basis, multiplication of the ARs is accomplished by addition of logarithms on the Bode diagram (which, we shall see, is equivalent to adding the slopes of the asymptotes of the individual curves to get the asymptote of the overall curve on log-log coordinates). The phase angles are added directly. The procedure is best illustrated by an example. cou9789x_ch15_285-322.indd 305 8/14/08 4:38:47 PM Confirming Pages 306 PART 4 FREQUENCY RESPONSE Example 15.6. function is Plot the Bode diagram for the system whose overall transfer 1 ( s ⫹ 1) ( s ⫹ 5) To put this in the form of two first-order systems in series, it is rewritten as 1 5 ( s ⫹ 1) ( 15 s ⫹ 1) The time constants are t1 ⫽ 1 and t 2 ⫽ responds to the steady-state gain. 1 . 5 (15.12) The factor 1 5 in the numerator cor- From Eqs. (15.12) and (15.10) AR overall ⫽ 1 5 w ⫹ 1 (w / 5)2 ⫹ 1 2 Hence, log AR overall ⫽ log w 2 1 1 1 ⫺ log w 2 ⫹ 1 ⫺ log ⫹ 1 5 2 2 5 ( ) or log AR overall ⫽ log 1 5 ⫹ log ( AR )1 ⫹ log ( AR )2 (15.13) where (AR)1 and (AR)2 are the ARs of the individual first-order systems, each with unity gain. Equation (15.13) shows that the overall AR is obtained, on logarithmic coordinates, by adding the individual ARs and a constant corresponding to the steady-state gain. The individual ARs must be plotted as functions of log w rather than log (wt) because of the different time constants. This is easily done by shifting the curves of Fig. 15–13 to the right or left so that the corner frequency falls at w ⫽ 1/t. Thus, the individual curves of Fig. 15–16 are placed so that the corner frequencies fall at w c1 ⫽ 1 and w c2 ⫽ 5 . These slopes of these curves are added to obtain the overall curve shown. Note that in this case the logarithms are negative and the addition is downward. To complete the AR curve, the factor log 15 should be added to the overall curve. This would have the effect of shifting the entire curve down by a constant amount. Instead of doing this, the factor 1 is incor5 porated by plotting the overall curve as AR overall / 15 instead of AR overall. This procedure is usually more convenient. Asymptotes have also been indicated on Fig. 15–16. The sum of the slopes of the individual asymptotes gives the slope of the overall asymptote, which is seen to be a good approximation to the overall curve. The overall asymptote has () cou9789x_ch15_285-322.indd 306 8/14/08 4:38:47 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 307 a slope of 0 below w ⫽ 1, ⫺1 for w between 1 and 5, and ⫺2 above w ⫽ 5. Its slope is obtained by simply adding the slopes of the individual asymptotes. 1 1 1 s+1 5 AR for overall curve, 1/5 AR for individual curves 0.5 1/5 s+1 1 s+1 5 0.1 1 s+1 Overall curve 0.05 Overall asymptote, Slope = −2 0.01 0 −45 −90 1/5 s + 1 1s + 1 5 −135 −180 1 s +1 5 1 s+1 0.1 0.5 1 5 10 FIGURE 15–16 Bode diagram for 0.2/[(s ⫹ 1)(0.2s ⫹ 1)]. To obtain the phase angle, the individual phase angles are plotted and added according to Eq. (15.11). The factor 1 has no effect on the phase angle, which 5 approaches ⫺180˚ at high frequency. Graphical Rules for Bode Diagrams Before we proceed to a development of the Bode diagram for other systems, it is desirable to summarize the graphical rules that were utilized in Example 15.5. Consider a number of systems in series. As shown in Examples 15.4 and 15.5, the overall AR is the product of the individual ARs, and the overall phase angle is the sum of the individual phase angles. Therefore, log ( AR overall ) ⫽ log ( AR )1 ⫹ log ( AR )2 ⫹ ⫹ log ( AR )n and cou9789x_ch15_285-322.indd 307 (15.14) foverall ⫽ f1 ⫹ f 2 ⫹ ⫹ fn 8/14/08 4:38:49 PM Confirming Pages 308 PART 4 FREQUENCY RESPONSE where n is the total number of systems. Therefore, the following rules apply to the true curves or to the asymptotes on the Bode diagram: 1. 2. 3. The overall AR is obtained by adding the individual ARs. For this graphical addition, an individual AR that is above unity on the frequency response diagram is taken as positive; an AR that is below unity is taken as negative. To understand this, recall that the logarithm of a number greater than 1 is positive and the logarithm of a number less than 1 is negative. If we are plotting the asymptotes of the individual curves, the asymptote of the overall curve is obtained by adding the slopes of the individual curves. The overall phase angle is obtained by addition of the individual phase angles. The presence of a constant in the overall transfer function shifts the entire AR curve vertically by a constant amount and has no effect on the phase angle. It is usually more convenient to include a constant factor in the definition of the ordinate. These rules will be of considerable value in later examples. Let us now proceed to develop Bode diagrams for other control system components. Second-Order System As shown in Example 15.2, the frequency response of a system with a second-order transfer function 1 G (s) ⫽ 2 2 t s ⫹ 2zts ⫹ 1 is given by Eq. (15.4), repeated here for convenience, AR ⫽ 1 (1 ⫺ w 2t 2 ) 2 Phase angle ⫽ tan⫺1 ⫹ ( 2zwt ) 2 (15.4) ⫺2zwt 1 ⫺ (wt )2 If wt is used as the abscissa for the general Bode diagram, it is clear that z will be a parameter. That is, there is a different curve for each value of z. These curves appear as in Fig. 15–17. The calculation of phase angle as a function of w from Eq. (15.4) requires careful attention. The calculation can be done most clearly with the aid of a plot of tan⫺1 x (or arctan x), as shown in Fig. 15–18. As wt goes from 0 to 1, we see from Eq. (15.4) that the argument of the arctan function goes from 0 to ⫺∞ and the phase angle goes from 0 to ⫺90˚ as shown by the branch from A to B in Fig. 15–18. As wt crosses unity from a value less than unity to a value greater than unity, the sign of the argument of the arctan function in Eq. (15.4) shifts from negative to positive. To preserve continuity in angle as wt crosses unity, the phase angle must go from ⫺90 to ⫺180˚ as wt goes from 1 to ⫹∞ and the branch of the arctan function goes from C to D (in Fig. 15–18). cou9789x_ch15_285-322.indd 308 8/14/08 4:38:50 PM Confirming Pages CHAPTER 15 10 INTRODUCTION TO FREQUENCY RESPONSE 309 = 0.1 5 0.2 0.3 0.5 0.6 0.8 2 1 AR 0.5 1.0 2.0 0.2 0.1 Asymptote 0.05 0.02 0 0 0.5 .3 −45 2.0 = 0.1 0.8 −90 −135 −180 0.1 0.2 0.5 FIGURE 15–17 Block diagram for second-order system 1.0 2.0 5.0 10.0 1 . t 2 s 2 ⫹ 2zts ⫹ 1 The arctan function available in calculators and computers normally covers the principal branches of the arctan function, shown as BAE in Fig. 15–18. For this reason, one must be very careful in calculating the phase angle with Eq. (15.4). If a calculator programmed for the principal branches of the arctan function is used and the argument is positive, one obtains the correct phase angle by subtracting 180˚ from the answer given by the calculator. Notice that for wt ⫽ 1, the phase angle is ⫺90˚, independent of z. This verifies that all phase curves intersect at ⫺90˚, as shown in Fig. 15–17. We may now examine the amplitude curves obtained from Eq. (15.4). For wt << 1, the AR, or gain, approaches unity. For wt → ∞, the only significant term in the denominator is (wt )4 , and the AR becomes asymptotic to the line AR ⫽ cou9789x_ch15_285-322.indd 309 1 (wt )2 8/14/08 4:38:51 PM Confirming Pages 310 PART 4 FREQUENCY RESPONSE tan−1x 270° 180° 90° E −3 B −2 −1 0° A 1 2 3 x −90° C −180° D −270° FIGURE 15–18 Use of plot of tan⫺1 x for computing phase angle of second-order system. This asymptote has slope ⫺2 and intersects the line AR ⫽ 1 at wt ⫽ 1. The asymptotic lines are indicated on Fig. 15–17. For z ⱖ 1, we have shown that the secondorder system is equivalent to two first-order systems in series. The fact that the AR for z ⱖ 1 (as well as for z < 1) attains a slope of ⫺2 and phase of ⫺180˚ is, therefore, consistent. Figure 15–17 also shows that, for z < 0.707, the AR curves attain maxima in the vicinity of wt ⫽ 1. This can be checked by differentiating the expression for the AR with respect to wt and setting the derivative to zero. The result is (wt )max ⫽ 1 ⫺ 2z 2 z ⬍ 0 . 707 (15.15) for the value of wt at which the maximum AR occurs. The value of the maximum AR, obtained by substituting (wt) max into Eq. (15.4), is ( AR )max ⫽ 1 2z 1 ⫺ z 2 z ⬍ 0 . 707 A plot of the maximum AR versus z is given in Fig. 15–19. The frequency at which the maximum AR is attained is called the resonant frequency and is obtained from Eq. (15.15), wr ⫽ cou9789x_ch15_285-322.indd 310 1 1 ⫺ 2z 2 t (15.16) 8/14/08 4:38:51 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 311 0.8 0.6 0.4 0.2 0 0 2 4 6 Maximum AR 8 10 FIGURE 15–19 Maximum AR versus damping for a second-order system. The phenomenon of resonance is frequently observed in our everyday experience. A vase may vibrate when the stereo is playing a particular note. As a car decelerates, perceptible vibrations may occur at particular speeds. A suspension bridge oscillates violently when soldiers march across, stepping at a certain cadence. It may be seen that AR values exceeding unity are attained by systems for which z < 0.707. This is in sharp contrast to the first-order system, for which the AR is always less than unity. The curves of Fig. 15–17 for z < 1 are not simple to construct, particularly in the vicinity of the resonant frequency. Fortunately, almost all second-order control system components for which we will want to construct Bode diagrams have z > 1. That is, they are composed of two first-order systems in series. Actually, the curves of Fig. 15–17 are presented primarily because they are useful in analyzing the closed-loop frequency response of many control systems. Transportation Lag As shown by Eq. (15.6), the frequency response for G(s) ⫽ e⫺ts is f ⫽ ⫺wt rad AR ⫽ 1 or f ⫽ ⫺57 . 2958 wt deg In this expression, w is in radians and 57.2958 is the number of degrees in 1 rad. There is no need to plot the AR since it is constant at 1.0. On logarithmic coordinates, the phase angle appears as in Fig. 15–20, where wt is used as the abscissa to make the figure general. The transportation lag contributes a phase lag, which increases without bound as w increases. Note that it is necessary to convert wt from radians to degrees to prepare Fig. 15–20. Proportional Controller A proportional controller with transfer function Kc has amplitude ratio Kc and phase angle zero at all frequencies. No Bode diagram is necessary for this component. cou9789x_ch15_285-322.indd 311 8/14/08 4:38:52 PM Confirming Pages 312 PART 4 FREQUENCY RESPONSE 0 Phase, deg −45 −90 −135 −180 −225 −270 0.1 0.2 0.5 1 2 , frequency 5 10 FIGURE 15–20 Phase characteristics of transportation lag. Proportional-Integral Controller This component has the ideal transfer function 1 G ( s ) ⫽ Kc 1 ⫹ t Is Accordingly, the frequency response is given by AR ⫽ G ( jw ) ⫽ K c 1 ⫹ 1 1 ⫽ Kc 1 ⫹ t I jw (wt I )2 1 1 ⫽ tan⫺1 ⫺ Phase ⫽ G ( jw ) ⫽ 1 ⫹ wt I t I jw The Bode plot of Fig. 15–21 uses wtI as the abscissa. The constant factor Kc is included in the ordinate for convenience. Asymptotes with a corner frequency of wc ⫽ 1/tI are indicated. The verification of Fig. 15–21 is recommended as an exercise for the reader. Proportional-Derivative Controller The transfer function is G ( s ) ⫽ K c (1 ⫹ t D s ) The reader should show that this has amplitude and phase behavior that are just the inverse of the first-order system 1 ts ⫹ 1 Hence, the Bode plot is as shown in Fig. 15–22. The corner frequency is wc ⫽ 1/tD. cou9789x_ch15_285-322.indd 312 9/5/08 9:55:09 AM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 313 100 Amplitude ratio Kc 10 1 Phase 0.1 0 −45 −90 0.01 0.1 1 10 100 1 10 100 I FIGURE 15–21 Bode diagram for PI controller. 100 Phase, deg Amplitude ratio Kc 10 1 0.1 90 45 0 0.01 0.1 D FIGURE 15–22 Bode diagram for PD controller. cou9789x_ch15_285-322.indd 313 8/14/08 4:38:53 PM Confirming Pages 314 PART 4 FREQUENCY RESPONSE This system is important because it introduces phase lead. Thus, it can be seen that using PD control for the tank temperature control system of Example 15.4 would decrease the phase lag at all frequencies. In particular, 180˚ of phase lag would not occur until a higher frequency. This may exert a stabilizing influence on the control system. In Chap. 16 we look in detail at designing stabilizing controllers using Bode diagram analysis. It is appropriate to conclude this chapter with a summarizing example. Example 15.7 Plot the Bode diagram for the open-loop transfer function of the control system of Fig. 15–23. This system might represent PD control of three tanks in series, with a transportation lag in the measuring element. U + R 1 + Kc (1 + − B Ds) s+1 + 2 s +1 10 C Kc = 10 D = 1/2 s e− 10 FIGURE 15–23 Block diagram of control system for Example 15.7. The open-loop transfer function is G (s) ⫽ 10 ( 0 . 5s ⫹ 1) e⫺s / 10 ( s ⫹ 1)2 ( 0.1s ⫹ 1) The individual components are plotted as dashed lines in Fig. 15–24. Only the asymptotes are used on the AR portion of the graph. Here it is easiest to plot the factor (s ⫹ 1)⫺2 as a line of slope ⫺2 through the corner frequency of 1. For the phase-angle graph, the factor (s ⫹ 1)⫺1 is plotted and added in twice to form the overall curve. The overall curves are obtained by the graphical rules previously presented. For comparison, the overall curves obtained without derivative action (i.e., by not adding in the curves corresponding to 0.5s ⫹ 1) are also shown. Note that on the asymptotic AR diagram, the slopes of the individual curves are added to obtain the slope of the overall curve. cou9789x_ch15_285-322.indd 314 8/14/08 4:38:54 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 315 5 1 s+1 2 1 1 1 (s + 1)2 0.5 ARi or AROverall 1 10 Slope = − 2 Slope = − 1 0.1 10 Overall curve with derivative action 0.05 Overall curve without derivative action 0.01 s+1 0.05 0.1 0.5 1 Slope =−2 5 10 50 (a) 90 45 1 s+1 2 0 1 1 −45 Overall curve without derivative action −135 −225 −270 s+1 s+1 −90 −180 1 10 Overall curve with derivative action 1 e− 10 s 1 Curve corresponding to s + 1 has been added twice to form overall curve 0.05 0.1 0.5 1 5 10 50 (b) FIGURE 15–24 Block diagram for Example 15.7: (a) Amplitude ratio; (b) phase angle. cou9789x_ch15_285-322.indd 315 8/14/08 4:38:55 PM Confirming Pages 316 PART 4 FREQUENCY RESPONSE 15.3 APPENDIX—GENERALIZATION OF SUBSTITUTION RULE The substitution rule (jw for s) is valid for transfer functions that yield stable systems. An nth-order linear system is characterized by an nth-order differential equation an d nY dt n ⫹ an ⫺1 d n ⫺1Y dt n ⫺1 ⫹ ⫹ a1 dY ⫹ a0Y ⫽ X ( t ) dt (15.17) where Y is the output variable and X(t) is the forcing function or input variable. For specific cases of Eq. (15.17), refer to Eq. (4.5) for a first-order system and Eq. (7.5) for a second-order system. If X(t) is sinusoidal X ( t ) ⫽ A sin w t the solution of Eq. (15.17) will consist of a complementary solution and a particular solution of the form Y p ( t ) ⫽ C1 sin w t ⫹ C2 cos w t (15.18) If the system is stable, the roots of the characteristic equation of Eq. (15.17) all lie to the left of the imaginary axis, and the complementary solution will vanish exponentially in time. Then Yp is the quantity previously defined as the sinusoidal or frequency response. If the system is not stable, the complementary solution grows exponentially, and the term frequency response has no physical significance because Yp(t) is inconsequential. The problem now is the evaluation of C1 and C2 in Eq. (15.18). Since we are interested in the amplitude and phase of YP(t), Eq. (15.18) is rewritten as Y p ⫽ D1 sin (w t ⫹ D2 ) (15.19) as was done previously [compare to Eq. (4.25) and related equations]. It will be convenient to change X(t) and Yp(t) from trigonometric to exponential form, using the identity sin q ⫽ Thus, X (t ) ⫽ e jq ⫺ e⫺ jq 2j ( A jw t e ⫺ e⫺ jw t 2j ) (15.20) and from Eq. (15.19) Yp ( t ) ⫽ cou9789x_ch15_285-322.indd 316 ( D1 j (w t ⫹ D2 ) e ⫺ e⫺ j (w t ⫹ D2 ) 2j ) (15.21) 8/14/08 4:38:59 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 317 Substitution of Eqs. (15.20) and (15.21) into Eq. (15.17) yields D1e j (w t ⫹ D2 ) n n ⫺1 ⫹ ⫹ a1 ( jw ) ⫹ a0 a jw ⫹ an ⫺1 ( jw ) n( ) 2j − D1e j (w t ⫹ D2 ) n n ⫺1 ⫹ ⫹ a1 (⫺ jw ) ⫹ a0 a ⫺ jw ) ⫹ an ⫺1 (⫺ jw ) n( 2j ⫽ ( A jw t ⫺ e⫺ jw t e 2j ) (15.22) The coefficients of ejwt on both sides of Eq. (15.22) must be equal. Hence, n n ⫺1 D1e jD2 an ( jw ) ⫹ an ⫺1 ( jw ) ⫹ ⫹ a1 ( jw ) ⫹ a0 ⫽ A (15.23) Equation (15.23) will be satisfied if and only if 1 n ⫺1 an ( jw ) ⫹ an ⫺1 ( jw ) 2 ⫹ ⫹ a1 ( jw ) ⫹ a0 1 n ⫺1 an ( jw ) ⫹ an ⫺1 ( jw ) n ⫹ ⫹ a1 ( jw ) ⫹ a0 ⫽ D1 A (15.24) ⫽ D2 But D1/A and D2 are the AR and phase angle of the response, respectively, as may be seen from Eq. (15.19) and the forcing function. Furthermore, from Eq. (15.17) the transfer function relating X and Y is Y (s) 1 ⫽ n n ⫺1 X (s) ⫹ ⫹ a1s ⫹ a0 an s ⫹ an ⫺1s (15.25) Equations (15.24) and (15.25)* establish the general result. * In writing this equation, it is assumed that X and Y have been written as deviation variables, so that initial conditions are zero. SUMMARY In this chapter the concept of frequency response was discussed. A graphical tool, the Bode diagram, was introduced to assist in the analysis of the frequency response characteristics of various system elements. Bode plots were constructed manually as well as with MATLAB. We will use this tool in Chap. 16 in the design of control systems. cou9789x_ch15_285-322.indd 317 8/14/08 4:39:00 PM Confirming Pages 318 PART 4 FREQUENCY RESPONSE PROBLEMS 15.1. For each of the following transfer functions, sketch the gain versus frequency, asymptotic Bode diagram. For each case, find the actual gain and phase angle at w ⫽ 10. Note: It is not necessary to use log-log paper; simply rule off decades on rectangular paper. (a) 100 (10s ⫹ 1)( s ⫹ 1) 10 s (b) ( s ⫹ 1)( 0.1s ⫹ 1)2 (c) ( 0.1s ⫹ 1)(10s ⫹ 1) (d) s ⫹1 s ⫺1 ( 0.1s ⫹ 1)(10s ⫹ 1) (e) (10s ⫹ 1)2 ( f ) (10 ⫹ s)2 15.2. A temperature bath in which the temperature varies sinusoidally at various frequencies is used to measure the frequency response of a temperature-measuring element B. The apparatus is shown in Fig. P15.2. A standard thermocouple A, for which the time constant is 0.1 min for the arrangement shown in the sketch, is placed near the element to be measured. The response of each temperature-measuring element is recorded simultaneously on a two-channel recorder. The phase lag between the two chart records at different frequencies is shown in the table. From these data, show that it is reasonable to consider element B as a first-order process and calculate the time constant. Describe your method clearly. A FIGURE P15–2 cou9789x_ch15_285-322.indd 318 B Frequency, cycles/min Phase lag of B behind A, deg 0.1 0.2 0.4 0.8 7.1 12.9 21.8 28.2 1.0 1.5 2.0 3.0 4.0 29.8 26.0 23.6 18.0 14.2 8/14/08 4:39:01 PM Confirming Pages CHAPTER 15 INTRODUCTION TO FREQUENCY RESPONSE 319 15.3. Plot the asymptotic Bode diagram for the PID controller 1 G ( s ) ⫽ Kc 1 ⫹ t D s ⫹ t I s where Kc ⫽ 10, tI ⫽ 1, and tD ⫽ 100. Label corner frequencies and give slopes of asymptotes. 15.4. One way of experimentally measuring the frequency response is to plot the output sine wave versus the input sine wave. The results of such a plot look like Fig. P15–4. This is the sinusoidal deviation in output versus sinusoidal deviation in input and appears as an ellipse centered at the origin. Show how to obtain the AR and phase lag from this plot. Output Input FIGURE P15–4 15.5. For the transfer function shown below, sketch carefully the gain versus frequency portion of the asymptotic plot of the Bode diagram. Determine the actual (exact) value of gain and phase angle at w ⫽ 1. Determine the phase angle as w → ∞. G (s) ⫽ 2 ( 0 . 1s ⫹ 1) s 2 (10 s ⫹ 1) Indicate very clearly the slopes of the asymptotic Bode diagram of G(s). 15.6. (a) Plot accurately and neatly the Bode diagram for the process shown in Fig. P15–6, using log-log paper for gain versus frequency and semilog paper for phase versus frequency. Plot the frequency as radians per minute. (b) Find the amplitude ratio and phase angle for Y/X at w ⫽ 1 rad/min and w ⫽ 4 rad/min. X 1 7 2s + 1 5s + 1 e−0.5s Y FIGURE P15–6 cou9789x_ch15_285-322.indd 319 8/14/08 4:39:02 PM Confirming Pages 320 PART 4 FREQUENCY RESPONSE 15.7. For the system shown in Fig. P15–7, determine accurately the phase angle in degrees between Y(t) and X(t) for w ⫽ 0.5. Determine the lag between the input wave and the output wave. X = 2 sin t 3 9s2 + 0.5s + 1 Y FIGURE P15–7 15.8. (a) For the transfer function given below, sketch carefully the asymptotic approximation of gain versus frequency. Show details such as slopes of asymptotes. G (s) ⫽ (1 ⫹ s )2 s (b) Find the actual (exact) value of gain and phase angle for w ⫽ 1 and for w ⫽ 2. 15.9. Derive expressions for amplitude ratio and phase angle as functions of w for the transfer function G(s) ⫽ 1/(s2 ⫺ 1). 15.10. The data given in the following table represent experimental, frequency response data for a process consisting of a first-order process and a transportation lag. Determine the time constant and the transportation lag parameter. Write the transfer function for the process, giving numerical values of the parameters. cou9789x_ch15_285-322.indd 320 Frequency, cycles/min Gain Phase angle, deg 0.01 0.02 0.04 0.06 0.08 1.0 1.0 1.0 1.0 1.0 0.0 ⫺2.0 ⫺6.0 ⫺7.0 ⫺8.5 0.10 0.15 0.20 0.30 0.40 1.0 1.0 1.0 1.0 0.98 ⫺11.0 ⫺17.0 ⫺23.0 ⫺36.0 ⫺48.0 0.60 0.80 1.00 1.50 2.00 0.94 0.88 0.83 0.71 0.61 ⫺73.0 ⫺96.0 ⫺122.0 ⫺180.0 ⫺239.0 4.00 6.00 8.00 10.00 20.00 40.00 0.37 0.26 0.20 0.16 0.080 0.041 — — — — — — 8/14/08 4:39:03 PM Confirming Pages CHAPTER 15 CAPSULE SUMMARY Substitution Rule: Substitute (jw for s) in the transfer function to obtain a complex number. The magnitude of the resulting complex number is the amplitude ratio AR, while the angle of the complex number is the phase angle shift of the output response. This rule is valid for transfer functions that yield stable systems. Input Output Process A sin ( t) + Transients A · AR · sin( t + ) Input Phase amplitude angle Output amplitude 2 1.5 Input Output 1 AR Signal/A 0.5 Time for transients to decay 0 −0.5 −1 −1.5 −2 0 1 2 3 4 Time 5 Phase lag 6 7 8 FIGURE 15–2 Characteristics of a steady-state sinusoidal response. Amplitude ratio AR ⫽ output amplitude ⫽ G ( jw ) ⫽ R input amp litude Phase angle f ⫽ G ( jw ) 321 cou9789x_ch15_285-322.indd 321 8/14/08 4:39:04 PM Confirming Pages 322 PART 4 FREQUENCY RESPONSE Key Features to Note About the Frequency Response of the Process • • • • • After transients die out, the output is a sine wave. Input frequency ⫽ output frequency ⫽ w. In general, the output is attenuated, that is, AR < 1. The output is shifted in time (it lags the input by the phase angle). Amplitude ratio and phase angle are both functions of frequency. Bode diagrams are plots of the amplitude ratio and the phase angle of the steady response as functions of the frequency of the input sine wave. Characteristics of Bode plots for some common transfer functions Transfer function Amplitude ratio AR K 1 ⫹ ts K 1 2 2 1⫹w t Slope on Bode plot v ⬍ vc Slope on Bode plot v ⬎ vc Phase angle (deg) e Corner frequency vc tan⫺1 (⫺wt) f → ⫺90˚ as w → ∞ 1 t 0 ⫺1 First-order system K (1 ⫹ t s) First-order lead K 1 ⫹ w 2t 2 tan⫺1 (wt) f → ⫹90˚ as w → ∞ 1 t 0 1 1 ts 1 wt ⫺90 0 ⫺1 ⫺1 wt ⫹90 0 1 1 1 t 0 ⫺2 0 — 0 Integral ts Derivative K t 2 s 2 ⫹ 2z ts ⫹ 1 K 1 (1 ⫺ w t ) 2 2 2 Second-order system exp(⫺ts) Transportation lag cou9789x_ch15_285-322.indd 322 1 ⫹ ( 2zwt ) 2 ⫺2zwt tan⫺1 1 ⫺ w 2t 2 f → ⫺180 ⬚ as w → 180 ⫺wt p ⬁ 8/14/08 4:39:05 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE T he purpose of this chapter is twofold. First, it is indicated that the stability of a control system can usually be determined from the Bode diagram of its open-loop transfer function. Then methods are presented for rational selection of controller parameters based on this Bode diagram. The material presented here is one of the more useful design aspects of the subject of frequency response. 16.1 TANK TEMPERATURE CONTROL SYSTEM It was indicated in the discussion following Example 15.4 that the control system of Fig. 16–1 might offer stability problems because of excessive phase lag. To review, this system represents proportional control of tank temperature with a delay in the feedback 1 is the process sensitivity 1/(wC), which gives the ultimate change loop. The factor 600 in tank temperature per unit change in heat input Q. 1 1 ⬚F ⫽ ⫽ Btu wC lb Btu 600 600 1 m in min lb⭈⬚ F The proportional gain Kc, in Btu per minute per degree of temperature error, is to be specified by the designer. The open-loop transfer function for this system is G (s) ⫽ ( Kc / 600 ) e⫺0.0396 s 0.203s ⫹ 1 (16.1) 323 cou9789x_ch16_323-350.indd 323 8/22/08 6:08:20 PM Confirming Pages 324 PART 4 FREQUENCY RESPONSE U = T i′ R Q wC Kc /wC + 1 = Kc /600 − 0.203s + 1 B = Tm ′ C = T′ e−0.0396s FIGURE 16–1 Control system for stirred-tank heater of Example 15.4. The Bode diagram for G(s) is plotted in Fig. 16–2. As usual, the constant factor Kc/600 is included in the definition of the ordinate for AR. Asymptote Kc /wC Amplitude ratio 1 0.5 0.2 0.1 0 e−0.0396s −45 1 0.203s + 1 Phase, deg −90 e−0.0396s 0.203s + 1 −135 −180 −225 43 −270 1 2 5 10 20 50 100 FIGURE 16–2 Bode diagram for open-loop transfer function of control system for stirred-tank heater: ( K c / wC ) e⫺t 2s / (t 1s ⫹ 1) . Block diagram is shown in Fig. 16–1. cou9789x_ch16_323-350.indd 324 8/22/08 6:08:21 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 325 At the frequency of 43 rad/min, the phase lag is exactly 180° and AR/(Kc/600) ⫽ 0.12. Kc Kc Kc AR ⫽ ( 0 . 12 ) ⫽ ⫽ 600 600 / 0 . 12 5000 Therefore, if a proportional gain of 5000 Btu/(min · °F) is used, AR ⫽ 5000 ⫽1 5000 This is the AR between the signals e and B. Note that it is dimensionless, as e and B both have the units of temperature. The control system is redrawn for Kc ⫽ 5000 in Fig. 16–3a, with the loop opened. That is, the feedback signal B is disconnected from the comparator. Imagine that a set point disturbance R ⫽ sin 43t is applied to the opened loop. Then, since the open-loop AR ⫽ 1 and the phase lag ⫽ 180°, B ⫽ sin(43t ⫺ 180 ⬚) ⫽ ⫺sin 43t Now imagine that, at some instant in time, R is set to zero and simultaneously the loop is closed. Figure 16–3b indicates that the closed loop continues to oscillate indefinitely. This oscillation is theoretically sustained even though both R and U are zero. U=0 + + = sin 43t R = sin 43t 8.33 − + 1 0.202s + 1 C = sin (43t−83°) Loop open Before closing loop B = −sin 43t e−0.0396s (a) U=0 + R=0 = sin 43t + 8.33 − + 1 0.202s + 1 C = sin (43t−83°) After closing loop B = −sin 43t e−0.0396s (b) FIGURE 16–3 Sustained closed-loop oscillation. cou9789x_ch16_323-350.indd 325 8/22/08 6:08:22 PM Confirming Pages 326 PART 4 FREQUENCY RESPONSE Now suppose Kc is set to a slightly higher value and the same experiment repeated. This time, the signal e is amplified slightly each time it passes around the loop. Thus, if Kc is set to 5001, after the first time around the loop the signal e becomes (5001/5000) sin 43t. After the second time, it is (5001/5000)2 sin 43t, etc. The phase angle is not affected by changing Kc. We thus conclude that, for Kc > 5000, the response is unbounded, since it oscillates with increasing amplitude. By using the definition of stability presented in Chap. 13, it is concluded that the control system is unstable for Kc > 5000 because it exhibits an unbounded response to the bounded input described above. (The bounded input is zero in this case, for U ⫽ R ⫽ 0.) The condition Kc > 5000 corresponds to AR ⬎ 1 for the open-loop transfer function, at the frequency 43 rad/min, where the open-loop phase lag is 180°. This argument is not rigorous. We know the response B only if e remains constant in amplitude because of the definition of frequency response. If, however, the change in Kc is very small, so that e is amplified infinitesimally, then B will closely approximate the frequency response. While this does not prove anything, it shows that we are justified in suspecting instability and that closer investigation is warranted. A rigorous proof of stability requires application of the Nyquist stability criterion [see Coughanowr and Koppel (1965) or Kuo (1987)], which uses the theory of complex variables. For our purposes, it is sufficient to proceed with heuristic arguments. 16.2 THE BODE STABILITY CRITERION It is tempting to generalize the results of the analysis of the tank temperature control system to the following rule. A control system is unstable if the open-loop frequency response exhibits an AR exceeding unity at the frequency for which the phase lag is 180°. This frequency is called the crossover frequency. The rule is called the Bode stability criterion. Actually, since the discussion of Sec. 16.1 was based on heuristic arguments, this rule is not quite general. It applies readily to systems for which the gain and phase curves decrease continuously with frequency. However, if the phase curve appears as in Fig. 16–4, the more general Nyquist criterion must usually be used to determine stability. Other exceptions may occur. Fortunately, most process control systems can be analyzed with the simple Bode criterion, and it therefore finds wide application. Application of the criterion requires nothing more than plotting the open-loop frequency response. This may be based on the theoretical transfer function, if it is available, as we have done for the tank temperature system. If the theoretical system dynamics are not known, the frequency response may be obtained experimentally. To do this, the open-loop system is disturbed with a sine wave input at several frequencies. At each frequency, records of the input and output waves are compared to establish the AR and phase lag. The results are plotted as a Bode diagram. This experimental technique is illustrated in greater detail in Chap. 18. For the remainder of this chapter, we accept the Bode stability criterion as valid and use it to establish the control system design procedure. cou9789x_ch16_323-350.indd 326 8/22/08 6:08:23 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 327 Phase, deg 0 −180 Frequency FIGURE 16–4 Phase behavior of a complex system for which the Bode criterion is not applicable. 16.3 GAIN AND PHASE MARGINS Let us consider the general problem of selecting Gc(s) for the system of Fig. 16–5. Suppose the open-loop frequency response, when a particular controller Gc(s) is tried, is as shown in the Bode diagram of Fig. 16–6. The crossover frequency, at which the phase lag is 180°, is noted as wco on the Bode diagram. At this frequency, the AR is A. If A exceeds unity, we know from the Bode criterion that the system is unstable and that we have made a poor selection of Gc(s). In Fig. 16–6 it is assumed that A is less than unity and therefore the system is stable. U + R + Gc G1 + − G2 C H FIGURE 16–5 Block diagram for a general control system. It is necessary to ascertain to what degree the system is stable. Intuitively, if A is only slightly less than unity, the system is “almost unstable” and may be expected to behave in a highly oscillatory manner even though it is theoretically stable. (Again, heuristic arguments are used. This statement is self-evident to the reader who has studied Chap. 14, where it is shown that the roots of the characteristic equation vary continuously with system parameters. Proof of the statement requires the Nyquist stability criterion.) cou9789x_ch16_323-350.indd 327 8/22/08 6:08:23 PM Confirming Pages PART 4 FREQUENCY RESPONSE Amplitude ratio 328 1 Gain margin A Phase angle, deg 0 Phase margin −180 co FIGURE 16–6 Open-loop Bode diagram for a typical control system. Furthermore, the constant A is determined by the physical parameters of the system, such as time constants. These can be only estimated and may actually change slowly with time because of wear or corrosion. Hence, a design for which A is close to unity does not have an adequate safety factor. To assign some quantitative measure to these considerations, the concept of gain margin (GM) is introduced. Using the nomenclature of Fig. 16–6, Gain margin ⫽ 1 1 ⫽ AR φ ⫽⫺180⬚ A Typical specifications for design are that the gain margin should be greater than 1.7. This means that the AR at crossover could increase by a factor of 1.7 over the design value before the system became unstable. AR φ⫽⫺180⬚ ⫽ 1 1 ⫽ ⫽ 0 . 59 GM 1.7 So, for GM ⫽ 1.7, AR ⫽ 0.59 at the crossover frequency. The design value of the gain margin is really a safety factor that maintains the AR a “safe distance” away from AR ⫽ 1 at wco. As such, its value varies considerably with the application and designer. A gain margin of unity or less indicates an unstable system. Another margin frequently used for design is the phase margin. As indicated in Fig. 16–6, it is the difference between 180° and the phase lag at the frequency for which the gain is unity. Phase margin (PM) ⫽ 180 ⬚ ⫺ f LAG AR⫽1 cou9789x_ch16_323-350.indd 328 8/22/08 6:08:23 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 329 The phase margin therefore represents the additional amount of phase lag required to destabilize the system, just as the gain margin represents the additional gain for destabilization. Typical design specifications are that the phase margin must be greater than 30°. A negative phase margin indicates an unstable system. Example 16.1. Find a relation between relative stability (see below) and the phase margin for the control system of Fig. 16–7. A proportional controller is to be used. This block diagram corresponds to the stirred-tank heater system, for which the block diagram has been given in Fig. 12–17. The particular set of constants is t ⫽ tm ⫽ 1 1 ⫽1 wC These are to be regarded as fixed, while the proportional gain Kc is to be varied to give a satisfactory phase margin. U R + + Kc − 1 s+1 + C 1 s+1 FIGURE 16–7 Block diagram for Example 16.1. The closed-loop transfer function for this system is given by Eq. (12.17), rewritten for our particular case as C Kc s ⫹1 ⫽ 2 2 1 ⫹ K c t 2 s ⫹ 2t 2z 2 s ⫹ 1 R where t2 ⫽ 1 1 ⫹ Kc z2 ⫽ (16.2) 1 1 ⫹ Kc Since the closed-loop system is second-order, it can never be unstable. The shape of the response of the closed-loop system to a unit step in R must resemble the curves of Fig. 7–3. The meaning of relative stability is illustrated by Fig. 7–3. The lower z2 is made, the more oscillatory and hence the “less stable” will be the response. Therefore, a relationship between phase margin and z2 will give the relation between phase margin and relative stability. cou9789x_ch16_323-350.indd 329 8/22/08 6:08:24 PM Confirming Pages FREQUENCY RESPONSE Amplitude ratio Kc PART 4 0.062 Phase angle, 330 −152 1 Slope = −2 =4 FIGURE 16–8 Open-loop Bode diagram for system of Example 16.1, G ⫽ 1/(s ⫹ 1)2. To find this relation the open-loop Bode diagram is prepared and is shown in Fig. 16–8. The simplest way to proceed from this diagram is as follows: Consider a typical frequency w ⫽ 4. If the open-loop gain were 1 at this frequency, then since the phase angle is ⫺152°, the phase margin would be 28°. To make the open-loop gain 1 at w ⫽ 4, it is required that Kc ⫽ 1 ⫽ 16 . 1 0 . 062 Then z2 ⫽ 1 ⫽ 0 . 24 1 ⫹ Kc Hence, a point on the curve of z2 versus phase margin is z 2 ⫽ 0 . 24 phase margin ⫽ 28 ⬚ Other points are calculated similarly at different frequencies, and the resulting curve is shown in Fig. 16–9. From this figure it is seen that z2 decreases with decreasing phase margin and that if the phase margin is less than 30°, then z2 is less than 0.26. From Fig. 7–3, it can be seen that the response of this system for z2 < 0.26 is highly oscillatory, hence relatively unstable, compared with a response for the system with phase margin 50° and z2 ⫽ 0.4. For the particular system of Example 16.1, it was shown that the response became more oscillatory as the phase margin was decreased. This result generalizes to more complex systems. Thus, the phase margin is a useful design tool for application to systems of higher complexity, where the transient response cannot be easily determined and a plot such as Fig. 16–9 cannot be made. To repeat, the rule of thumb is that the phase margin must be greater than 30°. cou9789x_ch16_323-350.indd 330 8/22/08 6:08:25 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 331 0.6 2 0.4 0.2 0 0 30 60 Phase margin 90 FIGURE 16–9 Damping versus phase margin for system of Fig. 16–7. A similar statement can be made about the gain margin. As the gain margin is increased, the system response generally becomes less oscillatory, hence more stable. A control system designer will often try to make both the gain and phase margins equal to or greater than specified minimum values, typically 1.7 and 30°. Note that, for the case of Example 16.1, the gain margin is always infinite because the phase lag never quite reaches 180°. However, the phase margin requirement of 30° necessitates that z2 > 0.26, hence Kc < 14, which means that 1 an offset of 15 [see Eq. (16.2)] must be accepted. This illustrates the importance of considering both margins. The reader should refer to Fig. 16–6 to see that both margins exist simultaneously. Example 16.2. Specify the proportional gain Kc for the control system of Fig. 15–23, reproduced as Fig. 16–10. U + R+ Kc (1 + Ds) 1 + s+1 2 − B C s +1 10 Kc = 10 D = 1/2 s e− 10 FIGURE 16.10 Block diagram of control system for Example 15.6. The gain is to be specified for the two cases: Proportional gain tD Case 1 (PD control) Kc 0.5 min G (s) ⫽ Case 2 (P control only) Kc 0 min G (s) ⫽ cou9789x_ch16_323-350.indd 331 Open-loop transfer function K c ( 0 . 5s ⫹ 1) e⫺s / 10 ( s ⫹ 1)2 ( 0 . 1s ⫹ 1) K c e⫺s / 10 ( s ⫹ 1)2 ( 0 . 1s ⫹ 1) 8/22/08 6:08:25 PM Confirming Pages 332 PART 4 FREQUENCY RESPONSE We’ll use MATLAB to generate the exact Bode diagrams (we sketched the asymptotes in Chap. 15) for the two cases. We first need to generate the transfer function in MATLAB. For case 1, the MATLAB commands to produce the Bode diagram are as follows: num=[0.5 1] ; % the coefficients of the polynomial in the numerator of G(s); den=conv([1 2 1],[0.1 1]) %conv multiplies the 2 polynomials in the denominator of G(s); den = 0.1000 1.2000 2.1000 1.0000 % thus the resulting polynomial in the denominator is 0 . 1s 3 + 1 . 2s 2 + 2 . 1s + 1 sys=tf(num,den,'iodelay',0.1) Transfer function: 0.5 s + 1 exp(-0.1*s) * 0.1 s^3 + 1.2 s^2 + 2.1 s + 1 % Now we can generate the Bode diagram(Fig.16–11) using mybode (discussed in Chap. 15). mybode(sys) Amplitude 100 10−2 10−4 −1 10 100 101 102 101 102 Frequency (rad/time) Phase (deg) 0 −200 −400 −600 −800 10−1 100 Frequency (rad/time) FIGURE 16–11 Bode diagram for Example 16.2, case 1 PD control. cou9789x_ch16_323-350.indd 332 8/22/08 6:08:27 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 333 We can examine the data used to generate Fig. 16–11 as follows. Solution Case 1. Consider first the gain margin. The crossover frequency for the curve with derivative action is 8.62 rad/min. At this frequency, the open-loop gain is 0.0445 if the value of 1 Kc is unity. (Including the factor of 10 in the ordinate is actually equivalent to plotting the case Kc ⫽ 1.) Therefore, according to the Bode criterion, the value of Kc necessary to destabilize the loop is 1/0.0445, or 22.5. To achieve a gain margin of 1.7, Kc must be taken as 22.5/1.7, or 13.2. To achieve proper phase margin, note that the frequency for which the phase lag is 150° (phase margin is 30°) is 5.52 rad/min. At this frequency, a value for Kc of 1/0.0815, or 12.3, will cause the open-loop gain to be unity. Since this is lower than 13.2, we use 12.3 as the design value of Kc. The resulting gain margin is then 1.83. The numerical data for the Bode diagram can be output to Excel and also calculated in Excel by using the formulas for magnitude and phase angle. Goal Seek can also be used in Excel to determine the frequency and magnitude at the phase angles of interest (⫺180° and ⫺150°). With Goal Seek, you have Excel vary the value of the frequency until the phase angle equals ⫺180° or ⫺150°. The Excel formulas to calculate the magnitude and phase angle are: Magnitude =(SQRT(frequency^2/4+1))/(frequency^2+1)/ (SQRT((0.1*frequency)^2+1)) Phase Angle =(2*ATAN(-frequency)+ATAN(-0.1*frequency) +ATAN(0.5*frequency)+(-frequency/10))*180/3.14159 (The data exported from the MATLAB bode function are shown in the Fig. 16–12 for comparison purposes.) frequency magnitude magnitude phase angle phase angle Using Goal Seek ... matlab Excel matlab Excel frequency magnitude phase angle 0.1 0.9913 0.9913 ⫺9.7047 ⫺9.7047 8.6223706 0.0445 ⫺180.0000 0.1184 0.9878 0.9878 ⫺11.4688 ⫺11.4736 5.5290187 0.0815 ⫺150.0000 ... ... ... ... ... 4.8187 0.097 0.0970 ⫺142.4299 ⫺142.4304 5.7029 0.0783 0.0783 ⫺151.8055 ⫺151.8054 6.7494 0.0627 0.0627 ⫺162.339 ⫺162.3387 7.988 0.0496 0.0496 ⫺174.1712 ⫺174.1712 9.4539 0.0388 0.0388 ⫺187.4274 ⫺187.4278 11.1887 0.03 0.0300 ⫺202.2378 ⫺202.2378 ... ... ... ... ... 60.3237 0.0014 0.0014 ⫺516.2162 ⫺516.2168 71.3935 0.001 0.0010 ⫺581.0807 ⫺581.0813 84.4947 0.0007 0.0007 ⫺657.369 ⫺657.3697 100 0.0005 0.0005 ⫺747.2471 ⫺747.2477 FIGURE 16–12 Use of Excel to calculate the magnitude and frequency for Bode diagrams. cou9789x_ch16_323-350.indd 333 8/22/08 6:08:29 PM Confirming Pages 334 PART 4 FREQUENCY RESPONSE Case 2. Proceeding exactly as in case 1 but using the curve for no derivative action, we found that Kc ⫽ 6.7 is needed for satisfactory gain margin and Kc ⫽ 5.14 for satisfactory phase margin. Hence Kc is taken as 5.14, and the resulting gain margin is 2.2. The MATLAB code for generating this Bode diagram (Fig. 16–13) is num=[1] num = 1 den=[0.1 1.2 2.1 1.0] den = 0.1000 1.2000 2.1000 1.0000 sys=tf(num,den,'iodelay',0.1) Transfer function: 1 exp(-0.1*s) * 0.1 s^3 + 1.2 s^2 + 2.1 s + 1 mybode(sys) Amplitude 100 10−2 10−4 10−6 −1 10 100 101 102 101 102 Frequency (rad/time) 0 Phase (deg) −200 −400 −600 −800 −1000 10−1 100 Frequency (rad/time) FIGURE 16–13 Bode diagram for Example 16.2, case 2, P control only. To see the advantage of adding derivative control in this case, note from Fig. 16–10 that the final value of C for a unit step change in U is 1/(1 ⫹ Kc) for any value of tD. The addition of the derivative action allows an increase in the value of Kc from 5.14 to 12.3 while maintaining approximately the same relative stability in terms of gain and phase margins. This reduces the offset from 16 percent of the change in U to 7.5 percent of the change in U. cou9789x_ch16_323-350.indd 334 8/22/08 6:08:29 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 335 The reader is cautioned that the values of Kc selected in this way should be regarded as initial approximations to the actual values, which give “optimal” control of the system of Fig. 16–10. More will be said about this matter later in this chapter in conjunction with the two-tank chemical-reactor control system of Chap. 10. Thus far, nothing has been said about upper limits on the gain and phase margins. Referring to Example 16.1 and Fig. 7–3, we see that if z2 is too large, the response is sluggish. In fact, Fig. 7–3 suggests that for the system of Fig. 16–7 one should choose a value of z2 low enough to give a short rise time without causing excessive response time and overshoot. In other words, one wants the most rapid response that has sufficient relative stability. The results of Example 16.1 generalize to many systems of higher complexity, in terms of margin. Hence, the designer frequently chooses the controller so that either the gain or phase margin is equal to its lowest acceptable value and the other margin is (probably) above its lowest acceptable value. This was the procedure followed in Example 16.2. In almost every situation, the designer faces this conflict between speed of response and degree of oscillation. In addition, if integral action is not used, the amount of the offset must be considered. The concepts of gain and phase margin are useful in selecting Kc for proportional action. However, for additional modes of control such as PD, these concepts are difficult to apply in practice. Consider the selection of Kc and tD in Example 16.2. For a different value of tD the derivative contribution is shifted to the right or left on the Bode diagram of Fig. 15–24. This means that a different value of Kc will provide the proper margins. A typical design procedure is to select the value of tD for which the value of Kc resulting in a 30° phase margin is maximized. The motivation for this choice is that the offset will be minimized. However, the procedure is clearly trial and error. In the case of three-mode control, there are two parameters, tI and tD, which must be varied by trial to meet various design criteria. Fortunately, for this case and others there are simple rules for directly establishing values of the control parameters that usually give satisfactory gain and phase margins. These are the Ziegler-Nichols rules, which we develop next. 16.4 ZIEGLER-NICHOLS CONTROLLER SETTINGS Consider selection of a controller Gc for the general control system of Fig. 16–5. We first plot the Bode diagram for the final control element, the process, and the measuring element in series G1G2H(jw). It should be emphasized that the controller is omitted from this plot. Suppose the diagram appears as in Fig. 16–6. As noted on the figure, the crossover frequency for these three components in series is wco. At the crossover frequency, the overall amplitude ratio is A, as indicated. According to the Bode criterion, then, the gain of a proportional controller which would cause the system of Fig. 16–5 to be on the verge of instability is 1/A. We define this quantity to be the ultimate gain Ku. Thus Ku ⫽ cou9789x_ch16_323-350.indd 335 1 A (16.3) 8/22/08 6:08:29 PM Confirming Pages 336 PART 4 FREQUENCY RESPONSE TABLE 16.1 Ziegler-Nichols controller settings Type of control Proportional Proportional-integral (PI) Proportional-integral-derivative (PID) sI Gc(S) Kc Kc 0.5Ku 1 Kc 1 ⫹ t I s 0.45Ku Pu 1.2 1 Kc 1 ⫹ ⫹ t D s tIs 0.6Ku Pu 2 sD Pu 8 The ultimate period Pu is defined as the period of the sustained cycling that would occur if a proportional controller with gain Ku were used. From the discussion of Fig. 16–3, we know this to be Pu ⫽ 2p radians / cycle sec [⫽] [⫽] w co radians / sec cy cle (16.3a) The factor of 2p appears, so Pu will be in units of time per cycle rather than time per radian. It should be emphasized that Ku and Pu are easily determined from the Bode diagram of Fig. 16–6. The Ziegler-Nichols settings for controllers are determined directly from Ku and Pu according to the rules summarized in Table 16.1. Unfortunately, specifications of Kc and tD for PD control cannot be made using only Ku and Pu . In general, the values 0.6 Ku and Pu /8, which correspond to the limiting case of no integral action in a threemode controller, are too conservative. That is, the resulting system will be too stable. There exist methods for this case which are in principle no more difficult to use than the Ziegler-Nichols rules. One of these is selection of tD for maximum Kc at 30° phase margin, which was discussed above. Another method, which utilizes the step response and avoids trial and error, is presented in Chap. 18. The reasoning behind the Ziegler-Nichols selection of values of Kc is relatively clear. In the case of proportional control only, a gain margin of 2 is established. The addition of integral action introduces more phase lag at all frequencies (see Fig. 16–20); hence a lower value of Kc is required to maintain roughly the same gain margin. Adding derivative action introduces phase lead. Hence, more gain may be tolerated. This was demonstrated in Example 16.2. However, by and large the Ziegler-Nichols settings are based on experience with typical processes and should be regarded as first estimates. Example 16.3. Using the Ziegler-Nichols rules, determine Kc and tI for the control system shown in Fig. 16–14. For this problem, the computation will be done without plotting a Bode diagram; however, the reader may wish to do the problem with such a diagram. We first obtain the crossover frequency by applying the Bode stability criterion. 180 ⫺180 ⬚ ⫽ ⫺tan⫺1w ⫺ (1 . 02 ) (w ) p cou9789x_ch16_323-350.indd 336 8/22/08 6:08:30 PM Confirming Pages CHAPTER 16 R + − CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE Kc 1 + 1 is 1 (s + 1) e−1.02s 337 C FIGURE 16–14 Block diagram for Example 16.3. The value 180/π ⫽ 57.3 converts radians to degrees. Solving this equation by trial and error gives for the crossover frequency wco ⫽ 2 rad/min. The amplitude ratio AR at the crossover frequency for the open loop can be written as AR ⫽ K c 1 1⫹w (1) ⫽ 2 Kc 2 . 24 where we have used Eq. (15.7) for the first-order system and the fact that the amplitude ratio for a transport lag is 1. According to the Bode criterion, the AR is 1.0 at the crossover frequency when the system is on the verge of instability. Inserting AR ⫽ 1 into the above equation and solving for Kc gives Kcu ⫽ 2.24. From the Ziegler-Nichols rules of Table 16.1, we obtain K c ⫽ 0 . 45K cu ⫽ ( 0 . 45) ( 2 . 24 ) ⫽ 1 . 01 and tI ⫽ Pu 2p / w co ⫽ 1.2 1.2 ⫽ 2p / 2 ⫽ 2 . 62 min 1.2 Example 16.4. Using the Ziegler-Nichols rules, determine controller settings for various modes of control of the two-tank chemical-reactor system of Chap. 10. The block diagram is reproduced in Fig. 16–15. U R + − + Gc + 1 (s + 1)(2s + 1) C e−s/2 FIGURE 16–15 Block diagram for two-tank chemical-reactor system. cou9789x_ch16_323-350.indd 337 8/22/08 6:08:31 PM Confirming Pages 338 PART 4 FREQUENCY RESPONSE For convenience, the process gain K and the controller gain Kc are combined into an overall gain K1. The equivalent controller transfer function is regarded as 1 Gc ⫽ K1 1 ⫹ ⫹ t D s tIs where K1 (as well as tI and tD) is to be selected by the Ziegler-Nichols rules. The required value of Kc is then easily determined as Kc ⫽ K1 K where K ⫽ 0.09 for the present case (see Chap. 10.) The Bode diagram for the transfer function without the controller e⫺0.5s ( s ⫹ 1) ( 2s ⫹ 1) is prepared by the usual procedures and is shown in Fig. 16–16. From this figure, it is found that w co ⫽ 1 . 56 rad / min 1 K1u ⫽ ⫽ 6.9 0 . 145 Pu ⫽ 2p ⫽ 4 . 0 min / cycle 1 . 56 (16.4) Hence, the Ziegler-Nichols control constants determined from Table 16.1 and Eq. (16.4) are given in Table 16.2. A plot comparing the open-loop frequency responses including the controller for the three cases, using the controller constants of Table 16.2, is given in Fig. 16–17. This figure shows quite clearly the effect of the phase lead due to the derivative action. The resulting gain and phase margins are listed in Table 16.3. From this table it may be seen that the margins are adequate and generally conservative. TABLE 16.2 Control constants for Example 16.4 cou9789x_ch16_323-350.indd 338 sI Control K1 P 3.5 PI 3.1 3.3 PID 4.2 2.0 sD 0.50 8/22/08 6:08:32 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 339 1 0.5 Amplitude ratio Asymptote 0.145 0.1 0.05 0.01 0 e−0.5s Phase angle −45 1 2s + 1 −90 1 s+1 −135 −180 −225 0.1 0.5 1 1.56 5 10 FIGURE 16–16 Bode diagram for e⫺0.5s/(s ⫹ 1)(2s ⫹ 1). Note that to obtain the Bode diagram for systems including the PID controller, the controller transfer function is rewritten as 1 t Dt I s 2 ⫹ t I s ⫹ 1 Kc 1 ⫹ ⫹ t D s ⫽ K c tIs tIs (16.5) TABLE 16.3 Gain and Phase margins for Example 16.4 Control P PI PID cou9789x_ch16_323-350.indd 339 Gain margin Phase margin 2.0 1.9 2.6 45° 33° 34° 8/22/08 6:08:33 PM Confirming Pages 340 PART 4 FREQUENCY RESPONSE PI, PID 1 P Amplitude ratio K1 0.5 0.1 PID P, PI 0.05 0.01 0 Phase angle −45 P PI −90 PID −135 −180 −225 0.1 0.5 1 5 10 FIGURE 16–17 Open-loop Bode diagrams for various controllers with system of Fig. 16–15. This is second-order in the numerator and has integral action in the denominator. In general, the numerator factors into first-order factors; hence it contributes two curves similar to that of Fig. 15–22 to the overall diagram. For the ZieglerNichols settings it is seen from Table 16.1 that tI ⫽ 4tD. Making this substitution into Eq. (16.5) gives Gc ⫽ K c 4t D2 s 2 ⫹ 4t D s ⫹ 1 K c ( 2t D s ⫹ 1)2 ⫽ 4t D s 4t D s (16.6) and shows that the numerator is equivalent to two PD components in series. This AR is represented by a high-frequency asymptote of slope ⫹2 passing through cou9789x_ch16_323-350.indd 340 8/22/08 6:08:33 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 341 the frequency w ⫽ 1/2tD and a low-frequency asymptote on the line AR ⫽ 1. It should be emphasized that these special considerations apply only to the ZieglerNichols settings. In the general case, the two time constants obtained by factoring the numerator of Eq. (16.5) will be different. The Bode plot of the denominator follows from 1 1 ⫽ ⫺ 90 ⬚ t I jw wt I The gain is a straight line of slope ⫺1 passing through the point (AR ⫽ 1, w ⫽ 1/tI). The phase lag is 90° at all frequencies. Plotting of the overall Bode diagram for the PID case to check the results of Fig. 16–17 is recommended as an exercise for the reader. Transient Responses For instructive purposes, the two-tank reactor system of Fig. 16–15 was simulated using MATLAB. Responses of C(t) to a unit-step change in R(t) are shown in Fig. 16–18. These responses were obtained using the Ziegler-Nichols controller settings determined in Example 16.4. The responses to a step load change were also obtained using MATLAB. These are the curves of Fig. 9–9 that were discussed in Chap. 9 to illustrate the function of the various modes of control. A load change for this system corresponds to a change in the inlet concentration of reactant to tank 1 (refer to Fig. 10–1). As process control engineers, we would be more interested in controlling against this kind of disturbance than against a set point change because the set point or desired product concentration is likely to remain relatively fixed. In other words, this is a regulator problem and the curves of Fig. 9–9 are those we would use to determine the quality of control. PID PI C (t) 1 P 0 0 2 4 6 8 10 Time 12 14 16 FIGURE 16–18 Closed-loop response to step change in set point for control system of Fig. 16–15 using various control modes. However, the step change in set point is frequently used to test control systems despite the fact that the system will be primarily subject to load changes during actual operation. The reason for this is the existence of well-established terminology used cou9789x_ch16_323-350.indd 341 8/22/08 6:08:34 PM Confirming Pages 342 PART 4 FREQUENCY RESPONSE TABLE 16.4 Parameters for response of control system of Fig. 16–15 with Z-N settings Response time, min Period of oscillation, min Offset Overshoot Decay ratio Rise time, min P 0.49 0.26 1.3 10.4 5.0 0.21 PI 0.46 0.29 1.5 11.8 5.5 0 PID 0.42 0.05 0.9 4.9 5.0 0 Control to describe the step response of the underdamped second-order system. This terminology, which was presented in Chap. 7, is used to assign quantitative measure to responses that are not truly second-order, such as those of Fig. 16–18. Of course, the terminology can be applied only to responses that resemble damped sinusoids. Values of the various parameters determined for the responses of Fig 16–18 are summarized in Table 16.4. Offset, realized only with proportional control, is included for completeness. It can be seen from Fig. 16–18 and Table 16.4 that addition of integral action eliminates offset at the expense of a more oscillatory response. When derivative action is also included, the response is much faster (lower rise time) and much less oscillatory (lower response time). The large overshoots realized in all three cases are characteristic of systems with relatively large time delays. In this case the controller is receiving information about the concentration in the second reactor that was true 12 min ago. This is to be compared with the reactor time constants of 1 and 2 min. Hence, it is not surprising that the system overshoots before the controller can take sufficient action. Figure 16–19 is presented for two purposes: (1) to illustrate that the ZieglerNichols controller settings should be regarded as first guesses rather than fixed values and (2) to show the effects of changing the various controller settings. These figures are transient responses to step changes in set point for the three-mode PID control. They show the effects of individually varying the three control parameters Kc, tI, and tD. As an example of the use of these figures, suppose that it is decided that the maximum overshoot that can be tolerated is 25 percent. Figure 16–19a shows that overshoot may be reduced by decreasing Kc at the expense of a considerably more sluggish response. From Fig. 16–19b, we see that overshoot may be reduced by increasing tI (decreasing integral action) at a lesser expense in speed of response. Thus, for tI ⫽ 5 min, the overshoot is reduced to 20 percent without a serious sacrifice in speed. The overshoot cannot be significantly reduced by changing tD, as can be seen from Fig. 16–19c. However, the speed of response may be significantly increased by increasing the derivative action (sometimes at the expense of greater oscillation before the response has settled, as indicated by a higher decay ratio and a lower period). From this brief study of these figures, it may be concluded that, to decrease overshoot without seriously slowing the response, a combination of changes should be made. A possible cou9789x_ch16_323-350.indd 342 8/22/08 6:08:34 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 343 Kc = 6.3 C (t) 4.2 2.1 Z-N I= D= 2.0 0.5 (a) I= 1.0 Z-N C (t) 2.0 5.0 Kc = 4.2 D = 0.5 (b) D= 0.5 0.2 C (t) 1.0 Z-N Kc = 4.2 I = 2.0 Time (c) FIGURE 16–19 Effect of varying controller settings on system response. (Z-N indicates response using Ziegler-Nichols settings.) combination, which should be tried, is to reduce Kc slightly and to increase tI and tD moderately. These changes would probably be tried on the actual reactor system when it is put into operation. Such adjustments from the preliminary settings are usually made by experienced control engineers, using trial procedures that are more art than science. For this reason, we leave the problem of adjustment at this point. SUMMARY In this chapter we used frequency response tools, developed in Chap. 15, to design control systems. We introduced the concepts of Bode stability criterion as well as gain margin and phase margin for determining appropriate controller settings to obtain the desired system response, while maintaining system stability. We also studied the use of Ziegler-Nichols controller settings as initial estimates for controller tuning. cou9789x_ch16_323-350.indd 343 8/22/08 6:08:35 PM Confirming Pages 344 PART 4 FREQUENCY RESPONSE PROBLEMS 16.1. Calculate the value of gain Kc needed to produce continuous oscillations in the control system shown in Fig. P16–1 when (a) n is 2 (b) n is 3 Do not use a graph for this calculation. R + 2 (2s + 1)n Kc − C FIGURE P16–1 16.2. (a) Plot the asymptotic Bode diagram |B/e| versus w for the control system shown in Fig. P16–2. (b) The gain Kc is increased until the system oscillates continuously at a frequency of 3 rad/min. From this information, calculate the transportation lag parameter td. R + Kc − 1 10s + 1 + − 1 s+1 C 1 s+1 B e− ds FIGURE P16–2 16.3. The frequency response for the block Gp in Fig. P16–3 is given in the following table: f, cycles/min 0.06 0.08 0.10 0.15 0.20 0.30 0.40 0.60 0.80 cou9789x_ch16_323-350.indd 344 Gain 1.60 1.40 1.20 0.84 0.61 0.35 0.22 0.11 0.066 Phase angle, deg ⫺68 ⫺88 ⫺105 ⫺145 ⫺177 ⫺235 8/22/08 6:08:35 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 345 Block Gp contains a distance velocity lag e⫺ts with t ⫽ 1 (this transfer function is included in the data given in the table). (a) Find the value of Kc needed to produce a phase margin of 30° for the system if tI ⫽ 0.2. (b) Using the value of Kc found in part (a) and using tI ⫽ 0.2, find the percentage change in the parameter t to cause the system to oscillate continuously with constant amplitude. R + ( ) Kc 1 + 1s i − Gp C FIGURE P16–3 16.4. The system shown in Fig. P16–4 is controlled by a proportional controller. The concentration of salt in the solution leaving the tank is controlled by adding a concentrated solution through a control valve. Controller C1 = 25 lb/ft3 1 ft3/min of water Holdup volume 3 ft3 C FIGURE P16–4 The following data apply: 1. Concentration of concentrated salt solution C1 ⫽ 25 lb salt/ft3 solution. 2. Controlled concentration C ⫽ 0.1 lb salt/ft3 solution. 3. Control valve: The flow through the control valve varies from 0.002 to 0.0006 ft3/min with a change of valve-top pressure from 3 to 15 psi. This relationship is linear. 4. Distance velocity lag: It takes 1 min for the solution leaving the tank to reach the concentration-measuring element at the end of the pipe. 5. Neglect lags in the valve. (a) Draw a block diagram of the control system. Place in each block the appropriate transfer function. Calculate all the constants and give the units. (b) Using a frequency-response diagram and the Ziegler-Nichols rules, determine the settings of the controller. (c) Using the controller settings of part (b), calculate the offset when the set point is changed by 0.02 unit of concentration. cou9789x_ch16_323-350.indd 345 8/22/08 6:08:35 PM Confirming Pages 346 PART 4 FREQUENCY RESPONSE 16.5. The stirred-tank heater system shown in Fig. P16–5 is controlled by a PI controller. The following data apply: Flow rate w of liquid through the tanks: 250 lb/min Holdup volume of each tank: 10 ft3 Density of liquid: 50 lb/ft3 Final control element: A change of 1 psi from the controller changes the heat input q by 100 Btu/min. The final control element is linear. psi Final control element w w PI controller Kc, psi/°F I, min w w T Thermocouple FIGURE P16.5 (a) Draw a block diagram of the control system. Show in detail such things as units and numerical values of the parameters. (b) Determine the controller settings by the Ziegler-Nichols rules. (c) If the control system is operated with proportional mode only, using the value of Kc found in part (b), determine the flow rate w at which the system will be on the verge of instability and oscillate continuously. What is the frequency of this oscillation? 16.6. The transfer function of a process and measurement element connected in series is given by e⫺0.4 s ( 2s ⫹ 1)2 (a) Sketch the open-loop Bode diagram (gain and phase) for a control system involving this process and measurement lag. (b) Specify the gain of a proportional controller to be used in this control system. 16.7. (a) For the control system shown in Fig. P16–7, determine the transfer function C/U. (b) For Kc ⫽ 2 and tD ⫽ 1, find C(1.25) and the offset if U(t) ⫽ u(t), a unit step. (c) Sketch the open-loop Bode diagram for Kc ⫽ 2 and tD ⫽ 1. For the upper part of the diagram (AR versus w), show the asymptotic approximation. Include the transfer function for the controller in the open-loop Bode diagram. (d) From the Bode diagram, what do you conclude about the stability of the closed-loop system? U R + Kc (1 + Ds) + + 2 s C − FIGURE P16–7 cou9789x_ch16_323-350.indd 346 8/22/08 6:08:36 PM Confirming Pages CHAPTER 16 CONTROL SYSTEM DESIGN BY FREQUENCY RESPONSE 347 16.8. The proportional controller of the temperature control system shown in Fig. P16–8 is properly tuned to give a good transient response for a standard set of operating conditions. If changes are made in the operating conditions, the control system may become more or less stable. If the changes listed below are made separately, determine whether the system becomes more stable, less stable, or remains the same. Try to use the Bode stability criterion and sketches of frequency response graphs to solve this problem. 1. Controller gain increases. 2. Length of pipe between measuring element and tank increases. 3. Measuring element is inserted in tank. 4. Integral action is provided in controller. 5. A larger valve is used (i.e., one with a higher Cv value). Controller Steam Water FIGURE P16–8 16.9. For each control system shown in Fig. P16–9, determine the characteristic equation of the closed-loop response and determine the value of Kc that will cause the system to be on the verge of instability (i.e., find the ultimate gain Kcu). If possible, use the Routh test. Note that the feedback element for system B is an approximation to e⫺2s. System A: R + Kc − 1 (8s + 1)2 C 1 (8s + 1)2 C e−2s System B: R + Kc − 1−s 1+s FIGURE P16–9 cou9789x_ch16_323-350.indd 347 8/22/08 6:08:37 PM Confirming Pages 348 PART 4 FREQUENCY RESPONSE 16.10. (a) For the system shown in Fig. P16–10 determine the value of Kc that will give 30° of phase margin. (b) If a PI controller with tI ⫽ 2 is used in place of the proportional controller, determine the value of Kc for 30° of phase margin. R + 1 (s + 1)2 Kc − C FIGURE P16–10 16.11. A stirred-tank heating process and its block diagram are shown in Fig. P16–11. The control system is tuned by the Ziegler-Nichols method, and the ultimate frequency wu is 2 rad/min. (a) Determine the value of Kc by the Ziegler-Nichols method of tuning. (b) What is the length of the pipe between the tank and the measuring element? (c) What are the gain margin and the phase margin for the control system when Kc is set to the Ziegler-Nichols value found in part (a)? Controller Energy w = 100 lb/min R + Kc − 100 wC s+1 e− ds C FIGURE P16–11 Data on process: Density of fluid, r ⫽ 62 lb / ft 3 Heat capacity , C of fluid ⫽ 1. 0 Btu / (lb⭈ ⬚ F ) Inside diameter of pipe ⫽ 2. 0 in cou9789x_ch16_323-350.indd 348 8/22/08 6:08:37 PM Confirming Pages CHAPTER 16 CAPSULE SUMMARY THE BODE STABILITY CRITERION A control system is unstable if the open-loop frequency response exhibits an AR exceeding unity at the frequency for which the phase lag is 180°. This frequency is called the crossover frequency. The rule is called the Bode stability criterion. Gain Margin and Phase Margin These quantities are defined to provide a safety margin in the design and selection of controllers. Design rules of thumb: Set gain margin ⫽ 1.7 and phase margin ⫽ 30°. Choose the more conservative of the criteria (the one that yields the lower value of Kc). In general, the system becomes less oscillatory and more stable as GM and PM are increased. Note GM ⭐ 0 and PM ⭐ 0 are unstable. Amplitude ratio Gain margin GM ⫽ 1 1 ⫽ A ARf ⫽⫺180⬚ Phase margin (PM) ⫽ 180⬚ ⫺fLAG AR⫽1 1 Gain margin A Phase angle, deg 0 Phase margin −180 co 349 cou9789x_ch16_323-350.indd 349 8/22/08 6:08:38 PM Confirming Pages 350 PART 4 FREQUENCY RESPONSE Suggested Initial Controller Tuning Parameters Ziegler-Nichols controller settings Type of control Proportional Proportional-integral (PI) Proportional-integral-derivative (PID) cou9789x_ch16_323-350.indd 350 s1 Gc(g) Kc Kc 0.5Ku 1 Kc 1 ⫹ t I s 0.45Ku Pu 1.2 1 Kc 1 ⫹ ⫹ t D s tIs 0.6Ku Pu 2 sD Pu 8 8/22/08 6:08:38 PM Confirming Pages PART V PROCESS APPLICATIONS cou9789x_ch17_351-390.indd 351 8/22/08 6:13:13 PM cou9789x_ch17_351-390.indd 352 8/22/08 6:13:13 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES U p to this point, the control systems considered have been single-loop systems involving one controller and one measuring element. In this chapter, several multiloop systems are described; these include cascade control, feedforward control, ratio control, Smith predictor control, and internal model control. The first three have found wide acceptance in industry. Smith predictor control has been known for about forty years, but it was considered impractical until the modern microprocessor-based controllers provided the simulation of transport lag. Internal model control, based on a rigorous mathematical foundation and an accurate model of the process, has been the subject of research for the past twenty years. The controller hardware and instrumentation for all these systems are readily available from manufacturers. Since this chapter is quite long, the reader may wish to select the type of advanced control strategy that is of particular interest. The descriptions of the five strategies are independent and need not be read in the order presented. 17.1 CASCADE CONTROL To provide motivation for the study of cascade control, consider the single-loop control of a jacketed kettle as shown in Fig. 17–1a. The system consists of a kettle through which water, entering at temperature Ti, is heated to To by the flow of hot oil through a jacket surrounding the kettle. The temperature of the water in the kettle is measured and transmitted to the controller, which in turn adjusts the flow of hot oil through the jacket. This control system is satisfactory for controlling the kettle temperature; however, if the temperature of the oil supply should drop, the kettle temperature can undergo a large prolonged excursion from the set point before control is again established. The reason is that the controller does not take corrective action until the effect of the drop in oil 353 cou9789x_ch17_351-390.indd 353 8/22/08 6:13:13 PM Confirming Pages 354 PART 5 PROCESS APPLICATIONS supply temperature has worked itself through the system of several resistances to reach the measuring element. To prevent the sluggish response of kettle temperature to a disturbance in oil supply temperature, the control system shown in Fig. 17–1b is proposed. In this system, which includes two controllers and two measuring elements, the output of the primary controller is used to adjust the set point of a secondary controller, which is used to control the jacket temperature. Under these conditions, the primary controller indirectly adjusts the jacket temperature. If the oil temperature should drop, the secondary control loop will act quickly to maintain the jacket temperature close to the value determined by the set point that is adjusted by the primary controller. This system shown in Fig. 17–1b is called a cascade control system. The primary controller is also referred to as the master controller, and the secondary controller is referred to as the slave controller. w Ti Water Tj Ts Hot oil To (a) Primary controller w Ti Water Set point Secondary controller Hot oil Tj To (b) FIGURE 17–1 (a) Single-loop control of a jacketed kettle; (b) cascade control of a jacketed kettle. cou9789x_ch17_351-390.indd 354 8/22/08 6:13:13 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 355 A simplified block diagram of the single-loop system is shown in Fig. 17–2a. Figure 17–2b, which is a block diagram representation of the cascade control system, shows clearly that an inner loop has been added to the conventional control system. L, oil supply temperature R + Gc Gv + + Gjacket Gkettle To − Gm (a) L R + + Gc Gc2 1 − Gv + + Gjacket Tj Gkettle To − Gm1 Gm2 (b) FIGURE 17–2 Block diagram: (a) single-loop conventional control; (b) cascade control. Analysis of Cascade Control To develop the closed-loop transfer functions for a cascade control system, consider the general block diagram shown in Fig. 17–3. In this diagram, the load disturbance U enters between two blocks of the plant, and the inner loop encloses this load disturbance. To determine the transfer function C/R, the inner loop is reduced to one block by the method shown in Chap. 11. The result is shown in Fig. 17–3b, and the block diagram of Fig. 17–3b can be used to give the result C Gc1 Ga G3 ⫽ R 1 ⫹ Gc1 Ga G3 H1 where Ga ⫽ (17.1) Gc2 G1G2 1 ⫹ Gc2 G1G2 H 2 To obtain the transfer function relating output to load C/U, the block diagram of Fig. 17–3a is rearranged by placing the transfer function Gc2 G1 in the feedback paths of the primary and secondary loops; the new arrangement is shown in Fig. 17–4a. Since cou9789x_ch17_351-390.indd 355 8/22/08 6:13:14 PM Confirming Pages 356 PART 5 PROCESS APPLICATIONS U=0 R + + Gc Gc2 1 + G1 + G2 G3 C − − H2 H1 (a) Ga R + Gc Gc2G1G2 1 − 1 + Gc2G1G2H2 G3 C H1 (b) FIGURE 17–3 Block diagram for cascade control for set point change. R ⫽ 0 for the case under consideration, the block diagram can be redrawn as shown in Fig. 17–4b. This diagram, which has the same form as the one in Fig. 17–3a, can now be reduced to the form shown in Fig. 17–4c. Application of the rules of Chap. 11 to Fig. 17–4c finally gives C G3 Ga ⫽ U G1Gc2 1 ⫹ Ga Gc1 H1G3 (17.2) where Ga is the same as given in Eq. (17.1). Example 17.1. To compare conventional control with cascade control, consider the conventional control system of Fig. 17–5a in which a third-order process is under PI control. A cascade version of this single-loop control system is shown in Fig. 17–5b in which an inner loop having proportional control encloses the load disturbance U. To obtain a response of the conventional control system for use in comparison with the response of the cascade system, the block diagram of Fig. 17–5a was simulated on a computer. The values of Kc and tI were chosen by trial and error to give the response to a step change in set point shown as curve I of Fig. 17–6; this response, which has a decay ratio of about 1 , was obtained with Kc ⫽ 2.84 and tI ⫽ 5. 4 cou9789x_ch17_351-390.indd 356 8/22/08 6:13:15 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 357 U + + R=0 + − + G2 G3 C − Gc G1 H2 2 Gc G1 Gc 2 H1 1 (a) U + + − G2 G3 C − Gc G1 H2 2 Gc G1 Gc 2 1 H1 (b) U G2 + − 1 + Gc G1G2H2 G3 C 2 Gc Gc G1H1 1 2 (c) FIGURE 17–4 Block diagram for cascade control for load change. The Ziegler-Nichols settings (Kc ⫽ 3.65 and tI ⫽ 3.0) gave a set point response that was too oscillatory. Having obtained satisfactory controller settings (Kc ⫽ 2.84 and tI ⫽ 5.0), we show the response of the system to a step change in U of 4 units as curve II of Fig. 17–7. The load response for no control (i.e., Kc ⫽ 0) is also shown as curve I for comparison. The cascade control system of Fig. 17–5b was also simulated to obtain a load response. The controller gain K c2 of the inner loop was chosen arbitrarily to be 10.0. This value was chosen to be high to obtain a fast-responding inner loop, a desirable situation for cascade control. Because of the introduction of the inner loop, the dynamics of the control system have changed, and it is necessary to tune the primary controller parameters for a good response to a step change in set point. By trial and error, primary controller settings of K c1 ⫽ 1 . 0 and cou9789x_ch17_351-390.indd 357 8/22/08 6:13:15 PM Confirming Pages 358 PART 5 PROCESS APPLICATIONS U + R Kc 1 + − + 1 s+1 1 Is 1 s+1 + 1 s+1 C 1 (a) U R + Kc1 1 + − 1 Is + 1 s+1 Kc2 − + + 1 s+1 1 s+1 C 1 1 (b) FIGURE 17–5 Block diagrams for Example 17.1: (a) single-loop conventional control; (b) cascade control. I Conventional control II Cascade control C 1 0.5 0 5 10 15 20 t FIGURE 17–6 Responses to step change in set point for single-loop control and cascade control for Example 17.1. I: Conventional control with Kc ⫽ 2.84 and tI ⫽ 5; II: cascade control with Kc1 ⫽ 1.0, tI ⫽ 0.63, and Kc2 ⫽ 10. tI ⫽ 0.63 were found that produced the response to a unit step in set point, shown as curve II in Fig. 17–6. The use of Ziegler-Nichols settings produced a less desirable response. Using the controller parameters found from the step change in set point ( K c1 ⫽ 1 . 0,t I ⫽ 0 . 63), the response of the cascade system to a step change in load of 4 units was obtained and is shown as curve III of Fig. 17–7. As shown in Fig. 17–7, the load response for the cascade control system is far superior to the load response of the conventional control system. The maximum deviation of the cascade response has been reduced by a factor of about 4, and the frequency of oscillation has nearly doubled. cou9789x_ch17_351-390.indd 358 8/22/08 6:13:16 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 359 4 I No control 3 2 C II Conventional PI control 1 III Cascade control 0 0 5 10 15 20 t FIGURE 17–7 Responses to step change in set point for single-loop control and cascade control for Example 17.1. I: No control; II: conventional control with Kc ⫽ 2.84 and tI ⫽ 5; III: cascade control with Kc1 ⫽ 1.0, tI ⫽ 0.63, and Kc2 ⫽ 10. Generalizations Cascade control is especially useful in reducing the effect of a load disturbance that moves through the control system slowly. The inner loop has the effect of reducing the lag in the outer loop, with the result that the cascade system responds more quickly with a higher frequency of oscillation. Example 17.2 will illustrate this effect of cascade control. The choice of control action and tuning of the primary and secondary controllers for a cascade control system must be given careful consideration. The control action for the inner loop is often proportional with the gain set to a high value. The rationale for the use of proportional control rather than two- or three-mode control is that tuning is simplified and any offset associated with proportional control of the inner loop can be handled by the presence of integral action in the primary controller. The gain of the secondary controller should be set to a high value to give a tight inner loop that responds quickly to load disturbance; however, the gain should not be so high that the inner loop is unstable. Although the primary control loop can provide stable control even when the inner loop is unstable, it is considered unwise to have an unstable inner loop because the system will go unstable if the primary controller is placed in manual operation or if there is a break in the outer loop. The action for the primary controller is generally PI or PID. The integral action is needed to reduce offset when sustained changes in load or set point occur. The problem of adjusting a primary controller is essentially the same as for a single-loop control system. Since the addition of the inner loop can change the dynamics of the outer loop significantly, the primary controller must be retuned when the inner loop is closed or when the secondary controller settings are changed. Microprocessor-based controllers available today can implement cascade control very easily. cou9789x_ch17_351-390.indd 359 8/22/08 6:13:16 PM Confirming Pages 360 PART 5 PROCESS APPLICATIONS Example 17.2. The claim is often made that cascade control gives a better response than conventional control because the lags in the outer loop are reduced. To illustrate this benefit, consider the conventional control and the cascade control of a thirdorder plant in Fig. 17–8a and b. The inner loop of the cascade system surrounds two of the first-order blocks in the plant. To simplify the discussion, the load disturbance is not shown since we are interested only in the closed-loop dynamics. The equivalent single-loop control system of the cascade system, shown in Fig. 17–8c, was obtained by the usual method for reducing a loop to a single block. + R Kc 1 (s + 1)2 1 s+1 − C (a) Kc 2 + R + Kc − 1 (s + 1)2 10 1 − 1 s+1 C (b) R + 10/11 Kc 1 − 1 11 s2 2 11 + s+1 1 s+1 (c) FIGURE 17–8 Block diagram for Example 17.2. Comparing Fig. 17–8a with c shows that the use of cascade control has replaced a second-order critically damped system represented by the first two blocks of the plant [1/(s ⫹ 1)2] with this underdamped second-order system K 2 2 t s ⫹ 2zts ⫹ 1 where K⫽ 10 11 t ⫽ 1 11 z ⫽ 1 11 This second-order underdamped system, for which t and z are small, responds much faster than the critically damped second-order transfer function of the first two blocks of the open-loop system. Consequently, the cascade system will respond faster with a higher frequency of oscillation, as we have already seen in the simulated response of Fig. 17–6. cou9789x_ch17_351-390.indd 360 8/22/08 6:13:17 PM Confirming Pages CHAPTER 17 17.2 ADVANCED CONTROL STRATEGIES 361 FEEDFORWARD CONTROL If a particular load disturbance occurs frequently in a control process, the quality of control can often be improved by the addition of feedforward control. Consider the composition control system shown in Fig. 17–9a in which a concentrated stream of control reagent containing water and solute is used to control the concentration of the stream leaving a three-tank system. The stream to be processed passes through a preconditioning stirred tank where composition fluctuations are smoothed out before the outlet stream is mixed with control reagent. A three-tank system has been chosen for ease of computation in a numerical example that follows. PI controller Control reagent Control valve ci c Preconditioning tank Tank1 Tank2 Tank3 (a) Ci R + − 1 5s + 1 1 (s + 1)3 1 Is 1 (s + 1)3 Kc 1 + + + C (b) FIGURE 17–9 Composition control system: (a) physical process; (b) block diagram. In the conventional feedback control system shown in Fig. 17–9a, the measurement of composition in the third tank is sent to a controller, which generates a signal that opens or closes the control valve, which in turn supplies concentrated reagent to the first tank. The block diagram corresponding to the control system of Fig. 17–9a is shown in Fig. 17–9b. [In Fig. 17–9a, concentration is denoted by c (lowercase letter). In the block diagram of the process in Fig. 17–9b, the symbol for concentration is denoted by C (capital letter) to denote a deviation variable. This use of symbols follows the procedure established in Chap. 4.] To obtain some specific control system responses, numerical values of the time constants of the tanks have been chosen as shown in Fig. 17–9b. To study the response of this control system, the block diagram shown in Fig. 17–9b was simulated on a computer. cou9789x_ch17_351-390.indd 361 8/22/08 6:13:17 PM Confirming Pages 362 PART 5 PROCESS APPLICATIONS The values of Kc and tI were chosen by trial and error to give the response to a step change in set point shown in curve II of Fig. 17–10; this response, which has a decay ratio of about 14 , was obtained with Kc ⫽2.84 and tI⫽5.0. The Ziegler-Nichols settings (Kc⫽3.65 and tI ⫽3.0) give a set point response shown as curve I of Fig. 17–10, which is too oscillatory. 3 I Ziegler-Nichols C 2 II Improved response 1 0 5 10 15 20 25 t Responses to a step change in set point for PI control. Curve I: Ziegler-Nichols settings: Kc = 3.65, I = 3.0; curve II: settings for improved response: Kc = 2.84, I = 5.0. FIGURE 17–10 Having obtained satisfactory settings for the controller (Kc ⫽ 2.84, tI ⫽ 5.0), we obtained the response of the system to a step change in Ci of 10 units, shown as curve I in Fig. 17–11. Note that the response is oscillatory and has a long tail. This response illustrates the fact that the feedback control system does not begin to respond until the load disturbance has worked its way through the forward loop and reaches the measuring element, with the result that the composition can move far from the set point during the transient. 2 I C 1 0 −1 −2 −3 10 IV III 15 20 Responses to a step change in load for t feedforward-feedback control. Curve I: PI control with Kc = 2.84, I = 5.0 II Curve II: FF control with Kc = 2.84, I = 5.0, Gf = −1/(5s + 1) Curve III: FF control with Kc = 2.84, I = 5.0, Gf = −1 Curve IV: FF control with Kc = 2.84, I = 5.0, Gf = −0.5 FIGURE 17–11 If the change in load disturbance Ci can be detected as soon as it occurs in the inlet stream, this information can be fed forward to a second controller that adjusts the control valve in such a way as to prevent any change in the outlet composition from the set point. A controller that uses information fed forward from the source of the load disturbance is called a feedforward controller. The block diagram that includes the feedforward controller Gf as well as the feedback controller Gc is shown in Fig. 17–12. cou9789x_ch17_351-390.indd 362 8/22/08 6:13:18 PM Confirming Pages CHAPTER 17 Ci G1 Gp 1 5s + 1 1 (s + 1)3 ADVANCED CONTROL STRATEGIES 363 Feedforward controller Gf(s) Gc R + E 1 Kc 1 + 1s Gp + + 1 (s + 1)3 + + C − Feedback controller FIGURE 17–12 Control system with feedforward and feedback controllers. Analysis of Feedforward Control The response of C to changes in Ci and R can be written from Fig. 17–12 as follows: C ( s ) ⫽ G1 ( s ) G p ( s ) Ci ( s ) ⫹ G f ( s ) G p ( s ) Ci ( s ) ⫹ Gc ( s ) G p ( s ) E ( s ) (17.3) where E(s) ⫽ R(s) ⫺ C(s). To determine the transfer function of Gf (s) that will prevent any change in the control variable C from its set point R, which is 0, we solve Eq. (17.3) for Gf (s) with C ⫽ 0 and R ⫽ 0. The result is G f ( s ) ⫽ ⫺G1 ( s ) (17.4) For the example under consideration in Fig. 17–12, G f (s) ⫽ ⫺1 5s ⫹ 1 (17.5) This transfer function can be implemented easily with existing control hardware. If the load response of the control system in Fig. 17–12, with Gf (s) given by Eq. (17.5), were obtained for a step change in Ci, there would be no deviation of C from the set point (i.e., perfect control). This response is shown as curve II in Fig. 17–11, which of course is a horizontal line at C ⫽ 0. Rather than use the Gf (s) of Eq. (17.5) in the feedforward controller, one can try using only the constant term of Gf (s), that is, G f ( s ) ⫽ ⫺1 The response for Gf ⫽ ⫺1 gives curve III in Fig. 17–11; this response has a very large undershoot before the feedback controller returns C to the set point. If we try using Gf (s) ⫽ ⫺0.5, we obtain curve IV of Fig. 17–11; the undershoot is less in this case, cou9789x_ch17_351-390.indd 363 8/22/08 6:13:18 PM Confirming Pages 364 PART 5 PROCESS APPLICATIONS but the response is still unsatisfactory. As shown by curves III and IV, omitting the dynamic part of Gf (s) can give very poor results. The success of using a feedforward controller depends on accurate knowledge of the process model, a luxury that may not be available in many applications. Implementing Feedforward Transfer Functions In applications of feedforward control, Gf (s) may take the form of a lead expression, such as Gf (s) ⫽ 1 ⫹ tf s. When this occurs, it is necessary to approximate 1 ⫹ tf s by a lead-lag expression, such as G f (s) ⫽ 1⫹t fs 1 ⫹ bt f s where  ⬍⬍ 1 (see Prob. 10.3 for an example of this). To see how Gf (s) takes the form of a lead expression, consider the load disturbance ci of Fig. 17–9 to enter tank 2. Since no change in concentration occurs in the stream entering the preconditioning tank, we may eliminate it from the diagram for the case under consideration to obtain the diagram in Fig. 17–13. Control reagent ci Tank 1 Tank 2 Tank 3 FIGURE 17–13 Composition control with disturbance to second tank. Adding feedforward control and feedback control to the system in Fig. 17–13 gives the block diagram of Fig. 17–14. The diagram shown in Fig. 17–14 is the same as that in Fig. 17–12 with the exception that the disturbance Ci enters tank 2 instead of the preconditioning tank. As shown previously, the response of C to a change in Ci and R can be written directly from Fig. 17–14 as C ( s ) ⫽ G1 ( s ) Ci ( s ) ⫹ G f ( s ) G p ( s ) Ci ( s ) ⫹ Gc ( s ) G p ( s ) E ( s ) (17.6) where E(s) ⫽ R(s) ⫺ C(s). cou9789x_ch17_351-390.indd 364 8/22/08 6:13:19 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 365 G1 1 (s + 1)2 Ci Gf R + E Gp + Gc + PI 1 (s + 1)3 + + C − FIGURE 17–14 Feedforward-feedback control for system in Fig. 17–13. In order for C not to change from the set point R, which is 0, we solve Eq. (17.6) for Gf (s) with C ⫽ 0 and R ⫽ 0 to obtain G f (s) ⫽ ⫺ G1 ( s ) Gp (s) (17.7) Introducing the expressions for G1 (s) and Gp (s) from Fig. 17–14 into Eq. (17.7) gives G f ( s ) ⫽ ⫺( s ⫹ 1) (17.8) It is not practical to implement ⫺(s ⫹ 1). To see this, consider the response of ⫺(s ⫹ 1) to a step change as shown in Fig. 17–15. X = 1s t 0 −1 − (s + 1) Y Y −∞ FIGURE 17–15 Step response for ⫺(s ⫹ 1). There is no hardware that will produce an impulse as shown in Fig. 17–15; however, one can approximate ⫺(s ⫹ 1) by means of a lead-lag transfer function of the form Y (s) X (s) where cou9789x_ch17_351-390.indd 365 ⫽⫺ t fs ⫹1 bt f s ⫹ 1 (17.9)  ⬍⬍ 1 8/22/08 6:13:20 PM Confirming Pages 366 PART 5 PROCESS APPLICATIONS If we let b ⫽ 0.1 and tf ⫽ 1 for the control system under consideration, we obtain as an approximation to Eq. (17.8) s ⫹1 G f (s) ⫽ ⫺ (17.10) 0 . 1s ⫹ 1 The response of this transfer function to a step input is shown in Fig. 17–16. The effect of this transfer function ⫺(s ⫹ 1)/(0.1s ⫹ 1) on the output of the feedforward controller for a step change in load is to give a sudden drop in flow followed by a fast exponential increase in the flow to a steady-state flow of ⫺1. Note that for the parameters chosen for the transfer functions in Fig. 17–14, a unit increase in Ci must eventually be compensated by a unit decrease in the signal from the feedforward controller if there is to be no change in the process output. The sudden, initial drop in flow may be too abrupt for the control hardware, in which case the output would saturate. In practice, b can be increased (perhaps to 0.5) to reduce the magnitude of the initial drop. t 0 X = 1s s+1 − 0.1s + 1 −1 Y Y −Y = (10 − 1) e−t/0.1 + 1 −10 FIGURE 17–16 Step response for ⫺(s ⫹ 1)/(0.1s ⫹ 1). The effect of using Gf (s) ⫽ ⫺ (s ⫹ 1)/(0.1s ⫹ 1) with feedback control is shown in Fig. 17–17. The responses shown, which were obtained by simulation, are for a unitstep change in Ci. Curve I is for the case of feedback control only with Kc ⫽ 2.84 and tI ⫽ 5.0. Curve II is for feedforward-feedback control using Eq. (17.10) for Gf (s) and Kc ⫽ 2.84 and tI ⫽ 5.0. One can see that the overshoot for the feedforward-feedback response has been reduced significantly. I C 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 −0.05 II 0 5 10 t FIGURE 17–17 Comparison of conventional feedback control with feedforward-feedback control for system in Fig. 17–14. Curve I: PI control with Kc ⫽ 2.84 and tI ⫽ 5. Curve II : Feedforward-feedback control with Kc ⫽ 2.84, tI ⫽ 5, and Gf ⫽ ⫺(s ⫹ 1)/(0.1s ⫹ 1). cou9789x_ch17_351-390.indd 366 8/22/08 6:13:21 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 367 Tuning Rules for Feedforward Feedback Control In the practical application of feedforward control, one does not have a block diagram with transfer functions as shown in Figs. 17–12 and 17–14. For such a practical situation, one can still tune the feedforward controller by introducing a step change in the disturbance that enters the feedforward controller (Ci in Fig. 17–14) and then applying some tuning rules. The rules to be discussed here are from a training film on feedforward control produced by Foxboro Co. (1978). Feedforward Rules In describing these rules, reference will be made to the general block diagram for a feedforward-feedback system shown in Fig. 17–18. It is assumed that Gf (s) will be a lead-lag transfer function of the form G f (s) ⫽ K f ( T1s ⫹ 1) T2 s ⫹ 1 (17.11) where Kf ⫽ steady-state gain of feedforward controller T1, T2 ⫽ time constants of dynamic part of feedforward controller Ci G1 Gf Mf f + R + Gc + + Gp + C − FIGURE 17–18 Feedforward-feedback control system. Commercial microprocessor-based controllers provide this lead-lag transfer function. The tuning rules listed below are explained with the help of Fig. 17–19. In that figure, a unit step is selected for the disturbance Ci, and Kf has been taken as ⫺1. In practice, Kf will, of course, depend on the particular process being controlled. 1. 2. 3. 4. Remove the control action in Gc(s) by setting the controller to manual. Set the feedforward controller to the computed steady-state gain (Kf) necessary to compensate ultimately for a step change in Ci. This means that the dynamic portion of Gf (s) will be removed, and only the constant term (Kf) will remain. Make a step change in Ci and observe the open-loop transient of C. The general shapes of the response to be expected are shown in Fig. 17–19. If the response shown in Fig. 17–19a occurs, lead must predominate in Gf (s) of Eq. (17.11) (i.e., T1 > T2). If the response of Fig. 17–19b occurs, lag must predominate in Gf (s) (i.e., T1 < T2). The values of T1 and T2 in Eq. (17.11) are found by use of the information in Table 17.1. The value of Kf in Eq. (17.11) has been obtained in step 2. cou9789x_ch17_351-390.indd 367 8/22/08 6:13:22 PM Confirming Pages 368 PART 5 PROCESS APPLICATIONS Ci 1 Ci 1 0 0 t t 0 Mf f −1 t t 0 Mf f −1 C tp C 0 0 tp t t (a) (b) FIGURE 17–19 Open-loop response to determine lead-lag time constants in feedforward tuning rules: (a) Lead must predominate in Gf ; (b) lag must predominate in Gf. TABLE 17.1 Tuning parameters for feedforward control Predominant mode T1 T2 Lead 1.5tp 0.7tp Lag 0.7tp 1.5tp G f (s) ⫽ K f T1s ⫹ 1 T2 s ⫹ 1 Example 17.3 will help clarify the use of these tuning rules. Example 17.3. Use of feedforward tuning rules. Apply the feedforward tuning rules to the system in Fig. 17–14. Since this example is concerned with the application of the tuning rules to a system for which a mathematical model is not generally available, the reader should assume that the transfer functions for G1(s) and Gp(s) in Fig. 17–14 are unknown. The determination of Gf (s) is to be obtained solely by information from open-loop transients. We must first determine the steady-state gain Kf for the system of Fig. 17–14. If a step change in Ci is made, C will undergo a transient and eventually level out at a steady-state value. If the controller parameters are properly selected, the value of C at the end of the transient will be the same as it was before the transient occurred. By computation or experiment, one can determine the value of Kf needed to obtain no change in C. For the system in Fig. 17–14, one can see that Kf of Eq. (17.11) must be equal to ⫺1. cou9789x_ch17_351-390.indd 368 8/22/08 6:13:22 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 369 We must now apply the feedforward tuning rules to obtain T1 and T2 in Eq. (17.11). After removing the feedback controller action [Gc(s)], we have the equivalent diagram shown in Fig. 17–20. A unit-step change in Ci produces the transient for C shown as curve I in Fig. 17–22. Comparing the 2 shape of the transient with those of Fig. 17–19, we see that lead must predominate in Gf (s). The peak value occurs at tp ⫽ 2. Applying the rules in Table 17.1 gives T1 ⫽ 1 . 5t p ⫽ 3 T2 ⫽ 0 . 7t p ⫽ 1 . 4 The feedforward controller transfer function is therefore G f (s) ⫽ ⫺( 3s ⫹ 1) 1.4s ⫹ 1 (17.12) G1(s) Ci + Kf = − 1 Gp(s) + C FIGURE 17–20 Open-loop feedforward test to determine parameters for Gf. It is of interest to show the response of C for feedforward only when the feedforward transfer function of Eq. (17.12) is used. MATLAB and Simulink can be used to simulate the result for a unit-step change in Ci (Fig. 17–21). It is shown as curve II in Fig. 17–22. −1 Step Kf 3s + 1 1.4s + 1 up 1 s3 + 3s2+ 3s + 1 G2 ++ 1 s2+ 2s + 1 G1 −1 Kf 2 Example 17.3 To workspace Scope 1 s3 + 3s2+ 3s + 1 G5 ++ 1 s2+ 2s + 1 G4 FIGURE 17–21 Simulink model for simulating Example 17.3. cou9789x_ch17_351-390.indd 369 8/22/08 6:13:23 PM Confirming Pages 370 PART 5 PROCESS APPLICATIONS 0.3 0.25 I 0.2 0.15 C 0.1 II 0.05 0 −0.05 −0.1 −0.15 −0.2 0 2 4 6 8 10 Time 12 14 16 18 20 FIGURE 17–22 Open-loop response for step change in Ci for Example 17.3. Curve I: Gf ⫽ ⫺1, curve II: Gf ⫽ ⫺(3s ⫹ 1)/(1.4s ⫹ 1). When the Gf(s) of Eq. (17.12) is used and the controller parameters for Gc(s) are Kc ⫽ 2.84 and tI ⫽ 5.0, the feedforward-feedback response to a unit-step change in Ci is shown as curve II in Fig. 17–23. For comparison, the response for feedback control only is also shown in Fig. 17–23. 0.4 I C 0.2 0 II 0 2 4 6 8 t −0.2 FIGURE 17–23 Comparison of conventional feedback control with feedforward-feedback control for Example 17.3. Curve I: PI control with Kc ⫽ 2.84 and tI ⫽ 5. Curve II: Feedforward-feedback control with Kc ⫽ 2.84, tI ⫽ 5, and Gf ⫽ ⫺(3s ⫹ 1)/(1.4s ⫹ 1). 17.3 RATIO CONTROL An important control problem in the chemical industry is the combining of two or more streams to provide a mixture having a desired ratio of components. Examples of this mixing operation include the blending of reactants entering a chemical reactor or for the injection of a fuel/air mixture into a furnace. cou9789x_ch17_351-390.indd 370 8/22/08 6:13:24 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 371 Figure 17–24a shows an example of a ratio control system. It depicts a control system for blending two liquid streams A and B to produce a mixed stream having the ratio Kr in units of mass B/mass A. Stream A, which is uncontrolled, is used to adjust the flow of stream B so that the desired ratio is maintained. The measured signal for stream A is multiplied by the desired ratio Kr to provide a signal that is the set point for the flow control loop for stream B. The parameter Kr can be adjusted to the desired value. Control hardware is available to perform the multiplication of two control signals. qA Fluid A Flow-measuring element Set point Kr Controller Flow-measuring element qB Fluid B PB = supply pressure (a) PB QA Gm1 Measured variable Kr + Gc Set point − Gv Gp 1 + + QB Gm2 (b) FIGURE 17–24 (a) Ratio control system; (b) block diagram for ratio control (set point ⫽ Gm1 K r Q A ). A block diagram of the ratio control system is shown in Fig. 17–24b. In a flow control loop, the dynamic elements consist of the controller, the flow-measuring element, and the control valve. For incompressible fluids, there is no lag between the change in valve position and the corresponding flow rate. For this reason, the transfer function between the valve and the measurement of flow rate is simply unity. The block diagram also shows a transfer function Gp that relates the flow rate of B to the supply pressure of B. A transfer function Gm1 is also shown that represents the dynamic lag of the flowmeasuring element for stream A. cou9789x_ch17_351-390.indd 371 8/22/08 6:13:26 PM Confirming Pages 372 PART 5 PROCESS APPLICATIONS From the block diagram, the flow of B may be written QB ⫽ Gp Gm1 K r Gc Gv QA ⫹ PB 1 ⫹ Gc Gv Gm2 1 ⫹ Gc Gv Gm2 The control action for a flow control system is usually PI. The integral action is needed to eliminate offset and thereby establish a precise ratio of the mixed streams of A and B. Derivative action is usually avoided in flow control because the signal from a flowmeasuring element is inherently noisy. The presence of derivative action would amplify the noise and give poor control. An example system from the LOOP-PRO software package (Cooper, 2005) is the control of the air/fuel ratio being fed to a process furnace. A description of the process from “Practical Process Control” that accompanies the software is as follows. A furnace burns natural gas to heat a process liquid. The measured process variable is the temperature of the process liquid as it exits the furnace. To maintain temperature, controllers adjust the feed rate of combustion air and fuel to the firebox using a ratio control strategy. The flow rate of the process liquid acts as a load disturbance to the process. . . . For the furnace, the independent stream is the combustion air flow rate and the dependent stream is the fuel flow rate. Note that while air flow rate is considered the independent stream for ratio control, its flow rate is specified by the temperature controller on the process liquid exiting the furnace. The process diagram is shown in Fig. 17–25. Excess 02(%) 4.37 Liquid Flow (L/min) Liquid Temp (°C) 60.0 198.4 (Disturbance) TC Fuel Flow (m3/min) Minimum Ratio (10/1) 2.0 198.4 X FT Ratio (Air/Fuel) HS FC Controller Output (m3/min) 12.0 RS Air Flow (m3/min) Set Point (°C) FT 24.0 FC 24.0 (Set Point) Screen shot from LOOP-PRO "Furnace Air/Fuel Ratio" Case Study (Cooper, 2005) FIGURE 17–25 cou9789x_ch17_351-390.indd 372 8/30/08 2:57:51 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 373 This computer simulation by Cooper is an excellent means of exploring the behavior of a ratio control system. Additionally, he points out some interesting environmental and safety considerations that factor into the design of such a control system, such as • Avoiding unburned hydrocarbons and carbon monoxide in the flue gases (too fuel-rich) • Avoiding excess energy losses (too much air, a fuel-lean situation) • Avoiding an explosive situation if the airflow fails (fail-safe considerations) This interesting example provides some insights into the application of ratio control in a real-world situation. 17.4 DEAD-TIME COMPENSATION (SMITH PREDICTOR) Processes that contain a large transport lag [exp (⫺t D s )] can be difficult to control because a disturbance in set point or load does not reach the output of the process until tD units of time have elapsed. The control strategy to be described here, which is also known as dead-time compensation, attempts to reduce the deleterious effect of transport lag. Dead-time compensation, which is also referred to as a Smith predictor, was first described by O. J. M. Smith (1957). Consider the single-loop control system of Fig. 17–26 in which the process transfer function Gp (s) is to be modeled by G p ( s ) ⫽ G (s ) e⫺t D s (17.13) U R + Gc(s) + + Gp(s) C − FIGURE 17–26 Control system. The right side of Eq. (17.13) is the product of a transport lag [exp(– tD s)] and a transfer function G(s), which has minimum phase characteristics, such as 1/(t s ⫹ 1). For convenience in developing the dead-time compensation method, only a change in set point R will be considered. If a step change is made in R, the disturbance will not break through and appear at C until tD units of time elapse. Up to time tD, no control action occurs, with the result that the overall closed-loop response will be sluggish and cou9789x_ch17_351-390.indd 373 8/22/08 6:13:27 PM Confirming Pages 374 PART 5 PROCESS APPLICATIONS generally unsatisfactory. To overcome this difficulty, Smith suggested that Gp (s) be modeled according to Eq. (17.13) and that additional feedback paths be inserted into Fig. 17–26 as shown in Fig. 17–27a. If Gp (s) is modeled exactly by Eq. (17.13), a close study of Fig. 17–27a shows that the signals entering comparator A will be identical; as a result, the signals cancel and cause the output of comparator A to be zero. The net effect is to completely eliminate the outer feedback path; this simplification is shown in Fig. 17–27b. R + + − M Gc (s) Gp C − G(s) s e− D − + C1 Comparator A (a) R + Gc (s) − C 1 M Gp(s) C G(s) (b) FIGURE 17–27 (a) Dead-time compensation (Smith predictor) block diagram; (b) equivalent diagram for part (a) when G p ⫽ G (s)e⫺t D s . The system of Fig. 17–27b is now much easier to control because the transport lag is not present in the loop. Of course, in the real system the transport lag is still present; we have eliminated it in a mathematical sense from the feedback path by the additional feedback paths of Fig. 17–27a and the assumption that the process transfer function Gp (s) can be modeled exactly as shown in Eq. (17.13). To achieve the simplification suggested by Fig. 17–27b, we must now face reality and realize that the signal C1 in Fig. 17–27b is not available to feed back. Only the signal C can be measured and fed back to the controller. In terms of controller hardware implementation, the diagram of Fig. 17–27a is redrawn in Fig. 17–28a to show which portion of the diagram will be implemented with controller hardware. Figure 17–28b, which is another way to represent Fig. 17–28a, is a form sometimes presented in the literature for dead-time compensation. The reader may legitimately ask whether hardware exists to actually implement what is shown within the dotted lines in Fig. 17–28. Until the appearance of microprocessor-based controllers, the answer was no. However, today many commercially available controllers provide dead-time compensation [exp(⫺ tD s)] and G(s) in the form of a first-order lag [1/(ts ⫹ 1)]. cou9789x_ch17_351-390.indd 374 8/22/08 6:13:28 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 375 The recommended procedure for applying dead-time compensation is as follows: 1. Model Gp (s) by using a first-order plus dead-time (FOPDT) model 1 e⫺t D s ts ⫹ 1 2. In this step, we have chosen G(s) of Fig. 17–28a to be first-order. Many processes in chemical engineering can be modeled by a first-order lag with dead time. By means of appropriate hardware, implement the controller portion of Fig. 17–28a or b. If Gp (s) can be exactly modeled by a first-order process with dead time, the response of the control system in Fig. 17–28 will be equivalent to the response obtained for the system in Fig. 17–27b in which the loop involves the control of a first-order process. In most practical situations, there will be some mismatch between Gp (s) and its first-order with dead-time model. The greater the mismatch, the greater the deterioration in control response from the ideal situation of Fig. 17–27b. The application of the dead-time compensation technique and the effect of mismatch between Gp (s) and its model will be illustrated in Example 17.4. Controller R + + − M Gc(s) Gp(s) C Gp(s) C − (1 − e − Ds) G(s) (a) Controller R + + − + + M Gc(s) − G(s) G(s) e − Ds (b) FIGURE 17–28 Hardware implementation of dead-time compensation. cou9789x_ch17_351-390.indd 375 8/22/08 6:13:28 PM Confirming Pages 376 PART 5 PROCESS APPLICATIONS Example 17.4. Dead-time compensation. Consider the control system shown in Fig. 17–29 in which the process is fourth-order; thus 1 4 G p (s ) ⫽ s ⫹ 1 In a practical situation, we would not know the transfer function of the process. In this example, we have taken the process model to be fourth-order to provide a system sufficiently complex to show considerable transfer lag. R + Kc Gp = 1 s+1 4 C − FIGURE 17–29 Control system for Example 17.4. One can show for the system in Fig. 17–29 that the ultimate gain and the corresponding period are K cu ⫽ 4 . 0 and Pu ⫽ 2p. Using the Ziegler-Nichols rules, one gets Kc ⫽ 2.0. The response for a unit-step change in set point for Kc ⫽ 2 is shown in curve I of Fig. 17–32. Notice that the decay ratio is about 14 . We now use the dead-time compensation method to control the process in Fig. 17–29. If we fit the step response of (s ⫹ 1)⫺4 to a first-order with dead-time model, we obtain 1 . 03 e⫺1.79 s 2 . 62s ⫹ 1 This model was obtained from a unit-step response using a least-squares fit procedure. We can now draw the diagram for the dead-time compensation system as shown in Fig. 17–30. Gp(s) 1 R= s + + − 1 s+1 Kc − 1−e−1.79s 4 C 1.03 2.62s + 1 FIGURE 17–30 Dead-time compensation for Example 17.4. cou9789x_ch17_351-390.indd 376 8/22/08 6:13:29 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 377 The system shown in Fig. 17–30 was simulated by computer to compare the responses of the two control systems as shown in Fig. 17–32. The Simulink block diagram for the simulation is shown in Fig. 17–31. simout To workspace Gain +− Step +− −+ Transfer Fcn1 1.03 2.62s + 1 Transport delay = 1.79 Step1 +− +− 1 s4 + 4s3 + 6s2 + 4s + 1 4 Transfer Fcn Scope 1.03 2.62s + 1 Transfer Fcn2 2 Gain1 1 s4 + 4s3 + 6s2 + 4s + 1 Transfer Fcn3 FIGURE 17–31 Simulink model for Example 17.4. Using a Kc of 2.0 (the Ziegler-Nichols value) for the conventional control, we see from curve I that the response is quite oscillatory and has an offset of 0.333 as required for this value of gain. Using a Kc of 4.0 for the dead-time compensation, we see that the response is less oscillatory and the offset is 0.20. Note that if a Kc of 4.0 were applied to the conventional control system, the system would be on the verge of instability since a Kc of 4.0 is the ultimate gain. 1.0 0.90 II Dead-time compensation, Kc = 4 0.80 0.70 C 0.60 I Conventional control, Kc = 2 0.50 0.40 0.30 0.20 0.10 0.00 t 0 15 30 FIGURE 17–32 Comparison of response for conventional control with response for dead-time compensation for Example 17.4. cou9789x_ch17_351-390.indd 377 8/22/08 6:13:30 PM Confirming Pages 378 PART 5 PROCESS APPLICATIONS In conclusion, the dead-time compensation has permitted the use of a higher value of Kc, reduced the offset, and produced a less oscillatory response. The dead-time compensation response shown in Fig. 17–32 can be improved by adding integral action to the controller and tuning the controller parameters. To successfully apply dead-time compensation to the control of a process, one must have an accurate model of the process, such as a first-order with dead-time model. The parameters in this model (t and tD) can be considered as controller parameters along with the controller parameters of Gc (s). For the case of dead-time compensation with proportional control in Example 17.4, we actually have three controller parameters: Kc, tD, and t. If the process dynamics [Gp (s)] changes, all three parameters may need adjustment to achieve good control. 17.5 INTERNAL MODEL CONTROL Internal model control (IMC), which is based on an accurate model of the process, leads to the design of a control system that is stable and robust. A robust control system is one that maintains satisfactory control in spite of changes in the dynamics of the process. In applying the IMC method of control system design, the following information must be specified: • • • • Process model Model uncertainty Type of input (step, ramp, etc.) Performance objective (integral square error, overshoot, etc.) In many industrial applications for control systems, none of the above items is available, with the result that the system usually performs in a less than optimum manner. Determining the mathematical model and its uncertainty can be a difficult task. When the process is not sufficiently understood to obtain a mathematical model by applying fundamental principles, one must obtain a model experimentally. A discussion on the modeling of a process is presented in Chap. 18. The choice of a performance objective is subjective and often arbitrary. In the IMC method, the integral square error is implied. A simple description of the IMC method will be presented here. The interested reader is advised to consult the book by Morari and Zafiriou (1989) for a full treatment of internal model control. The literature on IMC is difficult to understand without a good foundation in control theory and mathematics. A full treatment of IMC is beyond the scope of this text. It is hoped that the simple treatment given here will stimulate interest in this important area of process control. Internal Model Control Structure A block diagram of an IMC system is shown in Fig. 17–33a. Notice that the diagram is similar to the diagram for the Smith predictor method shown in Fig. 17–27a. In this diagram, G is the transfer function of the process and Gm is the model of the process. Although G and Gm are called the transfer functions of the process, they actually include cou9789x_ch17_351-390.indd 378 8/22/08 6:13:30 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 379 the valve and the process. The transfer function of the measuring element is taken as 1.0. The portion of the diagram that is implemented by the computer includes the IMC controller and the model; this portion is surrounded by the dotted boundary. To compare the IMC structure of Fig. 17–33a with the conventional control structure, the diagram of Fig. 17–33a has been rearranged as shown in Fig. 17–33b. For convenience, the transfer function through which the load U passes has been omitted. We show only the output from the load block (U1). Computer G1 IMC controller Comparator1 + Set point R Load U M GI Process + + G U1 C − Comparator 2 + Model − Gm B (a) U1 + R + − M GI G + + C + Gm (b) U1 Gc R + GI − M 1 − GIGm G + C (c) FIGURE 17–33 Internal model control structures: (a) basic structure; (b) alternate structure; (c) structure equivalent. We may use the structure in Fig. 17–33b to relate the IMC controller to the conventional controller. Replacing the inner loop of Fig. 17–33b with a single block gives the structure shown in Fig. 17–33c. Since this structure is the conventional single-loop cou9789x_ch17_351-390.indd 379 8/22/08 6:13:30 PM Confirming Pages 380 PART 5 PROCESS APPLICATIONS control structure, we can identify the single controller block as Gc. After one designs the IMC controller (GI) by the method to be described, one can determine the equivalent conventional controller Gc by the relation Gc ⫽ GI 1 ⫺ GI Gm (17.14) For the structure shown in Fig. 17–33a, one can show that C ⫽ U1 ⫹ GGI ( R ⫺ U1 ) 1 ⫹ GI ( G ⫺ Gm ) (17.15) If the model exactly matches the process (i.e., Gm ⫽ G), the only signal entering comparator 1 in Fig. 17–33a is U1. (The signals from G and Gm are equal and cancel each other in going through comparator 2.) Since U1 is not the result of any processing by the transfer functions in the forward loop, U1 is not a feedback signal but an independent signal that is equivalent to R in its effect on the output C. In fact, there is no feedback when G ⫽ Gm, and we have an open-loop system as shown in Fig. 17–34. In this case the stability of the control system depends only on GI and Gm. If GI and Gm are stable, the control system is stable. U1 + + GI R Gm + C − U1 FIGURE 17–34 IMC structure when model matches the process (Gm ⫽ G). Ideally, we should like to have C track R without lag when only a set point change occurs (i.e., U1 ⫽ 0). For this to occur, we see from Fig. 17–34 or Eq. (17.15) that GI G ⫽ 1, or since G ⫽ Gm, we may write GI Gm ⫽ 1. Solving for GI gives GI ⫽ 1 Gm (17.16) Equation (17.16) simply states that the IMC controller should be the inverse of the transfer function of the process model. Keep in mind that Eq. (17.16) is based on the assumption that the model exactly matches the process. For the case of only a change in load U1 (i.e., R ⫽ 0), we should like to have the output C remain unchanged (i.e., C ⫽ 0). For this to occur, we see again from Fig. 17–34 or Eq. (17.15) that GI Gm ⫽ 1; this leads to the same result as given by Eq. (17.16). cou9789x_ch17_351-390.indd 380 8/22/08 6:13:31 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 381 Even if there is no mismatch between the model and the process, the application of Eq. (17.16) will usually lead to a transfer function that cannot be implemented because it will be unstable, requires prediction, or requires pure differentiation. For example, if Gm ⫽ 1/(ts ⫹ 1), the application of Eq. (17.16) gives GI ⫽ ts ⫹ 1 This result is equivalent to an ideal PD controller, which cannot be implemented because of the derivative term. If Gm ⫽ e⫺ts/(t1s ⫹ 1), we obtain GI ⫽ (t 1s ⫹ 1) ets The term ets, which represents prediction, cannot be implemented. If Gm ⫽ (1 ⫺ s)/ [(1 ⫹ s)(ts ⫹ 1)], then GI ⫽ (1 ⫹ s ) (ts ⫹ 1) 1⫺s The term 1 ⫺ s in the denominator means that a pole is in the right half-plane, which leads to an unstable controller. With such difficulties of implementation of the internal model controller, one might ask if any practical result can be obtained. These difficulties can be overcome by application of the following simplified procedure. Design of IMC Controllers In using these rules, only a step change in disturbance is considered. The procedure for disturbances other than a step response is more complicated and beyond the scope of the limited discussion presented here. 1. Separate the process model Gm into two terms Gm ⫽ Gma Gmm (17.17) where Gma is a transfer function of an all-pass filter. An all-pass filter is one for which Gma ( jw ) ⫽ 1 for all w. Examples are e⫺t d s and (1 ⫺ s)/(1 ⫹ s). The Gmm is a transfer function that has minimum phase characteristics. A system has nonminimum phase characteristics if its transfer function contains zeros in the right half-plane or transport lags, or both. Otherwise, a system has minimum phase characteristics. For a step change in disturbance (R ⫽ 1/s or U1 ⫽ 1/s), GI is determined by GI ⫽ 1 Gmm (17.18) For a disturbance other than a step change, obtaining GI is more complicated and the reader is referred to Morari and Zafiriou (1989). The results of applying Eq. (17.18) will yield a transfer function that is stable and does not require prediction; however, it will have terms that cannot be implemented because they require pure differentiation (e.g., ts ⫹ 1). cou9789x_ch17_351-390.indd 381 8/22/08 6:13:31 PM Confirming Pages 382 PART 5 PROCESS APPLICATIONS 2. To obtain a practical IMC controller, one multiplies GI in step 1 by a transfer function of a filter f(s). The simplest form recommended by Morari and Zafiriou is given by 1 (17.19) f (s ) ⫽ ( λs ⫹ 1)n where l is a filter parameter and n is an integer. The practical IMC controller GI can now be expressed as f GI ⫽ (17.20) Gmm 3. The value of n is selected large enough to give a result for GI that does not require pure differentiation. For the simple treatment of IMC design presented here, l will be considered as a tunable parameter. In the full treatment of IMC given by Morari and Zafiriou, l can be related to the model uncertainty. In practice, model uncertainty may not be available, in which case one is forced to treat l as a tunable parameter. If one wants to obtain the conventional controller transfer function Gc, use is made of Eq. (17.14), with GI obtained from Eq. (17.20). For many simple process models, Gc turns out to be equivalent to a PID controller multiplied by a first-order transfer function; thus 1 1 Gc ⫽ K c 1 ⫹ t D s ⫹ t I s t 1s ⫹ 1 (17.21) where Kc, tD, tI, and t1 are functions of l and the parameters in GI and Gm. The examples that follow will illustrate the application of this simplified procedure for designing an IMC controller. Example 17.5. Internal model control. Design an IMC controller for the process which is first-order: Gm ⫽ K ts ⫹ 1 For this case Gma ⫽ 1 and Gmm ⫽ K / (ts ⫹ 1) . Applying Eq. (17.18) gives GI ⫽ 1 ts ⫹ 1 ⫽ Gmm K To be able to implement this transfer function, let f(s) ⫽ 1/(l s ⫹ 1). The IMC controller becomes GI ⫽ cou9789x_ch17_351-390.indd 382 1 ts ⫹ 1 K ls ⫹ 1 8/22/08 6:13:32 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 383 This result is a lead-lag transfer function that can be easily implemented with computer-based controllers. We may now obtain Gc from Eq. (17.14). Gc ⫽ GI 1 ⫺ GI Gm Introducing the expressions for GI and Gm into this equation gives ts ⫹ 1 K ( l s ⫹ 1) 1 ts ⫹ 1 t Gc ⫽ ⫽ ⫽ 1⫹ ts ⫹ 1 K ts lK Kl s 1⫺ K ( l s ⫹ 1) ts ⫹ 1 This result is in the form of a PI controller: 1 Gc ⫽ K c 1 ⫹ t I s Kc ⫽ t lK tI ⫽ t Although this design procedure results in the equivalence of a PI controller, only one parameter (l) must be used to tune the controller. This is a distinct advantage over the use of a conventional controller in which both Kc and tI must be tuned. Example 17.6. Internal model control. Design an IMC controller for a process which is first-order with transport lag G⫽K e⫺t d s ts ⫹ 1 In the model of this process, use as an approximation to the transport lag a firstorder Padé approximation [see Eq. (7.48)]; thus e⫺t d s ⫽ 1 ⫺ (t d / 2 ) s 1 ⫹ (t d / 2 ) s The model becomes Gm ⫽ K 1 ⫺ (t d / 2 )s 1 1 ⫹ ( t d / 2 ) s ts ⫹ 1 For this model, Gma ⫽ 1 ⫺ (t d / 2 )s 1 ⫹ (t d / 2 )s (an all-pass filter ) and K ts ⫹ 1 Following the same steps as used in Example 17.5, we obtain for the IMC controller 1 ts ⫹ 1 GI ⫽ K ls ⫹ 1 Gmm ⫽ cou9789x_ch17_351-390.indd 383 8/22/08 6:13:33 PM Confirming Pages 384 PART 5 PROCESS APPLICATIONS It is instructive to see the form Gc takes for this example. Applying Eq. (17.14) gives Gc ⫽ GI ⫽ 1 ⫺ GI Gm ts ⫹ 1 K ( l s ⫹ 1) 1⫺ K 1 ⫺ (t d / 2 ) s ts ⫹ 1 K ( l s ⫹ 1) 1 ⫹ (t d / 2 ) s (ts ⫹ 1) This may be reduced algebraically to the form given by Eq. (17.21) with Kc ⫽ 2t ⫹ t d 2 (l ⫹ t d ) tI ⫽ t ⫹ tD ⫽ t1 ⫽ td 2 tt d 2t ⫹ t d lt d 2 (l ⫹ t d ) The response of this first-order with transport lag system for several values of l and for K ⫽ 1, t ⫽ 1, and t d ⫽ 1 is given in Fig. 17–35. The values of Kc, tI, td, and t1 obtained from the above relations are shown in Table 17.2. Notice that once a model is accepted, the tuning of the modified IMC controller [Eq. (17.21)] depends only on the choice of l. For the range of l used, Fig. 17–35 shows that the step response is only slightly oscillatory for all values of l, and the fastest response is for l ⫽ 0.5. Also notice that l affects only Kc and t1. This example shows that the design of a controller by the IMC method is a straightforward procedure and leads to a controller that requires the adjustment of only one parameter, l. 2.0 1.80 1.60 = 0.5 1.0 1.5 1.40 C 1.20 1.00 2.0 0.80 0.60 0.40 0.20 0.00 t 0 5 10 FIGURE 17–35 Response for IMC designed controller of Example 17.6. cou9789x_ch17_351-390.indd 384 8/22/08 6:13:34 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 385 TABLE 17.2 IMC derived controller settings for Example 17.6 l Kc tI tD t1 0.5 1.0 1.5 0.33 0.167 1.0 0.75 1.5 0.33 0.25 1.5 0.60 1.5 0.33 0.30 2.0 0.5 1.5 0.33 0.33 It is instructive to compare the response for the IMC derived controller with the response for a PI controller using Ziegler-Nichols (Z-N) settings. The responses, which are given in Fig. 17–36, show that for this particular example the controller using Z-N settings produces a response with less overshoot and a higher frequency of oscillation than the controller designed by the IMC method. These two examples show clearly how the parameters of the conventional controller Gc are related to the parameters of the model and the filter. The treatment of internal model control presented here has been limited to single-input, single-output continuous systems for which the disturbance is a step change. Furthermore, we have not discussed the use of model uncertainty in selecting the filter parameters. Internal model control has been extended to sampled-data control systems and to multiple-input, multiple-output systems. IMC is an approach to the design of control systems that considers the process model as an essential part of the control system design. Computer-based controllers have the capability of implementing many of the control algorithms designed by the IMC method. There is no longer a need to be tied to the classical control algorithms. 2 1.8 1.6 1.4 C 1.2 I 1.0 II .8 .6 .4 .2 0.0 0 5 t FIGURE 17–36 Comparison of response for IMC controller and conventional controller for Example 17.6: Curve I: IMC derived controller with l ⫽ 1.0; curve II: PI controller with Ziegler-Nichols settings (Kc ⫽ 1.02 and tI ⫽ 2.84). cou9789x_ch17_351-390.indd 385 8/22/08 6:13:36 PM Confirming Pages 386 PART 5 PROCESS APPLICATIONS SUMMARY In this chapter, we have examined five advanced control strategies. The first three on cascade control, feedforward-feedback control, and ratio control are advanced only in the sense that each strategy is more complex than the single-loop systems we have encountered up to this chapter. These three strategies are used extensively in industry, and computer-based controllers can implement them easily. The other strategies, on Smith prediction and internal model control (IMC), are less likely to be used in industry and are closely related in their block diagram structure. Three of the strategies—feedforward-feedback, Smith prediction, and IMC—are dependent on accurate models of the processes for their application. Cascade control is especially useful in reducing the effect of a load disturbance that is located far from the control variable and that moves through the system slowly. The presence of the inner control loop reduces the lag in the outer loop, with the result that the cascade system responds more quickly to a load disturbance. If a particular load disturbance occurs frequently, the quality of control can often be improved by applying feedforward control. Ideally the transfer function of the feedforward controller is obtained from knowledge of the model of the process. In cases where the feedforward controller transfer function requires prediction (for example, tf s ⫹ 1), one must be satisfied with an approximation of the feedforward controller, which takes the form of a lead-lag transfer function. When a model of the process does not exist, the feedforward controller can be tuned after doing some open-loop step tests that relate the control variable to the load disturbance. To provide for load disturbances that cannot be measured or anticipated, feedforward control is always combined with feedback control in a practical situation. Ratio control is widely used in industry in the blending of two component streams (A and B) to produce a mixed stream of desired composition (i.e., ratio of components). Ratio control is essentially a flow control problem in which the flow measurement of stream A (the wild stream) is used to compute the set point for the flow of stream B so that the desired ratio of components will be obtained. The Smith predictor control scheme (dead-time compensation) was developed to improve the control of a system having a large transport lag. The method is based on a model of the process that is first-order with dead time. By introducing inner loops that contain elements of the transfer function of the model, the control system is transformed ideally to one without transport lag, a system that is much easier to control. This ideal situation occurs when the process and the model are in exact agreement. In reality, the success of the Smith predictor strategy depends on the degree of agreement between process and model. Internal model control resembles the Smith predictor strategy in terms of the structure of the block diagram. To apply the IMC method, one must have an accurate model of the process, the model uncertainty, the type of disturbance (step, ramp, etc.), and the performance objective (integral of square error). The method, which is based on a rigorous mathematical foundation, leads to an IMC controller that is the best that can be designed in terms of the performance objective. The IMC structure can be reduced to a conventional control structure in which the conventional controller is related to the IMC controller and the parameters of the model. For many simple processes with simple disturbance (impulse, step, etc.), the equivalent conventional controller based on the IMC design method turns out to be the equivalent of a PID controller. cou9789x_ch17_351-390.indd 386 8/22/08 6:13:36 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 387 PROBLEMS 17.1. (a) Obtain Gf for the feedforward-feedback system shown in Fig. P17–1 so that C does not change when a disturbance in Ci occurs. Would there be any problem in implementing this Gf? (b) If Gf is to be a lead-lag transfer function T1s ⫹ 1 T2 s ⫹ 1 determine T1 and T2 by the Foxboro rule. How do you determine whether lead or lag is to predominate? Use tp ⫽ 1.0 in the Foxboro rule. (c) When feedforward-feedback control is present, sketch the response C(t) when Ci ⫽ 1/s and when Gf from part (a) is used. (d) Repeat part (c) when Gf from part (b) is used. Only a rough sketch that suggests the transient response is expected in this case. (e) Determine C(t) when Ci ⫽ 1/s, Gf ⫽ ⫺1, and the feedback loop is broken at A-A. Obtain the numerical value of C(t) at t ⫽ 0.5, 1.0, and 1.5. Ci = 1 s+1 1 s Gf R=0 + K=1 − A + A + + 1 (s + 1)2 + C FIGURE P17–1 cou9789x_ch17_351-390.indd 387 8/22/08 6:13:37 PM Confirming Pages CHAPTER 17 CAPSULE SUMMARY CASCADE CONTROL Cascade control is used for improved disturbance rejection. It employs two controllers, a primary controller and a secondary controller. The primary controller provides the set point for the secondary controller. Primary set point + R − Disturbance L Secondary Final control Secondary controller element + process Tf + Gc2 Gv Gjacket Primary controller + Gc1 − Secondary process variable Primary process Gkettle To Gm1 Primary process variable Gm2 FEEDFORWARD CONTROL Feedforward control is usually used in conjunction with feedback control to improve disturbance rejection. A model of the process is used to predict the effect of a disturbance on the process. This information instructs the controller on how to preemptively take action to negate the effect of the disturbance. Ci Disturbance G1 Feedforward controller Gf Feedback controller R + Gc + + Process Gp + + C − 388 cou9789x_ch17_351-390.indd 388 8/22/08 6:13:37 PM Confirming Pages CHAPTER 17 ADVANCED CONTROL STRATEGIES 389 RATIO CONTROL Ratio control is a control strategy used to control the blending of two components in a desired ratio. qA Fluid A Flow-measuring element Kr Set point Controller Flow-measuring element qB Fluid B PB = supply pressure SMITH PREDICTOR CONTROL Sometimes referred to as dead-time compensation, this control strategy is used to counteract the deleterious effects of dead time on process control. A model of the process is used to predict and negate the effect of dead time. Process Controller + + Gc(s) R − M Gp C − Process model Dead time G(s) s e− D + + C1 INTERNAL MODEL CONTROL Internal model control resembles the Smith predictor strategy in terms of the structure of the block diagram. To apply the IMC method, one must have an accurate model of the process, the model uncertainty, the type of disturbance (step, ramp, etc.), and the performance objective (integral of square error). The method, which is based on a rigorous mathematical foundation, leads to an IMC controller that is the best that can be designed in terms of the performance objective. The IMC structure can be reduced to a conventional control structure in which the conventional controller is related to cou9789x_ch17_351-390.indd 389 8/22/08 6:13:38 PM Confirming Pages 390 PART 5 PROCESS APPLICATIONS the IMC controller and the parameters of the model. For many simple processes with simple disturbance (impulse, step, etc.), the equivalent conventional controller based on the IMC design method turns out to be the equivalent of a PID controller. Computer G1 IMC controller Comparator1 Set point R Load U + GI M Process + + G U1 C − + Model Gm Comparator 2 − B cou9789x_ch17_351-390.indd 390 8/22/08 6:13:38 PM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION T he selection of a controller type (P, PI, PID) and its parameters (Kc, tI, tD) is intimately related to the model of the process to be controlled. The adjustment of the controller parameters to achieve satisfactory control is called tuning. The selection of the controller parameters is essentially an optimization problem in which the designer of the control system attempts to satisfy some criterion of optimality, the result of which is often referred to as “good” control. The process of tuning can vary from a trial-and-error attempt to find suitable control parameters for good control to an elaborate optimization calculation based on a model of the process and a specific criterion of optimal control. In many applications, there is no model of the process, and the criterion for good control is only vaguely defined. A typical criterion for good control is that the response of the system to a step change in set point or load have minimum overshoot and a one-quarter decay ratio. Other criteria may include minimum rise time and minimum settling time. In the first part of this chapter, some of the widely used tuning rules for continuous controllers will be presented. In the second part of the chapter, methods for determining the model of a process from experimental tests will be described. Determining the model of a process experimentally is referred to as process identification. 18.1 CONTROLLER TUNING Before we present tuning rules, some discussion of the effect of each mode in a PID controller on the transient response of a controlled process will be instructive. Selection of Controller Modes Consider a typical loop as shown in Fig. 18–1 in which the process is second-order and the measuring element is a transport lag. (The transfer function of the valve is taken 391 cou9789x_ch18_391-422.indd 391 8/22/08 10:17:31 AM Confirming Pages 392 PART 5 PROCESS APPLICATIONS U R + + Gc 1 (10s + 1)(s + 1) + − B C e−s FIGURE 18–1 Typical control system used to study the effect of controller modes on load responses shown in Fig. 18–2. as 1.) Load responses for this process for four types of controllers (P, PD, PI, PID) are shown in Fig. 18–2. For each response curve, the process was subjected to a unit-step change in load (U ⫽ 1/s), and the controller parameters were selected by tuning rules to be presented later. Regardless of the specific tuning rules used, the responses shown in Fig. 18–2 are typical of well-tuned controllers for systems found in industry. The nature of the response for each type controller will now be described. (The reader should also refer to Figs. 9–10 and 16–18 to reinforce this discussion.) C 0.20 PD P 0.10 PI 0.00 −0.10 PD 0 10 t 20 FIGURE 18–2 Load response of a typical control system using various modes of control (process shown in Fig. 18–1). PROPORTIONAL CONTROL. As shown in Fig. 18–2, proportional control produces an overshoot followed by an oscillatory response, which levels out at a value that does not equal the set point; this ultimate displacement from the set point is the offset. PROPORTIONAL-DERIVATIVE CONTROL. For this case the response exhibits a smaller overshoot and a smaller period of oscillation compared to the response for proportional control. The offset that still remains is less than that for proportional control. PROPORTIONAL-INTEGRAL CONTROL. In this case, the response has about the same overshoot as proportional control, but the period is larger; however, the response returns to the set point (offset ⫽ 0) after a relatively long settling time. The most beneficial influence of the integral action in the controller is the elimination of offset. cou9789x_ch18_391-422.indd 392 8/22/08 10:17:32 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 393 PROPORTIONAL-INTEGRAL-DERIVATIVE CONTROL. As one might expect, the use of PID control combines the beneficial features of PD and PI control. The response has lower overshoot and returns to the set point more quickly than the responses for the other types of controllers. From the nature of the responses just described, we can make the following generalizations. Integral action, which is present in PI and PID controllers, eliminates offset. The addition of derivative action speeds up the response by contributing to the controller output a component of the signal that is proportional to the rate of change of the process variable. For simple, low-order (first- or second-order) processes that can tolerate some offset, P or PD control is satisfactory. For processes that cannot tolerate offset and are of low order, PI control is required. For processes that are of high order (those with transport lag or many first-order lags in series), PID control is needed to prevent large overshoot and long settling time. There is probably little justification to select a P or PD controller for most processes. The PI controller is often the choice because it eliminates offset and requires only two parameter adjustments. Tuning a PID controller is more difficult because three parameters must be adjusted. The presence of derivative action can also cause the controller output to be very jittery if there is much noise in the signals. We now turn our attention to some of the criteria for good control that are used to judge whether a control system is well tuned. Criteria for Good Control Before we can be satisfied with the response of a control system for a choice of control parameters, we must have some concept of what we want as an ideal response. Most operators of processes know what they want in the form of a response to a change in set point or load. For example, a response that gives minimum overshoot and 14 decay ratio is often considered as a satisfactory response. In many cases, tuning is done by trial and error until such a response is obtained. To compare different responses that use different sets of controller parameters, a criterion that reduces the entire response to a single number, or a figure of merit, is desirable. The figure of merit provides a means of “keeping score” for the different control parameters, and as we shall see, the low “score” generally wins. It is dangerous, however, to rely solely on the score to determine the best choice for the control parameters. The control system designer should examine the nature of the response in conjunction with the requirements for the process to determine the “best” choice of settings. One criterion that is often used to evaluate a response of a control system is the integral of the square of the error (ISE) with respect to time. The definition of ISE is as follows: Integral of the square of the error (ISE) ISE ⫽ ⬁ 2 ∫0 e dt (18.1) where e is the usual error (i.e., set point – control variable). For a stable system for which there is no offset [i.e., e(⬁) ⫽ 0], Eq. (18.1) produces a single number as a figure of merit. cou9789x_ch18_391-422.indd 393 8/22/08 10:17:33 AM Confirming Pages 394 PART 5 PROCESS APPLICATIONS The objective of the designer is to obtain the minimum value of ISE by proper choice of control parameters. A response that has large errors and persists for a long time will produce a large ISE. For the cases of P and PD control, where offset occurs, the integral given by Eq. (18.1) does not converge. In these cases, one can use a modified integrand, which replaces the error r(t) ⫺ c(t) by c(⬁) ⫺ c(t). Since c(⬁) ⫺ c(t) does approach zero as t goes to infinity, the integral will converge and serve as a figure of merit. Two other criteria often used in process control are defined as follows: Integral of the absolute value of error (IAE) IAE ⫽ ⬁ ∫0 | e | dt (18.2) Integral of time-weighted absolute error (ITAE) ITAE ⫽ ⬁ ∫0 | e | t dt (18.3) Each of the three figures of merit, given by Eqs. (18.1), (18.2), and (18.3), has different purposes. The ISE will penalize (i.e., increase the value of ISE) the response that has large errors, which usually occur at the beginning of a response, because the error is squared. The ITAE will penalize a response that has errors that persist for a long time. The IAE will be less severe in penalizing a response for large errors and treat all errors (large and small) in a uniform manner. The ISE figure of merit is often used in optimal control theory because it can be used more easily in mathematical operations (e.g., differentiation) than the figures of merit, which use the absolute value of error. In applying the tuning rules to be discussed in the next section, these figures of merit can be used in comparing responses that are obtained with different tuning rules. Example 18.1. For the control system shown in Fig. 18–1, determine the ISE, ITAE, and IAE for a unit-step load disturbance. Let the controller be a PI controller with Kc ⫽ 6 and tI ⫽ 4. Simulink can be used to easily simulate this process and calculate the figures of merit. The Simulink block diagram is shown in Fig. 18–3. The response generated using the model of Fig. 18–3 of the process for these controller settings is shown in Fig. 18–4. The figures of merit alone tell us nothing unless they are compared with figures for other controller settings. Remember, the goal is generally to obtain the lowest value. Once we have a Simulink model such as this prepared, it is relatively easy to check a variety of controller settings. We compare some other controller settings and their figures of merit shortly. 18.2 TUNING RULES Ziegler-Nichols (Z-N) Rules These rules were first proposed by Ziegler and Nichols (1942), who were engineers for a major control hardware company in the United States (Taylor Instrument Co.). Based cou9789x_ch18_391-422.indd 394 8/22/08 10:17:34 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 395 1 Out1 ITAE 1 s Product Integrator 2 Out2 1 u s IAE Absolute value Integrator2 3 Out3 1 u2 s ISE Error squared Integrator1 19.09 Clock Kc = 6, +− I=4 +− PID Set point, R = 0 ITAE value 1.515 IAE value 0.1357 ISE value 1 10s2 + 11s +1 Process Scope Load, U = 1/s Time Delay = 1 FIGURE 18–3 Simulink model for Example 18.1 0.25 0.2 0.15 0.1 0.05 0 −0.05 −0.1 0 10 20 30 40 50 60 70 80 90 100 FIGURE 18–4 Process response for Example 18.1 on their experience with the transients from many types of processes, they developed a closed-loop tuning method still used today in one form or another. The method is described as a closed-loop method because the controller remains in the loop as an active controller in automatic mode. This closed-loop method will be contrasted with an cou9789x_ch18_391-422.indd 395 8/22/08 10:17:34 AM Confirming Pages 396 PART 5 PROCESS APPLICATIONS open-loop tuning method to be discussed later. We have already discussed the ZieglerNichols rules in Chap. 16 as a natural consequence of our study of frequency response. Ziegler and Nichols did not suggest that the ultimate gain Kcu and ultimate period Pu be computed from frequency response calculations based on the model of the process. They intended that Kcu and Pu be obtained from a closed-loop test of the actual process. When the rules were first proposed, frequency response methods and process models were not generally available to the control engineers. The rules are presented below and are in the form that one would use for actual application to a real process. 1. 2. After the process reaches steady state at the normal level of operation, remove the integral and derivative modes of the controller, leaving only proportional control. On some PID controllers, this requires that the integral time tI be set to its maximum value and the derivative time tD to its minimum value. On computer-based controllers, the integral and derivative modes can be removed completely from the controller. Select a value of proportional gain Kc, disturb the system, and observe the transient response. If the response decays, select a higher value of Kc and again observe the response of the system. Continue increasing the gain in small steps until the response first exhibits a sustained oscillation. The value of gain and the period of oscillation that correspond to the sustained oscillation are the ultimate gain Kcu and the ultimate period Pu. Some very important precautions to take in applying this step of the tuning method are given in the next section. 3. From the values of Kcu and Pu found in step 2, use the Ziegler-Nichols rules given in Table 18.1 to determine controller settings (Kc, tI, tD). (This table is the same as Table 16.1.) Although variations in the tuning rules given in Table 18.1 are used by industry, the same approach of using Kcu and Pu to obtain controller parameters is used. The Ziegler-Nichols rules generally provide conservative (and safe) controller settings. The Z-N settings should be considered as only approximate settings for satisfactory control. Fine-tuning of the controller settings is usually required to get an improved control response. TABLE 18.1 Ziegler-Nichols controller settings Gc(s) Kc Proportional (P) Kc 0.5Ku 1 Kc 1 ⫹ t I s 0.45Ku Pu 1.2 1 Kc 1 ⫹ ⫹ t D s tIs 0.6Ku Pu 2 Proportional-integral (PI) Proportional-integral-derivative (PID) cou9789x_ch18_391-422.indd 396 sI Type of control sD Pu 8 8/22/08 10:17:35 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 397 The experimental determination of Kcu and Pu described in step 2 can be replaced by a computation using frequency response methods if an accurate model of the process, valve, and measuring element is known. This type of calculation was done in Chap. 16. PRECAUTIONS TO TAKE IN APPLYING THE Z-N METHOD. Some discussion is needed to avoid some pitfalls in applying step 2 of the Z-N method to obtain Kcu and Pu. These precautions are concerned with the type and size of the disturbance that induces the response and with the avoidance of using a limit cycle as the indication that the system is on the threshold of instability. The simplest way to introduce a disturbance is to move the set point away from the control variable for a short time and then return the set point to its original value. This procedure, which is equivalent to introducing a pulse function in the error, causes the system to respond and yet stay within a narrow band surrounding the normal operating point of the process. An alternate type of disturbance would be to introduce a small step change in set point. If step changes in set point are used to induce transients, the successive step changes should alternate around the normal operating point of the process. It is also important to make the disturbance as small as possible, especially as the gain of the controller is increased, so that the valve and other components do not exceed their physical limits. When the valve moves to its limits during a closed-loop transient, we say that the valve saturates. Under these conditions, a sustained oscillation occurs, which is called a limit cycle. The limit cycle that is caused by saturation is a nonlinear phenomenon, which will be covered in Chap. 25 on nonlinear control. If a limit cycle occurs, the gain that produces it and the period of the cycle should not be used in the Ziegler-Nichols rules. Since the limit cycle will appear to the observer to be the same as a sustained oscillation when the system is on the verge of instability, the novice will often mistakenly use the information derived from the limit cycle (controller gain and period) to obtain controller settings. A simple way to know if one has a limit cycle is to observe the swing in pressure to the valve. If the limits of the valve (e.g., 3 to 15 psig) are reached repeatedly during the oscillatory response, one has a limit cycle and the controller gain and period should not be used to determine controller settings. It is for this reason step 2 states that Kc should be increased in small steps until the response first exhibits a sustained oscillation. To appreciate the use of step 2 of the tuning method, one should have some laboratory experience in tuning a real process, or at least a computer simulation of a process. The experienced operator can develop some shortcuts to finding the ultimate gain and ultimate period. Cohen and Coon (C-C) Rules The next method of tuning to be discussed is an open-loop method, in which the control action is removed from the controller by placing it in manual mode and an open-loop transient is induced by a step change in the signal to the valve. This method was proposed by Cohen and Coon (1953) and is often used as an alternative to the Z-N method. Fig. 18–5 shows a typical control loop in which the control action is removed and the loop opened for the purpose of introducing a step change (M/s) to the valve. The step response is recorded at the output of the measuring element. The step change to the cou9789x_ch18_391-422.indd 397 8/22/08 10:17:35 AM Confirming Pages 398 PART 5 PROCESS APPLICATIONS valve is conveniently provided by the output from the controller, which is in manual mode. The response of the system (including the valve, process, and measuring element) is called the process reaction curve; a typical process reaction curve exhibits an S shape, as shown in Fig. 18–6. After we present the Cohen and Coon method of tuning, the basis for their recommendations will be discussed. The C-C method is summarized in the following steps: M/s R=0 U=0 + + Gv Gc − + Gp C Loop opened B H To recorder FIGURE 18–5 Block diagram of a control loop for measurement of a process reaction curve. 1. 2. 3. After the process reaches steady state at the normal level of operation, switch the controller to manual. In a modern controller, the controller output will remain at the same value after switching as it had before switching. (This is called “bumpless” transfer.) With the controller in manual, introduce a small step change in the controller output that goes to the valve and record the transient, which is the process reaction curve (Fig. 18–6). Draw a straight line tangent to the curve at the point of inflection, as shown in Fig. 18–6. The intersection of the tangent line with the time axis is the apparent transport lag Td; the apparent first-order time constant T is obtained from Bu (18.4) S where Bu is the ultimate value of B at large t and S is the slope of the tangent line. The steady-state gain that relates B to M in Fig. 18–5 is given by T ⫽ Bu (18.5) M 4. Using the values of Kp, T, and Td from step 3, the controller settings are found from the relations given in Table 18.2. Notice in Table 18.2 that all the controller settings are a function of the dimensionless group Td /T, the ratio of the apparent transport lag to the apparent time constant. Also Kc is inversely proportional to Kp. Kp ⫽ cou9789x_ch18_391-422.indd 398 8/22/08 10:17:36 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 399 B Bu Tangent line, slope = 0 0 Td Bu =S T t M input 0 0 t FIGURE 18–6 Typical process reaction curve showing graphical construction to determine first-order with transport lag model. TABLE 18.2 Cohen-Coon controller settings Type of control Parameter setting Proportional (P) Kc ⫽ 1 T Td 1 ⫹ K p Td 3T Proportional-integral (PI) Kc ⫽ 1 T 9 Td ⫹ K p Td 10 12T t I ⫽ Td Proportional-derivative (PD) Kc ⫽ 1 T 5 Td ⫹ K p Td 4 6T t D ⫽ Td Proportional-integral-derivative (PID) cou9789x_ch18_391-422.indd 399 Kc ⫽ 30 ⫹ 3Td / T 9 ⫹ 20Td / T 6 ⫺ 2Td / T 22 ⫹ 3Td / T 1 T 4 T ⫹ d K p Td 3 4T t I ⫽ Td 32 ⫹ 6Td / T 13 ⫹ 8Td / T t D ⫽ Td 4 11 ⫹ 2Td / T 8/22/08 10:17:37 AM Confirming Pages 400 PART 5 PROCESS APPLICATIONS The rationale for the C-C tuning method begins with the representation of the S-shaped process reaction curve by a first-order with transport lag model; thus Gp (s) ⫽ K pe⫺Td s Ts ⫹ 1 (18.6) Using the system expressed by Eq. (18.6), Cohen and Coon obtained by theoretical means the controller settings given in Table 18.2. Their computations required that the response have 1 decay ratio, minimum offset, minimum area under the load-response 4 curve, and other favorable properties. In applying the C-C tuning method, an important task is the graphical construction, shown in Fig. 18–6, which reduces the process reaction curve to first-order with the transport lag model given by Eq. (18.6). To understand the basis for the graphical procedure, consider the response of the transfer function of Eq. (18.6) to a step change in input; the resulting transient is shown in Fig. 18–7. After t ⫽ Td, the response is a firstorder response. The point of inflection of the curve in Fig. 18–7 occurs at t ⫽ Td, and the slope of the tangent line at this point is related to the time constant by the relation S ⫽ Bu T Solving for T gives the expression in Eq. (18.4). The response after t ⫽ Td, shown in Fig. 18–7, was also presented in Fig. 4–7. The attempt to model the process reaction curve by the method shown in Fig. 18–6 is crude and does not give a very good fit. Finding the point of inflection and drawing a tangent line at this point are quite difficult, especially if the data for the process reaction curve are not accurate and if they scatter. A better method for fitting the process reac- B Bu Tangent line, slope = 0 0 Td Bu T t M Input 0 0 t FIGURE 18–7 Step response for a first-order with transport lag model. cou9789x_ch18_391-422.indd 400 8/22/08 10:17:40 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 401 TABLE 18.3 Controller settings for the system of Fig. 18–1. Control type Parameter Closed-loop method (Z-N method) Open-loop method (C-C method) Kc Kc tI Kc tD Kc tI tD 6.4 5.8 5.6 11.4* 1.0* 7.7 3.4 1.6 8.1 7.0 4.4 9.8 0.43 10.5 3.9 0.59 P PI PD PID * Obtained by design for 30⬚ phase margin and maximum Kc. tion curve to a first-order with transport lag model is to perform a least-squares fit of the data. Some computer software, such as the LOOP-PRO package (see www.controlstation .com), provide an easy means of performing the fitting process. An example to be presented later will study the effect of the type of model-fitting procedure on the selection of controller parameters. To illustrate the two methods of controller tuning just presented, the system shown in Fig. 18–1 was simulated using Simulink as shown above in Example 18.1. Table 18.3 gives the values of the controller parameters obtained by applying both tuning methods (Z-N and C-C). Fig. 18–8 shows the resulting transients. Since the Z-N method does 0.20 0.15 0.10 0.05 0.00 −0.05 −0.10 0.20 0.15 0.10 0.05 0.00 −0.05 −0.10 Z- N c C- C Proportional 0 10 (a) c t 20 Z- N Proportional-integral C- C 0 10 (c) t 20 0.20 0.15 0.10 0.05 0.00 −0.05 −0.10 0.20 0.15 0.10 0.05 0.00 −0.05 −0.10 c C-C Z-N 0 C-C Proportional-derivative 10 (b) t 20 c Proportional-integral-derivative Z-N C-C 0 10 (d) t 20 FIGURE 18–8 Comparison of load responses for the system of Fig. 18–1 using controller settings obtained by the Ziegler-Nichols (Z-N) method and the Cohen-Coon (C-C) method. cou9789x_ch18_391-422.indd 401 8/22/08 10:17:41 AM Confirming Pages 402 PART 5 PROCESS APPLICATIONS not give a rule for a PD controller, the settings listed for a PD controller under the Z-N heading of Table 18.3 were obtained by using a theoretical frequency response calculation in which the design was based on 30⬚ phase margin and a maximum Kc. No general conclusions can be drawn about the relative merits of the two tuning methods from the results shown in Fig. 18–8, since these results apply to one specific example. About all that can be said is that for this specific example, both methods give reasonable first guesses of the control parameters. Example 18.2. For the control system shown in Fig. 18–1, shown again in Fig. 18–9, determine the “best” PI controller settings as indicated by each of the figures of merit (ISE, ITAE, and IAE). Compare the responses of those settings with the ones obtained from Z-N and C-C tuning. We can use the same Simulink diagram that we used in Example 18.1. This time, however, it will be advantageous to write a MATLAB m-file to vary the controller parameters to minimize a particular figure of merit. 1 Out1 ITAE 1 s Product Integrator 2 Out2 1 u s IAE Absolute value Integrator2 3 Out3 1 2 u s ISE Error squared Integrator1 Clock 19.09 ITAE value 1.515 IAE value 0.1357 ISE value P, I +− ++ PID Set point, R = 0 1 10s2 + 11s +1 Process Scope Load, U = 1/s Time delay = 1 FIGURE 18–9 Simulink model for Example 18.2. The Simulink block diagram is saved as fig18_1.mdl. The m-file that we use to call the Simulink diagram is called merit_score.m. We make use of the MATLAB optimization function FMINSEARCH to vary the controller parameters for us and to find the minimum figure of merit of interest. The comment statements in the m-file describe how it’s used. An example of the MATLAB command that performs the optimization is cou9789x_ch18_391-422.indd 402 8/22/08 10:17:41 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 403 [zn,fval,exitflag]=FMINSEARCH('merit_score',[5.8 1],OPTIMSET ('MaxIter',100)) m-file: merit_score.m function merit_score=merit_score(Varg); % this m-file is for use with the FMINSEARCH function to vary P & I to % minimize the ITAE, IAE or ISE for the control system in Fig. 18.1: % % issue the FMINSEARCH command from the MATLAB workspace % be sure that the PID controller in simulink model, fig18_1.mdl, % has the P and I values set to the variables P & I % also be sure to declare P and I global in the MATLAB workspace % the 5.8 is the initial guess for P and 1 for I (these are the ZN values) % %[zn,fval,exitflag]= FMINSEARCH('merit_score',[5.8 1],options) %fval is the final value of the objective function %set the max iterations with: %options = OPTIMSET('MaxIter',100) % If EXITFLAG is: % 1 then FMINSEARCH converged with a solution X. % 0 then the maximum number of iterations was reached. % %note that the ITAE is outport 1 on the simulink diagram, fig18_1 %IAE is outport 2 and ISE is outport 3 on fig18_1 %the column specified in the y variable is the outport. . . % y(:,1)=ITAE % y(:,2)=IAE % y(:,3)=ISE % global P I P=Varg(1); I=Varg(2); [t,x,y]=sim('fig18_1',50); merit_score=max(y(:,1)); The results of running this optimization to minimize each figure of merit for PI control are shown in Table 18.4 and Fig. 18–10. TABLE 18.4 Comparison of controller settings for Example 18.2 using different tuning methods Parameter Z-N method C-C method Minimum ITAE Minimum IAE Minimum ISE Kc tI 5.8 5.6 7.0 4.4 5.5 6.0 6.6 6.7 8.7 7.9 cou9789x_ch18_391-422.indd 403 8/22/08 10:17:41 AM Confirming Pages 404 PART 5 PROCESS APPLICATIONS 0.25 Min ITAE 0.2 Z-N 0.15 Min ISE Response 0.1 C-C 0.05 Min IAE 0 0 5 10 15 20 25 Time 30 35 40 45 50 FIGURE 18–10 Plot of system response for Example 18.2 using different tuning methods. The controller settings for all the settings are very similar. The C-C settings are the most oscillatory for this system. Z-N and minimum ITAE have the largest initial peak, but settle the most rapidly. Let’s look at some additional examples. Example 18.3. For the control system shown in Fig. 18–11, determine controller settings for a PI controller using the Z-N method and the C-C method. This problem will be instructive because the transfer function of the model is already in the form of first-order with transport lag, which is the form used by Cohen and Coon to derive their tuning rules. C-C method. Since the transfer function of the plant is in the form of Eq. (18.6), we obtain T and Td immediately without having to draw a tangent line through cou9789x_ch18_391-422.indd 404 8/22/08 10:17:42 AM Confirming Pages CHAPTER 18 + 1 s +1 1 Kc 1 + R = u(t) CONTROLLER TUNING AND PROCESS IDENTIFICATION 1s − e−s 405 C FIGURE 18–11 Process for Example 18.3. the point of inflection; i.e., T ⫽ 1 and Td ⫽ 1. We also observe from the block diagram that Kp ⫽ 1. Substituting these values into the appropriate equations of Table 18.2 gives Kc ⫽ T Td 1 1 0.9 ⫹ ⫽ 0 . 9 ⫹ ⫽ 0 . 9 8 3 K pTd 12T 1 12 and t I ⫽ Td 30 ⫹ 3 Td / T 30 ⫹ 3 ⫽ ⫽ 1 . 14 9 ⫹ 20Td / T 9 ⫹ 20 Using these values for Kc and tI, the step response shown in Fig. 18–12 was obtained by simulation. c C– C 1.50 1.00 Z– N 0.50 0.00 0 5 t 10 FIGURE 18–12 Response to unit-step change in set point for the system in Fig. 18–11 (Example 18.3). Z-N method. Application of the Bode criterion from Chap. 16 gives w co ⫽ 2 . 03 or K cu cou9789x_ch18_391-422.indd 405 2p ⫽ 3 . 09 w co ⫽ 2 . 26 Pu ⫽ 8/22/08 10:17:43 AM Confirming Pages 406 PART 5 PROCESS APPLICATIONS The details for obtaining these results will not be given here since this type calculation was covered in depth in Chap. 16. Applying the Z-N rules for PI control from Table 18.1 gives K c ⫽ 0 . 45 K cu ⫽ (0 . 45 )(2 . 26 ) ⫽ 1 . 02 and tI ⫽ Pu 3 . 09 ⫽ ⫽ 2 . 58 1.2 1.2 The step response for these controller settings is shown in Fig. 18–12. The ISE value for each response was calculated out to a sufficiently long time (10 units of time) for the integral to converge; the results are as follows: C-C response : ISE ⫽ 1 . 54 at t ⫽ 10 Z-N response : ISE ⫽ 1 . 49 at t ⫽ 10 Although the ISE values are nearly the same, the transient for the Z-N settings is better than the transient for the C-C settings. The Z-N transient has much less overshoot. The lesson to be learned from this example is that the comparison of two transients based on only one criterion (in this case, the ISE) may be misleading in the selection of the best transient. It is also important to judge the quality of a transient by its actual appearance. Note that for this example, in which there is a relatively large transport lag (Td ⫽ 1), much of the contribution to the ISE occurs from t ⫽ 0 to t ⫽ 1, during which time the ISE reaches 1.0. This value of the ISE at t ⫽ 1 is the same, regardless of the tuning method used because the transport lag causes error to be constant from t ⫽ 0 to t ⫽ 1. Example 18.4. For the control system shown in Fig. 18–13, determine the controller settings for a PI controller using the Z-N method and the C-C method. In this problem, the process reaction curve must be modeled by the method shown in Fig. 18–6. U= R=0 + KC 1+ − 1 is 1 s 1 (s + 1) 4 C FIGURE 18–13 Process for Example 18.4. C-C method. Since the transfer function of the plant is given as 1/(s ⫹ 1)4, we can obtain the values of Td and T for use in the C-C method analytically. A unitstep response for the plant transfer function is c (t ) ⫽ 1 ⫺ cou9789x_ch18_391-422.indd 406 ( 16 t 3 ⫹ 12 t 2 ⫹ t ⫹ 1) e⫺t 8/22/08 10:17:44 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 407 From this result one can readily obtain the first and second derivatives; thus c ( t ) ⫽ c( t ) ⫽ 1 3 ⫺t t e 6 1 ⫺t e 6 (3t 2 ⫺ t3 ) The location of the inflection point on the transient c(t) is obtained by setting the second derivative to zero: 0 ⫽ 1 ⫺t e 6 ( 3t 2 ⫺ t 3 ) Solving for t gives as the root of interest in this problem t ⫽ 3. Knowing that the point of inflection occurs at t ⫽ 3, we can compute the slope of the tangent line through this point to be S ⫽ c(3) ⫽ 1 3 ⫺3 (3 )e 6 ⫽ 0 . 224 We can now determine Td as shown in Fig. 18–14. c 1.0 Slope = 0.224 0.9 0.8 0.7 0.6 0.5 0.4 Point of inflection 0.3 0.2 0.1 0.0 0 Td t 5 10 FIGURE 18–14 Process reaction curve for Example 18.4. From the expression for c(t), we obtain the value of c at the inflection point to be c(3) ⫽ 0.353. The value of t where the tangent line intersects the t axis is obtained from the slope S; thus 0 . 353 ⫺ 0 ⫽ S ⫽ 0 . 224 3 ⫺ Td Solving for Td gives Td ⫽ 1 . 42 cou9789x_ch18_391-422.indd 407 8/22/08 10:17:44 AM Confirming Pages 408 PART 5 PROCESS APPLICATIONS Solving for T from Eq. (18.4) gives T ⫽ Bu 1 ⫽ ⫽ 4 . 46 S 0 . 224 Having found Td and T, we can apply the appropriate equations from Table 18.2 to get K c ⫽ 2 . 91 t I ⫽ 2 . 86 The transient for these settings that was obtained by simulation is shown as curve C-C1 in Fig. 18–15. To our surprise, it is unstable. Z-N method. We apply the Z-N method for a PI controller, and we obtain the following results: Kcu ⫽ 4, Pu ⫽ 2p, Kc ⫽ 1.8, and tI ⫽ 5.23. 1.5 1 CC1 Response CC2 Z-N 0.5 0 −0.5 −1 0 5 10 15 20 25 Time 30 35 40 45 50 FIGURE 18–15 Comparison of transients produced by different tuning methods for Example 18.4 (shown in Fig. 18–13). Z-N: Ziegler-Nichols method; CC1: Cohen-Coon graphical method; CC2: Cohen-Coon method using least-squares fit of data. The transient for this set of controller parameters is also shown in Fig. 18–15. We see that the response is stable and well damped. The lesson learned in this example is that the application of a tuning method may not produce a satisfactory transient. Fine-tuning of these first guesses is usually needed. Before we abandoned the C-C method for this example, the process reaction curve was fitted to a first-order plus dead-time (FOPDT) model by means cou9789x_ch18_391-422.indd 408 8/22/08 10:17:45 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 409 of a least-squares fitting procedure using LOOP-PRO and Simulink software. The process was simulated in Simulink and disturbed with a unit step occurring at time t ⫽ 1.0. The response data were imported into Control Station and fit to a FOPDT model to produce the following results (see Fig. 18–16). Td ⫽ 1 . 79 T ⫽ 2 . 65 K P ⫽ 1 . 03 LOOP-PRO: Design Tools Model: First Order Plus Dead Time (FOPDT) Process variable 1.00 0.75 0.50 0.25 Manipulated variable 0.00 1.00 0.75 0.50 0.25 0.00 0.0 10.0 12.5 7.5 Time Gain (K) = 1.03, Time Constant (T1) = 2.65, Dead Time (TD) = 1.79 Goodness of Fit: R-Squared = 0.9972, SSE = 0.0271 2.5 5.0 15.0 FIGURE 18–16 LOOP-PRO fit of FOPDT model to process for Example 18.4. Applying the C-C method for these FOPDT values gives K c ⫽ 1 . 374 and t I ⫽ 2 . 548 Notice that the value of Kc is now considerably less than the value obtained from the fitting procedure shown in Fig. 18–14. This leads to the expectation that the response will now be stable. This expectation is fulfilled as shown by the transient labeled C-C2 in Fig. 18–15. Additionally, we can minimize ISE, ITAE, or IAE to find initial controller parameters as shown in Table 18.5. The response of the process to these controller settings is shown in Fig. 18–17. The qualitative nature of these responses is similar to that for the Z-N settings and the CC2 settings shown in Fig. 18–15. cou9789x_ch18_391-422.indd 409 8/22/08 10:17:46 AM Confirming Pages 410 PART 5 PROCESS APPLICATIONS TABLE 18.5 Controller settings for Example 18.4 using various tuning methods Z-N method C-C method Kc tI 1.8 5.23 2.91 2.86 ITAE 24.2 IAE ISE Parameter C-C using LOOP-PRO FOPDT fit Minimum ITAE Minimum IAE Minimum ISE 1.374 2.548 1.18 3.28 1.66 4.16 2.45 5.51 8431 30.7 19.37 21.64 41.33 2.93 116 3.34 2.96 2.79 3.311 0.819 284 0.910 1.06 0.816 0.67 0.6 ITAE 0.5 Response 0.4 0.3 ISE 0.2 IAE 0.1 0 −0.1 0 5 10 15 20 25 Time 30 35 40 45 50 FIGURE 18–17 Comparison of transients produced by minimizing the figures of merit for Example 18.4 (shown in Fig. 18–13), ITAE: minimum ITAE; ISE: minimum ISE; IAE: minimum IAE. 18.3 PROCESS IDENTIFICATION Up to this point, the processes used in our control systems have been described by transfer functions that were derived by applying fundamental principles of physics and chemical engineering (e.g., Newton’s law, material balance, heat transfer, fluid mechanics, reaction kinetics) to well-defined processes. In practice, many of the industrial processes to be controlled are too complex to be described by the application of fundamental principles. Either the task requires too much time and effort, or the fundamentals of the process are not understood. By means of experimental tests, one can identify the dynamical nature of such processes and from the results obtain a process model which is at least satisfactory for use in designing control systems. The experimental determination of the dynamic behavior of a process is called process identification. cou9789x_ch18_391-422.indd 410 8/22/08 10:17:46 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 411 The need for process models arises in many control applications, as we have seen in the use of tuning methods. Process models are also needed in developing feedforward control algorithms, self-tuning algorithms, and internal model control algorithms. Some of these advanced control strategies were discussed in Chap. 17. Process identification provides several forms that are useful in process control; some of these forms are Process reaction curve (obtained by step input) Frequency response diagram (obtained by sinusoidal input) Pulse response (obtained by pulse input) We have already encountered the need for process identification in applying the tuning methods presented earlier in this chapter. In the case of the Z-N method, the procedure obtained one point on the open-loop frequency response diagram when the ultimate gain was found. (This point corresponds to a phase angle of ⫺180⬚ and a process gain of 1/Kcu at the crossover frequency wco.) In the case of the C-C method, the process identification took the form of the process reaction curve. Step Testing As already described in the application of the C-C tuning method, a step change in the input to a process produces a response, which is called the process reaction curve. It is important that no disturbances other than the test step enter the system during the test; otherwise the transient will be corrupted by these uncontrolled disturbances and will be unsuitable for use in deriving a process model. Additionally, the step must be large enough to produce a change in the process that is much larger than any noise present in the process. (If the noise band is 0.5⬚F, the step should move the process at least 10 times this amount, or at least 5⬚F. This corresponds to a signal-to-noise ratio of 10.) For many processes in the chemical industry, the process reaction curve is an S-shaped curve, as shown in Fig. 18–6. For systems that produce an S-shaped process reaction curve, a general model that can be fitted to the transient is a second-order plus dead time (SOPDT) model: Gp (s) ⫽ Y (s) K pe⫺Td s ⫽ (T1s ⫹ 1) ( T2 s ⫹ 1) X ( s ) (18.7) This model is an extension of the one used in the C-C tuning method, in which there was only one first-order term, the FOPDT model. Computer software is available to fit a variety of models to process data using a least-squares technique. LOOP-PRO is one of the easiest and most efficient. Cooper (2005) makes the following points regarding the successful use of computer software to fit process models to data for the purposes of controller tuning. 1. 2. 3. The process must be at steady state before data collection begins. The first data point in the file must be the steady-state value. The test must be such that the dynamics caused by the test dominate the process noise. cou9789x_ch18_391-422.indd 411 8/22/08 10:17:46 AM Confirming Pages 412 PART 5 PROCESS APPLICATIONS 4. 5. There are no disturbance changes present during the test. The change in the process variable must be solely due to the changes introduced in the controller output. Finally, does the model appear to visually fit the process data? We will now work an example to demonstrate the use of such software to fit models to process data. Example 18.5. Use LOOP-PRO software to obtain a process model for the two tanks in series process shown in Fig. 18–18. Fit a FOPDT model and a SOPDT model to the data. Inlet flow (L/min) 17.8 Controller output (%) 70.0 Lower tank level (m) 3.99 Pumped flow (L/min) 2.0 LC Outlet flow (L/min) 15.8 (Disturbance) FIGURE 18–18 Screen shot from LOOP-PRO software for gravity-drained tanks. The process is a pair of gravity-drained tanks. Flow enters the upper tank and drains into the lower tank. A level controller monitors the level in the lower tank and adjusts the flow into the upper tank to maintain the lower level at the desired set point. The process disturbance is a pumped flow (not dependent on tank level) from the lower tank. We perform a step test on the inlet flow rate to the upper tank and record the data. In this test, the controller output was stepped from 70 to 80 percent, and the process response was recorded (Fig. 18–19). cou9789x_ch18_391-422.indd 412 8/22/08 10:17:47 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 413 LOOP-PRO: Case Studies Process Variable/Set point Process: Gravity Drained Tank Cont.:Manual Mode 5 4 3 Controller Output 2 80 70 60 50 −3 0 3 6 9 12 15 18 21 24 27 30 Time (mins) FIGURE 18–19 LOOP-PRO gravity-drained tanks step test for Example 18.5. cou9789x_ch18_391-422.indd 413 8/22/08 10:17:47 AM Confirming Pages 414 PART 5 PROCESS APPLICATIONS LOOP-PRO has the built-in capability to fit a FOPDT or SOPDT model (as well as some others) to the process data that we have generated. We use the Design Tools feature of the software and fit an FOPDT model to the data as shown in Fig. 18–20. LOOP-PRO: Design Tools Model: First Order Plus Dead Time (FOPDT) Process variable 5.0 G= 4.5 0.1316 exp(−0.6469s) 1.7 3 s + 1 4.0 5 10 15 Time Gain (K) = 0.1316, Time Constant (T1) = 1.73, Dead Time (TD) = 0.6469 Goodness of Fit: R-Squared = 0.9993, SSE = 0.0424 20 FIGURE 18–20 LOOP-PRO design tools for data analysis. FOPDT fit of step test data for Example 18.5. The FOPDT model fits the data very well. It captures the essential features of the response. The most noticeable difference is located at the time when the process variable begins to change. The process, being second-order (two tanks in series), rises as an S-shaped curve. The FOPDT, after the dead time, rises with a nonzero slope. There are two indications of the goodness of fit: R2 ⫽ 0.9993 is very close to 1.00 (which would indicate a perfect fit). The sum of the squares of the errors, SSE, is defined as SSE ⫽ N ∑ ( Measured Datai ⫺ Model Datai )2 i ⫽1 (18.8) For this case SSE ⫽ 0.0424, which indicates a very close fit. We can also fit a SOPDT model to the process data as shown in Fig. 18–21. The SOPDT model fits the data nearly perfectly (R2 ⫽ 1, SSE ⫽ 0.0019), as we would expect, since the process is in fact a second-order process. cou9789x_ch18_391-422.indd 414 8/22/08 10:17:48 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 415 LOOP-PRO: Design Tools Model: Second Order Plus Dead Time (SOPDT) 5.25 Process variable 5.00 4.75 4.50 4.25 4.00 0 5 10 15 Time 20 25 30 K = 0.1311, T1 = 1.04, T2 = 1.04, TD = 0.2208 Goodness of Fit: R-Squared = 1.00, SSE = 0.0019 FIGURE 18–21 LOOP-PRO design tools for data analysis. SOPDT fit of step-test data for Example 18.5. Pulse and Doublet Testing Pulse testing is similar to step testing; the only difference in the experimental procedure is that a pulse disturbance is used in place of a step disturbance. The pulse is introduced as a variation in valve top pressure as was done for step testing (see Fig. 18–5). In applying the pulse, the open-loop system is allowed to reach steady state, after which the valve top pressure is displaced from its steady-state value for a short time and then returned to its original value. The response is recorded at the output of the measuring element (B in Fig. 18–5). Usually the pulse shape is rectangular in experimental work, but other well-defined shapes are also used. An example of a pulse change in a process variable is shown in Fig. 18–22. 60 50 50 FIGURE 18–22 Sample pulse input. cou9789x_ch18_391-422.indd 415 8/22/08 10:17:49 AM Confirming Pages 416 PART 5 PROCESS APPLICATIONS This brief outline describing pulse testing may appear deceptively simple. In practice, the data on the response must be very accurate and noise-free for the method to succeed. This means that the recorder used to measure the response must be very sensitive. The selection of the pulse height and width is also critical. If the pulse height and width are too small, the disturbance to the system will be too small to produce a transient that can be measured accurately by the recorder. If the pulse height is too large, the system may be operating too far from the linear range of interest. The proper pulse height and width can be determined by some preliminary open-loop experiments. The pulse test is the least disruptive to plant operation among the process identification methods we have considered. The pulse disturbance does not cause the process output to depart far from its normal operating point. A doublet test is two pulse tests performed in rapid succession and in opposite direction as shown in Fig. 18–23. (Cooper, 2005). For example, we might step the input to the process from 50 to 60 to 40 to 50. This is the equivalent of two pulse tests 50-60-50 and 50-40-50 performed back to back. The process variable does not need to come to steady state for either pulse. It merely needs to show a clear response to the first pulse. 60 50 50 40 FIGURE 18–23 Sample doublet test input. The advantages of the doublet test are that it starts from and returns to the design level of operation relatively quickly, it produces data above and below the design level, and from a process standpoint, it minimizes the amount of time that the process is away from the design level, and hence off-spec production (Cooper, 2005). Example 18.6. Consider a process with an unknown transfer function such as that shown in Fig. 18–24. We wish to perform some process identification testing on it to determine a FOPDT model for use in controller tuning. Compare the results of pulse and doublet testing for this process. Manipulated variable Black box process Process variable FIGURE 18–24 “Black-box” process for Example 18.6. cou9789x_ch18_391-422.indd 416 8/22/08 10:17:49 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 417 For calculation purposes we will use G ⫽ 2/[(s ⫹ 1)(s ⫹ 2)(0.5s ⫹ 1)] for the transfer function of the black box. Pulse test. We will disturb the process by using a pulse input for the manipulated variable, as shown in Fig. 18–22, and use LOOP-PRO to determine the parameters for a FOPDT model (see Fig. 18–25). LOOP-PRO: Design Tools Model: First Order Plus Dead Time (FOPDT) File Name: _ _pulse_data.txt Process variable 70 65 60 55 Manipulated variable 50 60.0 57.5 55.0 52.5 50.0 0 10 20 30 40 50 60 70 Time Gain (K) = 1.99, Time Constant (T1) = 2.38, Dead Time (TD) = 0.5784 Goodness of Fit: T-Squared = 0.9994, SSE = 7.03 80 FIGURE 18–25 FOPDT fit of black-box process response to a pulse input (Example 18.6). The resulting FOPDT model from a least-squares fit of the pulse data is G ⫽ 1 . 99 exp (⫺0 . 5784 s ) 2 . 38 s ⫹ 1 (18.9) Doublet test. The doublet test on the black-box process is shown in Fig. 18–26. The FOPDT model from a least-squares fit of the doublet data is G ⫽ 2 . 09 exp(⫺0 . 7458 s ) 2 . 70 s ⫹ 1 (18.10) If we use the model, Eq. (18.9), obtained from the pulse test, and perform a doublet test on it to compare with our actual black-box process response, we obtain Fig. 18–27. Notice that the pulse model deviates from the process response, mainly on the second half of the doublet. The SSE for the pulse model on the doublet cou9789x_ch18_391-422.indd 417 8/22/08 10:17:50 AM Confirming Pages 418 PART 5 PROCESS APPLICATIONS LOOP-PRO: Design Tools Model: First Order Plus Dead Time (FOPDT) File Name: doublet_data.txt Process variable 70 60 50 40 30 Manipulated variable 60 55 50 45 40 0 10 20 30 40 50 60 70 Time Gain (K) = 2.09, Time Constant (T1) = 2.70, Dead Time (TD) = 0.7458 Goodness of Fit: R-Squared = 0.9981, SSE = 29.71 80 FIGURE 18–26 FOPDT fit of black-box process response to a doublet input (Example 18.6). test is 159.15 compared to 29.71 for the doublet model. Since we would like the model to capture process characteristics for deviations above and below the design operating level, we would probably be better off using the FOPDT model obtained from the doublet testing to determine our initial tuning parameters. Frequency Testing We have shown in the section on frequency response that a process having a transfer function G(s) can be represented by a frequency response diagram (or Bode plot) by taking the magnitude and phase angle of G(jw). This procedure can be reversed to obtain G(s) from an experimentally determined frequency response diagram. The procedure requires that a device be available to produce a sinusoidal signal over a range of frequencies. We describe such a device as a sine wave generator. In frequency testing of an industrial process, a sinusoidal variation in pressure is applied to the top of the control valve so that the manipulated variable can be varied sinusoidally over a range of frequencies. The block diagram that applies during frequency testing is the same as that of Fig. 18–5 with the step input (M/s) replaced by a sinusoidal signal. The sine wave generator used to test electronic devices operates at frequencies that are too high for many slow-moving chemical processes. For frequency testing of chemical processes, special low-frequency generators must be built that can produce a sinusoidal variation cou9789x_ch18_391-422.indd 418 8/22/08 10:17:52 AM Confirming Pages CHAPTER 18 CONTROLLER TUNING AND PROCESS IDENTIFICATION 419 LOOP-PRO: Design Tools Model: First Order Plus Dead Time (FOPDT) File Name: Process variable 70 60 50 40 Manipulated variable 30 60 55 50 45 40 0 10 20 30 40 Time 50 60 70 80 Gain (K) = 1.99, Time Constant (T1) = 2.38, Dead Time (TD) = 0.5785 Goodness of Fit: R-Squared = 0.9897, SSE = 159.15 FIGURE 18–27 A doublet test on the FOPDT model derived from a pulse input test (Example 18.6). in pressure to a control valve. To preserve the sinusoidal signal in the flow of manipulated variable through the valve, the valve must be linear. In the 1960s when frequency response methods were first introduced to chemical engineers as a means for process identification, several chemical and petroleum companies constructed mobile units containing low-frequency sine wave generators and recorders that could be moved to processing units in a plant for the purpose of frequency testing. The great disadvantage of frequency testing is that it takes a long time to collect frequency response data over a range of frequencies that can be used to construct frequency response plots. The time is especially long for chemical processes, often having long time constants measured in minutes or even hours. The frequency test at a given frequency must last long enough to make sure that the transients have disappeared and only the ultimate periodic response is represented by the data. Frequency testing usually ties up plant equipment too long to be recommended as a means of process identification. Step testing, pulse testing, and doublet testing take much less time and can usually provide satisfactory process identification. cou9789x_ch18_391-422.indd 419 8/22/08 10:17:52 AM Confirming Pages 420 PART 5 PROCESS APPLICATIONS SUMMARY In the practical application of process control, some methods for tuning and process identification are needed. The selection of controller modes depends on the process to be controlled. Proportional control is simple, but the response exhibits offset. The derivative action in PD control makes it possible to increase the controller gain with the result that the response has less offset and responds more quickly compared to proportional control. To eliminate offset, integral action must be present in the controller in the form of PI and PID control. PI control often causes the response to have large overshoot and a slow return to the set point especially for high-order processes. The presence of derivative action in a PID controller gives less overshoot and a faster return to the set point, compared to the response for PI control. To compare the quality of control on a numerical basis, several criteria that integrate some function of the error with respect to time have been proposed. These include the integral of the square of the error (ISE), the integral of the absolute error (IAE), and the integral of the time-weighted absolute error (ITAE). In the first part of this chapter two well-known tuning methods are presented: the Ziegler-Nichols (Z-N) method (a closed-loop method) and the Cohen-Coon (C-C) method (an open-loop method). These two methods were applied to several examples and the transients for each compared. The lesson to be learned through these examples is that the controller parameters obtained from a tuning rule should be considered as first guesses; fine online tuning is usually needed to get a satisfactory transient. The Z-N and C-C methods actually require information about the process model. The Z-N method is based on the ultimate gain at the crossover frequency, which is equivalent to knowing one point on the open-loop frequency response diagram. The C-C method requires the use of an open-loop step response (process reaction curve). In the advanced control strategies discussed in Chap. 17, a process model is often needed to apply the strategy. When a process model cannot be found by application of theoretical principles, one must obtain a model experimentally. The experimental approach to obtaining a model is called process identification. The three methods of process identification discussed in this chapter are step testing, frequency testing, and pulse testing. The frequency method is seldom used because of the time it takes to test a system over a wide range of frequencies. Step testing is easy to apply and ties the process up for only enough time to obtain one transient. Pulse testing is also simple to apply, but the analysis of the input-output data requires extensive calculations that must be done by a computer. PROBLEMS 18.1. Calculate C-C and Z-N parameters for PI control for the process in Example 18.5, using the pulse FOPDT model and the doublet FOPDT model. Compare the results for a step change in load for a closed-loop PI control system. Use the ITAE figure of merit for comparison. cou9789x_ch18_391-422.indd 420 8/22/08 10:17:53 AM Confirming Pages CHAPTER 18 CAPSULE SUMMARY Figures of merit for comparing control parameters (computed from process transients) Figure of merit Equation ⬁ 2 Purpose Integral of the square of the error (ISE) ISE ⫽ ∫0 e Integral of the absolute value of error (IAE) IAE ⫽ ∫0 | e | dt Integral of time-weighted absolute error (ITAE) ITAE ⫽ dt Penalizes large errors, particularly those that occur initially ⬁ Treats all errors (large and small) the same ⬁ Penalizes errors that persist for a long time ∫0 | e | t dt First-order plus dead-time (FOPDT) model Gp (s) ⫽ Y (s) K p e⫺Td s ⫽ (T1s ⫹ 1) X ( s ) Second-order plus dead-time (SOPDT) model Gp (s) ⫽ Y (s) K p e⫺Td s ⫽ (T1s ⫹ 1) (T2 s ⫹ 1) X ( s ) Ziegler-Nichols controller settings sI Type of control Gc(s) Kc Proportional (P) Kc 0.5Ku 1 Kc 1 ⫹ t I s 0.45Ku Pu 1.2 1 Kc 1 ⫹ ⫹ t D s tIs 0.6Ku Pu 2 Proportional-integral (PI) Proportional-integral-derivative (PID) sD Pu 8 421 cou9789x_ch18_391-422.indd 421 8/22/08 10:17:53 AM Confirming Pages 422 PART 5 PROCESS APPLICATIONS Cohen-Coon controller settings Type of control Parameter setting Proportional (P) Kc ⫽ 1 T Td 1 ⫹ K p Td 3T Proportional-integral (PI) Kc ⫽ 1 T 9 Td ⫹ K p Td 10 12T t I ⫽ Td Proportional-derivative (PD) Kc ⫽ 1 T 5 Td ⫹ K p Td 4 6T t D ⫽ Td Proportional-integral-derivative (PID) cou9789x_ch18_391-422.indd 422 Kc ⫽ 30 + 3Td / T 9 ⫹ 20Td / T 6 ⫺ 2Td / T 22 ⫹ 3Td / T 1 T 4 Td ⫹ K p Td 3 4T t I ⫽ Td 32 ⫹ 6Td / T 13 ⫹ 8Td / T t D ⫽ Td 4 11 ⫹ 2Td / T 8/22/08 10:17:54 AM Confirming Pages CHAPTER 19 CONTROL VALVES O ne of the basic components of any control system is the final control element, which comes in a variety of forms depending on the specific control application. The most common type of final control element in chemical processing is the pneumatic control valve, which regulates the flow of fluids. Some other types include the variable-speed pump and the power controller (used in electrical heating). Since the pneumatic control valve is so widely used in chemical processing, this chapter will be devoted to the description, selection, and sizing of control valves. 19.1 CONTROL VALVE CONSTRUCTION The control valve is essentially a variable resistance to the flow of a fluid in which the resistance, and therefore the flow, can be changed by a signal from a process controller. As shown in Fig. 19–1, the control valve consists of an actuator and a valve. The valve itself is divided into the body and the trim. The body consists of a housing for mounting the actuator and connections for attachment of the valve to a supply line and a delivery line. The trim, which is enclosed within the body, consists of a plug, a valve seat, and a valve stem. The actuator moves the valve stem as the pressure on a spring-loaded diaphragm changes. The stem moves a plug in a valve seat to change the resistance to flow through the valve. When a valve is supplied by the manufacturer, the actuator and the valve are attached to each other to form one unit. For most actuators, the motion of the stem is proportional to the pressure applied on the diaphragm. In general, this type of actuator can be used for functions other than moving a valve stem. For example, it can be used to adjust dampers, variable-speed drives, rheostats, and other devices. As the pressure to the valve varies over its normal 423 cou9789x_ch19_423_442.indd 423 8/22/08 2:19:23 PM Confirming Pages 424 PART 5 PROCESS APPLICATIONS range of operation (3 to 15 psig), the range of motion of the stem varies from a fraction of an inch to several inches depending on the size of the actuator. Manufacturers provide a range of actuators for various valve sizes. The valves available vary over a wide range of sizes. The size is usually referred to by the size of the end connectors. For example, a 1-in valve would have connectors (threaded or flanged) to fit into a 1-in pipeline. In general, the larger the valve size, the larger the flow capacity of the valve. For the control valve shown in Fig. 19–1, an increase in signal pressure above the diaphragm exerts a force on the diaphragm and back plate, which causes the stem to move down; this causes the cross-sectional area for flow between the plug and the seat to decrease, thereby reducing or throttling the flow. Such valve action as shown in Fig. 19–1 is called air-to-close (AC) action. The reverse action, air-to-open (AO), can be accomplished by designing the actuator so that pressure is applied to the underside of the diaphragm, for which case an increase in pressure to the valve raises the stem. An alternate method to reverse the valve action is to leave the actuator as shown in Fig. 19–1 and to invert the plug on the stem and place it under the valve seat. In general, selection of the type of valve (AO or AC) is made based on safety considerations. We would like the valve to fail in a safe position for the process in the event of a loss of air pressure. For example, if the control valve is controlling the inlet flow of cooling water to a cooling jacket on an exothermic chemical reactor, we would like the valve to fail in the open position so that we do not lose coolant flow. Thus, we would select an airto-close (AC) valve. Air signal Air signal Diaphragm Diaphragm ACTUATOR Packing ACTUATOR Packing Stem Stem VALVE VALVE Supply Delivery Plug Seat Air-to-Close Valve Supply Delivery Plug Seat Air-to-Open Valve FIGURE 19–1 Single-seated control valves. The valve shown in Fig. 19–1 is single-seated, meaning the valve contains one plug with one seating surface. For a single-seated valve, the plug must open against the full pressure drop across the valve. If the pressure drop is large, this means that a larger, more expensive actuator will be needed. To overcome this problem, valves are also constructed with double seating as shown in Fig. 19–2. In this type valve, two plugs are attached to the valve stem, and each one has a seat. The flow pattern through the cou9789x_ch19_423_442.indd 424 8/22/08 2:19:23 PM Confirming Pages CHAPTER 19 CONTROL VALVES 425 valve is designed so that the pressure drop across the seat at A tends to open the plug and the pressure drop across the seat at B tends to close the plug. This counterbalancing of forces on the plugs reduces the effort needed to open the valve with the result that a smaller, less expensive actuator is needed. Stem Packing A B FIGURE 19–2 Double-seated control valve. In a double-seated valve, it is difficult to have tight shutoff. If one plug has tight closure, there is usually a small gap between the other plug and its seat. For this reason, single-seated valves are recommended if the valve is required to be shut tight. In many processes, the valve is used for throttling flow and is never expected to operate near its shutoff position. For these conditions, the fact that the valve has a small leakage at shutoff position does not create a problem. 19.2 VALVE SIZING To specify the size of a valve in terms of its capacity, the following equation is used: q ⫽ Cv f ( x ) where ∆pvalve sg (19.1) q flow rate, gal/min f (x) fraction of maximum ﬂow ( 1 for fully open) x fractional stem position (i.e., fraction open) ∆ pvalve pressure drop across the valve, psi sg speciﬁc gravity of ﬂuid at stream temperature relative to water; for water sg 1 Cv factor associated with capacity of valve Equation (19.1) applies to the flow of an incompressible, nonflashing fluid through the valve. Manufacturers rate the size of a valve in terms of the factor Cv. Sometimes Cv is defined as the flow (gal/min) of a fluid of unit specific gravity (water) through a fully open valve, across which a pressure drop of 1.0 psi exists. This verbal definition is, of course, obtained directly from Eq. (19.1) by letting f (x) 1, ∆ pvalve 1, and sg 1. Equation (19.1) is based on the well-known Bernoulli equation for determining the pressure drop across valves and resistances. It is important to emphasize that Cv must be cou9789x_ch19_423_442.indd 425 8/22/08 2:19:26 PM Confirming Pages 426 PART 5 PROCESS APPLICATIONS determined from Eq. (19.1) using the units listed. Since so many valves in use are rated in terms of Cv, Eq. (19.1) is of practical importance; however, some industries now are defining a valve coefficient Kv defined by the equation q ⫽ Kv ∆pvalve sg q flow rate, m3/h ∆ pvalve pressure drop across valve, kgf/cm2 sg specific gravity of fluid relative to water The relation between Kv and Cv is where K v ⫽ 0 . 856Cv For gases and steam, modified versions of Eq. (19.1) are used in which Cv is still used as a factor. Manufacturers of valves provide brochures, nomographs, and computer programs for sizing valves for use with gases and steam. In general, as the physical size of a valve body (i.e., size of pipe connectors) increases, the value of Cv increases. For a sliding stem and plug type of control valve, the value of Cv is roughly equal to the square of the pipe size multiplied by 10. Using this rule, a 3-in control valve should have a Cv of about 90, with units corresponding to those of Eq. (19.1). This implies the capacity of the fully open valve is 90 gal/min with a pressure drop of 1 psi. Example 19.1 A valve with a Cv rating of 4.0 is used to throttle the flow of glycerine for which sg 1.26. Determine the maximum flow through the valve for a pressure drop of 100 psi. q ⫽ 4.0 100 ⫽ 35 . 6 gal / min 1 . 26 The coefficient Cv varies with the design of the valve (shape, size, roughness) and the Reynolds number for the flow through the valve. This relationship is analogous to the relationship between the friction factor and roughness and the Reynolds number for flow through a pipe. For relatively nonviscous fluids, Cv in Eq. (19.1) can be taken as a constant for a valve of given size and type. The reason for this is that at high Reynolds numbers, the friction factor changes very little with flow rate. Except for very viscous fluids, the flow through a valve, which involves sudden contraction and expansion, is in the turbulent regime of fluid flow; turbulence in the valve exists even if the flow in the supply pipe is near the critical Reynolds number of 2100. Consequently, for relatively nonviscous fluids, Eq. (19.1) is satisfactory for sizing a valve for any fluid. For the control of flow of very viscous fluids, such as tar or molasses, the value of Cv found from Eq. (19.1) must be multiplied by a correction factor that depends on viscosity, density, flow rate, and valve size (i.e., on the Reynolds number). Methods for determining the viscosity correction factor are provided by manufacturers for their valves. If one does not apply the correction factor for a very viscous fluid, the value of Cv will be too low and the valve will be undersized. cou9789x_ch19_423_442.indd 426 8/22/08 2:19:26 PM Confirming Pages CHAPTER 19 19.3 CONTROL VALVES 427 VALVE CHARACTERISTICS f, fraction of maximum flow The function of a control valve is to vary the flow of fluid through the valve by means of a change of pressure to the valve top. The relation between the flow through the valve and the valve stem position (or lift) is called the valve characteristic, which can be conveniently described by means of a graph as shown in Fig. 19–3 where three types of characteristics are illustrated. 1.0 III 0.5 0 I II 1.0 0.5 0 x, fraction of maximum lift FIGURE 19–3 Inherent valve characteristics (pressure drop across valve is constant). I: Linear, II: increasing sensitivity (e.g., equal-percentage valve), and III: decreasing sensitivity (e.g., square root valve). In general, the flow through a control valve for a specific fluid at a given temperature can be expressed as q ⫽ f1 ( L , p0 , p1 ) (19.2) where q volumetric flow rate L valve stem position (or lift) p0 upstream pressure p1 downstream pressure The inherent valve characteristic is determined for fixed values of p0 and p1, for which case Eq. (19.2) becomes (19.3) q ⫽ f2 ( L ) or, in other words, the flow is a function of the valve stem position. For convenience, let f⫽ q qmax and x⫽ L Lmax where qmax maximum flow when valve is fully open (stem is at its maximum lift Lmax) x fraction of maximum lift f fraction of maximum flow In general, f will be a function of x, which we will denote as f (x). Equation (19.3) may be written as q L f ⫽ f (x) ⫽ (19.4) Lmax qmax cou9789x_ch19_423_442.indd 427 8/22/08 2:19:27 PM Confirming Pages 428 PART 5 PROCESS APPLICATIONS The types of valve characteristics can be defined in terms of the sensitivity of the valve, which is simply the fractional change in flow to the fractional change in stem position for fixed upstream and downstream pressures; mathematically, sensitivity may be written df Sensitivity ⫽ dx In terms of valve characteristics, valves can be divided into three types: decreasing sensitivity (square root or quick-opening), linear, and increasing sensitivity (equal-percentage). These types are shown in Fig. 19–3 where the fractional flow f (x) is plotted against fractional lift x. For the decreasing sensitivity type, the sensitivity (or slope) decreases with increasing flow. For the linear type, the sensitivity is constant and the characteristic curve is a straight line. For the increasing sensitivity type, the sensitivity increases with flow. Valve characteristic curves, such as the ones shown in Fig. 19–3, can be obtained experimentally for any valve by measuring the flow through the valve as a function of lift (or valve-top pressure) under conditions of constant upstream and downstream pressures. Two types of valves that are widely used are the linear valve and the equalpercentage valve. Linear Valves The linear valve is one for which the sensitivity is constant and the relation between flow and lift is linear. For the linear valve, the mathematical relationship is df ⫽a dx (19.5) where a is a constant. Assuming that the valve is shut tight when the lift is at lowest position, we have f 0 at x 0. For a single-seated valve that is not badly worn, the valve can be shut off for x 0. Recall that the definitions of x and f require that f 1 at x 1. Integrating Eq. (19.5) and introducing the limits f 0 at x 0 and f 1 at x 1 give 1 ∫0 df ⫽ 1 ∫0 a dx Integrating this equation and inserting limits give a ⫽1 For a 1, Eq. (19.5) can now be integrated to give f (x) ⫽ x linear valve (19.6) Equal-Percentage Valves For the equal-percentage valve, the defining equation is df ⫽ bf dx cou9789x_ch19_423_442.indd 428 (19.7) 8/22/08 2:19:28 PM Confirming Pages CHAPTER 19 CONTROL VALVES 429 where b is constant. Integration of this equation gives f ∫f 0 df ⫽ f x ∫ b dx (19.8) 0 or ln f ⫽ bx f0 (19.9) where f0 is the flow at x 0. Rearranging gives f ⫽ f0 e b x (19.10) The term b can be expressed in terms of f0 by inserting f 1 at x 1 into Eq. (19.9). The result is 1 b ⫽ ln f0 If we substitute for b in Eq. (19.10), we obtain f ⫽ f0 e x ln (1 / f0 ) ln 1 / f ln f ⫽ f0 e ( 0 ) ⫽ f0 e ( 0 ) ⫺x x ⫽ f0 f0⫺x ⫽ f01⫺ x which is of the form f ⫽ a x ⫺1 equal-percentage valve (19.11) where a 1/f0 is a constant. Equation (19.9) shows that a plot of f versus x on semilog coordinates gives a straight line. A convenient way to determine if a valve is of the equal-percentage type is to plot the flow versus lift on semilog coordinates. The relation expressed by Eq. (19.9) is the basis for calling the valve characteristic logarithmic. The basis for calling the valve characteristic equal percentage can be seen by rearranging Eq. (19.7) into the form df ⫽ b dx f or ∆f ⫽ b ∆x f In this form it can be seen that an equal fractional (or percentage) change in flow ∆ f/f occurs for a specified increment of change in stem position ∆ x, regardless of where the change in stem position occurs along the characteristic curve. In integrating Eq. (19.7), the flow was assumed to be f0 at x 0. Mathematically this is necessary, because f0 cannot be taken as zero at x 0 because the term on the left side of Eq. (19.9) becomes infinite. In practice, there may be some leakage (hence f0 0) when the stem is at its lowest position for a double-seated valve or for a valve in which the plug and seat have become worn. For some valves, especially large ones, the valve manufacturer intentionally allows some leakage at minimum lift (x 0) to prevent binding and wearing of the plug and seat surfaces. For a valve that does shut tight and is also classified as an equalpercentage valve, the equal-percentage characteristic will not be followed when the valve is nearly shut. In practice, the control valve serves as a throttling valve and is not intended to be wide open or completely closed during normal operation. cou9789x_ch19_423_442.indd 429 8/22/08 2:19:28 PM Confirming Pages 430 PART 5 PROCESS APPLICATIONS To express the range over which an equal-percentage valve will follow the equalpercentage characteristic, the term rangeability is used. Rangeability is defined as the ratio of maximum flow to minimum controllable flow over which the valve characteristic is followed. f ( x )max f ( x )min, controllable Rangeability ⫽ For example, if f0 is 0.02, the rangeability is 50. It is not uncommon for a control valve to have a rangeability as high as 50. In practice, the ideal characteristics for linear and equal-percentage valves are only approximated by commercially available valves. These discrepancies cause no difficulty because the inherent characteristics are changed considerably when the valve is installed in a line having resistance to flow, a situation that usually prevails in practice. The inherent valve characteristics are shown in Table 19.1. In the next section, the effect of line loss on the effective valve characteristic will be discussed. TABLE 19.1 Inherent valve characteristics Sensitivity Valve type df dx Relationship Linear Constant f (x) x Equal-percentage Increasing f (x) ax1 Square root or quick-opening Decreasing f (x ) ⫽ x Effective Valve Characteristic When a valve is placed in a line that offers resistance to flow, the inherent characteristic of the valve will be altered. The relation between flow and stem position (or valve-top pressure) for a valve installed in a process line is called the effective valve characteristic. pv Water p0 supply L p1 FIGURE 19–4 Control valve with supply line. cou9789x_ch19_423_442.indd 430 8/22/08 2:19:29 PM Confirming Pages CHAPTER 19 CONTROL VALVES 431 Consider a linear control valve attached to the end of a pipeline that delivers water to an open tank. A diagram of the system is shown in Fig. 19–4. If the pipe is of large diameter relative to the size of the control valve, the pressure drop in the line will be negligible and the full pressure drop p0 p1 will be across the valve as the lift varies between 0 and 1 (from fully closed to fully open). In this case a plot of flow versus lift will give a linear relation as shown by curve I of Fig. 19–5. This curve is for the flow of water at 5°C through a control valve for which Cv 4.0 and the overall pressure drop p0 p1 is 100 psi. To show the effect of line loss, curve II is constructed for the same conditions as curve I, with the exception that 100 ft of 1.0-in (inside diameter) pipe is used to supply the valve. Example 19.2 will give the detailed calculations used to obtain the results in Fig. 19–5. For 100 ft of pipe, the plot of flow versus lift gives curve II, shown in Fig. 19–5, in which the curve falls away or droops from the linear relation that holds for no line loss. Since line loss is proportional to the square of the velocity, the line loss is very small when the valve is nearly closed, for which case the total pressure drop is across the valve. For this reason, curves I and II in Fig. 19–5 are close together at low rates. A rule often followed in industrial application of control valves is that the pressure drop across the wide-open valve should be greater than 25 percent of the pressure drop across the closed valve. A valve not selected according to this rule will lose its effectiveness to control at high flow rates. 40 I q, gal/min 30 II 20 10 0 0 0.2 0.4 0.6 x, fraction lift 0.8 1.0 FIGURE 19–5 Effect of line loss on control valve characteristics from Example 19.2. I: No pressure drop in supply line to valve, II: pressure drop present in supply line to valve. Example 19.2. Determine the flow versus lift relation for the linear control valve installed in the flow system of Fig. 19–4. The fluid is water at 5°C. The following data apply. Pipe length Inside pipe diameter Density of water Viscosity of water Cv of control valve Total pressure drop p0 p1 ∆ ptotal cou9789x_ch19_423_442.indd 431 100 ft 1 in 62.4 lb/ft3 1.5 cP 4.0 100 psi 8/22/08 2:19:30 PM Confirming Pages 432 PART 5 PROCESS APPLICATIONS If there is no line loss, as is the case for a large-diameter line, then ∆ ptotal ∆ pvalve, and the maximum flow can be calculated from Eq. (19.1): q ⫽ Cv f ( x ) ∆pvalve 100 ⫽ 4 . 0(1 . 0) ⫽ 40 . 0 gal / min sg 1 To determine the flow/lift relation for the case of line loss, we arbitrarily start the calculation with a flow of 30 gal/min. The pressure drop in the 100 ft-pipe can be calculated from the well-known expression from fluid mechanics ∆ppipe ⫽ 32 f LρQ 2 (19.12) 144p 2 gc D 5 where ∆ ppipe pressure loss in line, psi gc ⫽ 32 . 174 lbm ⭈ ft / s2 lbf r density of fluid, lbm/ft3 f Fanning friction factor, dimensionless Q volumetric flow through pipe, ft3/s L pipe length, ft D inside pipe diameter, ft The Fanning friction factor is a function of the Reynolds number and the pipe roughness. Equation (19.11) and a correlation for the Fanning friction factor can be found in the literature (Perry and Green, 1999). We now calculate the Reynolds number Re: Dur Re ⫽ m Replacing the velocity u with Q/[(p/4)D2] gives Re ⫽ Q ⫽ 30 gal / min 4Qr pmD (60 s/min ) ( 7.48 gal / ft3 ) ⫽ 0 . 066 8 (19.13) ft 3 ft 3 ⫽ 240 . 6 s h ( 4 ) ( 240.6 ft3 / h ) ( 62.4 lb / ft3 ) = 63, 220 lb / (ft ⭈ h) (p ) ft (1.50 cP ) 2.42 cP For this value of Reynolds number and for smooth pipe, the Fanning friction factor f is 0.005. Equation (19.12) may now be used to calculate the line loss: Re ⫽ ( ) 1 12 ∆ppipe 2 32 ) ( 0 . 005) (100 ) ( 62 . 4 ) ( 0 . 0668 ) ( ⫽ 5 (144 ) (p 2 ) (32.2) ( 121 ) ⫽ 24 . 2 psi Therefore, ∆ pvalve 100 psi 24.2 psi 75.8 psi. cou9789x_ch19_423_442.indd 432 8/22/08 2:19:31 PM Confirming Pages CHAPTER 19 CONTROL VALVES 433 We next calculate the flow through the wide-open valve that gives a pressure drop of 75.8 psi: qmax ⫽ Cv ∆pvalve 75 . 8 ⫽ 4.0 ⫽ 34 . 8 gal / min sg 1 f (x) x for linear , f ( x ) ⫽ x ⫽1 for max. fl o w Recall that we arbitrarily started this calculation with a flow rate of 30 gal/min. Now, since a flow of 34.8 gal/min through the wide-open valve produces a pressure drop of 75.8 psi, we conclude that the valve must be partially closed since we have ∆ pvalve 75.8 psi with only 30 gal/min of flow. Since the valve is linear, we can calculate the lift x as follows: q 30 f (x) ⫽ x ⫽ ⫽ ⫽ 0 . 86 qmax 34 . 8 Linear valve We can generalize this calculation to generate the data necessary to produce Fig. 19–5. x ⫽ q qmax ⫽ q Cv ∆pvalve / sg q ⫽ 4 ∆pvalve We also know that ∆ptotal ⫽ 100 psi ⫽ ∆ppipe + ∆pvalve We can observe that the pressure drop through the pipe is proportional to the square of the flow rate through the pipe [Eq. (19.12)]. Therefore, we can write ( ∆ppipe )old ( ∆ppipe )new ⫽ 2 qold 2 qnew qold ⫽ qnew 2 We have one data point that we have already calculated: q 30 gal/min and ∆ pvalve 24.2 psi, so 24 . 2 psi 30 gpm ⫽ ∆ppipe q 2 and q ∆ppipe ⫽ 30 gpm 2 ( 24.2 psi) From the expression for the total pressure drop, we can write q ∆pvalve ⫽ ∆ptotal ⫺ ∆ppipe ⫽ 100 psi ⫺ 30 gpm cou9789x_ch19_423_442.indd 433 2 ( 24.2 psi) 8/22/08 2:19:31 PM Confirming Pages 434 PART 5 PROCESS APPLICATIONS Finally, we can substitute this back into the expression for x. x ⫽ q ⫽ 4 ∆pvalve q q 4 100 psi ⫺ 30 gpm 2 ( 24.2 psi) Using this equation, we can generate the effective characteristic curve (curve II) shown in Fig. 19–5. The data are summarized in Table 19.2. TABLE 19.2 Effective characteristic for a linear valve with supply line loss Curve I pressure drop in valve only Flow q, gal/min Stem position x Valve pressure drop, psi 0 10 20 30 30.5 31 31.5 32 32.5 33 33.5 34 34.5 35 35.5 36 36.5 37 37.5 38 38.5 39 39.5 40 0.00 0.25 0.50 0.75 0.76 0.78 0.79 0.80 0.81 0.83 0.84 0.85 0.86 0.88 0.89 0.90 0.91 0.93 0.94 0.95 0.96 0.98 0.99 1.00 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Curve II pressure drop in valve and pipe Stem position x Valve pressure drop, psi 0.00 0.25 0.53 0.86 0.88 0.90 0.92 0.94 0.96 0.98 1.00 100 97 89 76 75 74 73 72 72 71 70 Example 19.3. A control valve is to be installed in the flow system of Fig. 19–4. The valve is supplied by water at 5°C through 200 ft of pipe having an inside diameter of 1.0 in. The total pressure drop p0 p1 is 100 psi. When the valve is wide open, the flow is to be 30 gal/min. Determine Cv for the valve. Plot the effective characteristic curve for the valve as flow versus lift. Do this problem for a linear valve and for an equal-percentage valve. The equal-percentage valve has an a 33.3. cou9789x_ch19_423_442.indd 434 8/22/08 2:19:32 PM Confirming Pages CHAPTER 19 CONTROL VALVES 435 Linear valve. To obtain the pressure drop in the line, use is made of Eqs. (19.12) and (19.13) as was done in Example 19.2. From Eq. (19.13), we obtain the Reynolds number as follows: 30 ft 3 ft 3 ⫽ 0 . 0668 ⫽ 240 . 6 (60)(7 . 48) s h q ⫽ 4qr Re ⫽ ⫽ pm D ( 4 ) ( 240.6 ft3 / h ) ( 62.4 lb / ft3 ) lb / ( ft ⭈ h ) (p ) ( 112 ft ) (1.50 cP ) 2.42 cP ⫽ 63, 2 20 From a correlation for the Fanning friction factor, we obtain f 0.005. From Eq. (19.12), the line loss is calculated to be ∆p ⫽ (32) ( 0.005) ( 200 ) (62.4 ) ( 0.0668)2 5 (144 ) (p 2 ) (32.2) ( 121 ) ⫽ 48. 4 psi ∆pvalve ⫽ 100 ⫺ 48 . 4 ⫽ 51. 6 psi From knowledge of the maximum flow through the wide-open valve (30 gal/min) and ∆ pvalve, we calculate Cv from Eq. (19.1) as follows: Cv ⫽ qmax ∆pvalve / sg 30 ⫽ 4 . 18 51 . 6 ⫽ Now that we have Cv, we can calculate the stem position x needed for various flow rates q. As we showed in Example 19.2, the pressure drop through the pipe is proportional to the square of the flow rate. Since the total pressure drop is constant, we can calculate the pressure drop across the valve for any flow rate. ∆pvalve ⫽ ∆ptotal ⫺ ∆ppipe q ⫽ 100 psi ⫺ 30 gpm 2 ( 48.4 psi) We can generalize the rest of the solution procedure for this problem as follows: 1. Pick a value for q (less than qmax, 30 gal/min in this case). q 2. Find ∆pvalve ⫽ 100 psi ⫺ 30 gpm 2 ( 48.4 psi) 3. Find qmax ⫽ Cv f ( x ) cou9789x_ch19_423_442.indd 435 q ∆pvalve ∆pvalve ⫽ ( 4.18 ) (1.0 ) ⫽ 4.18 100 psi ⫺ 30 gpm sg 1 2 ( 48.4 psi ). 8/22/08 2:19:33 PM Confirming Pages 436 PART 5 PROCESS APPLICATIONS 4. Find f ( x ) ⫽ q q ⫽ . 2 qmax 4 . 18 100 ⫺ ( q / 30 ) ( 48 . 4 ) 5. Find x f (x) for a linear valve. As a numerical example, 1. Choose q 20 gal/min. 20 2. ⌬pvalve ⫽ 100 psi ⫺ 30 gpm 2 ( 48.4 psi) ⫽ 7 8. 5 psi 3. qmax ⫽ 4 . 18 78 . 5 ⫽ 37 gpm 20 ⫽ 0 . 54 37 5. x f (x) 0.54 for a linear valve 4. f ( x ) ⫽ For other flow rates, one can repeat this procedure to obtain values of x. The results are shown in Table 19.3 and in Fig. 19–6. The latter also shows the inherent characteristic of the linear valve for comparison with the effective characteristic of the valve when line loss is present. Equal-percentage valve. Calculation of the effective characteristic will now be made for an equal-percentage valve having the same Cv of 4.18 as calculated for the linear valve in the first part of this example. The procedure (steps 1 to 5) that we outlined above for the linear valve applies to this case as well, except that we must modify step 5 for an equal-percentage valve. The relationship between stem position and fraction of maximum flow is given by Eq. (19.11). f ( x ) ⫽ a x ⫺1 (19.11) Solving Eq. (19.11) for x, we get x ⫽ 1⫹ ln f ( x ) ln a For this example, a 33.3, and we modify step 5 from above to determine x as follows: 5. Find x ⫽ 1 ⫹ ln f ( x ) ln f ( x ) ln f ( x ) ⫽ 1⫹ ⫽ 1⫹ ln a ln 33 . 3 3 . 51 For a flow rate of 20 gal/min, we found f (x) 0.54 above. For the equal-percentage valve, we can now find the stem position. x ⫽ 1⫹ ln(0 . 54 ) ⫽ 0 . 82 3 . 51 For other values of flow, corresponding values of x are calculated for the equalpercentage valve, and the results are shown in Table 19.3 and Fig. 19–6. cou9789x_ch19_423_442.indd 436 8/22/08 2:19:34 PM Confirming Pages CHAPTER 19 CONTROL VALVES 437 TABLE 19.3 Effective characteristics for a linear and an equal percentage valve (Example 19.3) Q Ppipe Pvalve Qmax x x x (gal/min) (psi) (psi) (gal/min) Linear Equal-percentage Linear, inherent 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0.22 0.86 1.94 3.44 5.38 7.74 10.54 13.77 17.42 21.51 26.03 30.98 36.35 42.16 48.40 99.78 99.14 98.06 96.56 94.62 92.26 89.46 86.23 82.58 78.49 73.97 69.02 63.65 57.84 51.60 41.76 41.62 41.39 41.07 40.66 40.15 39.54 38.82 37.98 37.03 35.95 34.73 33.35 31.79 30.03 0.05 0.10 0.14 0.19 0.25 0.30 0.35 0.41 0.47 0.54 0.61 0.69 0.78 0.88 1.00 0.13 0.33 0.45 0.53 0.60 0.66 0.70 0.75 0.79 0.82 0.86 0.89 0.93 0.96 1.00 0.07 0.13 0.20 0.27 0.33 0.40 0.47 0.53 0.60 0.67 0.73 0.80 0.87 0.93 1.00 30 25 Q, gal/min 20 15 10 Inherent linear 5 Effective equal % Effective linear 0 0.00 0.20 0.60 0.40 x, fraction lift 0.80 1.00 FIGURE 19–6 Comparison of effective characteristics for linear and equal-percentage valves from Example 19.3. cou9789x_ch19_423_442.indd 437 8/22/08 2:19:35 PM Confirming Pages 438 PART 5 PROCESS APPLICATIONS BENEFIT OF AN EQUAL-PERCENTAGE VALVE. It is often stated in the control literature that the benefit derived from an equal-percentage valve arises from its inherent nonlinear characteristic that compensates for the line loss to give an effective valve characteristic that is nearly linear. A study of Fig. 19–6 shows that in this example an equal-percentage valve overcompensates for line loss and produces an effective characteristic that is not linear, but is bowed in the opposite direction to that of the effective characteristic of the linear valve. In summary, neither valve in this example produces an effective characteristic that is linear. One can show that as the line loss increases, the linear valve will depart more from the ideal linear relation and the equal-percentage valve will move more closely toward the linear relation. In practice, a valve designated as linear will not give a linear characteristic exactly as defined in this chapter. To achieve a truly linear characteristic would require very careful design and precision machining of the valve plug and seat. The same comment can be made for an equal-percentage valve, as defined by Eq. (19.11). To know the effective characteristic of a valve, one must test it experimentally. 19.4 VALVE POSITIONER The operation of an ideal air-to-open control valve is shown in Fig. 19–7a. Any given air pressure signal to the valve results in a unique stem position x. The friction in the packing and guiding surfaces of a control valve often causes a control valve to exhibit hysteresis, as shown in Fig. 19–7b. When the air pressure increases to the valve top, the stem position increases along the lower curve. When the air pressure decreases, the stem position decreases along the upper curve. At the moment the air pressure signal reverses, the stem position stays in the last position until the dead band H is exceeded, after which the pressure begins to decrease or increase along the paths shown by the arrows. If the valve is subjected to a slow periodic variation in pressure, a typical path taken by the stem position is shown by the closed curve ABCDA in Fig. 19–7b. No hysteresis Hysteresis D x, stem position x, stem position B C Closing A B Opening H A pv, signal to valve (a) pv, signal to valve (b) FIGURE 19–7 Control valve hysteresis. cou9789x_ch19_423_442.indd 438 8/22/08 2:19:36 PM Confirming Pages CHAPTER 19 CONTROL VALVES 439 The net result of this behavior is that the same air pressure signal to the valve top corresponds to two different stem positions, depending upon whether the valve is in the process of opening or closing. The hysteresis described in the previous paragraph should be distinguished from the dynamic lag of a valve discussed in Chap. 9. The dynamic lag discussed in Chap. 9 is caused by the volume of space above the valve diaphragm, the resistance to flow of air to the valve top, and the inertia of the valve stem and plug; such a lag is expressed by a first-order or second-order transfer function. On the other hand, hysteresis, which is caused by the friction between the stem and the packing, is a nonlinear phenomenon and cannot be expressed by a transfer function. A valve can exhibit both dynamic lag and hysteresis. The presence of hysteresis in the valve can cause the controlled signal to exhibit an oscillation or ripple called a limit cycle. Since this limit cycle is usually considered objectionable and contributes to the wear of the valve, a method is needed to eliminate it. Since the limit cycle is a nonlinear phenomenon related to the hysteresis, controller tuning is not a solution to the problem. To reduce the deleterious effect of hysteresis and to also speed up the response of the valve, one can attach to the control valve a positioner that acts as a high-gain proportional controller that receives a set point signal from the primary controller and a measurement from the valve stem position. In this sense, the addition of a valve positioner introduces a form of cascade control, which was discussed in a previous chapter. A sketch of a control valve with a positioner attached is shown in Fig. 19–8. The positioner, bolted to the valve actuator, has an arm that is clamped to the valve stem to detect the stem position. Notice that the valve positioner shown in Fig. 19–8 has the usual connections for a controller: a set point that calls for a desired stem position in the form of a signal from the primary controller pc, a measurement in the form of stem position x, and a pneumatic output in the form of a pressure to the valve top pv. Some positioners are now electronic microprocessor-based controllers, while others are still pneumatically based. Valve positioners are especially important for speeding up the valve motion and eliminating hysteresis and valve stem friction. Output to valve pv Valve positioner Signal from controller pc Valve position indicator pv x Arm attached to stem to sense valve position FIGURE 19–8 Control valve with positioner. (Compare with Fig. 19–1.) cou9789x_ch19_423_442.indd 439 8/22/08 2:19:36 PM Confirming Pages 440 PART 5 PROCESS APPLICATIONS SUMMARY The control valve is a component of a control system often overlooked in a course on process control. In this chapter, the description, selection, and sizing of pneumatic control valves were presented. Valves may be of the pressure-to-close or the pressureto-open type; the selection of the type is often related to safety considerations. If the air pressure fails, the valve should return to a position that ensures safe operating conditions for a process. The flow capacity of a valve is based on an equation relating flow to the square root of the pressure drop across the valve; the proportionality constant Cv in this equation is a measure of the valve’s capacity for flow—the larger Cv, the larger the flow. Valves are classified according to their inherent flow characteristics such as linear or equal-percentage. A linear valve produces a flow (for constant pressure drop across the valve) that is proportional to the valve stem position, which in turn is proportional to the valve-top pressure. The presence of a long, small-diameter line supplying a valve causes the pressure drop across the valve to decrease with the increase of flow, for a fixed overall pressure drop across the system. If the pressure drop in the line is excessive, the characteristic of the linear valve will become nonlinear and in terms of control theory, the steady-state gain Kv of the valve decreases with flow. As a result of the change in valve gain, the controller in the loop must be readjusted for different flow rates to maintain the same degree of stability. To overcome this limitation of the linear valve, an equal-percentage (or logarithmic) valve is available for which the gain of the valve increases with flow rate. Such a valve compensates for the line loss and produces an effective characteristic that approaches a linear relation. The basis for the name equal-percentage (or logarithmic) is related to one form of the mathematical expression that describes the valve. In this form, an equal-percentage change in flow occurs for a specified change in stem position, regardless of the stem position. To eliminate hysteresis, which can produce cycling and cause wear of the valve plug and seat, a valve positioner may be attached to a control valve. The positioner also speeds up the motion of the valve in response to a signal from the controller. PROBLEMS 19.1. A linear control valve having a Cv of 0.1 is connected to a source of water. If the pressure drop across the valve is 400 psi and if the pneumatic pressure to the valve top is 12 psig, what is the flow rate through the valve? The valve goes from completely shut to completely open as the valve-top pressure varies from 3 to 15 psig. 19.2. (a) Under what conditions would an equal-percentage valve be used instead of a linear valve? (b) What are some reasons to use a valve positioner? cou9789x_ch19_423_442.indd 440 8/22/08 2:19:36 PM Confirming Pages CHAPTER 19 CAPSULE SUMMARY Air signal Air signal Diaphragm Diaphragm ACTUATOR Packing ACTUATOR Packing Stem Stem VALVE VALVE Supply Delivery Plug Seat Air-to-Close Valve Supply Delivery Plug Seat Air-to-Open Valve FIGURE 19–1 Single-seated control valves. Control valves can be air-to-open or air-to-close. The choice is usually made so that the valve fails in the safe position upon loss of air signal pressure. Valve sizing equation where q ⫽ Cv f ( x ) ∆pvalve sg (19.1) q flow rate, gal/min x fractional stem position (i.e., fraction open) ∆ pvalve pressure drop across valve, psi f (x) fraction of maximum flow ( 1 for fully open) Cv factor associated with capacity of valve sg specific gravity of fluid (water sg 1) Equation (19.1) applies to the flow of an incompressible, nonflashing fluid through a fully open valve. Factor Cv can be defined as the flow (gal/min) of a fluid of unit specific gravity (water) through a fully open valve, across which a pressure drop of 1.0 psi exists. Therefore, the bigger Cv, the larger the valve. 441 cou9789x_ch19_423_442.indd 441 8/22/08 2:19:37 PM Confirming Pages 442 PART 5 PROCESS APPLICATIONS Sensitivity df dx Valve type Linear (I) Equal-percentage (II) Square root or quick-opening (III) Constant Increasing Decreasing Relationship f(x) x f(x) ax1 f (x ) ⫽ x f, fraction of maximum flow A valve positioner is a device that can be attached to a valve that drives the valve to the desired position, in spite of friction or hysteresis. 1.0 III 0.5 0 I II 1.0 0.5 0 x, fraction of maximum lift FIGURE 19–3 Inherent valve characteristics (pressure drop across valve is constant). I: Linear, II: increasing sensitivity (e.g., equal-percentage valve), and III: decreasing sensitivity (e.g., square root valve). cou9789x_ch19_423_442.indd 442 8/22/08 2:19:37 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES T o investigate theoretically the control of a process, it is necessary first to know the dynamic character of the process that is being controlled. In previous chapters, the processes have been very simple for the purpose of illustrating control theory. Many physical processes are extremely complicated, and it requires considerable effort to construct a mathematical model that will adequately simulate the dynamics of the actual system. In this chapter, we analyze several complex systems to indicate some of the types of problems that can be encountered. In these examples, the technique of linearization, first presented in Chap. 5, is applied to a function of several variables. One example leads to a multiloop control system. In Sec. 20.3 distributed-parameter systems are discussed. 20.1 CONTROL OF A STEAM-JACKETED KETTLE The dynamic response and control of the steam-jacketed kettle shown in Fig. 20–1 are to be considered. The system consists of a kettle through which water flows at a variable rate w lb/time. The entering water is at temperature Ti, which may vary with time. The kettle water, which is well agitated, is heated by steam condensing in the jacket at temperature Tv and pressure pv. The temperature of the water in the kettle is measured and transmitted to the controller. The output signal from the controller is used to change the stem position of the valve, which adjusts the flow of steam to the jacket. The major problem in this example is to determine the dynamic characteristics of the kettle. The kettle is actually a nonlinear system, and to obtain a linear model, a number of simplifying assumptions are needed. 443 cou9789x_ch20_443-474.indd 443 8/22/08 12:27:43 PM Confirming Pages 444 PART 5 PROCESS APPLICATIONS Computer/Controller Temperature-measuring element w Water Ti Control valve Tv Jacket Pv wall wc Tc Condensate m To wv Steam w To FIGURE 20–1 Control of a steam-jacketed kettle. Analysis of Kettle The following assumptions are made for the kettle: 1. 2. 3. 4. 5. 6. The heat loss to the atmosphere is negligible. The holdup volume of water in the kettle is constant. The thermal capacity of the kettle wall, which separates steam from water, is negligible compared with that of the water in the kettle. The thermal capacity of the outer jacket wall, adjacent to the surroundings, is finite, and the temperature of this jacket wall is uniform and equal to the steam temperature at any instant. The kettle water is sufficiently agitated to result in a uniform temperature. The flow of heat from the steam to the water in the kettle is described by the expression q ⫽ U ( Tv ⫺ To ) where q ⫽ flow rate of heat, Btu/(h · ft2) U ⫽ overall heat-transfer coefficient, Btu/(h · ft2 · °F) Tv ⫽ steam temperature, °F To ⫽ water temperature, °F The overall heat-transfer coefficient U is constant. 7. 8. 9. cou9789x_ch20_443-474.indd 444 The heat capacities of water and the metal wall are constant. The density of water is constant. The steam in the jacket is saturated. 9/4/08 10:44:56 AM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 445 The assumptions listed here are more or less arbitrary. For a specific kettle operating under a particular set of conditions, some of these assumptions may require modification. The approach to this problem is to make an energy balance on the water side and another energy balance on the steam side. To aid the development of the transfer functions, a schematic diagram of the kettle is shown in Fig. 20–2. The symbols used throughout this analysis are defined as follows: Ti temperature of inlet water, °F To temperature of outlet water, °F Tv temperature of jacket steam, °F Tc temperature of condensate, °F w flow rate of inlet water, lb/time wv flow rate of steam, lb/time wc flow rate of condensate from kettle, lb/time m mass of water in kettle, lb m1 mass of jacket wall, lb V volume of jacket steam space, ft3 C heat capacity of water, Btu/(lb · °F) C1 heat capacity of metal in jacket wall, Btu/(lb · °F) A cross-sectional area for heat exchange, ft2 t time Hv specific enthalpy of steam entering, Btu/lb Hc specific enthalpy of condensate leaving, Btu/lb Uv specific internal energy of steam in jacket, Btu/lb rv density of steam in jacket, lb/ft3 m1 Steam wv A Tv Pv To m wc Tc Water w Ti w To FIGURE 20–2 Schematic diagram of kettle. An energy balance on the water side gives wC ( Ti To ) UA ( Tv To ) mC dTo dt (20.1) In Eq. (20.1), the terms C, U, A, and m are constants. The first term in Eq. (20.1) is nonlinear, since it contains the product of flow rate and temperature, that is, wTi and wTo. cou9789x_ch20_443-474.indd 445 9/4/08 10:46:33 AM Confirming Pages 446 PART 5 PROCESS APPLICATIONS To obtain a transfer function from Eq. (20.1), these nonlinear terms must be linearized. Before continuing the analysis, we digress briefly to discuss the general problem of linearization of a function of several variables. Recall from Chap. 5 [Eq. (5.39)] that a function of two variables z(x, y) can be expanded around an operating point xs, ys by means of a Taylor series expansion: z z ( xs , ys ) z x ( x xs ) xs , ys z ( y ys ) y xs , ys (20.2) h igher-order terms in x xs and y ys The subscript s stands for steady state. In control problems, the operating point (xs, ys), around which the expansion is to be made, is selected at steady-state values of the variables before any disturbance occurs. Linearization of the function z consists of retaining only the linear terms, on the basis that the deviations (e.g., x xs) will be small. Thus, z ⬵ zs z xs ( x xs ) z ys ( y ys ) (20.3) where z xs and z ys are the partial derivatives in Eq. (20.2). If z is a function of three or more variables, the linearized form is the same as that of Eq. (20.3) with an additional term for each variable. The linearization expressed by Eq. (20.3) may be applied to the terms wTi and wTo in Eq. (20.1) to obtain wTi ws Tis ws ( Ti Tis ) Tis ( w ws ) (20.4) wTo ws Tos ws ( To Tos ) Tos ( w ws ) (20.5) and Notice that for these cases the nonlinear terms are wTi and wTo. The first partial derivatives, evaluated at the operating point, are ( wTis ) Tis w ws ,Tis ( wTis ) ws w ws ,Ti and so on. Introducing Eqs. (20.4) and (20.5) into Eq. (20.1) gives the following linearized equation: [(Ti s cou9789x_ch20_443-474.indd 446 Tos ) ( w ws ) ws ( Ti To )] C UA ( Tv To ) mC dTo dt (20.6) 8/22/08 12:27:44 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 447 At steady state, dTo/dt 0, and Eq. (20.1) can be written wsC ( Tis Tos ) UA ( Tvs Tos ) 0 (20.7) Subtracting Eq. (20.7) from Eq. (20.6), introducing the deviation variables Ti Ti Tis To To Tos Tv Tv Tvs W w ws and rearranging give the result ) ) d To C ( Tis Tos ) W ws Ti To UA Tv To mC dt ( ( (20.8) Taking the transform of Eq. (20.8) and solving for To (s) give To ( s ) K1 K2 K3 Ti ( s ) Tv ( s ) W(s) t ws 1 t ws 1 t ws 1 (20.9) wsC UA wsC UA K2 UA wsC C ( Tos Tis ) K3 UA wsC mC tw UA wsC K1 where From Eq. (20.9), we see that the response of To to Ti, Tv, or W is first-order with a time constant tw. The steady-state gains (Ks) in Eq. (20.9) are all positive. The following energy balance can be written for the steam side of the kettle: w v Hv in by flow w c Hc out by flow UA ( Tv To ) + o u t by convection Vd ( rvU v ) dt accumula t ion in steam dTv m1C1 t d (20.10) accumulation in kettle wall Notice that we have made use of assumption 4 in writing the last term of Eq. (20.10), which implies that the metal in the outer jacket wall is always at the steam temperature. cou9789x_ch20_443-474.indd 447 8/22/08 12:27:45 PM Confirming Pages 448 PART 5 PROCESS APPLICATIONS A mass balance on the steam side of the kettle yields wv wc V d rv dt (20.11) Combining Eqs. (20.10) and (20.11) to eliminate wc gives wv ( H v Hc ) (U v Hc ) V Vr v d rv dTv m1C1 UA ( Tv To ) dt dt dU v dt (20.12) The variables rv, Uv, Hv, and Hc are functions of the steam and condensate temperatures and can be approximated by expansion in Taylor series and linearization as follows: r v r vs a ( Tv Tvs ) U v U vs f ( Tv Tvs ) H v H vs g ( Tv Tvs ) (20.13) Hc Hcs s ( Tc Tcs ) dv dTv s dU v dTv s dH v g dTv s dHc s dTc s a where f The parameters a, f, g, and s in these relationships can be obtained from the steam tables once the operating point is selected. For example, if the operating point is at 212°F and the deviation in steam temperature is 10°F, we obtain the following estimate of g from the steam tables: Tvs 212 F H vs 1150 . 4 Btu / lb Hv 1154.1 at Tv 222°F Hv 1146.6 at Tv 202°F g ⬇ 1154 . 1 1146 . 6 0 . 375 222 202 Hv 1150.4 0.375(Tv 212) cou9789x_ch20_443-474.indd 448 8/22/08 12:27:46 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 449 In a similar manner, the properties of saturated steam can be used to evaluate a, f, and s. Introducing the relationships of Eq. (20.13) into Eq. (20.12) and assuming the condensate temperature Tc to be the same as the steam temperature Tv give the following result: H vs Hcs (g s ) ( Tv Tvs ) wv f m1C1 (U vs Hcs ) ( 2f s ) ( Tv Tvs ) r vs a aV dTv aV UA(Tv To ) dt (20.14) Some of the terms in Eq. (20.14) can be neglected. The term (g s ) ( Tv Tvs ) can be dropped because it is negligible compared with H vs Hcs . For example, for steam at atmospheric pressure, a change of 10°F gives a value of (g s ) ( Tv Tvs ) of about 7 Btu/lb while H vs Hcs is 970 Btu/lb. Similarly, the term ( 2f s ) ( Tv Tvs ) can be neglected. For example, this term is about 4 Btu/lb for a change in steam temperature of 10°F for steam at about 1 atm pressure; the term U vs Hcs is 897 Btu/lb under these conditions. Also, the term fr vs /a is about 15 Btu/lb and can be neglected. Discarding these terms, writing the remaining terms in deviation variables, and transforming yield Tv ( s ) where K5 1 To ( s ) Wv ( s ) t vs 1 t vs 1 (20.15) Tv Tv Tvs Wv Wv Wvs K5 tv H vs Hcs UA (U vs Hcs )aV m1C1 UA From Eq. (20.15), we see that the steam temperature Tv depends on the steam flow rate Wv and the water temperature To. The combination of Eqs. (20.9) and (20.15) gives the dynamic response of the water temperature to changes in water flow rate, inlet water temperature, and steam flow rate. These equations are represented by a portion of the block diagram of Fig. 20–4. Before completing the analysis of the control system, we must consider the effect of valve stem position on the steam flow rate. cou9789x_ch20_443-474.indd 449 8/22/08 12:27:47 PM Confirming Pages 450 PART 5 PROCESS APPLICATIONS Analysis of Valve The flow of steam through the valve depends on three variables: steam supply pressure, steam pressure in the jacket, and the valve stem position, which we assume to be proportional to the pneumatic valve-top pressure p. For simplicity, assume the steam supply pressure to be constant with the result that the steam flow rate is a function of only the two remaining variables; thus wv f ( p, pv ) (20.16) Because of the assumption that the steam in the jacket is always saturated, we know that pv is a function of Tv; thus pv g ( Tv ) (20.17) This functional relation can be obtained from the saturated steam tables. Equations (20.16) and (20.17) can be combined to give wv f [ p, g ( Tv )] f1 ( p, Tv ) The function f1( p, Tv) is in general nonlinear, and if an analytic expression is available, the function can be linearized as described previously. The flow of steam through a control valve can often be represented by the relationship wv A0Cv ps pv where (20.18) ps supply pressure of steam pv pressure downstream of valve A0 cross-sectional area for flow of steam through valve Cv constant of valve For a linear valve, A0 is proportional to stem position, and the stem position is proportional to the valve-top pressure p; under these conditions, Eq. (20.18) takes the form wv Cv p ps pv (20.19) For this example, however, we assume that an analytic expression is not available. The linearized form of f1( p, Tv) can be obtained by making some experimental tests on the valve. If the valve-top pressure is fixed at its steady-state (or average) value and wv is measured for several values of Tv (or pv), a curve such as the one shown in Fig. 20–3a can be obtained. If the steam temperature Tv (or pv) is held constant and the flow rate is measured at several values of valve-top pressure, a curve such as that shown in Fig. 20–3b can be obtained. These two curves can now be used to evaluate the partial derivatives in the linear expansion of f1( p, Tv) as we now demonstrate. cou9789x_ch20_443-474.indd 450 8/22/08 12:27:48 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 451 Expanding wv about the operating point ps , Tvs and retaining only the linear terms give w v w vs wv p (p ps ) ps ,Tvs wv Tv ( Tv Tvs ) ps ,Tvs This equation can be written in the form Wv K v P where 1 Tv Rv (20.20) Wv wv wvs P p ps Tv Tv Tvs Kv wv p ps ,Tvs w 1 v Tv Rv ps ,Tvs The coefficients Kv and 1/Rv in Eq. (20.20) are the slopes of the curves of Fig. 20–3 at the operating point ps , Tvs . This follows from the definition of a partial derivative. Notice that 1/Rv has been defined as the negative of the slope so that Rv is a positive quantity. The experimental approach described here for obtaining a linear form for the flow characteristics of a valve is always possible in principle. However, it must be emphasized that the linear form is useful only for small deviations from the operating point. If the operating point is changed considerably, the coefficients Kv and 1/Rv must be reevaluated. Notice that, in writing Eq. (20.20), we have assumed the valve to have no dynamic lag between p and stem position. This assumption is valid for a system having large time constants, such as a steam-jacketed kettle, as was demonstrated in Chap. 9. Slope = ∂wv 1 =− ∂Tv s Rv Slope = wv wv wvs wvs ∂wv = Kv ∂p s p Tvs Tv (a) p = ps ps p (b) Tv = Tvs wv pv or Tv (c) FIGURE 20–3 Linearization of valve characteristics from experimental tests. cou9789x_ch20_443-474.indd 451 8/22/08 12:27:49 PM Confirming Pages 452 PART 5 PROCESS APPLICATIONS Block Diagram of Control System We have now completed the analysis of the kettle and valve. A block diagram of the control system, based on Eqs. (20.9), (20.15), and (20.20), is shown in Fig. 20–4. G1 K1 T i′ w s+1 − G3 K3 w s+1 W + − 1 Rv G5 R + − Gc P Kv − Wv + K5 v s+1 G2 + T ′v + K2 w s+1 + + T ′o G4 1 s+1 v H FIGURE 20–4 Block diagram for control of steam-jacketed kettle. The controller action is not specified but merely denoted by Gc in the block diagram. Also, the feedback element is denoted as H. From Fig. 20–4, we see that the steamjacketed kettle is a multiloop control system. Furthermore, the loops overlap. The block diagram can be used to obtain the overall transfer function between any two variables by applying the methods of Chap. 12. After considerable algebraic manipulation, the following result is obtained: To Gc G2G5 K v G1 ( 1 G5 / Rv ) G3 ( 1 G5 / Rv ) R Ti W (20.21) D (s) D (s) D (s) where D(s) 1 G5/Rv Gc G2 G5 Kv H G2 G4. The terms G1, G2, G3, G4, G5, Gc, and H are defined in Fig. 20–4. For example, if Gc Kc and H 1, one obtains from Eq. (20.21) the transfer function To K 2 2 R t s 2zts 1 cou9789x_ch20_443-474.indd 452 (20.22) 8/22/08 12:27:50 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES K where t2 2zt 453 Kc K v K 2 K5 D1 t vt w D1 t v t w K 5t w / Rv D1 D1 1 K5 Kc K v K 2 K5 K 2 Rv It is seen that the response of the control system is second-order when proportional control is used and the measuring element does not have dynamic lag. Notice that the parameters K, t2, and 2z t in Eq. (20.22) are positive. This follows from the fact that the parameters Kc, Kv, K2, K5, Rv, tv, and tw are all positive and that K2 < 1. When a block diagram of a control system becomes very complicated, such as the one in this example, it is convenient to simulate the control system with a software package such as Simulink. 20.2 DYNAMIC RESPONSE OF A GAS ABSORBER Another example of a complex system is the plate absorber shown in Fig. 20–5. The reader who has not studied gas absorption may find this subject presented in any textbook on chemical engineering unit operations; for example, see McCabe, Smith, and Harriott (2004). Water L3 x3 V2 y2 Plate 2 Bubble cap Downcomer weir L1 x1 V1 y1 L2 x2 Plate 1 V0 y0 Airammonia FIGURE 20–5 Bubble-cap absorber. cou9789x_ch20_443-474.indd 453 8/22/08 12:27:51 PM Confirming Pages 454 PART 5 PROCESS APPLICATIONS In this process, air containing a soluble gas such as ammonia is contacted with fresh water in a two-plate column in order to remove part of the ammonia from the gas. The action of gas bubbling through the liquid causes thorough mixing of the two phases on each plate. During the mixing process, ammonia diffuses from the bubbles into the liquid. In an industrial operation, many plates may be used; however, for simplicity, we consider only two plates in this example, since the basic principles are unaffected by the number of plates. Our problem is to analyze the system for its dynamic response. In other words, we want to know how the concentrations of liquid and gas change as a result of change in inlet composition of the gas stream to be treated. We will consider the inlet flow of the liquid phase to be constant. Throughout the analysis, the following symbols are used: Ln flow of liquid leaving nth plate, mol/min Vn flow of gas leaving nth plate, mol/min xn concentration of liquid leaving nth plate, mole fraction NH3 yn concentration of gas leaving nth plate, mole fraction NH3 Hn holdup (or storage) of liquid on nth plate, mole To avoid too many complicating details, the following assumptions will be used: 1. 2. 3. 4. The temperature and total pressure throughout the column are uniform and do not vary with changes in flow rate. The entering gas stream is dilute (say 5 mol % NH3) with the consequence that we can neglect the decrease in total molar flow rate of gas as ammonia is removed. Likewise, we can assume that the molar flow rate of liquid does not increase as ammonia is added. The plate efficiency is 100 percent, which means that the vapor and liquid streams leaving a plate are in equilibrium. Such a plate is called an ideal equilibrium stage. The equilibrium relationship is linear and is given by the expression yn mxn∗ b (20.23) where m and b are constants that depend on the temperature and total pressure of the system and xn∗ is the concentration of liquid in equilibrium with gas of concentration yn. For an ideal plate xn xn∗ (If the efficiency of the plate is not 100 percent, we can introduce an individual tray efficiency of the Murphree type, defined as En xn xn 1 xn∗ xn 1 where xn∗ is the concentration of the liquid in equilibrium with gas of composition ∗ yn. Notice that for an ideal plate En 1 and xn xn . In general, the efficiency cou9789x_ch20_443-474.indd 454 8/22/08 12:27:51 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 5. 6. 7. 455 of a plate depends on the design of the plate, the properties of the gas and liquid streams, and the flow rates. We could include efficiency in our mathematical model; however, to do so would greatly increase the complexity of the problem. To account properly for the variation in efficiency with flow rates would require empirical relationships for a specific plate design. In this list of assumptions, the one which is most likely to be invalid for a practical process is that the plate is an ideal equilibrium stage.) The holdup of liquid Hn on each plate is constant and independent of flow rate. Furthermore, the holdup is the same for each plate, that is, H1 H2 H. The inlet liquid flow is constant. This assumption, in addition to assumptions 5 and 2, leads us to the conclusion that the liquid flows throughout the column are constant and equal, that is, L3 L2 L1 L. The holdup of gas between plates is negligible. As a consequence of this assumption and assumption 2, the flow rate of gas from each plate is the same and equal to the entering gas flow rate; that is, V0 V1 V2 V Analysis We begin the analysis of this process by writing an ammonia balance around each plate. A mass balance on ammonia around plate 1 gives H dx1 Lx2 Vy0 Lx1 Vy1 dt (20.24) This last equation states that the accumulation of NH3 on plate 1 is equal to the flow of NH3 into the plate minus the flow of NH3 out of the plate. Notice that V, L, and H do not have subscripts because of assumptions 5 through 7. A mass balance on ammonia around plate 2 gives H dx2 Vy1 Lx2 Vy2 dt (20.25) The last equation does not contain a term Lx3, since we have assumed that x3 0 (pure water). For an ideal plate xn x*n, and the equilibrium relation of Eq. (20.23) becomes yn mxn b Substituting the equilibrium relationship into Eqs. (20.24) and (20.25) gives H and cou9789x_ch20_443-474.indd 455 dx1 Lx2 Lx1 Vm ( x0 x1 ) dt H dx2 Vm ( x1 x2 ) Lx2 dt 8/22/08 12:27:52 PM Confirming Pages 456 PART 5 PROCESS APPLICATIONS where x0 ( y0 b)/m is the composition of liquid that would be in equilibrium with the entering gas of composition y0. Solving these last two equations for the derivatives gives dx1 L Vm ( x2 x1 ) ( x0 x1 ) dt H H (20.26) dx2 Vm L ( x1 x2 ) x2 dt H H (20.27) The analysis has resulted in two first-order differential equations. The forcing functions in this process, which must be specified as functions of t, are the inlet gas concentration x0 ( y0 b)/m and the inlet liquid flow rate L3 L. We can now solve for x1(t) and x2(t). Equations (20.26) and (20.27) can be written dx1 ax1 bx2 cx0 dt (20.28) dx2 cx1 ax2 dt (20.29) L Vm H H L b H Vm c H where a At steady state, dx1/dt dx2/dt 0, and Eqs. (20.30) and (20.31) can be written 0 ax1s bx2s cx0 s (20.30) 0 cx1s ax2s (20.31) Subtracting these steady-state equations from Eqs. (20.28) and (20.29) and introducing the deviation variables X1 x1 x1s , X 2 x2 x2 s , and X 0 x0 x0 s give dX1 aX1 bX 2 cX 0 dt (20.32) dX 2 cX1 aX 2 dt (20.33) Notice that X0 Y0 /m because cou9789x_ch20_443-474.indd 456 8/22/08 12:27:52 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 457 X 0 x0 x0s y0 b y0 b y0 y0 s Y0 X0 − s m m m m Equations (20.32) and (20.33) can be transformed to give sX1 aX1 bX 2 cX 0 sX 2 cX1 aX 2 We now have two algebraic equations and three unknowns (X1, X2, and X0). Solving this pair of equations to eliminate X1 and replacing X2 by Y2/m and X0 by Y0/m give the transfer function ( ) c 2 / a 2 bc Y2 ( s ) Y0 ( s ) 1 / a 2 bc s 2 2a / a 2 b c s 1 ) ( ( ) (20.34) This result shows that the response of outlet gas concentration to a change in inlet gas concentration is second-order. Equation (20.34) is of the standard second-order form K/(t2s2 2z t s 1) with the parameters t2 1 a bc 2 2zt and 2a a bc 2 Solving these two equations to eliminate t gives z 1 1 bc / a 2 Writing a and b in terms of the original system parameters (L, H, V, m) gives ( L / H ) (Vm / H ) z 1 2 ( L / H Vm / H ) −1 / 2 Simplifying this expression gives Vm / L z 1 2 (1Vm / L ) −1/ 2 Since Vm/L > 0, we see that z > 1, meaning that the response is overdamped. If the analysis is repeated for a gas absorber containing n plates, it will be found that the response between inlet gas concentration and outlet gas concentration is nth-order. cou9789x_ch20_443-474.indd 457 8/22/08 12:27:53 PM Confirming Pages 458 PART 5 PROCESS APPLICATIONS 20.3 DISTRIBUTED-PARAMETER SYSTEMS Heat Conduction into a Solid In Chap. 4, the analysis of the mercury thermometer was based on a “lumped parameter” model. At that time, reference was made to a distributed-parameter model of the thermometer. To illustrate the difference between a lumped-parameter system and a distributed-parameter system, consider a slab of solid conducting material of infinite thickness, as shown in Fig. 20–6. Let the input to this system be the temperature at the left face (x 0), which is some arbitrary function of time. The output will be the temperature at the position x L. For convenience, we may consider this system to represent the response of a bare thermocouple embedded in a thick wall, as the surface of the wall experiences a variation in temperature. The conductivity k, heat capacity C, and density r of the conducting material are constant, independent of temperature. Initially (t < 0), the slab is at a uniform steady-state temperature. Therefore in deviation variables, which will be used henceforth, the initial temperature is zero. The cross-sectional area of the slab is A. T(0,t ) 0 x x+ x L FIGURE 20–6 Heat conduction in a solid. ANALYSIS. In this problem the temperature in the slab is a function of position and time and is indicated by T(x,t). The temperature at the surface is indicated by T(0,t), and that at x L by T(L,t). To derive a differential equation that describes the heat conduction in the slab, we first write an energy balance over a differential length ∆ x of the slab. This energy balance can be written Flow of heat into left face by conduction Flow of heat out Rate of accumulation of right face of in ternal energy in by condd uction volume element (20.35) The flow of heat by conduction follows Fourier’s law: q k where cou9789x_ch20_443-474.indd 458 T x (20.36) q heat flux by conduction ì T/ì x temperature gradient k thermal conductivity 9/4/08 10:46:35 AM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 459 Applying Eq. (20.36) to Eq. (20.35) gives Ak T x x T Ak x t [C rA ∆x ( T Tr )] x ∆x (20.37) where Tr is the reference temperature used to evaluate internal energy. If we divide both sides by the volume of the element, and rearrange, we obtain Ak T x x ∆x Ak T x k x A ∆x T x x ∆x k T x Cr x ∆x [(T Tr )] t (20.38) Now, if we let the volume of the element shrink to zero, we obtain the fundamental equation describing conduction in a solid k 2T T rC 2 t x This is often written as a 2 T T 2 t x (20.39) where a k/(r C) is defined as the thermal diffusivity. Several points are worth mentioning at this time. In this analysis, we have allowed the capacity for storing heat (r CA per unit length of x) and the resistance to heat conduction [1/(kA) per unit length of x] to be “spread out” or distributed uniformly throughout the medium. This distribution of capacitance and resistance is the basis for the term distributed parameter. The analysis has also led to a partial differential equation, which in general is more difficult to solve than the ordinary differential equation that results from a lumped-parameter model. TRANSFER FUNCTION. We are now in a position to derive a transfer function from Eq. (20.39). First notice that since T is a function of both time t and position x, a transfer function may be written for an arbitrary value of x. In this problem, the temperature is to be observed at x L; hence the transfer function will relate T(L,t) to the temperature at the left surface T(0,t) which is taken as the forcing function. Equation (20.39) will be solved by the method of Laplace transforms. Taking the Laplace transform of both sides of Eq. (20.39) with respect to t gives a∫ 2 T 2 0 x ( x , t ) est dt T ∫0 t ( x , t ) est dt (20.40) Consider first the integral on the left side of Eq. (20.40). Interchanging the order of integration and differentiation results in cou9789x_ch20_443-474.indd 459 8/22/08 12:27:55 PM Confirming Pages 460 PART 5 PROCESS APPLICATIONS ∫ 2 T 2 0 x ( x , t ) est dt 2 x 2 ∫0 T ( x , t ) est dt d 2T ( x , s ) dx 2 (20.41) where T ( x, s ) is the Laplace transform of T(x,t). This interchange is allowed for most functions of engineering interest (Churchill,1972). In this chapter the overbar is often used to indicate the Laplace transform of a function of two variables. Note that the presence of x has no effect on the second integral of Eq. (20.41) because the integration is with respect to t. Also note that the derivative on the right side of Eq. (20.41) is taken as an ordinary derivative because T(x,s) will later be seen to be a function of only one independent variable x and a parameter s. Next consider the integral on the right side of Eq. (20.40). Again, the presence of x has no effect on the integration with respect to t, and the rule for the transform of a derivative may be applied directly to yield T ∫0 t ( x , t ) est dt sT ( x , s ) T ( x , 0 ) (20.42) where T(x,0) is the initial temperature distribution in the solid. Introducing the results of the transformation into Eq. (20.40) gives a d 2T ( x , s ) sT ( x , s ) T ( x , 0 ) dx 2 (20.43) The partial differential equation has now been reduced to an ordinary differential equation, which can usually be solved without difficulty. It should be clear that s in Eq. (20.43) is merely a parameter, with the result that this equation is an ordinary secondorder differential equation in the independent variable x. This follows because there are no derivatives with respect to s in Eq. (20.43). Since we have taken T(x,0) 0 for the example under consideration, Eq. (20.43) becomes d 2T s T 0 a dx 2 (20.44) Equation (20.44) is a linear differential equation and can be solved to give T A1e (s / a ) x A2e (s / a )x (20.45) The arbitrary coefficients A1 and A2 may be evaluated as follows: In order that T may be finite as x → , it is necessary that A2 0. Equation (20.45) then becomes T A1e (s / a ) x (20.45a) The transformed forcing function at x 0 is T ( 0, s ) , which can be substituted into Eq. (20.45a) to determine A1; then T ( 0, s ) A1e 0 cou9789x_ch20_443-474.indd 460 8/22/08 12:27:55 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 461 A1 T ( 0, s ) or Substituting A1 into Eq. (20.45a) gives T ( x, s ) e T ( 0, s ) (s / a )x (20.46) By specifying a particular value of x, say x L, the transfer function is T ( L, s ) e T ( 0, s ) (s / a )L (20.47) STEP RESPONSE. To illustrate the use of this transfer function, consider a forcing function that is the unit-step function; thus T ( 0, t ) u ( t ) for which case T ( 0, s ) 1 / s. Substituting this into Eq. (20.47) gives T ( L, s ) 1 e s (s / a )L (20.48) To obtain the response in the time domain, we must invert Eq. (20.48). A table of transforms gives the following transform pair: L { 1 e s (s / a )x } erfc x 4a t (20.49) where erfc x is the complementary error function of x defined as erfc x 1 2 p x u 2 ∫0 e du This function is tabulated in many textbooks and mathematical tables. Using this transform pair, Eq. (20.48) becomes T ( L , t ) erfc 1 a t 1 / 2 L erfc 2 4a t 2 L (20.50) A plot of T versus the dimensionless group a t/L2 is shown in Fig. 20–7. cou9789x_ch20_443-474.indd 461 8/22/08 12:27:56 PM Confirming Pages 462 PART 5 PROCESS APPLICATIONS 1.0 T 0.8 0.6 0.4 0.2 0 0 1 2 3 t L2 FIGURE 20–7 Response of temperature in the interior of a solid to a unit-step change in temperature at the surface. SINUSOIDAL RESPONSE. It is instructive to consider the response in temperature at x L for the case where the forcing function is a sinusoidal variation; thus T ( 0, t ) A sinw t Using the substitution rule of Chap. 15, in which s is replaced by jw, Eq. (20.47) becomes T ( L , jw ) e T ( 0 , jw ) ( jw / a ) L (20.51) To obtain the AR and phase angle requires that the magnitude and argument of the right side of Eq. (20.51) be evaluated. This can be done as follows: First write j in polar form (see Fig. 20–8); thus j e ( / 2) j 1 e (p / 2 ) j p hase ang le magnitude from which we get ( j e j (p / 2) )1 / 2 e jp / 4 1 (1 j ) 2 Im /2 Re FIGURE 20–8 Polar representation of j. cou9789x_ch20_443-474.indd 462 8/22/08 12:27:58 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES Substituting the positive form of 463 j into Eq. (20.51) gives T ( L , jw ) e T ( 0 , jw ) w / 2a L j w / 2a L e [Notice that the substitution of (1 j) / 2 into Eq. (20.51) leads to a result in which the AR is greater than 1 and the phase angle leads. This is contrary to the response of the physical system and is not considered as a useful solution.] From this form, we can write by inspection AR T ( L , jw ) e T ( 0 , jw ) Phase angle w / 2a L T ( L , jw ) w L T ( 0 , jw ) 2a (20.52) rad (20.53) From these results, it is seen that the AR approaches zero as w → and the phase angle decreases without limit as w → . Such a system is said to have nonminimum phase lag characteristics. With the exception of the transport lag, all the systems that have been considered up to now have given a limited value of phase angle as w → . These are called minimum phase systems and always occur for lumped-parameter systems. The nonminimum phase behavior is typical of distributed-parameter systems. Transport Lag as a Distributed-Parameter System We can demonstrate that the transport lag (distance-velocity lag) is, in fact, a distributedparameter system as follows: Consider the flow of an incompressible fluid through an insulated pipe of uniform cross-sectional area A and length L, as shown in Fig. 20–9a. The fluid flows at velocity v, and the velocity profile is flat. v To Ti (a) v To Ti (b) FIGURE 20–9 Obtaining the transfer function of a transport lag from a lumped-parameter model. cou9789x_ch20_443-474.indd 463 8/22/08 12:27:59 PM Confirming Pages 464 PART 5 PROCESS APPLICATIONS We know from Chap. 7 that the transfer function relating outlet temperature To to the inlet temperature Ti is To ( s ) e( L / v ) s Ti ( s ) Let the pipe be divided into n zones as shown in Fig. 20–9b. If each zone of length L /n is considered to be a well-stirred tank, then the pipe is equivalent to n noninteracting first-order systems in series, each having a time constant t L1 nv This expression for t is equivalent to that appearing in Eq. (8.10). Since the transfer function for flow through a tank was developed in Chap. 8, the analysis will not be repeated here. (Note that taking each zone to be a well-stirred tank is called lumping of parameters.) The overall transfer function for this lumped-parameter model is therefore n 1 To ( s ) 1 n Ti ( s ) t s 1 [( L / v ) / n ] s 1 To “distribute” the parameters, we let the size of the individual lumps go to zero by letting n → . n 1 To ( s ) lim Ti ( s ) n → [( L / v ) / n ] s 1 The thermal capacitance is now distributed over the tube length. It can be shown by use of calculus that the limit is e( L / v ) s which is the transfer function derived previously. This demonstration should provide some initial insight into the relationship between a distributed-parameter system and a lumped-parameter system and indicates that a transport lag is a distributed system. Heat Exchanger As our last example of a distributed-parameter system, we consider the double-pipe heat exchanger shown in Fig. 20–10. [The analysis presented here essentially follows that of W. C. Cohen and E. F. Johnson (1956). These authors also present the experimental results of frequency response tests on a double-pipe, steam-to-water heat exchanger.] The fluid that flows through the inner pipe at constant velocity v is heated by steam condensing outside the pipe. The temperature of the fluid entering the pipe and the steam temperature vary according to some arbitrary functions of time. The steam temperature varies with time, but not with position in the exchanger. The metal wall separating steam from fluid is assumed to have negligible thermal capacity for the purpose of the analysis. The heat transfer from the steam to the fluid depends on the heat-transfer coefficient on the steam side ho and the convective transfer coefficient on the water side hi. cou9789x_ch20_443-474.indd 464 8/22/08 12:27:59 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 465 The resistance of the metal wall is neglected. The goal of the analysis will be to find transfer functions relating the exiting fluid temperature T(L,t) to the entering fluid temperature T(0,t) and the steam temperature Tv(t). Steam Tv(t) v T(0, t) T(L, t) Condensate 0 L x x x+ x FIGURE 20–10 Double-pipe heat exchanger. The following symbols will be used in this analysis: T(x,t) fluid temperature Tv(t) steam temperature Tr reference temperature for evaluating enthalpy v fluid velocity r density of fluid C heat capacity of fluid Ai cross-sectional area for flow inside pipe Di inside diameter of inner pipe Do outside diameter of inner pipe hi convective heat-transfer coefficient inside pipe ho heat-transfer coefficient for condensing steam Ui overall heat-transfer coefficient based on inside area 1 1 / hi (1 / ho )( Di / Do ) ANALYSIS. We begin the analysis by writing a differential energy balance for the fluid inside the pipe over the volume element of length ∆ x (see Fig. 20–10). This balance can be stated as Heat transferred Flow of Flow of from steam enthalpy in enthalpy ou t to liquid Rate of accumulation of internal energy cou9789x_ch20_443-474.indd 465 (20.54) 9/4/08 10:46:35 AM Confirming Pages 466 PART 5 PROCESS APPLICATIONS The terms in this balance can be evaluated as follows: Flow of enthalpy in at x vAi rC(T Tr) T Flow of enthalpy out at x ∆x vAi rC T ∆x Tr x Heat transfer through film p Di Ui ∆ x(Tv T ) Accumulation of internal energy [ Ai r ∆x C (T Tr )] t Introducing these terms into Eq. (20.54) gives, after simplification, T T 1 v ( Tv T ) t x t (20.55) 1 pDiUi t Ai rC where We now have the differential equation that describes the dynamics of the system. As in previous problems, the dependent variables will be transformed to deviation variables. At steady state, the time derivative in Eq. (20.55) is zero, and it follows that 0 v dTs 1 ( Tvs Ts ) dx t (20.56) where the subscript s is used to denote the steady-state value. Note that to determine the steady-state temperature profile requires the solution of Eq. (20.56). We can rearrange this equation to 1 1 dTs Ts Tvs u ( t ) tv tv dx (20.57) Transforming gives ( sTs ( s ) Ts x 0 ) 1 1 Ts Tvs tv tv 1 s (20.58) Now, let Ts0 represent the steady-state inlet (at x 0) temperature of the water Ts and rearrange: Tvs /t v Ts0 Ts ( s ) s ( s 1 /t v) s 1 /t v x0 The steady-state temperature profile as a function of x may now be obtained by inverting cou9789x_ch20_443-474.indd 466 8/22/08 12:28:00 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES x Ts ( x ) Tvs ( Ts0 Tvs ) exp vt 467 (20.59) All equations for T to be derived below should be recognized as deviations from this expression. Subtracting Eq. (20.56) from Eq. (20.55) and introducing deviation variables give T T 1 Tv T v t x t ( ) (20.60) where T T Ts and Tv Tv Tvs . Assuming that the exchanger is initially at steady state, so that Ts (x,0) Ts and Tv ( 0 ) Tvs , Eq. (20.60) may be transformed with respect to t to yield 1 dT ( x , s ) Tv ( s ) T ( x , s ) t dx s 1 /t 1 dT ( x , s ) T( x, s ) Tv ( s ) dx v tv sT ( x , s ) v (20.61) Equation (20.61) is an ordinary first-order differential equation with the boundary condition T ( x , s ) T ( 0, s ) at x 0 . We can transform Eq. (20.61) with respect to x (using p as the transform variable) to yield pT ( p, s ) T ( 0, s ) s 1 /t 1 /t v T ( p, s ) Tv ( s ) v p Rearranging gives T ( p, s ) 1 /tv 1 Tv ( s ) T ( 0, s ) p [ p (s 1 / t ) / v ] p [ (s 1 / t ) / v ] This expression is of the form T ( p, s ) A 1 Tv ( s ) T ( 0, s ) p ( p B) pB (20.62) where A and B are constants defined as A 1/t v and B (s 1/t)/v. Equation (20.62) can now easily be inverted back to the (x,s) domain to yield T( x, s ) cou9789x_ch20_443-474.indd 467 ( ) 1 1 e[(ts 1) / t v] x Tv ( s ) e[(ts 1) / tv] x T ( 0, s ) ts 1 (20.63) 8/22/08 12:28:01 PM Confirming Pages 468 PART 5 PROCESS APPLICATIONS Let’s look at the expression for the liquid outlet temperature from the exchanger, the quantity of interest in this example. The transform of the liquid outlet temperature is T (L,s). If we substitute x L into Eq. (20.63), we get T( L, s ) ) ( 1 1 e[(t s 1) /t v] L Tv ( s ) e[(t s 1) /t v ] L T ( 0, s ) (20.64) ts 1 If we examine the exponent of the exponential term L td L t s 1 L s t d s v t v vt t we see a familiar term appear: td is the transport lag on the liquid side of the exchanger. We can rewrite Eq. (20.64) as T( L, s ) 1 s t d t d /t t d /t 1 es t d e e T ( 0, s ) Tv ( s ) e t s 1 constant constant T( L, s ) outlet temperature (output) 1 Kes t d t s 1 t s 1 transfer function relating outll et temperature to changes in steam temperatt ure Tv ( s ) steam temperature (input ) s t (Ke ) d T ( 0, s ) transfer function inlet temperature (disturbance ) relating outlet temperature to changes in inlet temperaturee (20.65) where K et d / t constant. If the steam temperature is held constant Tv (s) 0, the transfer function relating the outlet temperature to the inlet temperature is T( L, s ) Kest d T ( 0, s ) (20.66) If we set 1/t 0 (which corresponds to the case of no heat transfer U0 0), then K et d / t e 0 1 . The response is simply that of a transport lag. This is the physical situation for which case the wall separating cold fluid from hot fluid acts as a perfect insulator. We saw in Chap. 7 that this situation is represented by a transport lag. If the inlet temperature of the fluid entering the heat exchanger does not vary, the transfer function relating the exit fluid temperature to the steam temperature is T( L, s ) 1 Kest d ts 1 ts 1 Tv ( s ) cou9789x_ch20_443-474.indd 468 (20.67) 8/22/08 12:28:02 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 469 The response in the temperature of the fluid leaving the exchanger can be found for any forcing function, T(0,t) or Tv(t), by introducing the corresponding transforms into Eq. (20.65). Example 20.1 Consider the heat exchanger shown in Fig. 20–10. The water is flowing through a 40-ft-long, 3 -in 14 BWG tube at a rate of 2 gal/min. The 4 overall heat-transfer coefficient for the exchanger is Ui 100 Btu/(ft2 · h · °F). The steady-state inlet water temperature is 70°F, and the steady-state steam temperature is 212°F. (a) Determine the steady-state temperature profile of the tube-side liquid as a function of distance from the inlet of the exchanger. (b) For a 10°F step increase in the inlet temperature (steam temperature remains constant), determine the response of the exit temperature from the exchanger. (c) For a 10°F step increase in the steam temperature (inlet temperature remains constant), determine the response of the exit temperature from the exchanger. Solution Data: Do 0 . 75 in Di 0 . 584 in Ai flow area p Di2 p (0 . 584 / 12)2 0 . 00186 ft 2 4 4 3 ft 1 1 min gal 1 ft 2.4 v 2 min 7 . 48 gal 0 . 00186 ft 2 6 0 s s ( ) p [( 0 . 584 / 12 ) ft ] 100 Btu / ft 2 h F 1 p DiUi 131 . 7 h1 t Ai r C 0 . 00186 ft 2 62 . 4 lb / ft 3 1 Btu / ( lb F ) )( ( ) t 0 . 0 0 759 h 27 . 3 s (a) Making use of Eq. (20.59) gives x Ts ( x ) Tvs ( Ts0 Tvs ) exp vt x x 212 1 4 2 exp Ts ( x ) 212 ( 7 0 212 ) exp 65 . 52 ( 2 . 4 ft / s ) ( 27 . 3 s ) (20.59) If we plot this result from x 0 to x L 40 ft, we get the plot in Fig. 20–11. cou9789x_ch20_443-474.indd 469 8/22/08 12:28:03 PM Confirming Pages 470 PART 5 PROCESS APPLICATIONS Steady-State Temperature Profile 140 130 Temperature (°F) 120 110 100 90 80 70 60 0 5 10 15 20 25 Length (ft) 30 35 40 FIGURE 20–11 Steady-state temperature profile down length of heat exchanger tube for Example 20.1. This plot shows the temperature profile in the tube along the length of the exchanger. The initial steady-state temperature at x 40 ft, Ts(40), is 134.9°F. This is easily obtained by substituting x 40 into the above expression. (b) Using Eq. (20.66), we have T ( L , s ) Kes t d T ( 0, s ) (20.66) T ( L , s ) Kes t d T ( 0, s ) et d td T ( 0, s ) / t s t d e T ( 0, s ) L 40 ft 16 . 7 s v 2.4 ft s 10 step function of magnitude 10 s 5 . 42e27.3 s 10 10 T ( L , s ) e16.7 s / 27.3 s es ( 27.3 s ) 0 . 542e27.3 s s s s ( )( ) 5 . 42e16.7 s 10 10 T ( L , s ) e16.7 s / 27.3 s es (16.7 s ) 0 . 542e16.7 s s s s ( )( ) and finally, 0 T ( L , t ) 5 . 42u ( t 16 . 7 ) 5 . 42 t t 16 . 7 s 16 . 7 s So the response to a step change in the inlet temperature is delayed 16.7 s from the inlet change. The output increases by only 5.42°F (to 140.32°F) in response to a 10°F change in inlet temperature (the steady-state gain is 0.542). cou9789x_ch20_443-474.indd 470 8/22/08 12:28:04 PM Confirming Pages CHAPTER 20 THEORETICAL ANALYSIS OF COMPLEX PROCESSES 471 (c) The applicable equation is Eq. (20.67). T ( L , s ) Kest d 1 ts1 ts 1 Tv ( s ) 1 1 Kes t d Kes t d 10 T ( L , s ) Tv ( s ) ts 1 t s 1 s ts 1 ts 1 ( ) ( ) t / 27 . 3 ( t 16 . 7 ) / 27 . 3 10 (1 e ) u ( t ) 10 ( 0.542)(1 e ) u (t 16.7) T ( L , t ) 10 1 et / t u ( t ) 10 K 1 e( t t d ) / t u ( t t d ) 0 T ( L , t ) 10 1 et / 27.3 10 1 et / 27.3 5 . 42 1 e( t 16.7) / 27.3 ( ( ) ) ( ) t 0 t t 0 16 . 7 16 . 7 A plot of the output response T(L,t) 134.9°F T (L,t) is shown in Fig. 20–12. 140 Temperature (°F) 139 138 137 136 135 134 0 5 10 15 Time (s) 20 25 FIGURE 20–12 Plot of exit temperature from the heat exchanger as a function of time T(L,t). After the transport lag time (16.7 s) the outlet temperature from the exchanger remains constant. cou9789x_ch20_443-474.indd 471 8/22/08 12:28:04 PM Confirming Pages 472 PART 5 PROCESS APPLICATIONS SUMMARY In this chapter, we analyzed several complex systems, using material and energy balances to predict their dynamic responses to disturbances. The process is often called modeling, and the resulting set of equations is referred to as the mathematical model of the system. In general, the model is based on the physics and chemistry of the system. As systems are analyzed in greater detail and with fewer assumptions, the models that describe them become more complex, although more accurate. To predict the response of the system from the model requires that equations of the model be solved for some specific input disturbance. The only practical way to solve a complex model is to use computer simulation. The computer response will resemble that of the physical system if the model is accurate. PROBLEMS 20.1 For the heat exchanger described in Example 20.1, determine the output response for a 10°F change in both steam temperature and inlet temperature. 20.2 What is the physical reason for the outlet temperature increasing less than 10°F for a step increase in the inlet temperature of 10°F? (That is, why doesn’t the 10°F change propagate through the exchanger and ultimately appear at the outlet?) cou9789x_ch20_443-474.indd 472 8/22/08 12:28:05 PM Confirming Pages CHAPTER 20 CAPSULE SUMMARY In this chapter, several complex systems have been analyzed mathematically. The result of each analysis was a set of equations (algebraic and/or differential) that presumably describe the dynamic response of the system to one or more disturbances. The process of obtaining the set of equations is often called modeling, and the set of equations is referred to as the mathematical model of the system. In general, the model is based on the physics and chemistry of the system. For example, in the analysis of a heat exchanger, one may write that the heat flux through a wall is equal to a convective transfer coefficient times a temperature driving force. For a process not well understood, there is little chance that an accurate model can be obtained from the theoretical approach used here. For such systems, a direct dynamic test can be made. To do this, a known disturbance such as a pulse, step, or sinusoidal input is applied, and the response is recorded. This approach was discussed in Chap. 18. On the other hand, a model based on a theoretical analysis is extremely valuable, for it means that the system is well understood and that the effect of changes in system design and operation can be predicted. The analysis of a steam-jacketed kettle provided an example of a nonlinear system containing nonlinear functions of several variables. The problem was handled by linearizing these functions about an operating point and ultimately obtaining a block diagram of the system from which the transfer function of the control system could be obtained. Although this approach is relatively straightforward, the resulting linear model can only be used over a narrow range of variables. The analysis of the gas absorber gave some insight into the dynamic character of a typical multistage process that is widely used in the chemical process industries. A linear analysis of an n-plate column leads to n ordinary differential equations, which combine to give an overdamped nth-order response. Nonlinearities may be present in this system in such forms as a product of flow and concentration or a nonlinear equilibrium relationship. When the change of plate efficiency with flow is considered, the model of a gas absorber becomes even more complex. Most of the design techniques developed for multistage operations (e.g., gas absorption and distillation) have applied to steady-state operation. The dynamic analysis of such processes calls for dynamic parameters that are usually unavailable. For example, the liquid flow dynamics of trays used in distillation towers are relatively unknown. The discussion of distributed-parameter systems further illustrated the complexities that can arise in physical systems. The distributed-parameter systems lead to partial differential equations, which may be very difficult to solve for most of the 473 cou9789x_ch20_443-474.indd 473 8/22/08 12:28:05 PM Confirming Pages 474 PART 5 PROCESS APPLICATIONS forcing functions of practical interest. However, we saw that the response of distributed-parameter systems to sinusoidal forcing functions can be obtained directly by application of the substitution rule, in which s is replaced by jw. A distributedparameter system features nonminimum phase lag characteristics. This is in sharp contrast to the lumped-parameter systems for which the phase angle approaches a limit at infinite frequency. As systems are analyzed in greater detail and with fewer assumptions, the models that describe them become more complex, although more accurate. To predict the response of the system from the model requires that equations of the model be solved for some specific input disturbance. The only practical way to solve a complex model is to use computer simulation. The computer response will resemble that of the physical system if the model is accurate. cou9789x_ch20_443-474.indd 474 8/22/08 12:28:05 PM Confirming Pages PART VI STATE-SPACE METHODS cou9789x_ch21_475-497.indd 475 8/22/08 6:17:42 PM cou9789x_ch21_475-497.indd 476 8/22/08 6:17:42 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 21.1 INTRODUCTION Up to this point, we have described dynamic physical systems by means of differential equations and transfer functions. Another method of description, which is widely used in all branches of control theory, is the state-space method. In fact, other disciplines of engineering (e.g., electrical engineering) introduce the state-space description before the transfer function description. The reader who plans to go beyond an introductory course in control or read from other engineering disciplines should be familiar with state-space methods. In the chapters of Part 6 of the book, the state-space method will be developed and compared with the transfer function method. It is much easier to start with the transfer function method and then develop the state-space method. The mathematical background needed for the transfer function approach involves differential equations and Laplace transforms. The additional mathematical background needed for the state-space method involves matrix algebra. Nearly all students today receive information on matrices in their mathematics courses. For those who are rusty in this topic, it is recommended that they review some of the fundamental matrix operations. A brief review of matrix algebra is given in App. 21A. The transfer function approach is sufficient to calculate the response of linear control systems. The state-space approach is especially valuable in the field of optimal control of linear or nonlinear systems. The concepts developed in Part 6 of the book will be used in Part 7 on nonlinear control. 21.2 STATE VARIABLES A linear physical system can be described mathematically by • An nth-order differential equation • A transfer function 477 cou9789x_ch21_475-497.indd 477 8/22/08 6:17:42 PM Confirming Pages 478 PART 6 STATE-SPACE METHODS • n first-order differential equations • A matrix differential equation So far, we have used the first two mathematical representations for describing physical systems. The third and fourth representations are referred to as state variable descriptions. To illustrate these four methods of description, consider the familiar second-order process relating an output y to an input u. The four representations for this process are shown below. 1. An nth-order differential equation (n 2) t2 2. d2y dy 2zt y u dt dt 2 Transfer function. The transfer function corresponding to Eq. (21.1) is 1 Y (s) 2 2 U (s) t s 2zts 1 3. (21.1) (21.2) n first-order differential equations (n 2). Equation (21.1) can be expressed by two first-order differential equations. First, we need to make a variable change. Let’s represent y as x1 and y as x2. Thus, x1 y and x2 y (Note: y dy /dt and ÿ d2y/dt2.) Then, x1 y x2 (21.3) x2 x1 y (21.4) 1 2z 1 y y 2 u 2 t t t (21.5) and Solving Eq. (21.1) for ÿ yields y Substituting in terms of x1 and x2, we obtain x2 1 2z 1 x x2 2 u 2 1 t t t (21.6a) and x1 x2 (21.6b) where x1 and x2 are called the state variables (x1 y and x2 y ). Equations (21.6a) and (21.6b) are the system of two first-order differential equations that represent the second-order differential equation, Eq. (21.1). cou9789x_ch21_475-497.indd 478 8/22/08 6:17:42 PM Confirming Pages CHAPTER 21 4. STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 479 We shall see later that other choices for x1 and x2 are possible; at this point, however, the reader is asked to accept Eqs. (21.6a) and (21.6b) as a valid description of the second-order system under consideration. Matrix differential equation. Equations (21.6a) and (21.6b) can be written as one matrix differential equation as follows: x ⴝ Ax ⴙ bu (21.7) or 1 0 x1 x2 1 2z 2 t t x A 0 x1 x2 1 u t 2 scalar x b The representation given by Eqs. (21.6) and the representation given by Eq. (21.7) are exactly the same; Eq. (21.7) is in a more compact form. The state variables x1 and x2 are represented by the column vector x. The coefficients of the state variables on the right sides of Eqs. (21.6a) and (21.6b) are the elements of the matrix A. In this example, there is only one input or forcing term u, which is a scalar. Each term on the right side of Eq. (21.7) must be a vector containing two elements (i.e., a 2 1 matrix). For the expression given by Eq. (21.7) to agree with Eqs. (21.6a) and (21.6b), the coefficient of u must be a vector with the upper element zero. With some practice, the reader will be able to look at a matrix expression such as Eq. (21.7) and quickly see the equivalent set of differential equations. The output y in representation 1 or 2 often represents a physical variable of interest, such as the temperature of a process or the position of a mechanical system. The alternate state variable representation given by Eq. (21.3) or Eq. (21.7) contains two state variables, one of which is y and the other of which is the derivative of y (or y ). In this case only y may be of interest to the control engineer; y is available, but may not be of interest since it cannot always be measured easily. (For example, there is no easy way to measure the rate of change of temperature if y represents temperature.) State-Space Description In general, a physical system can be described by state variables as follows x1 f1 ( x1 , x2 , … , xn , u1 , u2 , … , um ) x2 = f2 ( x1 , x2 , … , xn , u1 , u2 , … , um ) (21.8) xn = fn ( x1 , x2 , … , xn , u1 , u2 , … , um ) where x1, x2, . . ., xn are n state variables and u1, u2, . . ., um are m inputs or forcing terms. The above set of equations may be written as a matrix expression as follows: x ⴝ f ( x, u ) cou9789x_ch21_475-497.indd 479 8/22/08 6:17:44 PM Confirming Pages 480 PART 6 STATE-SPACE METHODS If the system parameters vary with time, the vector f will contain explicit functions of time. An example for an element of f might be the expression on the right side of the following equation: x1 2tx1 x2 u1 u2 In this chapter, we shall be concerned with time-invariant systems for which xi is a linear combination of state variables and the coefficients are constant. For the timeinvariant case, we may write the general term xi in Eq. (21.8) as follows: xi ai1 x1 ai 2 x2 ain xn bi1u1 bim um (21.9) for i 1, 2, 3, . . ., n. The equivalent matrix expression for Eq. (21.9) is x1 a11 a12 x2 a21 a22 xn an1 an 2 a1n x1 b11 a2 n x2 b21 ann xn bn1 b12 b22 bn 2 b1m u1 b2 m u2 bnm um (21.10) Writing this in the more compact matrix form, we have x ⴝ Ax ⴙ Bu (21.11) In this expression, there are m different inputs where m n. Also, note that u in this expression is a vector quantity, not a scalar. The nature of the linear physical system expressed by Eq. (21.11) is completely stated by the matrices A and B. For the timeinvariant system, the elements of A and B are constants. The outputs of interest to the control engineer may differ from the state variables xi. The most general statement for relating the output to the state variables is y ⴝ Cx (21.12) where y is the vector of outputs (y1, y2, . . ., yp) chosen by the control engineer for some practical reason. The matrix C is a p n matrix containing constant elements. The way in which the matrix C is selected will be clarified in Example 21.1. In summary, the state-space description for a linear time-invariant system is given by Eqs. (21.11) and (21.12). Example 21.1. For the two-tank noninteracting liquid-level system shown in Fig. 21–1, obtain the state-space description as expressed by Eqs. (21.11) and (21.12). The output y of interest is the level in tank 2. Notice that streams enter both tanks. cou9789x_ch21_475-497.indd 480 A1 1 R1 0 . 5 A2 0 . 5 R2 2 3 8/22/08 6:17:44 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 481 u1 A1 h1 R1 u2 q1 A2 h2 R2 q2 FIGURE 21–1 Two-tank liquid-level system. For this example, let the state variables be the physical variables h1 and h2, which are the levels in tanks 1 and 2. These state variables are called physical variables because they can be easily measured or observed. (In another example, we shall consider a different set of state variables.) For the liquid-level system shown in Fig. 21–1 we may write dh1 h u1 1 dt R1 (21.13) dh2 h h u2 1 2 dt R1 R2 (21.14) dh1 1 1 h1 u1 dt R1 A1 A1 (21.15) dh2 1 1 1 h1 h2 u2 dt R1 A2 A2 R2 A2 (21.16) A1 A2 or These equations can be written as h ⴝ Ah ⴙ Bu where cou9789x_ch21_475-497.indd 481 1 0 R1 A1 h1 1 1 h2 R1 A2 R2 A2 h A 1 0 h1 A1 h2 1 0 A h 2 B (21.17) u1 u2 u 8/22/08 6:17:45 PM Confirming Pages 482 PART 6 STATE-SPACE METHODS If the output y is to be the level in tank 2 (h2), we have y ⴝ Ch h1 y1 0 1 h2 y C h and y y1 h2. In this case y is a scalar (i.e., a 1 1 matrix). The choice of output can be stated in many ways. Regardless of the choice, the output is related to the state variables by Eq. (21.12). To see how the matrix C depends on the choice of output, consider the following examples: If y is to be a scalar that is the arithmetic average of the levels in the two tanks, one can show that h1 y1 0 . 5 0 . 5 h2 y C h If the output is to be h1 and h2, one can show that y1 1 0 h1 0 1 h2 y2 y h C Selection of State Variables To the beginner, the selection of state variables may seem mysterious. The state variables of a system are the smallest set of variables that contain sufficient information to permit all future states to be determined. Although the number of state variables is fixed, the actual selection of these state variables is not unique. If possible, it is convenient to choose state variables that are directly related to physical variables which can be measured or observed (e.g., temperature, level, composition, position, velocity). For mechanical systems, transducers are available for measuring velocity; for this reason, velocity is considered a physical variable. On the other hand, since the measurement of rate of change of composition is not easily made, this variable is not usually considered a physical variable. In the control literature, the types of state variables have been classified as follows. 1. 2. cou9789x_ch21_475-497.indd 482 Physical variables. State variables are called physical variables when they are readily measured and observed (level, temperature, composition, etc.). Physical variables were discussed at the beginning of this chapter and illustrated for a liquid-level system in Example 21.1 where x1 h1 and x2 h2. Phase variables. State variables that are chosen to be the dependent variable and its successive derivatives are called phase variables. Phase variables were selected at the beginning of this chapter where x1 y and x2 y. 8/22/08 6:17:47 PM Confirming Pages CHAPTER 21 3. STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 483 Canonical variables. If the state variables are selected to be canonical variables, the result is that the matrix A is diagonal. At this point, it is sufficient to say that canonical variables are selected as state variables for ease in matrix computation. In general, the canonical variables are not readily identified with physical variables. In addition to the types of state variables listed above, any other legitimate set of variables can be selected. In Example 21.1, we used physical variables, namely, the levels in the tanks of the liquid-level system. In Examples 21.2 and 21.3, the method for selecting state variables will be shown. Example 21.2. For the two-tank liquid-level system of Example 21.1, shown in Fig. 21–1, obtain the state-space description as expressed by Eqs. (21.11) and (21.12) when phase variables are selected for the state variables. To simplify the problem, let u2 0; i.e., there is only one input u1. For the system shown in Fig. 21–1, one can show that H2 ( s ) R2 U1 ( s ) (t 1s 1) (t 2 s 1) (21.18) where t1 A1 R1 and t2 A2 R2. Introducing the parameters in Fig. 21–1 into Eq. (21.18) gives H2 ( s ) U1 ( s ) or H2 ( s ) U1 ( s ) ( 2 3 1 2 )( s 1 1 s 3 ) (21.19) 1 4 ( s 2 ) ( s 3) (21.20) To obtain the differential equation corresponding to Eq. (21.20), we cross-multiply to obtain ( s 2) ( s 3) H2 4U1 or (s 2 ) 5s 6 H 2 4U1 This may be expressed as the following differential equation: h2 5h2 6h2 4u1 (21.21) Let the state variables be the following phase variables: x1 h2 (21.22) x2 h2 (21.23) x1 x2 h2 (21.24) x2 h2 (21.25) We may now write cou9789x_ch21_475-497.indd 483 8/22/08 6:17:48 PM Confirming Pages 484 PART 6 STATE-SPACE METHODS Equation (21.21) becomes x2 5 x2 6 x1 4u1 (21.26) The system can be described by Eqs. (21.24) and (21.26): x1 x2 x2 6 x1 5 x2 4u1 (21.27a) (21.27b) In terms of a matrix expression, Eqs. (21.27) may be written x ⴝ Ax ⴙ bu1 1 x1 0 x2 6 5 x A x1 0 x2 4 u1 scalar x b If the output y is to be the level in tank 2, x1 y 1 0 x2 y C x MATLAB Solution of Example 21.2 MATLAB has the capability to convert from a transfer function model to a state-space model of a system. From Eq. (21.20), the transfer function that relates the height in tank 2 to changes in the inlet flow to tank 1 is H2 ( s ) U1 ( s ) 4 4 2 2 3 s s ( )( ) s 5s 6 (21.20) The MATLAB commands that create the state space representation are: num=[4]; den=[1 5 6]; [A,B,C,D] = tf2ss(num,den) A = -5 1 -6 0 B = 1 0 C = 0 4 D = 0 cou9789x_ch21_475-497.indd 484 8/22/08 6:17:50 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 485 tf2ss is a “transfer function to state space” command that generates the matrices for the state-space model. The generalized model that MATLAB uses is x ⴝ Ax ⴙ Bu y ⴝ Cx ⴙ Du Note that the output y is some function of the states x and/or the inputs u. MATLAB returns a canonical form for the state-space representation that is equivalent (but not identical) to the one that we developed using phase variables. The output y that is returned from the tf2ss routine is a scalar that represents the dependent variable of the original transfer function, h2 in this case. Let’s see if we can verify that the state-space model that MATLAB returns is a valid representation of the original system, Eq. (21.21). The MATLAB model is x1 5 6 1 0 x 2 x A x1 1 u1 (t ) x2 0 x B x1 y h2 0 4 0 u1 (t ) x2 y C x From the output equation, h2 4 x2 x2 0 . 25h2 and from the differential equations, x1 5 x1 6 x2 u1 (t ) x2 x1 and so we can write x1 x2 0 . 25h2 and finally, 0 . 25h2 5(0 . 25h2 ) 6(0 . 25h2 ) u1 (t ) h2 5h2 6h2 4u1 (t ) which we recognize as Eq. (21.21), the original system. Although the state-space representation is not the same as the representation we derived using phase variables, it is a valid model for the system, and the output from a computer simulation of these models for h2 would be identical. cou9789x_ch21_475-497.indd 485 8/22/08 6:17:51 PM Confirming Pages 486 PART 6 STATE-SPACE METHODS Gc R + E Kc Is Gp +1 Kp M Is ( 1s + 1)( 2s + 1) C − FIGURE 21–2 PI control system for Example 21.3. Example 21.3. For the PI control system shown in Fig. 21–2, obtain a statespace representation in the form of Eq. (21.7); thus x ⴝ Ax ⴙ br where r is a scalar (the set point). Let x1 c (21.28) x2 c x1 (21.29) With this choice of state variables, we have selected phase variables. From Fig. 21–2, we may write Kp K p /t 1t 2 C (s) M (s) (t 1s 1) (t 2 s 1) ( s 1/t 1 ) ( s 1/t 2 ) C (s) or where M (s) A Kp t 1t 2 a 1 t1 b 1 t2 A (s a)(s b) (21.30) Cross-multiplying Eq. (21.30) gives s 2 ( a b ) s ab C ( s ) AM ( s ) or, in the time domain, c ( a b ) c abc Am cou9789x_ch21_475-497.indd 486 (21.31) 8/22/08 6:17:52 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 487 From Eqs. (21.28), (21.29), and (21.31) we obtain x1 x2 (21.32) x2 abx1 ( a b ) x2 Am (21.33) We must now obtain the state variables associated with the PI controller. From Fig. 21–2, we obtain tIs 1 M (s) Kc tIs E (s) or t I sM ( s ) K ct I sE ( s ) K c E ( s ) In terms of the time domain, this expression becomes m K c e Kc e tI (21.34) From the signals entering and leaving the comparator, we may write e r c or, since x1 c, we may write and e r x1 (21.35a) e r x1 (21.35b) Combining Eqs. (21.34) and (21.35) gives Kc m K c r x1 ( r x1 ) tI x2 or m K c r K c x2 Kc Kc r x1 tI tI (21.36) At this stage, we are faced with the difficulty of having a derivative term on the right side of Eq. (21.36). In state-space representation, all variables on the right side must be state variables, not derivatives of state variables. One way to handle the present difficulty is to define a new state variable x3; let or cou9789x_ch21_475-497.indd 487 x3 m K c r (21.37) x3 m K c r (21.38) 8/22/08 6:17:54 PM Confirming Pages 488 PART 6 STATE-SPACE METHODS Combining Eqs. (21.38) and (21.36) leads to x3 Kc Kc x1 K c x2 r tI tI x3 a x1 K c x2 a r or (21.39) (21.40) where a Kc/tI. Summarizing the state variable equations given by Eqs. (21.32), (21.33), and (21.40) and using the definition of x3 in Eq. (21.37) give x1 x2 x2 abx1 ( a b ) x2 Ax3 AK c r x3 a x1 K c x2 a r where A Kp t 1t 2 a 1 t1 b 1 t2 a Kc tI Therefore, the reduced form x ⴝ Ax br is 1 0 0 x1 x ab ( a b ) A 2 a x3 0 K c A x1 0 x AK r (t ) 2 c x3 a b If m is required as a function of t, it can always be found by solving Eq. (21.37) for m; thus m x3 K c r SUMMARY State-space representation is an alternative to the transfer function representation of a physical system that we have used up to this point. A transfer function that relates an output variable to an input variable represents an nth-order differential equation. In the state-space representation, the nth-order differential equation is written as n first-order cou9789x_ch21_475-497.indd 488 8/22/08 6:17:55 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 489 differential equations in terms of n state variables. These n differential equations can also be written in a more compact form as a matrix differential equation x ⴝ Ax ⴙ B u For an nth-order dynamic system, the number of state variables is fixed at n, but the selection of the variables is not unique. Of the many sets of state variables that one can choose, we discussed three sets that are useful in control theory, namely, physical variables, phase variables, and canonical variables. The state-space representation gives all the dynamic details of a system (e.g., the dependent variable and its successive derivatives for the case of phase variables, for example, h1 and h2 from Example (21.3). Whether or not this detail is needed depends on the problem being solved. We shall see the value of state-space representation in multivariable control and in nonlinear control in later chapters. cou9789x_ch21_475-497.indd 489 8/22/08 6:17:56 PM Confirming Pages APPENDIX 21A ELEMENTARY MATRIX ALGEBRA The purpose of this section is to provide, in a convenient location, a review of some of the elementary operations of matrix algebra for use in state-space methods. It is expected that the reader has had some course work in linear algebra discussing the concepts of a vector and a matrix and the operations performed on them. VECTORS An n-dimensional column vector is an ordered series of elements (numbers) x1, x2, . . ., xn and is written as x1 x2 x xn Multiplication of a vector by a scalar lx results in a vector for which each element is multiplied by l. MATRICES A matrix is a rectangular array of elements (numbers) that takes the form a11 a21 A an1 a12 a1m a22 a2 m an 2 anm in which the elements are written aij. The subscript i refers to the ith row and j to the jth column. Matrix A is called an n m matrix where n is the number of rows and m is the number of columns. If n m, the matrix is called a square matrix. If m 1, the matrix is a column vector (n 1). If n 1, the matrix is a row vector (1 m). 490 cou9789x_ch21_475-497.indd 490 8/22/08 6:17:56 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 491 The transpose of a matrix AT is a matrix for which the rows and columns of the matrix A are interchanged. If the diagonal elements (aij for all i j) of a square matrix are 1 and all off-diagonal elements are 0, then the matrix is called a unit (or identity) 1 0 matrix and is given the symbol I. For example, a 2 2 unit matrix is given by . 0 1 If A AT for a square matrix, the matrix A is said to be symmetric. When two matrices are added (or subtracted), the corresponding elements are added (or subtracted); thus a11 b11 a21 b21 AB a n1 bn1 a12 b12 a22 b22 an 2 bn 2 a1m b1m a2 m b2 m anm bnm The product of two matrices C AB is a matrix whose elements are obtained by the expression cij m ∑ aik bkj for i 1, … , n and j 1, … , p k 1 where A is an n m matrix and B is an m p matrix. The matrix C is an n p matrix. The number of columns in A must equal the number of rows in B to perform this operation. INVERSE OF A MATRIX The inverse of a matrix is related to the concept of division for numbers. It can only be found for square matrices. The inverse of a number x is written 1/x or x1. The product of a number x and its inverse is equal to unity. The inverse of a matrix A is written A1, and the product of a matrix and its inverse is equal to the unit matrix; thus A1A I The expression used for matrix inversion takes the form adj A A1 A (21A.1) where |A| is the determinant of A and adj A is the adjoint of A. These two terms will now be described. The determinant of a matrix |A| is a scalar which is computed from the elements of the matrix as follows: A ai1 Ai1 ai 2 Ai 2 ain Ain cou9789x_ch21_475-497.indd 491 8/22/08 6:17:57 PM Confirming Pages 492 PART 6 STATE-SPACE METHODS n ∑ aij Aij A or for any i (21A.2) j 1 where Aij, the cofactor of element aij, is computed as Aij (1)i j Mij The determinant Mij is the minor of the element aij and is deﬁned as follows. If the row and column containing the element aij are deleted from a square matrix A, the determinant of the resulting matrix, which is an (n 1) (n 1) matrix, is the minor Mij. An alternate expression for the calculation of a determinant which uses the elements of a speciﬁc column and its cofactors is as follows: A n ∑ aij Aij for any j i 1 (21A.3) The determinant of a matrix with two equal rows or columns is zero. We now deﬁne the adjoint of a matrix. Let the matrix B be an n n matrix whose elements bij are the cofactors Aji of A, i.e., the transpose of the cofactor matrix. Matrix B is the adjoint of A; thus B adj A transpose of cofactor matrix, or A11 A12 adj A A1n A21 An1 A22 An 2 A2 n Ann Some useful properties of the inverse are ( AB )1 B1A1 ( A1 ) ( AT ) 1 ( A1 ) A T 1 The derivations of relationships presented here, as well as other properties of matrices, can be found in textbooks on linear algebra (see Anton, 1984). EXAMPLES 1. Evaluate the determinant of A for the following matrix. 2 3 5 A 1 0 1 2 1 0 cou9789x_ch21_475-497.indd 492 8/22/08 6:17:58 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 493 For this problem, we use Eq. (21A.2) with i 1 (i.e., use row 1). 0 1 1 1 1 0 A 2 3 5 1 0 2 0 2 1 2 [( 0 ) ( 0 ) (1) (1)] 3 [(1) ( 0 ) (1) ( 2 )] 5 [(1) (1) ( 0 ) ( 2 )] 2 (1) 3 (2 ) 5 (1) 9 MATLAB Solution: A=[2,3,5;1,0,1;2,1,0] A = 2 1 2 det(A) 3 0 1 5 1 0 ans = 9 2. Find the inverse of the matrix 2 3 A 1 4 A1 adj A A The determinant of A is A ( 2 ) ( 4 ) ( 3) (1) 5 The matrix of minors is 4 1 3 2 The cofactor matrix is 4 1 3 2 cou9789x_ch21_475-497.indd 493 8/22/08 6:17:59 PM Confirming Pages 494 PART 6 STATE-SPACE METHODS The adjoint of the matrix, which is the transpose of the cofactor matrix, is 4 3 adj A 2 1 therefore A1 45 35 1 4 3 1 2 2 5 1 5 5 MATLAB Solution A= [2,3;1,4] A = 2 1 3 4 A^-1 ans = 0.8000 -0.2000 -0.6000 0.4000 3. Obtain the inverse of the matrix 2 3 1 A 1 2 3 3 1 2 One can show that A 18 The cofactor matrix is 7 5 1 5 1 7 7 5 1 The adjoint matrix is A1 cou9789x_ch21_475-497.indd 494 7 1 5 7 1 5 5 7 1 7 1 5 1 7 1 5 18 7 1 5 8/22/08 6:17:59 PM Confirming Pages CHAPTER 21 STATE-SPACE REPRESENTATION OF PHYSICAL SYSTEMS 495 MATLAB Solution A=[2,3,1;1,2,3;3,1,2] A = 2 1 3 3 2 1 1 3 2 Aˆ-1 ans = 0.0556 -0.2778 0.3889 0.3889 0.0556 -0.2778 -0.2778 0.3889 0.0556 which is the same as we determined by hand... if we multiply each element by 18, the result will be obvious. ans*18 ans = 1.0000 -5.0000 7.0000 7.0000 1.0000 -5.0000 -5.0000 7.0000 1.0000 PROBLEMS 21.1. In the liquid-level process shown in Fig. P21–1, the three tanks are interacting. The process may be described by x Ax Bu x1 x x where 2 x3 and u1 u1 u u2 u2 A2 = 1 2 A1 = 1 x1 A3 = 1 x2 x3 R1 = 1 R4 R2 = 1 R3 R5 FIGURE P21–1 cou9789x_ch21_475-497.indd 495 8/22/08 6:18:00 PM Confirming Pages 496 PART 6 STATE-SPACE METHODS (a) If 1 0 3 A 2 3 2 0 1 3 determine values of R3, R4, and R5. If one of these values of R is negative, what is your interpretation? (b) Determine B. 21.2. For the system shown in Fig. P21–2, find A and b in x Ax bu The tanks are interacting. The following data apply: A1 1, A2 R1 1 , 2 R2 2, 1 , 2 R3 1 u A1 A2 x2 x1 R1 R2 R3 FIGURE P21–2 cou9789x_ch21_475-497.indd 496 8/22/08 6:18:01 PM Confirming Pages CHAPTER 21 CAPSULE SUMMARY A linear physical system can be described mathematically in several ways. For a secondorder differential equation t 2 d2y/dt2 2z t dy/dt y u, we may represent it by • A transfer function. Y (s) U (s) 1 2 2 t s 2zt s 1 • Two first-order differential equations x1 y x2 and x2 y x1 x2 2z 1 1 x1 x2 2 u t t2 t • A matrix differential equation. x ⴝ Ax ⴙ bu 1 0 x1 1 2z x 2 t 2 t x A y ⴝ 0 x1 x 1 u 2 t 2 sc a lar x b Cx The xi are the state variables, and the yi are the outputs. In control literature, there are generally three types of state variables used: 1. 2. 3. Physical variables (readily measured and observed, i.e., level, temperature, composition, etc.) Phase variables (the dependent variable and its successive derivatives) Canonical variables (state variables selected so that matrix A is diagonal, chosen for ease in matrix computation) SOME MATLAB COMMANDS tf2ss—converts a transfer function model to a state-space model. The state variables are canonical. det(A)—finds the determinant of square matrix A Aˆ-1 or inv(A)—finds the inverse of square matrix A 497 cou9789x_ch21_475-497.indd 497 8/22/08 6:18:02 PM Confirming Pages CHAPTER 22 TRANSFER FUNCTION MATRIX I n Chap. 21, we have seen that a linear dynamic system can be expressed in terms of the following equations x ⴝ Ax ⴙ Bu y ⴝ Cx (22.1) (22.2) where x ⫽ column vector of n state variables (x1, x2, . . ., xn) u ⫽ column vector of m inputs or forcing terms (u1, u2, . . ., um) y ⫽ column vector of p outputs (y1, y2, . . ., yp) A ⫽ n ⫻ n matrix of coefficients B ⫽ n ⫻ m matrix of coefficients C ⫽ p ⫻ n matrix of coefficients One of the objectives of this chapter is to show how one solves Eqs. (22.1) and (22.2) in a systematic manner. Before discussing the solution of the matrix differential equation of Eq. (22.1), we consider the scalar differential equation dx ⫽ Ax ⫹ Bu dt (22.3) In this equation all the terms are scalars. If we multiply both sides of Eq. (22.3) by e⫺At, we obtain, after rearrangement, e⫺ At dx ⫺ Axe⫺ At dt ⫽ Be⫺ At u(t ) dt 498 cou9789x_ch22_498-511.indd 498 8/23/08 9:32:47 AM Confirming Pages CHAPTER 22 TRANSFER FUNCTION MATRIX 499 We can recognize that the left-hand side is an exact differential, and we can rewrite this expression as ) ( d xe⫺ At ⫽ Be⫺ At u(t ) dt Integrating both sides from t ⫽ 0 to t ⫽ t, we obtain ∫0 d ( xe t ⫺ At ) ⫽ ∫tt =⫽0t Be⫺At u(t ) dτ xe⫺ At ⫺ x ( 0 ) ⫽ x ⫽ e At x ( 0 ) ⫹ e At ∫ t ⫽t t =0 t ⫽t ⫺ At ∫t = 0 Be u(t ) d t Be⫺ At u(t ) d t ⫽ e At x ( 0 ) ⫹ t ⫽t ∫t = 0 Be A( t ⫺t ) u(t ) d t The solution to Eq. (22.3) can thus be written as the sum of the complementary function and the particular integral as follows: x ( t ) ⫽ e At x ( 0 ) ⫹ ∫ e A( t ⫺t ) Bu(t ) d t 0 particular t solution (22.4) complementary solution Equation (22.4) is a well-known result that has been derived in many books on the solution of ordinary differential equations. 22.1 TRANSITION MATRIX Let us now turn our attention to the solution of the matrix differential equation x ⴝ Ax (22.5) This is Eq. (22.1) for the case of no inputs (i.e., u ⫽ 0). The initial conditions for Eq. (22.5) may be expressed as x(0). One can show that the solution to Eq. (22.5) with initial conditions x(0) is given by A2 2 A k k x ( t ) ⫽ I ⫹ At ⫹ t ⫹ ... ⫹ t x (0) k! 2! (22.6) The infinite series of matrix terms within the braces is given the symbol eAt. This symbol is chosen to recall that the infinite series of the scalar term eat is 1 ⫹ at ⫹ a2 2 ak k t ⫹ ... ⫹ t 2! k! Using the symbol eAt, we may write Eq. (22.6) as x ( t ) ⫽ e At x ( 0 ) cou9789x_ch22_498-511.indd 499 (22.7) 8/23/08 9:32:47 AM Confirming Pages 500 PART 6 STATE-SPACE METHODS The symbol eAt is an n ⫻ n matrix in which each element contains a power series of t. The solution to Eq. (22.1) can be shown to be x ( t ) ⫽ e At x ( 0 ) ⫹ t A ( t ⫺t ) ∫0 e Bu(t ) d t (22.8) Notice that Eq. (22.8) resembles Eq. (22.4), which is the solution for the scalar differential equation. Since eAt is awkward and perhaps misleading as to its nature, eAt is sometimes replaced by f (t); thus e ( t ) ⫽ e At ( transition matrix) (22.9) Either of the terms e (t) and eAt can be used for the transition matrix. In this book, we shall use eAt. Example 22.1. Solution of a matrix differential equation. Solve the following matrix differential equation 1 ⫺1 0 x ⫽ x ⫹ u (t ) 0 ⫺2 1 where u(t) is a unit-step function and ⫺1 x (0) ⫽ 0 One can show that e At e⫺t ⫽ 0 e⫺t ⫺ e⫺2 t e⫺2 t In the next section, the method used to obtain the elements of this matrix will be developed. Applying Eq. (22.8) gives e⫺t x(t ) ⫽ 0 e⫺t ⫺ e⫺2 t ⫺1 ⫹ e⫺2 t 0 ⫺( t ⫺t ) t e ∫0 0 e⫺( t ⫺t ) ⫺ e⫺2( t ⫺t ) 0 dt 1 e⫺2( t ⫺t ) t ⫽t or or cou9789x_ch22_498-511.indd 500 ⫺e⫺t e⫺( t ⫺t ) ⫺ 0 . 5e⫺2( t ⫺t ) x (t ) ⫽ ⫹ 0 . 5e⫺2( t ⫺t ) 0 t ⫽0 0 . 5 ⫺ 2e⫺t ⫹ 0 . 5e⫺2 t 0 . 5 ⫺ 0 . 5e⫺2 t x (t ) ⫽ 8/23/08 9:32:48 AM Confirming Pages CHAPTER 22 TRANSFER FUNCTION MATRIX 501 MATLAB Solution of Matrix Differential Equation Let’s resolve Example 22.1 using MATLAB. A=[−1,1;0,−2] A = −1 0 syms t 1 −2 %Declare t to be a symbolic variable. B=expm(A*t) %Calculate the transition matrix. B = [exp(−t), exp(−t)−exp(−2*t)] [0, exp(−2*t)] C=B*[−1;0] %This is the “particular” solution, the first term on the right-hand side. C = [ −exp(−t)] [ 0] syms tau %Declare tau to be symbolic. D=expm(A*(t−tau)) %Preparing the integrand D = [exp(−t+tau), −exp(−2*t+2*tau)+exp(−t+tau)] [0, exp(−2*t+2*tau)] E=D*[0;1] %Completing the integrand E = [−exp(−2*t+2*tau)+exp(−t+tau)] [exp(−2*t+2*tau)] F=int(E,tau,0,t) %Integrating the expression over tau, from tau=0 to tau=t F = [1/2+1/2*exp(−2*t)−exp(−t)] [1/2−1/2*exp(−2*t)] %The “complementary” solution G=F+C G = [1/2+1/2*exp(−2*t)−2*exp(−t)] [1/2−1/2*exp(−2*t)] %The complete solution . . . this checks with our hand calculation above. cou9789x_ch22_498-511.indd 501 8/23/08 9:32:49 AM Confirming Pages 502 PART 6 STATE-SPACE METHODS Determining eAt One method for determining the elements of the transition matrix eAt is to use Laplace transforms. Consider the matrix differential equation of Eq. (22.1) x ⫽ Ax ⫹ Bu If we take the Laplace transform of each side, we obtain sX ( s ) ⫺ x ( 0 ) ⫽ AX ( s ) ⫹ BU ( s ) sX ( s ) ⫺ AX ( s ) ⫽ x ( 0 ) ⫹ BU ( s ) or Solving for X(s) gives ( sI ⫺ A ) X ( s ) ⫽ x ( 0 ) ⫹ BU ( s ) (22.10) To obtain an expression for X(s), premultiply both sides of Eq. (22.10) by (sI ⫺ A)⫺1; thus ( sI ⫺ A )⫺1 ( sI ⫺ A ) X ( s ) ⫽ ( sI ⫺ A )⫺1 x ( 0 ) ⫹ ( sI ⫺ A )⫺1 BU ( s ) This equation becomes X ( s ) ⫽ ( sI ⫺ A )⫺1 x ( 0 ) ⫹ ( sI ⫺ A )⫺1 BU ( s ) (22.11) To obtain x(t) from Eq. (22.11), we take the inverse transform; thus { } { } x ( t ) ⫽ L⫺1 ( sI ⫺ A )⫺1 x ( 0 ) ⫹ L⫺1 ( sI ⫺ A )⫺1 BU ( s ) (22.12) By comparing Eqs. (22.8) and (22.12), we see that { e At ⫽ L⫺1 ( sI ⫺ A )⫺1 and 22.2 t At ⫺t ∫0 e } { (22.13) } Bu(t ) d t ⫽ L⫺1 ( sI ⫺ A )⫺1 BU ( s ) (22.14) TRANSFER FUNCTION MATRIX When x(0) ⫽ 0, a case frequently used in control applications, we obtain from Eq. (22.11) X ( s ) ⫽ ( sI ⫺ A )⫺1 BU ( s ) (22.15) X (s) ⫽ G (s) U (s) (22.16) This may be written cou9789x_ch22_498-511.indd 502 8/23/08 9:32:49 AM Confirming Pages CHAPTER 22 TRANSFER FUNCTION MATRIX 503 where G ( s ) ⫽ ( sI ⫺ A )⫺1 B ( transfer function matrix) (22.17) The term G(s) is called the transfer function matrix and serves the same purpose as the transfer function for the scalar case; namely, it relates a set of state variables X(s) to a set of inputs U(s). If we prefer to relate the output to the input as expressed by Eq. (22.2), we may proceed as follows. Taking the Laplace transform of both sides of Eq. (22.2) gives Y ( s ) ⫽ CX ( s ) (22.18) Combining Eqs. (22.15) and (22.18) gives Y ( s ) ⫽ C ( sI ⫺ A )⫺1 BU ( s ) We may now write where Y ( s ) ⫽ G1 ( s ) U ( s ) (22.19) G1 ( s ) ⫽ C ( sI ⫺ A )⫺1 B (22.20) The term G1 (s) in Eq. (22.20) is also a transfer function matrix that relates the output vector Y to the input vector U. Example 22.2. Determine the transfer function matrix for the two-tank liquid-level system shown in Fig. 22–1. Given: A1 ⫽ 1, A2 ⫽ 0 . 5, R1 ⫽ 0 . 5, R2 ⫽ 23 . u1 A1 h1 R1 q1 u2 A2 h2 R2 q2 FIGURE 22–1 Two-tank liquid-level system. cou9789x_ch22_498-511.indd 503 8/23/08 9:32:50 AM Confirming Pages 504 PART 6 STATE-SPACE METHODS As developed in Example 21.1 [Eq. (21.17)], this system is described by h ⫽ Ah ⫹ Bu where 0 ⫺2 A ⫽ 4 ⫺3 1 0 B ⫽ 0 2 From the definition of the transfer function matrix of Eq. (22.17), we write G ( s ) ⫽ ( sI ⫺ A )⫺1 B The inverse of sI ⫺ A is obtained as follows (see App. 21A for details on the inversion of a matrix): ( sI ⫺ A )⫺1 ⫽ adj ( sI ⫺ A ) sI ⫺ A 0 s ⫹ 2 sI ⫺ A ⫽ s ⫹ 3 ⫺4 Cofactor of 4 s ⫹ 3 sI ⫺ A ⫽ s ⫹ 2 0 We can now find the adjoint: 0 s ⫹ 3 adj ( sI ⫺ A ) ⫽ s ⫹ 2 4 The determinant of sI ⫺ A is s⫹2 0 ⫽ ( s ⫹ 2 ) ( s ⫹ 3) ⫺4 s⫹3 We can now determine the inverse of sI ⫺ A. ( sI ⫺ A )⫺1 0 s ⫹ 3 4 s ⫹ 2 ⫽ ( s ⫹ 2 ) ( s ⫹ 3) (22.21) 0 0 s ⫹ 3 s ⫹ 3 4 4 1 0 2 2 ⫹ s ( ) 2 ⫹ s ⫽ G (s) ⫽ ( s ⫹ 2 ) ( s ⫹ 3) 0 2 ( s ⫹ 2 ) ( s ⫹ 3) Simplifying this expression gives 1 s⫹2 G (s) ⫽ 4 ( s ⫹ 2 ) ( s ⫹ 3) cou9789x_ch22_498-511.indd 504 2 s ⫹ 3 0 8/23/08 9:32:51 AM Confirming Pages CHAPTER 22 TRANSFER FUNCTION MATRIX 505 From Eq. (22.16) we write H (s) ⫽ G (s) U (s) therefore 1 s⫹2 H1 ( s ) H ( s ) ⫽ 4 2 ( s ⫹ 2 ) ( s ⫹ 3) U1 ( s ) 2 U 2 ( s ) s ⫹ 3 0 (22.22) From Eq. (22.22), we obtain H1 ( s ) ⫽ and H2 ( s ) ⫽ 1 U1 ( s ) s⫹2 4 ( s ⫹ 2 ) ( s ⫹ 3) 2 U2 ( s ) s⫹3 U1 ( s ) ⫹ For given inputs, the above equations may be inverted to obtain h1 (t) and h2 (t). For the case of U1(s) ⫽ 1/s and U2(s) ⫽ 0, we get H1 ( s ) ⫽ 1 0.5 ⫽ s ( s ⫹ 2) s ( 0 . 5s ⫹ 1) H2 ( s ) ⫽ and 4 s ( s ⫹ 2 ) ( s ⫹ 3) Inversion of H1(s) and H2(s) gives ( h1 ( t ) ⫽ 0 . 5 1 ⫺ e⫺2 t h2 ( t ) ⫽ ) ( 2 4 2 ⫹ e⫺3t ⫺ 2e⫺2 t ⫽ 1 ⫹ 2e⫺3t ⫺ 3e⫺2 t 3 3 3 ) The results given above can be obtained, of course, by the methods presented earlier in this book. The transition matrix can be obtained by applying Eq. (22.13) to Eq. (22.21): e At 1 s ⫹ 2 ⫽ L⫺1 4 ( s ⫹ 2 ) ( s ⫹ 3) 1 s ⫹ 3 0 Inverting each term in the matrix gives e⫺2 t e At ⫽ ⫺2 t 4 e ⫺ e⫺3t ( ) 0 e⫺3t This matrix can be used in Eq. (22.8) to calculate h1 (t) and h2 (t). The result will be the same as obtained by inversion of Eq. (22.22). cou9789x_ch22_498-511.indd 505 8/23/08 9:32:51 AM Confirming Pages 506 PART 6 STATE-SPACE METHODS MATLAB Solution of Example 22.2 A = [−2,0;4, − 3] A = −2 4 0 −3 B = [1,0;0,2] B = 1 0 0 2 syms s t I = eye(2) %Set up the identity matrix I = 1 0 0 1 E = s*I − A E = [s + 2, 0] [−4, s + 3] F = Eˆ −1 F = [1/(s + 2), 0] [4/(s + 2)/(s + 3), 1/(s + 3)] G = F*B G = [1/(s + 2), 0] [4/(s + 2)/(s + 3), 2/(s + 3)] This is the same expression that we obtained by hand for G(s). Let’s check the transition matrix. H = ilaplace(F) H = [ exp(−2*t), 0] [ 4*exp(−2*t) − 4*exp(−3*t), exp(−3*t)] cou9789x_ch22_498-511.indd 506 %This is exp(At), the transition matrix 8/23/08 9:32:52 AM Confirming Pages CHAPTER 22 507 TRANSFER FUNCTION MATRIX This matches the value we arrived at above for the transition matrix. syms tau K = [exp(−2*(t − tau)),0; 4*exp(−2*(t − tau)) − 4*exp(−3*(t − tau)), exp(−3*(t − tau))] K = [ exp(−2*t + 2*tau), 0] [ 4*exp(−2*t + 2*tau) −4*exp(−3*t + 3*tau), exp(−3*t + 3*tau)] %This is exp(A(t−τ)) M = K*B M = [exp(−2*t + 2*tau), 0] [ 4*exp(−2*t + 2*tau) − 4*exp(−3*t + 3*tau), 2*exp(−3*t + 3*tau)] N = M*[1;0] %This is M*u, remember u1 = 1 and u2 = 0 N = [exp(−2*t + 2*tau)] [4*exp(−2*t + 2*tau) − 4*exp(−3*t + 3*tau)] P = int(N,tau,0,t) %The complementary solution. P = [1/2 − 1/2*exp(−2*t)] [2/3 − 2*exp(−2*t) + 4/3*exp(−3*t)] % The first element is H1 and the second is H2. This result matches the earlier result we obtained for H1 and H2. Simulink Modeling of Example 22.2 Simulink can model a state-space system. Using the model developed in Example 22.2, 0 h1 1 0 u1 h1 ⫺2 ⫽ 4 ⫺3 h ⫹ 0 2 u 2 h2 2 A B If we desire to examine the response in tank 2, we want the output y to be equal to h2. Therefore, h1 u1 y = [ 0 1] ⫹ [ 0 0 ] ⫽ outp p ut h2 u2 C D The Simulink block diagram for the model is shown in Fig. 22–2. cou9789x_ch22_498-511.indd 507 8/23/08 9:32:53 AM Confirming Pages 508 PART 6 STATE-SPACE METHODS y To workspace U1 x' = Ax+Bu y = Cx+Du 0 State-space Scope U2 FIGURE 22–2 Simulink model for Example 22.2. The simulation is quite easy to set up. Double-clicking on the state-space block pops up a menu that must be filled in using the values for the matrices A, B, C, and D (see Fig. 22–3). The completed menu box is shown below. The input signals u1 and u2 are multiplexed and fed into the state-space block. The upper signal is u1, a unit step, and the lower signal is u2, which is zero. The scope output is shown in Fig. 22–4. FIGURE 22–3 Simulink state-space menu block. cou9789x_ch22_498-511.indd 508 FIGURE 22–4 Simulink model output, Example 22.2. 8/23/08 9:32:53 AM Confirming Pages CHAPTER 22 TRANSFER FUNCTION MATRIX 509 The output from the model is also sent back to the MATLAB workspace, as y, where we can plot it in comparison to the analytical expression for h2. h2 ( t ) ⫽ ( 2 4 2 ⫹ e⫺3t ⫺ 2e⫺2 t ⫽ 1 ⫹ 2e⫺3t ⫺ 3e⫺2 t 3 3 3 ) The MATLAB commands required to generate a comparison plot are h2 = y.signals.values; t = y.time; h2analytical = (2/3) + (4/3)*exp(−3*t) − 2*exp(−2*t); plot(t,h2,t,h2analytical,’*’) Comparison of h2 calculated with Simulink and the analytical expression 0.7 0.6 0.5 h2 0.4 0.3 Simulink Analytical 0.2 0.1 0 0 0.5 1 1.5 2 Time 2.5 3 3.5 4 FIGURE 22–5 Comparison of h2 calculated using Simulink with analytical solution. From Fig. 22–5, we note that the solutions are identical. cou9789x_ch22_498-511.indd 509 8/23/08 9:32:53 AM Confirming Pages 510 PART 6 STATE-SPACE METHODS SUMMARY The matrix differential equation x ⫽ Ax ⫹ Bu used to describe a control system by the state-space method can be solved for the vector of state variables x by use of the transfer function matrix. It consists of a matrix of transfer functions that relate the state variables to the inputs. The transfer function matrix serves the same purpose in a multiple-input multiple-output system as the transfer function does for a single-input single-output system. The transfer function matrix is obtained from the matrix differential equation by application of Laplace transforms. PROBLEMS 22.1. Determine x(t) for the system x ⫽ Ax ⫹ Bu where e⫺5t e At ⫽ 0 0 x (0) ⫽ 1 ⫺e⫺2 t ⫹ e⫺5t e⫺2 t 3 u (t ) ⫽ 1 2 1 B ⫽ 0 ⫺ 4 22.2 (a) Solve Prob. 21.1, using the techniques in this chapter. (b) Solve Prob. 21.1, using MATLAB/Simulink. 22.3 (a) Solve Prob. 21.2, using the techniques in this chapter. (b) Solve Prob. 21.2, using MATLAB/Simulink. cou9789x_ch22_498-511.indd 510 8/23/08 9:32:54 AM Confirming Pages CHAPTER 22 CAPSULE SUMMARY For the general state-space model, x ⫽ Ax ⫹ Bu y ⫽ Cx ⫹ Du The solution was shown to be x ( t ) ⫽ e At x ( 0 ) ⫹ { t A( t ⫺t ) ∫0 e Bu(t ) d t } where e At ⫽ L⫺1 ( sI ⫺ A )⫺1 ⬅ transition matrix and thus, ∫0 e t At ⫺t { } Bu(t ) d t ⫽ L⫺1 ( sI ⫺ A )⫺1 BU ( s ) If x(0) ⫽ 0 (as is normally the case with deviation variables), X ( s ) ⫽ ( sI ⫺ A )⫺1 BU ( s ) X (s) ⫽ G (s) U (s) or G ( s ) ⫽ ( sI ⫺ A )⫺1 B ⬅ transfer function matrix (rell ates X to U) where G1 ( s ) ⫽ C ( sI ⫺ A )⫺1 B ⬅ transfer fun c tion matrix (relates Y to U) USEFUL MATLAB COMMANDS expm(A)—matrix exponential command. int—performs a symbolic integration. Aˆ -1 — inverts matrix A. x'= Ax + Bu y = Cx + Du > >—Simulink block that performs state-space simulation State-Space 511 cou9789x_ch22_498-511.indd 511 8/23/08 9:32:54 AM Confirming Pages CHAPTER 23 MULTIVARIABLE CONTROL U p to this point, the fundamentals of process dynamics and control have been illustrated by single-input single-output (SISO) systems. The processes encountered in the real world are usually multiple-input multiple-output (MIMO) systems. To explore these concepts, consider the interacting, two-tank liquid-level system in Fig. 23–1 where there is one input, the flow to tank 1 (m1), and one output, the level in tank 2 (h2). In this figure, h2 is related to m1 by a second-order transfer function. From the point of view of a SISO system, the relation between h2 and m1 may be represented by the block diagram in Fig. 23–1b. One may place a feedback control system around the open-loop system of Fig. 23–1b to maintain control of H2. m1 h1 h2 R2 R1 (a) M1 R2 2s2 + 2 s+1 H2 (b) FIGURE 23–1 Single-input single-output (SISO) system: (a) two-tank interacting level system and (b) block diagram for SISO system. 512 cou9789x_ch23_512-530.indd 512 8/22/08 1:16:00 PM Confirming Pages CHAPTER 23 MULTIVARIABLE CONTROL 513 Now consider the same process of Fig. 23–1 in which there are two inputs (m1 and m2) and two outputs (h1 and h2). This system is shown in Fig. 23–2a. A change in m1 alone will affect both outputs (h1 and h2). A change in m2 alone will also change both outputs. (Remember that this is an interacting process for which the level in tank 1 is affected by the level in tank 2.) The interaction between inputs and outputs can be seen more clearly by the block diagram of Fig. 23–2b. In this diagram, the transfer functions show how the change in one of the inputs affects both of the outputs. For example, if a change occurs in only M1, the responses of H1 and H2 are H1 ( s ) ⫽ G11 ( s ) M1 ( s ) H 2 ( s ) ⫽ G21 ( s ) M1 ( s ) The transfer functions in Fig. 23–2b will be worked out for a specific set of process parameters in Example 23.1. (If the tanks were noninteracting, G12 ⫽ 0, with the result that a change in flow to tank 2 would not affect H1.) If both H1 and H2 are to be controlled, a single control loop will not be sufficient; in this case two control loops are needed. The addition of control loops to the interacting system will be considered in Sec. 23.1. m2 m1 h2 h1 R1 R2 (a) M1 + G11 (s) H1 + G21 (s) G12 (s) M2 + G22 (s) + H2 (b) FIGURE 23–2 Multiple-input multiple-output (MIMO) system: (a) level process and (b) block diagram. cou9789x_ch23_512-530.indd 513 8/22/08 1:16:01 PM Confirming Pages 514 PART 6 STATE-SPACE METHODS 23.1 CONTROL OF INTERACTING SYSTEMS The problem of controlling the outputs of a MIMO system will be discussed by means of a 2 ⫻ 2 system shown in Fig. 23–3. The problem can be extended to the case of more than two pairs of inputs and outputs by the same procedure described here. The control objective is to control C1 and C2 independently, in spite of changes in M1 and M2 or other load variables not shown. Two control loops are added to the diagram of Fig. 23–3 as shown in Fig. 23–4. Each loop has a block for the controller, the valve, and the measuring element. In principle, the multiloop control system of Fig. 23–4 will maintain control of C1 and C2. However, because of the interaction present in the system, a change in R1 will also cause C2 to vary because a disturbance enters the lower loop through the transfer function G21. Because of interaction, both outputs (C1 and C2) will change if a change is made in either input alone. If G21 and G12 provide weak interaction, the two-controller scheme of Fig. 23–4 will give satisfactory control. In the extreme, if G12 ⫽ G21 ⫽ 0, we have no interaction and the two control loops are isolated from each other. + G11 M1 C1 + G21 G12 M2 + G22 + C2 FIGURE 23–3 MIMO system for two pairs of inputs and outputs Gm1 R1 + − E1 Gc11 Gv1 M1 G11 + G21 G12 R2 + E2 Gc22 Gv2 M2 G22 C1 + + + C2 − Gm2 FIGURE 23–4 Multiloop control system with two controllers. cou9789x_ch23_512-530.indd 514 8/22/08 1:16:02 PM Confirming Pages CHAPTER 23 MULTIVARIABLE CONTROL 515 To completely eliminate the interaction between outputs and set points, two more controllers (cross-controllers) are added to the diagram of Fig. 23–4 to give the diagram shown in Fig. 23–5. In principle, these cross-controllers can be designed to eliminate interaction. The following analysis, which is expressed in matrix form, will lead to the method of design for cross-controllers that will eliminate interaction. Response of Multiloop Control System From Fig. 23–5, we may write by direct observation the following relationships in the form of the matrix expression (23.1) C ⫽ G pM G11 G12 where G p = G21 G22 C1 C = C2 M1 M = M 2 We also may write from Fig. 23–5 M1 ⫽ Gv1Gc11E1 ⫹ Gv1Gc12 E2 (23.2) M 2 ⫽ Gv 2Gc 21E1 ⫹ Gv 2Gc 22 E2 (23.3) where Gv1 and Gv2 are the transfer functions for the valves. Equations (23.2) and (23.3) may be written in matrix form as M ⫽ G v Gc E Gv1 0 where G v = 0 Gv 2 (23.4) Gc11 Gc12 Gc = Gc 21 Gc 22 (valve matrix ) (controller matrix ) E1 E = E2 Gm1 R1 + − E1 Gc 11 Gv 1 + M1 G11 + Gc 12 R2 + E2 Gc 22 + G12 + Gv 2 M2 − Gm2 cou9789x_ch23_512-530.indd 515 + G21 Gc 21 G22 C1 + + + C2 FIGURE 23–5 Multiloop control system with two primary controllers and two crosscontrollers. 8/22/08 1:16:02 PM Confirming Pages 516 PART 6 STATE-SPACE METHODS From Fig. 23–5, we write directly E1 ⫽ R1 ⫺ Gm1C1 (23.5) E2 ⫽ R2 ⫺ Gm 2C2 (23.6) where E1 and E2 are the error signals from the comparators. Equations (23.5) and (23.6) can be written in matrix form E ⴝ R ⴚ G mC (23.7) 0 Gm1 (measuring element matrix) where G m ⫽ Gm 2 0 R1 R ⫽ R2 From Eqs. (23.1) and (23.4), we obtain C ⫽ G p G v Gc E (23.8) If we let Go ⫽ Gp Gv Gc, Eq. (23.8) becomes C ⫽ Go E (23.9) Combining Eqs. (23.7) and (23.9) gives C ⫽ Go R ⫺ GoG m C (23.10) We may now solve Eq. (23.10) for C to obtain C ⫽ [ I ⫹ G oG m ]⫺1 G o R (23.11) Notice that the closed-loop behavior expressed by this matrix equation is analogous to the closed-loop response of a SISO system, which may be written C (s) ⫽ Go ( s ) R (s) 1 ⫹ Go ( s ) Gm ( s ) (23.12) The matrix term ( I ⫹ Go Gm)⫺1 is equivalent to the scalar term 1/[1 ⫹ Go(s)Gm(s)]. R + E − Gc Gv M Gp C Gm FIGURE 23–6 Block diagram for MIMO control system in terms of matrix blocks. cou9789x_ch23_512-530.indd 516 8/22/08 1:16:03 PM Confirming Pages CHAPTER 23 MULTIVARIABLE CONTROL 517 A block diagram equivalent to the diagram for the MIMO control system in Fig. 23–5 is − shown in Fig. 23–6. In this diagram, the blocks are filled with the matrices in Eqs. (23.1), (23.4), Gm and (23.7). The double line indicates that more than one variable is being transmitted. Each block FIGURE 23–7 contains a matrix of transfer functions that relates Reduced diagram for MIMO control an output vector to an input vector. The diasystem where Go ⫽ Gp Gv Gc. gram can be simplified by multiplying the three matrices in the forward loop together and calling the result Go, as was done to obtain Eq. (23.9). The simplified diagram is shown in Fig. 23–7. R E + Go C Noninteracting Control In order for no interaction to occur between C and R in Fig. 23–5 (i.e., R1 affects only C1 and R2 affects only C2), the off-diagonal elements of [ I ⫹ Go Gm]⫺1 Go in Eq. (23.11) must be zero. Since I and Gm are diagonal, [ I ⫹ Go Gm]⫺1 Go will be diagonal if Go is diagonal. Multiplication of the matrices in the expression for Go is now shown: G o ⫽ G p G v Gc 0 Gc11 Gc12 G11 G12 Gv1 Go ⫽ G21 G22 0 Gv 2 Gc 21 Gc 22 The result of multiplying these matrices gives G11Gv1Gc11 ⫹ G12Gv 2Gc 21 G11Gv1Gc12 ⫹ G12Gv 2Gc 22 Go ⫽ (23.13) G21Gv1Gc11 ⫹ G22Gv 2Gc 21 G21Gv1Gc12 ⫹ G22Gv 2Gc 22 Setting the off-diagonal elements to zero and solving for Gc12 and Gc21 give Gc12 ⫽ ⫺ G12Gv 2Gc 22 G11Gv1 (23.14) G21Gv1Gc11 G22Gv 2 (23.15) Gc 21 ⫽ ⫺ Example 23.1 will give some experience with the computations involved in applying the theory developed so far in this chapter. Example 23.1. For the two-tank, interacting liquid-level system shown in Fig. 23–8, develop the block diagram for a MIMO system corresponding to Fig. 23–3. Material balances around tank 1 and tank 2 give the following differential equations: c1 ⫺ c2 c1 ⫺ A1c1 ⫽ m1 ⫺ R1 R3 (23.16) A2c2 ⫽ m2 ⫹ cou9789x_ch23_512-530.indd 517 c1 ⫺ c2 c2 ⫺ R1 R2 (23.17) 8/22/08 1:16:05 PM Confirming Pages 518 PART 6 STATE-SPACE METHODS m1 m2 A1 A2 c2 c1 R3 R1 R2 FIGURE 23–8 Process for Example 23.1: A1 ⫽ 1, A2 ⫽ 1 ,R 2 1 ⫽ 1 ,R 2 2 ⫽ 2, R3 ⫽ 1 . Introducing the parameters given in Fig. 23–8 into Eqs. (23.16) and (23.17) gives c1 ⫽ m1 ⫺ 3c1 ⫹ 2c2 (23.18) c2 ⫽ 2m2 ⫹ 4c1 ⫺ 5c2 (23.19) These equations may be written in matrix form as c ⴝ Ac ⴙ Bm where 2 ⫺3 A ⫽ 4 ⫺5 1 0 B ⫽ 0 2 We use Eq. (22.15) to obtain C ( s ) ⫽ ( sI ⫺ A )⫺1 BM ( s ) (23.20) Writing Eq. (23.20) in the form of Eq. (23.1) gives C ⫽ G pM where Gp ⫽ (s I ⫺ A)⫺1 B. After several steps involving the inversion of s I ⫺ A and multiplying the result of inversion by B, we get 4 s ⫹ 5 4 ) 2 s ⫹ 3 ( Gp ⫽ ( s ⫹ 1) ( s ⫹ 7 ) (23.21) The block diagram can now be drawn as shown in Fig. 23–3 with G11 ⫽ G21 ⫽ cou9789x_ch23_512-530.indd 518 s⫹5 ( s ⫹ 1) ( s ⫹ 7 ) 4 ( s ⫹ 1) ( s ⫹ 7 ) G12 ⫽ G22 ⫽ 4 ( s ⫹ 1) ( s ⫹ 7 ) 2(s ⫹ 3) ( s ⫹ 1) ( s ⫹ 7 ) 8/22/08 1:16:06 PM Confirming Pages CHAPTER 23 MULTIVARIABLE CONTROL 519 Notice that the diagonal elements of Gp(s) are of the form a (s ⫹ b ) ( s ⫹ 1) ( s ⫹ 7 ) 1.0 1.0 0.8 0.8 c1 0.6 0.6 c2 0.4 0.4 0.2 0.2 0.0 0 t 1.5 (a) c2 c1 0.0 3 0 1.5 (b) t 3 FIGURE 23–9 1 Open-loop response for Example 23.1. (a) M1 ⫽ , M 2 ⫽ 0 (step change to tank 1) and s 1 (b) M1 ⫽ 0, M 2 ⫽ (step change to tank 2). s These elements, which relate c1 to m1 and c2 to m2, will produce a second-order response to a step change in input that has a finite slope at the origin because of the numerator term s ⫹ b. In contrast, the off-diagonal elements have second-order transfer functions without numerator dynamics, for which case the step response will be second-order with zero slope at the origin. The responses of c1 and c2 for unit-step changes in m1 and m2 taken separately are shown in Fig. 23–9. Example 23.2. For the two-tank liquid-level system of Example 23.1, determine the controller transfer function matrix Gc needed to eliminate interaction. The primary controllers are to be proportional; i.e., Gc11 ⫽ K1 and Gc22 ⫽ K2. The diagram of the control system is shown in Fig. 23–10. r1 r2 m2 m1 Controller A1 c2 c1 R3 R2 R1 FIGURE 23–10 Process for Example 23.2. A1 ⫽ 1, A2 ⫽ cou9789x_ch23_512-530.indd 519 A2 1 ,R 2 1 ⫽ 1 ,R 2 2 ⫽ 2, R3 ⫽ 1, Gc11 ⫽ K1 , Gc 22 ⫽ K 2 . 8/22/08 1:16:07 PM Confirming Pages 520 PART 6 STATE-SPACE METHODS The block labeled Controller contains the four transfer functions that are the elements of Gc. In this problem, Gv is a unit diagonal matrix, i.e., Gv1 ⫽ Gv2 ⫽ 1. From Eqs. (23.14) and (23.15) we obtain Gc12 ⫽ ⫺ G12Gc 22 4 ( s ⫹ 1) ( s ⫹ 7 ) K2 ⫽⫺ G11 s⫹5 ( s + 1) ( s + 7 ) Gc12 ⫽ or Gc 21 ⫽ ⫺ ⫺4 K 2 s⫹5 (23.22) G21Gc11 4 ( s ⫹ 1) ( s ⫹ 7 ) K1 ⫽⫺ G22 2 ( s ⫹ 3) ( s ⫹ 1) ( s ⫹ 7 ) ⫺2 K1 (23.23) s⫹3 Having found the transfer functions for the cross-controllers, we can now determine the nature of the uncoupled response of c1 to a change in r1 and of c2 to a change in r2. Inserting Gv1 ⫽ Gv2 ⫽ 1 and the expressions for Gc12 and Gc21 from Eqs. (23.14) and (23.15) into Eq. (23.13) gives for Go Gc21 ⫽ or 0 G11Gc11 ⫹ G12Gc 21 Go ⫽ 0 G21Gc12 ⫹ G22Gc 22 (23.24) Inserting the appropriate elements of the Gp matrix [Eq. (23.21)] and the Gc matrix in Eq. (23.24) gives, after considerable simplification, K1 s ⫹ 3 Go ⫽ 0 2K 2 s ⫹ 5 0 decoupled system m (23.25) The block diagram for this decoupled MIMO system is shown in Fig. 23–11. Go R K1 s+3 E + − 0 Gm1 0 0 Gm2 0 2K2 C s+5 Gm FIGURE 23–11 Block diagram for decoupled system in Example 23.2. cou9789x_ch23_512-530.indd 520 8/22/08 1:16:08 PM Confirming Pages CHAPTER 23 MULTIVARIABLE CONTROL 521 Assuming that the measurement matrix Gm is a unit diagonal matrix, the diagram in Fig. 23–11 can be simplified to the unity feedback diagram of Fig. 23–12. Go R E + − K1 (s + 3) 0 0 2K2 s+5 C FIGURE 23–12 Simplified block diagram for Example 23.2. From Fig. 23–12, we may write directly C ⫽ Go E E ⫽ R⫺C therefore, C ⫽ G o R ⫺ G oC or 0 R1 Go11 0 C1 Go11 C1 ⫺ C 2 ⫽ 0 Go 22 C2 Go 22 R2 0 From this expression, we may write C1 ⫽ Go11 R1 ⫺ Go11C1 C2 ⫽ Go 22 R2 ⫺ Go 22C2 Solving for C1(s) gives Go11 R1 ( s ) 1 ⫹ Go11 Inserting Go11 from Eq. (23.25) gives K1 / ( s ⫹ 3) C1 ( s ) ⫽ R1 ( s ) 1 ⫹ K1 / ( s ⫹ 3) C1 ( s ) ⫽ (23.26) In a similar way, one can show that C2 ( s ) ⫽ 2 K 2 / ( s ⫹ 5) R2 ( s ) 1 ⫹ 2 K 2 / ( s ⫹ 5) (23.27) The result shows that the cross-controllers of Eqs. (23.22) and (23.23) give two separate noninteracting control loops, as shown in Fig. 23–13. The response of the control system of Fig. 23–10 is shown in Fig. 23–14 for a unit-step change in R1. In Fig. 23–14a, no cross-controllers are present in the matrix Gc. In Fig. 23–14b, cross-controllers having the transfer functions given by Eqs. (23.22) and (23.23) are present. As expected, for the case of no crosscontrollers, one sees from Fig. 23–14a that a request for a unit-step change in r1 causes both c1 and c2 to change. For the case where cross-controllers are present, we see from Fig. 23–14b that a change in r1 does not affect c2 as demanded by a decoupled system. cou9789x_ch23_512-530.indd 521 8/22/08 1:16:09 PM Confirming Pages 522 PART 6 STATE-SPACE METHODS + R1 − + R2 − K1 1 s+3 C1 K2 2 s+5 C2 FIGURE 23–13 Decoupled control system for Example 23.2 where primary controllers are proportional. 1.0 1.0 0.8 0.8 c1 0.6 c1 0.6 0.4 0.4 c2 0.2 0.2 0.0 t 0 1.5 (a) 3 c2 0.0 t 0 1.5 (b) 3 FIGURE 23–14 Response for control system in Example 23.2 for R1 ⫽ 1/s, R2 ⫽ 0, Gc11 ⫽ K1 ⫽ 4, Gc22 ⫽ K2 ⫽ 4. (a) No cross-controllers; (b) cross-controllers present. Example 23.3. Use MATLAB and Simulink to simulate the system shown in Example 23.2. Examine the system behavior for values of Kc (both K1 and K2) ⫽ 2, 4, 6, 8, and 10. The Simulink diagram for the process is shown in Fig. 23–5. The PID controller block for each loop is a proportional controller only, Gc11 ⫽ K1 and Gc22 ⫽ K2. We use a MATLAB m-file to call this model, using the following command: >>ex23_3driver The MATLAB m-file that calls the Simulink model is % The name of this file is ex23_3driver.m % The Simulink file for the model is example23_3 % The variable h is used to plot each line in different colors h(1,:)='g−'; h(2,:)='r−'; h(3,:)='b−'; cou9789x_ch23_512-530.indd 522 8/22/08 1:16:10 PM Confirming Pages CHAPTER 23 +− E1 PID ++ M1 GC11 R1 −2*k1 s+3 GC21 −4*k2 s+5 GC12 s2 ++ +− R2 M2 E2 E2 s+5 s2 + 8s + 7 G11 4 s2 + 8s + 7 G21 PID MULTIVARIABLE CONTROL ++ 1 Out1 C1 C1 Scope C2 4 + 8s + 7 G12 2s + 6 s2 + 8s + 7 G22 523 2 Out2 ++ C2 GC22 FIGURE 23–15 Simulink model for simulation of Example 23.2. h(4,:)='k − '; h(5,:)='g− −'; h(6,:)='r − −'; for i=1:5 K1=2*i; K2=2*i; % The variable y in the 'sim' statement is taken from the outport in the model [t,x,y] = sim('example23_3',3); plot(t,y(:,1),h(i,:)) hold on plot(t,y(:,2),h(i+1,:)) end grid title('c_1 and c_2 vs time for K_c = 2,4,6,8,10'); grid hold off The output from the m-file is the desired plot shown in Fig. 23–16. (Note that the text labels for the lines were added afterward using MATLAB plot editing commands by double-clicking the graph and filling in the appropriate menu boxes.) The MATLAB output shows that as we increase the controller gain, the height in tank 1 more nearly tracks the set point change (as we might expect), and the height in tank 2 is unaffected because of the decoupling crosscontrollers. cou9789x_ch23_512-530.indd 523 8/22/08 1:16:11 PM Confirming Pages 524 PART 6 STATE-SPACE METHODS c1 and c2 vs time for Kc = 2, 4, 6, 8, 10 0.8 K1 = 10 0.7 0.6 0.5 c1 K1 = 2 0.4 0.3 R1 = 1/s R2 = 0 0.2 0.1 Including cross-controllers 0 −0.1 0.5 0 1 1.5 Time c2 2 2.5 3 FIGURE 23–16 Solution of Example 23.2 using MATLAB/Simulink. To completely avoid the offset associated with proportional control, we can use PI controllers for the primary controllers for the decoupled system. To study the effect of PI controllers for the decoupled system, let 1 Gc11 ⫽ K1 1 ⫹ s and 1 Gc 22 ⫽ K 2 1 ⫹ s Note that tI ⫽ 1 for both controllers. For this case, the cross-controller transfer functions may be obtained from Eqs. (23.14) and (23.15); the results are Gc12 ⫽ ⫺4 K 2 ( s ⫹ 1) s ( s ⫹ 5) and Gc 21 ⫽ ⫺2 K1 ( s ⫹ 1) s ( s ⫹ 3) A simulation using these four controller transfer functions with K1 ⫽ K2 ⫽ 4 is shown in Fig. 23–17. From the transient response, we see that c1 moves toward the set point of 1.0 and that c2 does not change, as is expected for a decoupled system. Set Point 1.0 c1 0.8 0.6 0.4 0.2 c2 t 0.0 0 cou9789x_ch23_512-530.indd 524 1.5 3 FIGURE 23–17 Response of decoupled control system in Example 23.2 for PI primary controllers: Gc11 ⫽ Gc22 ⫽ 4(1 ⫹ 1/s), R1 ⫽ 1/s, R2 ⫽ 0. 8/22/08 1:16:11 PM Confirming Pages CHAPTER 23 23.2 MULTIVARIABLE CONTROL 525 STABILITY OF MULTIVARIABLE SYSTEMS Determining the stability for a multivariable control system, such as the one in Fig. 23–4 or Fig. 23–5, can be much more complicated than for a SISO system. The transfer function for the closed-loop response of a MIMO system is given by Eq. (23.11): C ⫽ [ I ⫹ G oG m ]⫺1 G o R To invert this expression, we write C ⫽ adj [ I ⫹ G oG m ] G o R I ⫹ GoGm (23.28) The numerator of this expression is an n ⫻ n matrix; the denominator is a nth-order polynomial. To simplify the following argument, let the matrix in Eq. (23.28) be 2 ⫻ 2. Let the elements of the numerator, after expansion, be written as follows: b11 ( s ) b12 ( s ) adj[ I ⫹ G oG m ] G o R ⫽ b 21 ( s ) b 22 ( s ) (23.29) Let the elements of Go Gm be written as follows: a 11 ( s ) a 12 ( s ) GoGm ⫽ a 21 ( s ) a 22 ( s ) (23.30) Expansion of the determinant in Eq. (23.28), using Eq. (23.30), is I ⫹ GoGm ⫽ or a 12 ( s ) 1 ⫹ a 11 ( s ) a 21 ( s ) 1 ⫹ a 22 ( s ) I ⫹ G oG m ⫽ [1 ⫹ a 11 ( s )] [1 ⫹ a 22 ( s )] ⫺ a 12 ( s ) a 21 ( s ) (23.31) Equation (23.31) is a polynomial expression, for which the order will depend on the order of the transfer functions in Go and Gm. Equation (23.28) can now be written in terms of the expansions shown in Eqs. (23.29) and (23.31) as follows: I C ⫽ I b11 ( s ) ⫹ GoGm b 21 ( s ) ⫹ GoGm b12 ( s ) I ⫹ GoGm b 22 ( s ) I ⫹ GoGm Since each term contains the polynomial | I ⫹ Go Gm| in the denominator, the stability of the multivariable system will depend on the roots of the polynomial equation I ⫹ GoGm ⫽ 0 cou9789x_ch23_512-530.indd 525 characteristic equation (23.32) 8/22/08 1:16:12 PM Confirming Pages 526 PART 6 STATE-SPACE METHODS Equation (23.32) is the characteristic equation of the multivariable system. Although Eq. (23.32) has been derived here for the case where Go Gm is a 2 ⫻ 2 matrix, one can show that Eq. (23.32) applies to the general MIMO system of Fig. 23–7 in which Go Gm is a matrix of any size (n ⫻ n). If the roots of the characteristic equation are in the left half of the complex plane, we know that the system is stable. One method to be used for examining the stability of a multivariable system is to apply the Routh test to the characteristic equation of Eq. (23.32). In practice, the characteristic equation can be of high order for a simple 2 ⫻ 2 multivariable control system. Example 23.4 illustrates the determination of stability for a multivariable control system. Example 23.4. For the control system of Example 23.2, which is shown in Fig. 23–10, determine stability for the case where Gc11 ⫽ K1, Gc22 ⫽ K2, and there are no cross-controllers present (i.e., Gc12 ⫽ Gc21 ⫽ 0); also let Gm and Gv be unit matrices. From Example 23.1, we have for the elements of Gp G11 ⫽ G21 ⫽ s⫹5 ( s ⫹ 1) ( s ⫹ 7 ) 4 ( s ⫹ 1) ( s ⫹ 7 ) G12 ⫽ G22 4 ( s ⫹ 1) ( s ⫹ 7 ) 2 ( s ⫹ 3) ⫽ ( s ⫹ 1) ( s ⫹ 7 ) Since Gv ⫽ I, Go ⫽ Gp Gc. Since Gm ⫽ I, the characteristic equation of Eq. (23.32) can now be written as I ⫹ G p Gc ⫽ 0 (23.33) Introducing the elements of the matrices Gp and Gc into Eq. (23.33) gives, after expansion of the determinant, [( s ⫹ 1) ( s ⫹ 7) ⫹ K1 ( s ⫹ 5)][( s ⫹ 1) ( s ⫹ 7) ⫹ 2K 2 ( s ⫹ 3)] ⫺ 16K1K 2 ⫽ 0 For given values of K1 and K2, this expression can be expanded into a fourthorder polynomial equation of the form s 4 ⫹ as3 ⫹ bs 2 ⫹ g s ⫹ ∆ ⫽ 0 (23.34) where a, b, g, and ∆ will include the gains K1 and K2. The Routh test can be applied to Eq. (23.34) to determine whether the system is stable. From this simple example, the reader can appreciate the algebraic tedium that may be needed to determine the stability of a multivariable system. One way to express the stability of this system is to plot the stability boundaries on a graph of K1 versus K2. The region within the boundaries gives the combinations of values of K1 and K2 for which the system is stable. Since the details of stability boundaries are beyond the scope of this chapter, the reader may consult Seborg, Edgar, and Mellichamp (2004) for examples of stability boundaries for multivariable systems. cou9789x_ch23_512-530.indd 526 8/22/08 1:16:13 PM Confirming Pages CHAPTER 23 MULTIVARIABLE CONTROL 527 SUMMARY Most of the systems encountered are multiple-input multiple-output (MIMO) systems. Such systems have several inputs and several outputs that are often interacting, meaning that a disturbance at any input causes a response in some of or all the outputs. This interaction in a MIMO system makes control and stability analysis of the system very complicated compared to that of a single-input single-output (SISO) system. A convenient way to describe a MIMO system is by means of a block diagram in which each block contains a matrix of transfer functions that relates an input vector to an output vector. It is often desirable to have a control system decoupled so that certain outputs can be controlled independently of other outputs. A systematic procedure was described for decoupling a control system by including cross-controllers along with the principal controllers. This approach to decoupling requires an accurate model of the system; the number of controllers (principal controllers and cross-controllers) increases rapidly with the number of inputs and outputs. A system represented by two inputs and two outputs requires as many as four controllers; a system of three inputs and three outputs requires as many as nine controllers; and so on. The characteristic equation for a multivariable control system, from which one can determine stability by examining its roots, can be of high order for a relatively simple system. Expressing stability boundaries in terms of controller parameters becomes complex because of the large number of controller parameters that can be adjusted. PROBLEMS 23.1. For the liquid-level system shown in Fig. P23–1 determine the cross-controller transfer functions that will decouple the system. Fill in each block of the diagram shown in Fig. 23–5 with a transfer function obtained from an analysis of the control system. The transfer function for each feedback measuring element is unity. The following data apply: A1 ⫽ 1 A2 ⫽ 0 . 5 R1 Res1 ⫽ 0 . 5 Res 2 ⫽ 2 3 Gc11 ⫽ K1 Gc 22 ⫽ K 2 R2 Gc m1 A1 x1 m2 Res1 A2 x2 Res2 cou9789x_ch23_512-530.indd 527 FIGURE P23–1 8/22/08 1:16:14 PM Confirming Pages 528 PART 6 STATE-SPACE METHODS The resistance on the outlet of a tank has been denoted by Res to avoid confusion with the symbol for set point (R). 23.2. (a) For the interacting liquid-level system shown in Fig. P23–2, draw very neatly a block diagram that corresponds to Fig. 23–4. Each block should contain a transfer function obtained from an analysis of the liquid-level system. There are no cross-controllers in this system. The transfer function for each feedback element is unity. The following data apply: A1 ⫽ 1 A2 ⫽ Proportional controller Res1 ⫽ 1 2 Res2 ⫽ 2 Res3 ⫽ 1 Proportional controller R1 K1 ≡ K11 = 1 1 2 K2 ≡ K22 R2 m1 m2 A1 A2 c1 c2 Res3 Res1 Res2 FIGURE P23–2 (b) Obtain the characteristic equation of this system in the form s n ⫹ a s n ⫺1 ⫹ b s n ⫺ 2 ⫹ . . . ⫽ 0 Obtain expressions for a, b, etc., in terms of K2 (K1 ⫽ 1). (c) How would you determine stability limits for this interacting control system? 23.3. Modify the Simulink model from Example 23.3 to simulate a PI controller for the process and verify Fig. 23–17. cou9789x_ch23_512-530.indd 528 8/22/08 1:16:15 PM Rev. Confirming Pages CHAPTER 23 CAPSULE SUMMARY SISO ⫽ single-input single-output system MIMO ⫽ multiple-input multiple-output system. Most real-life systems are MIMO. In MIMO systems, the variables can interact, making control difficult. A change in one variable by a controller attempting to drive the process variable to its desired set point can cause unwanted disturbances in other process variables. Decoupling process variable interactions is possible by introducing appropriate cross-controllers. MATLAB and Simulink are useful for simulating MIMO systems. 529 cou9789x_ch23_512-530.indd 529 8/30/08 3:31:23 PM cou9789x_ch23_512-530.indd 530 8/22/08 1:16:15 PM Confirming Pages PART VII NONLINEAR CONTROL cou9789x_ch24_531-552.indd 531 8/22/08 1:18:58 PM cou9789x_ch24_531-552.indd 532 8/22/08 1:18:58 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS I n previous chapters, we have confined our attention to the behavior of linear systems or to the analysis of linearized equations representative of nonlinear systems in the vicinity of the steady-state condition. While much useful information can be obtained from such analysis, it frequently is desirable or necessary to consider nonlinearities in control system design. No real physical system is truly linear, particularly over a wide range of operating variables. Hence, to be complete, a control system design should allow for the possibility of a large deviation from steady-state behavior and resulting nonlinear behavior. The purpose of Chaps. 24 and 25 is to introduce some of the tools that can be used for this purpose and to indicate some of the complications that arise when nonlinear systems are considered. 24.1 DEFINITION OF A NONLINEAR SYSTEM A nonlinear system is one for which the principle of superposition does not apply. Thus, by superposition, the response of a linear system to the sum of two inputs is the same as the sum of the responses to the individual inputs. This behavior, which allows us to characterize completely a linear system by a transfer function, is not true of nonlinear systems. As an example, consider a liquid-level system. If the outflow is proportional to the square root of the tank level, superposition does not hold and the system is nonlinear. If the tank will always operate near the steady-state condition, the square root behavior may be adequately represented by a straight line and superposition applied, as we have done before. On the other hand, if the tank level were to fall to one-half the steady-state value, we would no longer expect the transfer function derived on the linearized basis to apply. The analysis becomes more complicated, as we shall see in our introduction to the study of nonlinear systems. 533 cou9789x_ch24_531-552.indd 533 8/22/08 1:18:58 PM Confirming Pages 534 PART 7 NONLINEAR CONTROL 24.2 THE PHASE PLANE The analysis of nonlinear dynamic systems may often be conceptually simplified by changing to a coordinate system known as phase space. In this coordinate system, time no longer appears explicitly, it being replaced by some other property of the system. For example, consider the flight of a rocket to the moon. In a grossly oversimplified manner, we may describe this motion by a plot of the distance of the rocket from the moon versus time. If all goes well, we would like such a plot to resemble Fig. 24–1. Note the initial acceleration during launch and the final deceleration at landing. We may, however, also represent this motion by a plot of rocket velocity versus distance from the moon. This plot is shown in Fig. 24–2, where velocity is defined as d (distance from moon)/dt. Figure 24–2 is called a phase diagram of the rocket motion. Time now appears merely as a parameter along the curve of the rocket motion. It has been replaced as a coordinate by the rocket velocity. Although in the present example Fig. 24–2 may not be of significant advantage over Fig. 24–1, we shall find phase diagrams very helpful in the analysis of certain nonlinear control systems. Distance from moon Earth Landing Launch 0 Time FIGURE 24–1 Distance-time plot for moon rocket. Earth Velocity Distance from moon 0 Time FIGURE 24–2 Velocity-distance plot for moon rocket. To begin our study of phase diagrams, we convert the second-order system studied previously in Chap. 7 to the phase plane. cou9789x_ch24_531-552.indd 534 8/22/08 1:18:58 PM Confirming Pages CHAPTER 24 24.3 EXAMPLES OF NONLINEAR SYSTEMS 535 PHASE-PLANE ANALYSIS OF DAMPED OSCILLATOR The differential equation describing the motion of the system of Fig. 7–1 in response to a unit-step function is d 2Y dY ⫹ 2zt ⫹Y ⫽ 1 (24.1) 2 dt dt Equation (24.1) has previously been solved to yield the motion in the form of Y(t) versus t as shown in Fig. 7–3. For phase analysis, however, we want the motion in terms of velocity versus position, Y versus Y, where the dot notation is used to indicate differentiation with respect to t. Hence, we rewrite Eq. (24.1) as t2 dY ⫽ Y dt dY ⫺Y ⫺ 2ztY ⫹ 1 ⫽ dt t2 (24.2) It is sometimes convenient in phase-plane analysis to write the variables in terms of deviation about the final condition. In this case, the system will ultimately come to rest d2y dy equal 0 at steady state). Hence and at Y ⫽ 1 (because both time derivatives dt dt 2 we define X ⫽ Y ⫺1 X ⫽ Y Then Eq. (24.2) becomes dX ⫽ X dt dX ⫺X ⫺ 2ztX ⫽ dt t2 (24.3) These are now viewed as two simultaneous, first-order differential equations in the variables X and X (which, we recall, are phase variables). To solve Eqs. (24.3), we may use the methods presented in Chaps. 21 and 22. For this purpose, let X1 ⫽ X and X 2 ⫽ X . Equations (24.3) may be written in the form 1 0 X1 ⫽ ⫺1 ⫺2z X2 t 2 t or X1 where X ⫽ X2 cou9789x_ch24_531-552.indd 535 ⫽ AX X 1 0 A ⫽ ⫺1 ⫺2z t 2 t X1 X 2 (24.4) 8/22/08 1:18:59 PM Confirming Pages 536 PART 7 NONLINEAR CONTROL Equation (24.4) is in the standard form of a matrix differential equation [Eq. (21.7)]. Notice that the term bu of Eq. (21.7) is not present because no forcing term is present in Eqs. (24.3). Equation (24.4) may be solved by use of Eq. (22.7): X ( t ) ⫽ e At X ( 0 ) where { e At ⫽ L⫺1 ( sI ⫺ A )⫺1 (22.7) } (22.13) Following the usual steps required to solve these equations gives the result X1 ⫽ X ⫽ C1e s1 t ⫹ C2e s2 t X 2 ⫽ X ⫽ s1C1e s1 t ⫹ s2C2e s2 t C1 ⫽ where C2 ⫽ (24.5) s2 X o ⫺ X o s2 − s1 X o ⫺ s1 X o s2 − s1 and X0 and X 0 are the initial conditions; thus X0 ⫽ X(0) and X 0 ⫽ X ( 0 ) . The terms s1 and s2 are the roots of the characteristic equation | sI ⫺ A |⫽ 0 (24.6) Expanding this equation gives t 2 s 2 ⫹ 2zts ⫹ 1 ⫽ 0 This quadratic equation has two roots: s1, 2 ⫽ ⫺z ⫾ z 2 ⫺ 1 t If we take s2 as the root with the positive sign s2 ⫽ ⫺z ⫹ z 2 ⫺ 1 t the constants take the form C1 ⫽ C2 ⫽ t 2 z2 ⫺1 t 2 z2 ⫺1 ( s2 X0 ⫺ X 0 ) ( X 0 ⫺ s1 X 0 ) (24.7) Equations (24.5) and (24.7) together give X(t) and X ( t ) for all possible initial conditions X0 and X 0 . For a given set of initial conditions, we compute C1 and C2 from Eq. (24.7), and then each value of t in Eq. (24.5) yields a pair of values for X and X. cou9789x_ch24_531-552.indd 536 8/22/08 1:19:00 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS 537 These may be plotted as a point on an X -X diagram (i.e., a phase plane). The locus plane. As an of these points as t varies from zero to infinity will be a curve in the XX example, consider the case X 0 ⫽ ⫺1, X 0 ⫽ 0, z ⬍ 1 . The solution is already known to us in the form of X versus t (Chap. 7) and is replotted in Fig. 24–3 for convenience, together with a plot of X versus t. If these curves are replotted as X versus X, with t as a parameter, the result is as shown in Fig. 24–4. The reader should carefully compare Figs. 24–3 and 24–4 to be satisfied that they are indeed equivalent. The relationship between the two may be expressed by the statement that Fig. 24–3 is a parametric representation of Fig. 24–4. Having only the curve X versus t of Fig. 24–3, one can construct Fig. 24–4. X t . X FIGURE 24–3 Typical motion of second-order system. t t . X t=0 −1 X t=∞ at origin FIGURE 24–4 Phase plane corresponding to motion of Fig. 24–3. To explore the phase diagram concept further, note that division of the second of Eqs. (24.3) by the first yields dX ⫺X ⫺ 2ztX ⫽ (24.8) dX t 2 X in which the variable t has been eliminated. Equation (24.8) may be recognized as a homogeneous first-order differential equation. Hence, the substitution X ⫽ VX yields ( ⫺ 1 ⫺ 2ztV ⫹ t 2V 2 X dV ⫺1 ⫺ 2ztV ⫽ ⫺ V ⫽ dX t 2V t 2V cou9789x_ch24_531-552.indd 537 ) 8/22/08 1:19:01 PM Confirming Pages 538 PART 7 NONLINEAR CONTROL an equation which is separable in X and V. This can then be easily solved for V in terms of X. Finally, replacing V ⫽ X / X gives the solution for X versus X, or the equation for the curve of Fig. 24–4. The algebraic details of this rather tedious process are omitted. The point of the discussion is to emphasize further the equivalence between the description of the motion as X versus t or X versus X. A convenient feature of the phase diagram is that several motions, corresponding to different initial conditions, can be readily plotted on the same diagram. Thus, if we add to Fig. 24–4 a curve for the motion under the initial condition X 0 ⫽ 1, X 0 ⫽ 0, we obtain Fig. 24–5. This new trajectory represents the motion of the system after it is stretched 2 units and released from rest. (This follows from the definition X ⫽ Y ⫺ 1.) Furthermore, we have also interpolated in Fig. 24–5 to obtain the motion corresponding to X 0 ⫽ 0, X 0 ⫽ 1, As we shall see later, this interpolation is justified. Hence, it is evident that the phase diagram gives us the “big picture” of the motion of the underdamped U-tube manometer or a spring-mass-damper system. No matter where the system starts, it spirals to the condition X 0 ⫽ 1, X 0 ⫽ 0, the steady-state position. This spiral motion in the phase plane corresponds to the oscillatory nature of the X versus t curve of Fig. 24–3. . X0 = −1, X0 = 0 Interpolation . for X0 = 0, X0 = 1 x . X0 = 1, X0 = 0 FIGURE 24–5 Interpolation on the phase plane. Before we begin a more detailed study of the mechanics of phase analysis, it may be worthwhile to see how situations amenable to such analysis arise naturally in the physical world. The Damped Oscillator A classical example from mechanics is the damped oscillator, shown in Fig. 24–6. A block of mass W resting on a horizontal, frictionless table is attached to a linear spring. A viscous damper (dashpot) is also attached to the block. Assume that the system is free to oscillate horizontally under the influence of a forcing function F(t). The origin of the coordinate system is taken as the right edge of the block when the spring is in the relaxed or unstretched condition. At time zero, the block is assumed to be at rest at this origin. In effect, this assumption makes the displacement variable Y(t) a deviation variable. Also, the assumption that the block is initially at rest permits derivation of the second-order transfer function in its standard form. An initial velocity has the same effect as a forcing function. Hence, this assumption is in no way restrictive. cou9789x_ch24_531-552.indd 538 8/22/08 1:19:02 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS 539 Spring F(t) K W C Dashpot 0 Y FIGURE 24–6 The damped oscillator. Positive directions for force and displacement are indicated by the arrows in Fig. 24–6. Consider the block at some instant when it is to the right of Y ⫽ 0 and when it is moving toward the right (positive direction). Under these conditions, the position Y and the velocity dY/dt are both positive. At this particular instant, the following forces are acting on the block: 1. 2. 3. The force exerted by the spring (toward the left) of ⫺KY where K is a positive constant, called Hooke’s constant The viscous friction force (acting to the left) of ⫺C dY/dt, where C is a positive constant called the damping coefficient The external force F(t) (acting toward the right) Newton’s law of motion, which states that the sum of all forces acting on the mass is equal to the rate of change of momentum (mass ⫻ acceleration), takes the form W d 2Y dY ⫹ F (t ) 2 ⫽ ⫺KY ⫺ C gc dt dt (24.9) W d 2Y dY ⫹C ⫹ KY ⫽ F ( t ) gc dt 2 dt (24.10) Rearrangement gives where W ⫽ mass of block, lbm gc ⫽ 32.2 lbm · ft/(lbf · s2) C ⫽ viscous damping coefficient, lbf/(ft/s) K ⫽ Hooke⬘s constant, lbf/ft F(t) ⫽ driving force, a function of time, lbf Dividing Eq. (24.10) by K gives W d 2Y C dY F (t ) ⫹Y ⫽ 2 ⫹ gc K dt K dt K cou9789x_ch24_531-552.indd 539 (24.11) 8/22/08 1:19:03 PM Confirming Pages 540 PART 7 NONLINEAR CONTROL For convenience, this is written as t2 d 2Y dY ⫹ 2zt ⫹ Y ⫽ X (t ) 2 dt dt where (24.12) t2 ⫽ W gc K (24.13) 2zt ⫽ C K (24.14) F (t ) K (24.15) X (t ) ⫽ Solving for t and z from Eqs. (24.13) and (24.14) gives W gc K t ⫽ z ⫽ gcC 2 4WK s (24.16) dimensionless (24.17) If the block is motionless (dY/dt ⫽ 0) and located at its rest position (Y ⫽ 0) before the forcing function is applied, the Laplace transform of Eq. (24.12) becomes t 2 s 2Y ( s ) ⫹ 2zt sY ( s ) ⫹ Y ( s ) ⫽ X ( s ) (24.18) Example 24.1. Consider the motion of the damped oscillator in Fig. 24–6. Write the matrix differential equations that characterize the system. The following constants apply, and Y(0) ⫽ 0 and Y ( 0 ) ⫽ 0 . W lbf ⫽1 gc ft / s2 lbf K ⫽5 ft C ⫽ 2 lbf ft / s F ( t ) ⫽ 10 lbf From Eqs. (24.15) through Eq. (24.17), we can calculate X (t ) ⫽ 10 ⫽ 2 ft 5 t ⫽ z ⫽ cou9789x_ch24_531-552.indd 540 1 s 5 22 ⫽ ( 4 ) ( 5) 1 5 8/22/08 1:19:03 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS 541 The governing differential equation Eq. (24.12) can now be written as 1 d 2Y 2 dY ⫹ ⫹ Y ⫽ X (t ) 5 dt 2 5 dt Defining the following phase variables Y ⫽ Y1 dy ⫽ Y1 ⫽ Y2 dt the differential equation becomes 2zt 1 ⫺1 Y2 ⫽ 2 Y1 ⫺ 2 Y2 ⫹ 2 X t t t Summarizing these equations in matrix form yields 1 0 Y1 ⫺1 ⫽ − 2z ⫺ Y2 2 t t 0 1 Y1 0 Y1 ⫹ 1 X ( t ) ⫽ 0 X ( t ) (24.19) ⫹ Y2 ⫺5 ⫺2 Y2 5 2 t Example 24.2. Use MATLAB to solve the matrix differential equations developed in Example 24.1 for the damped oscillator, and plot the phase plane. We can use the MATLAB differential equation-solving routine ODE23 to solve the system of first-order differential equations represented by Eq. (24.19). The following m-file represents the model. function ydot⫽ex24_2(t,y) %function file for damped oscillator %position is the first variable and velocity is the second A=[0,1; −5, −2]; B=[0;5]; x=2; ydot=A*y+B*x; The corresponding MATLAB commands to solve the model are >> >> >> >> [t,y] = ode45(@ex24_2,[0,5],[0,0]); plot(t,y) figure plot(y(:,1),y(:,2)) The solution is returned in the y matrix. The first column is the displacement Y1, and the second column is the velocity Y2. Thus, a plot of y versus t yields the typical time plots, while the plot of the first column, y(:,1), versus the second column, y(:,2), is the phase-plane plot, with time as a parameter along the curve. (See Fig. 24–7) cou9789x_ch24_531-552.indd 541 8/22/08 1:19:04 PM Confirming Pages 542 PART 7 NONLINEAR CONTROL 3 2.5 Displacement (Y1) 2 1.5 1 0.5 Velocity (Y2) 0 −0.5 −1 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Time (a) Velocity and displacement for the damped oscillator in Example 24.2. Phase Plane for Damped Vibrator 3 2.5 2 Velocity (Y2) 1.5 1 0.5 0 −0.5 −1 0 0.5 1 1.5 2 Displacement (Y1) (b) Phase plane for the damped oscillator in Example 24.2. 2.5 FIGURE 24–7 Plot of Solution to Example 24.2. cou9789x_ch24_531-552.indd 542 8/22/08 1:19:05 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS 543 Notice that the phase plane plot starts at (0,0) and spirals into the steady state (2,0) at t ⫽ ⬁. It’s also clear from the original differential equation [Eq. (24.18)] that Y ⫽ 2 at steady state. 24.4 MOTION OF A PENDULUM Consider the pendulum of Fig. 24–8. As the pendulum is moving in the direction shown, there are two forces acting to oppose its motion. R Pivot Motion BR d dt mg sin θ mg FIGURE 24–8 Forces acting on a pendulum. These forces, which act tangentially to the circle of motion, are (1) the gravitational force mg sin θ and (2) the friction in the pivot, which we suppose to be proportional to the tangential velocity of the mass BR(dq /dt), where B is the proportionality constant. We shall assume the air resistance to be negligible and the rod to be of negligible mass. Application of Newton’s second law gives ⫺mR d 2q dq ⫽ mg sin q ⫹ BR 2 dt dt Rearrangement leads to d 2q dq + D + w n2 sin q ⫽ 0 2 dt dt where D ⫽ B m w n2 ⫽ (24.20) g R This equation resembles the equation for the motion of the damped oscillator system. However, the presence of the term involving sin q makes the equation nonlinear. Equation (24.20) has the following form in phase coordinates: dq ⫽ q dt d q ⫽ ⫺w n2 sin q ⫺ Dq dt cou9789x_ch24_531-552.indd 543 (24.21) 8/22/08 1:19:05 PM Confirming Pages 544 PART 7 NONLINEAR CONTROL and a phase diagram would be a plot of angular velocity q versus position q. At this point, we can gain some insight by simple analysis of Eq. (24.21) without actually obtaining a solution. Referring for the moment to the spring-mass-damper system of Fig. 24–3, we saw that the system ceased to oscillate when the point dX / dt ⫽ dX / dt ⫽ 0 was reached. That is, all curves stopped at the origin of Fig. 24–5. Since neither X nor X is changing with time, the motion ceases. Further examination of Eqs. (24.3) shows that X ⫽ X ⫽ 0 is the only point at which both dX/dt and dX / dt are zero. Thus, we see that the mass will come to rest only when the situation of zero displacement (remember that in this case X is a deviation variable defined relative to the final steady-state displacement) and zero velocity is reached. Now we can perform a similar analysis on Eqs. (24.21). We are asking the following question: At what point or points in the phase plane ( q versus q diagram) do both dq /dt and d q /dt become zero? From the first of these equations, we see that this can happen only when q ⫽ 0. The reader should not be lulled into a false sense of security at this point. It would be wise to disregard the fact that dq /dt and q are, in fact, the same quantity; q should be thought of as a coordinate in the phase plane, and dq /dt as the rate of change with time of the other coordinate. The virtue of making this distinction will become clear in the next example, a chemical reactor. Thus, the analysis leads to the conclusion that the motion will cease when the pendulum comes to rest in either of the positions shown in Fig. 24–9. In addition, it is clear from Eqs. (24.21) that if the pendulum stops at any other point, the motion continues. Of course, this analysis agrees with our physical intuition. However, we expect to find a distinction between the stability characteristics of the two equilibrium points, since the position at p is likely to be hard to attain and maintain. This distinction will be explored in greater detail in Chap. 25. = Two possible equilibrium positions for pendulum Pivot =0 FIGURE 24–9 Equilibrium positions for pendulum. By using this result in the second equation, it can be seen that it is also necessary that sinq ⫽ 0 cou9789x_ch24_531-552.indd 544 (24.22) 8/22/08 1:19:05 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS 545 Equation (24.22) is satisfied at any of the points q ⫽ np where n is a positive or negative integer or zero. However, from a physical standpoint, we can really distinguish between only two of these points, which we take as q ⫽ 0 and q ⫽ p. Thus, the positions q ⫽ 0, 2p, 4p, ⫺2p, etc., all look the same; i.e., the pendulum is hanging straight down. Similarly, the points q ⫽ p, 3p, etc., all correspond to the pendulum standing straight up. Example 24.3. Use MATLAB to plot the displacement and velocity of a pendulum as a function of time as well as the phase-plane plot, using the following numerical values: R ⫽1m kg B ⫽ 2 s p q (0) ⫽ 4 m ⫽ 1 kg m g ⫽ 9.8 2 s dq ⫽ 0 dt Using these values, Eq. (24.21) becomes q ⫽ y1 dq dy1 ⫽ ⫽ y2 dt dt d 2 dy2 ⫽ ⫽ ⫺9 . 8 sin y1 ⫺ 2 y2 2 dt dt The MATLAB m-file for the model is function ydot = ex24_3(t,y) %function file for pendulum %position is the first variable and velocity is the second ydot(1,1) = y(2); ydot(2,1) = −9.8*sin(y(1)) −2*y(2); The MATLAB commands to execute the file and plot the necessary graphs are >>[t,y] = ode45(@ex24_3,[0,10],[pi/4,0]); >> plot(t,y) >> figure >> plot(y(:,1),y(:,2)) Note that the phase plane starts at the initial condition of q ⫽ p / 4, q ⫽ 0 and spirals into the steady state of q ⫽ 0, q ⫽ 0 as we would expect (see Fig. 24–10). cou9789x_ch24_531-552.indd 545 8/22/08 1:19:06 PM Confirming Pages 546 PART 7 NONLINEAR CONTROL 1 Displacement (rad) 0.5 Velocity (rad/s) 0 −0.5 −1 −1.5 −2 0 1 2 3 4 5 6 7 8 9 Time (s) (a) Displacement and velocity of the pendulum in Example 24.3. 10 1 0.5 Velocity (rad/s) 0 −0.5 −1 −1.5 −2 −0.4 −0.2 0 0.2 0.6 0.8 0.4 Displacement (rad) (b) Phase plane for the pendulum in Example 24.3. 1 FIGURE 24–10 Plot of the solution for Example 24.3. cou9789x_ch24_531-552.indd 546 8/22/08 1:19:07 PM Confirming Pages CHAPTER 24 24.5 EXAMPLES OF NONLINEAR SYSTEMS 547 A CHEMICAL REACTOR*1 Feed: reactant + diluent Reaction A B Product Cooling water FIGURE 24–11 Schematic of exothermic chemical reactor. Consider the stirred-tank chemical reactor of Fig. 24–11. The contents of the reactor are assumed to be perfectly mixed, and the reaction taking place is A → B (24.23) RA ⫽ kC Ae⫺E / RT (24.24) which occurs at a rate where RA ⫽ moles of A decomposing per hour per cubic foot of reacting mixture k ⫽ reaction velocity constant, h⫺1 CA ⫽ concentration of A in reacting mixture, mol/ft3 E ⫽ activation energy, a constant, Btu/mol R ⫽ universal gas law constant T ⫽ absolute temperature of reacting mixture The reaction is exothermic; ∆ H Btu of heat is generated for each mole of A that reacts. Hence, to control the reactor, cooling water is supplied to a cooling coil. The actual reactor temperature is compared with a set point, and the rate of cooling water flow is adjusted accordingly. To indicate this control mathematically, we write that Q(T) Btu/h of heat is removed through the cooling coil. In Chap. 25 we make a more detailed analysis of the dynamic behavior of the reactor. For the present preliminary analysis, it is not necessary to look carefully at Q(T), and hence it is merely assumed that as T rises, more heat is removed in the coil. Let x A0 ⫽ mole fraction of A in feed stream x B0 ⫽ moll e fraction of B in feed stream * This example is based on the work of R. Aris and N. R. Amundson (1958). cou9789x_ch24_531-552.indd 547 8/22/08 1:19:07 PM Confirming Pages 548 PART 7 NONLINEAR CONTROL Then 1 ⫺ x A0 ⫺ x B0 is the fraction of inerts in the feed stream. A mole balance on A Molar rate of Molar flow rate Molar flow rate of A in with feed ⫺ of A out with product ⫺ A reacting Molar rate of accumulation ⫽ of A in reactor takes the form Frx A0 ⫺ Frx A ⫺ krVe⫺E / RT x A ⫽ rV dx A dt (24.25) where F ⫽ feed rate, ft3/h xA ⫽ mole fraction of A in reactor r ⫽ molar density of reacting mixture, mol/ft3 V ⫽ volume of reacting mixture, ft3 To arrive at Eq. (24.25), we have used Eq. (24.24) and made the following assumptions: 1. 2. 3. 4. The density of the reacting mixture is constant, unaffected by the conversion of A to B. The feed and product rates F are equal and constant. Together, 1 and 2 imply that V, the volume of reacting mixture, is constant. Perfect mixing occurs, so that xA is the same in the reactor and product stream. A similar mole balance may be derived for substance B. However, Eq. (24.23) shows that 1 mol of B appears for every mole of A destroyed. Hence x B ⫺ x B0 ⫽ x A0 ⫺ x A (24.26) Equation (24.26) permits us to circumvent the mass balance for xB, since knowing xA, we can calculate xB directly. The energy balance on the reactor (Sensible heat in feed) ⫺ (Sensible heat in p roduct) ⫹ (Heat generated by reaction) ⫺ (Hea t removed in cooling coil ) = (Energy accumulating in reactor) can be written as FrC p ( T0 ⫺ T ) ⫹ krV ( ∆H ) e⫺E / RT x A ⫺ Q ( T ) ⫽ rVC p dT dt (24.27) where T0 ⫽ temperature of feed stream T ⫽ temperature in reactor (and product, since reactor is well mixed) Cp ⫽ specific heat of reacting mixture cou9789x_ch24_531-552.indd 548 8/22/08 1:19:08 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS 549 In writing Eq. (24.27), it is assumed that 1. 2. 3. The specific heat of the reacting mixture is constant, unaffected by the conversion of A to B. The perfect mixing means that the temperatures of the reacting mixture and product stream are the same. The heat of reaction ∆H is constant, independent of temperature and composition. We remark here that these assumptions, as well as those made in Eq. (24.25), may be relaxed without affecting the conceptual aspects of the phase analysis. They are made only to keep the example as uncluttered as possible, without being trivial. Equations (24.25) and (24.27) may be rearranged to the system dx A F ⫽ ( x A0 ⫺ x A ) ⫺ ke⫺E / RT x A dt V Q (T ) k ( ∆H ) ⫺E / RT dT F ⫽ ( T0 ⫺ T ) ⫹ e xA ⫺ Cp rVC p dt V (24.28) As a typical application of this system of equations, we might consider starting up the reactor, initially filled with a mixture at composition xA(0) and temperature T(0). Suppose the feed rate, feed composition, feed temperature, and flow rate of cooling water are held constant and the reactor is operated in this manner until steady state is reached. To describe the transient behavior of the chemical reactor, one can solve Eqs. (24.28) by integrating them numerically, using a typical stepwise procedure such as the Euler or Runge-Kutta method. In MATLAB, ODE45, a differential equation solver, could be used. This will result in functions xA(t) and T(t) for values of t from zero to some value (if one exists) at which, for practical purposes, xA(t) and T(t) cease to change with t. Alternatively, we may consider a phase-plane analysis of Eqs. (24.28) and seek solutions in the form of xA versus T curves. Note that division of the first of Eqs. (24.28) by the second gives dx A ( F / V ) ( x A0 ⫺ x A ) ⫺ ke⫺E / RT x A ⫽ dT ( F / V ) ( T0 ⫺ T ) ⫹ [ k ( ∆H H ) / C p ] e⫺E / RT x A ⫺ Q ( T ) / r VC p (24.29) The parameter t has been eliminated in Eq. (24.29), which is simply a differential equation relating xA and T. As we shall see in Chap. 25, this phase-plane analysis of the chemical reactor offers significant advantages over the ordinary analysis. In the chemical reactor, we no longer have the special relationship among the phase variables that we had in both previous cases. For both the spring and pendulum problems, we more or less artificially changed a second-order differential equation to two first-order equations by introducing the phase variable X (or q ). This phase variable was directly related to the other phase variable X (or q ) by the equation dX X ⫽ dt cou9789x_ch24_531-552.indd 549 8/22/08 1:19:08 PM Confirming Pages 550 PART 7 NONLINEAR CONTROL For the chemical reactor, there is no such simple relation between xA and T. We can study the steady-state solutions to Eqs. (24.28) without solving the equations, much as was done in the case of the damped pendulum of the previous example. As before, we note that steady state requires that xA and T simultaneously cease to change with time, dx A dT ⫽ ⫽ 0 dt dt From Eqs. (24.28), this implies that F ( x A0 ⫺ x As ) ⫺ ke⫺E / RTs x As ⫽ 0 V Q ( Ts ) F k ( ∆H ) ⫺E / RTs ( T0 ⫺ Ts ) ⫺ x As ⫺ ⫽ 0 e rVC p Cp V (24.30) where x As and Ts are the steady-state values of xA and T. The first of Eqs. (24.19) can be solved for x As , yielding x As ⫽ x A0 1 1 ⫹ ( kV / F ) e⫺E / RTs (24.31) Substitution of (24.31) into the second of Eqs. (24.30) yields k ( ∆H ) x A0 / C p e E / RTs ⫹ kV / F ⫽ Q ( Ts ) F ⫹ ( Ts ⫺ T0 ) rVC p V (24.32) Equation (24.32) is implicit in Ts, the steady-state temperature. In physical terms, it expresses an equality between the heat generated by the reaction and the heat removed in the cooling coil and product stream. To emphasize this, we have arranged it so that the left side is the heat generation and the right side is the heat removal. Solution of Eq. (24.32) for Ts requires numerical values for the various parameters. Without going into this much detail at present, we may obtain some qualitative information. To do this, we sketch the right and left sides of this equation as functions of Ts. A typical shape for the left side is given by the sigmoidal curve of Fig. 24–12. (See Aris and Amundsen, 1958, p.121.) The unusual curvature, of course, is caused by E / RTs the e term in the denominator. To plot the right side, we must know Q(T). While we have avoided specifying the form of Q(T), we know it increases with T. If there were no control action, i.e., if the flow rate of cooling water were maintained constant regardless of T, then Q(T) would increase almost linearly with T. This is so because at constant water rate, the heat transfer in the coil is approximately proportional to the difference between T and the mean temperature of the cooling water. This latter temperature would not vary as rapidly as T at practical flow rates. However, since we expect to have control action, we know that the cooling water flow rate will be increased with increasing T. Therefore, Q(T) may be expected to increase faster than linearly with T, which means that the right side of Eq. (24.32) increases faster than linearly. Several typical curves of this right side are shown in Fig. 24–12. A solution of Eq. (24.32) requires that the graphs of the right and left sides intersect. As shown in Fig. 24–12, there may be one, two, or three such intersections, depending on the relative locations of the heat generation (left side) and heat removal (right side). This means that there may be one, two, or three possible steady states for the reactor. cou9789x_ch24_531-552.indd 550 8/22/08 1:19:09 PM Confirming Pages CHAPTER 24 EXAMPLES OF NONLINEAR SYSTEMS 551 Different possible locations of heat removal Heat generated or removed, Btu/h (a) (b) (c) (d ) Heat generated Ts FIGURE 24–12 Steady-state generation and removal functions for exothermic chemical reactor As we shall see in Chap. 25, the steady state actually attained by the reactor depends on initial conditions xA(0) and T(0). The steady-state temperature Ts is then the temperature at the pertinent intersection, and the steady-state composition can be determined from Eq. (24.31). We shall also see that some of the steady states are unstable. In fact, the low-temperature steady state for curve (c) of Fig. 24.12, occurring as a point of tangency, is to be regarded with suspicion. Practically speaking, a perfect tangency would not occur. Minor variations in operating conditions (i.e., noise), which occur continually in actual process operation, may shift the curve (c) slightly to the left or right, resulting in two or zero low-temperature intersections, respectively. SUMMARY In this chapter, we have introduced the concept of a phase analysis and some of its basic elements. We have seen how physical situations give rise naturally to phase solutions. Furthermore, we have had our first look at true nonlinear behavior. In so doing, we have come to at least one interesting conclusion: a nonlinear motion or control system response may have more than one steady-state solution. This was true for the chemical reactor and for the pendulum. In contrast, the linear motions and control system responses we studied in previous chapters had only one steady-state solution. In Chap. 25, we shall discover still more differences that render nonlinear analysis more difficult than linear analysis. PROBLEMS 24.1 Use MATLAB to analyze the damped oscillator shown in Fig. 24–6, using the following numerical values. Plot the displacement and velocity as a function of time as well as the phase plane for the sytem. W lbf lbf lbf ⫽1 C ⫽ 2 K ⫽ 2 2 / gc ft s ft ft / s (a) F(t) ⫽ 6 lbf, Y(0) ⫽ 0, and Y ( 0 ) ⫽ 0 (b) F(t) ⫽ 6 lbf, Y(0) ⫽ 2, and Y ( 0 ) ⫽ 0 (c) F(t) ⫽ 6 lbf, Y(0) ⫽ 0, and Y ( 0 ) ⫽ 2 24.2. Plot the phase planes for Prob. 24.1, parts (a) to (c), on the same set of axes. cou9789x_ch24_531-552.indd 551 8/22/08 1:19:10 PM Confirming Pages CHAPTER 24 CAPSULE SUMMARY In this chapter, we have introduced the concept of a phase analysis and some of its basic elements. We have seen how physical situations give rise naturally to phase solutions. Furthermore, we have had our first look at true nonlinear behavior. In so doing, we have come to at least one interesting conclusion: a nonlinear motion or control system response may have more than one steady-state solution. This was true for the chemical reactor and for the pendulum. In contrast, the linear motions and control system responses we studied in previous chapters had only one steady-state solution. In Chap. 25, we shall discover still more differences that render nonlinear analysis more difficult than linear analysis. X t Typical displacement versus time plot . X t Typical velocity versus time plot t . X t=0 −1 X t=∞ at origin Phase-plane plot with time as a parameter along the length of the curve 552 cou9789x_ch24_531-552.indd 552 8/22/08 1:19:10 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS T he advantages of the phase analysis introduced in Chap. 24 can be more fully appreciated after some acquaintance with the tools available for such analysis. To give a detailed exposition of all, or even most, of the aspects of this subject is not intended. Instead, this chapter strives to indicate its flavor and to stimulate further study. 25.1 PHASE SPACE In Chap. 24, we considered three examples for which the dynamic response can be described by two state variables. For the cases of the damped oscillator and the pendulum, the state variables were phase variables in which the dependent variable and its derivative (X, X or q , q ) were chosen as the state variables. For the exothermic chemical reactor, the state variables selected were temperature and composition (T, xA); these variables, which arose naturally in the analysis of the chemical reactor, were called physical variables in Chap. 21. In general, an nth-order dynamic system can be described by n state variables. The state variables (x1, x2, . . ., xn) can be located in a coordinate system called phase space. Each value of t, say t1, defines a point in this space: x1(t1), x2(t1), . . . , xn(t1). The solution curve is a locus of these points for all values of t. It is called a trajectory and connects successive states of the system. For the damped oscillator presented in Chap. 24, the coordinate system was a plane with an axis for each state variable; we shall refer to this coordinate system as a phase plane. Figure 24–5 is a typical phaseplane representation of a dynamic system. When the physical system is third-order, the coordinate system consists of three axes, one for each state variable. Of course, systems of fourth or higher order require treatment in space that is of too many dimensions to be visualized. The graphic aspects of phase-space representation are advantageous primarily in the case of two dimensions (the phase plane) and to a limited extent for three 553 cou9789x_ch25_553-578.indd 553 8/22/08 3:41:17 PM Confirming Pages 554 PART 7 NONLINEAR CONTROL dimensions. The bulk of practical use of phase-space analysis has been made in the two-dimensional autonomous (time-invariant) case: dx1 ⫽ f1 ( x1 , x2 ) dt dx2 ⫽ f2 ( x1 , x2 ) dt (25.1) For this reason, we largely confine our attention in the remainder of this study to systems that may be written in the form of Eqs. (25.1). As we have seen, there is no loss in conceptual generality, but we cannot expect the graphical aspects of the material we shall develop to generalize to higher-dimension phase space. The solution of the system (25.1) may be presented as a family of trajectories in the x2x1 plane. If we are given the initial conditions x1 ( t0 ) ⫽ x10 x2 ( t0 ) ⫽ x20 the initial state of the system is the point (x10, x20) in the x2x1 plane, and the trajectory may be traced from this point. By dividing the second of Eqs. (25.1) by the first, we obtain dx2 f2 ( x1 , x2 ) ⫽ dx1 f1 ( x1 , x2 ) (25.2) Now dx2/dx1 is merely the slope of a trajectory, since a trajectory is a plot of x2 versus x1 for the system. Hence, at each point in the phase plane (x1, x2), Eq. (25.2) yields a unique value for the slope of a trajectory through the point, namely, f2 (x1, x2) / f1(x1, x2). This last statement should be amended to exclude any point (x1, x2) at which f1(x1, x2) and f2 (x1, x2) are both zero. These important points are called critical points and will be examined in greater detail below. Since the slope of the trajectory at a point, say (x1, x2), is unique by Eq. (25.2), it is clear that trajectories cannot intersect except at a critical point, where the slope is indeterminate. Using Simulink/MATLAB to Plot a Phase Plane Let us illustrate the use of Simulink/MATLAB in phase-plane analysis with an example. Example 25.1. Find the trajectory of the system dx1 ⫽ x2 dt (25.3) dx2 ⫽ ⫺5 x1 ⫺ 2 x2 dt which passes through the point x1 ⫽ 1 x2 ⫽ 0 The Simulink diagram for simulating this system is shown in Fig. 25–1a. cou9789x_ch25_553-578.indd 554 8/22/08 3:41:18 PM Confirming Pages CHAPTER 25 x2 = dx1/dt 1 s Integrator ++ x1 −5 Gain EXAMPLES OF PHASE-PLANE ANALYSIS 555 ++ −5x1 1 x1 x1(0) To Workspace1 XY Graph −2x2 x2 To Workspace dx2/dt 1 s Integrator1 ++ x2 = dx1/dt −2 Gain1 0 x2(0) (a) Simulink diagram for Example 25.1. 1 (x1(0) = 1, x2(0) = 1) 0.5 (x10 = 0.6, x20 = 0) (x10 = 1, x20 = 0) (x1(0) = −0.6, x2(0) = 0) x2 0 −0.5 −1 −1.5 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 x1 (b) Phase plane for Example 25.1 from Simulink. FIGURE 25–1 Simulink solution for Example 25.1. cou9789x_ch25_553-578.indd 555 8/22/08 3:41:19 PM Confirming Pages 556 PART 7 NONLINEAR CONTROL Another trajectory, starting from the point (⫺0.6, 0), is shown on Fig. 25–1b. This serves to emphasize that interpolation is possible on the phase plane, and many trajectories representing various initial conditions are easily visualized or sketched. Analysis of Critical Points In the situations of most interest to us, Eq. (25.2) will represent the behavior of a (nonlinear) control system, as in Eq. (24.29). Therefore, we shall be interested in maintaining the system at or near a steady state. Since, from Eq. (25.1), a steady-state point is defined by f1 ( x1 , x2 ) ⫽ f2 ( x1 , x2 ) ⫽ 0 it is clear that the steady states are critical points. At the critical points, the slope of the trajectory is undefined; hence, many trajectories may intersect at these points. In Fig. 25–1 the origin is a critical point. It can be seen from the figure that, in this case, all trajectories spiral into the origin. Hence, this particular system is such that, no matter what the initial state (i.e., for any disturbance which is applied), the system returns to steady state at the critical point. The critical point of Fig. 25–1 is called a focus, because the trajectories spiral into it. This spiral motion of the trajectories corresponds to the oscillatory approach of the system to steady state. The oscillatory motion occurs because the system of Eqs. (25.3) is underdamped, as indicated by the characteristic equation sI ⫺ A ⫽ 0 or 1 s ⫺1 s 0 0 2 0 s ⫺ ⫺5 ⫺2 ⫽ 5 s ⫹ 2 ⫽ s ⫹ 2s ⫹ 5 ⫽ 0 ⫽ sI 1 2 5s ⫹ 25 s ⫹ 1 A When put into standard form, this characteristic equation has parameters t2 ⫽ t ⫽ 1 5 1 5 2zt ⫽ z ⫽ 2 5 1 5 Since z < 1, the system is underdamped. An overdamped system, such as that generated by the system dx1 dx2 ⫽ x2 ⫽ ⫺5 x1 ⫺ 6 x2 dt dt having characteristic equation s 2 ⫹ 6s ⫹ 5 ⫽ 0 so that t ⫽ cou9789x_ch25_553-578.indd 556 1 5 z ⫽ 3 5 8/22/08 3:41:20 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 557 TABLE 25.1 Classification of critical points Type of critical point Characteristic equation Pertinent values of z Nature of roots Sign of roots Stable node t2 s2 ⫹ 2zts ⫹ 1 ⫽ 0 z>1 Real Both ⫺ Stable focus t2 s2 ⫹ 2zts ⫹ 1 ⫽ 0 0<z<1 Complex Real parts both ⫺ Unstable focus t2s2 ⫹ 2zts ⫹ 1 ⫽ 0 ⫺1 < z < 0 Complex Real parts both ⫹ Unstable node t2s2 ⫹ 2zts ⫹ 1 ⫽ 0 z < ⫺1 Real Both ⫹ All Real One ⫹, one ⫺ Saddle point 2 2 t s ⫹ 2zts ⫺ 1 ⫽ 0 has a critical point such as that of Fig. 25–2a. Here the trajectories enter the critical point directly, without oscillation. This type of critical point is called a node. For comparison, a typical focus is sketched in Fig. 25–2b. In fact, other types of behavior may be exhibited by critical points of a second-order system, depending on the nature of the roots of the characteristic equation. These are summarized for linear systems in Table 25.1 and sketched in Fig. 25–2. The distinction between stable and unstable nodes or foci is (a) (b) (d) (c) (e) cou9789x_ch25_553-578.indd 557 FIGURE 25–2 Second-order critical points: (a) stable node, (b) stable focus, (c) unstable focus, (d) unstable node, (e) saddle point. 8/22/08 3:41:21 PM Confirming Pages 558 PART 7 NONLINEAR CONTROL made to indicate that the trajectories move toward the stable type of critical point and away from the unstable point. The saddle point arises when the roots of the characteristic equation are real and have opposite sign. In this case there are only two trajectories that enter the critical point, and after entering, the trajectories may leave the critical point (permanently) on either of two other trajectories. No other trajectory can enter the critical point, although some approach it very closely. This categorization of critical points according to the particular linear system is often of value in the analysis of nonlinear systems. The reason for this is that, in a sufficiently small vicinity of a critical point, a nonlinear system behaves approximately linearly. Thus, the system of Eq. (24.21) for the pendulum is nonlinear. It has two physically distinguishable steady states, corresponding to the pendulum pointing up or down. The nonlinear term sin q may be linearized around each steady state. Near the steady state at q ⫽ 0, sinq 艐 sinq q ⫽0 ⫹ d ( sinq ) (q ⫺ 0 ) ⫹ higher-order term d q q ⫽0 sinq 艐 0 ⫹ cosq q ⫽0 (q ⫺ 0 ) ⫹ higher-order te rms 1 sinq 艐 θ and near the steady state at q ⫽ p, a Taylor series yields sinq ⫽ ⫺(q ⫺ p ) Therefore, near q ⫽ 0, Eqs. (24.21) are closely approximated by the linear equations dq ⫽ q dt (25.4) dx ⫽ x dt (25.5) d q ⫽ ⫺w n2q ⫺ Dq dt and near q ⫽ p, by dx ⫽ w n2 x ⫺ Dx dt where x ⫽ q ⫺ p. These linearized versions of Eqs. (24.10) can be easily solved to determine the nature of the linear approximations to the critical points. Thus, the characteristic equation for Eqs. (25.4) is s 2 ⫹ Ds ⫹ w n2 ⫽ 0 (25.6) s 2 ⫹ Ds ⫺ w n2 ⫽ 0 (25.7) while that for Eqs. (25.5) is cou9789x_ch25_553-578.indd 558 8/22/08 3:41:21 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 559 As shown in Table 25.1, Eq. (25.6) yields a stable critical point, which may be a node or focus depending on the degree of damping. (Note that as the damping is increased, the behavior changes from focus to node, or from oscillatory to nonoscillatory.) On the other hand, Eq. (25.7) indicates a saddle point for the motion near q ⫽ p. These conclusions apply strictly only to the linearized phase equations, Eqs. (25.4) and (25.5). To compare them with the behavior of the true system of Eqs. (24.10), the actual phase diagram is sketched for a lightly damped case in Fig. 25–3. For simplicity, this diagram is extended beyond the range 0 ⱕ q ⱕ 2p even though this is the only region of physical significance. Actually, the section for 0 ⱕ q ⱕ 2p should be cut out and rolled into a cylinder so that the lines corresponding to q ⫽ 0 and q ⫽ 2p coincide. This phase cylinder would more realistically represent the motion of the pendulum. As seen from Fig. 25–3, the point at q ⫽ p is, indeed, a saddle point and the point q ⫽ 0 (or 2p) is a stable focus. If the system were more heavily damped, this latter point would be a stable node. =0 =2 =− = FIGURE 25–3 Phase portrait of lightly damped pendulum. A greater understanding of the saddle point may now be obtained by analyzing the q ⫽ p point in terms of what we know to be the physical behavior of the pendulum at this point. That is, the point may be approached from either of two directions. When the pendulum is at the point, an infinitesimal disturbance will cause it to fall in either of two directions. Other trajectories narrowly miss this point, indicating that just the right initial velocity must be imparted to the pendulum at a given initial point to cause it to stop in the q ⫽ p position. In summary, it can be concluded that in this case the linearized equations give valuable, accurate information about the behavior of the nonlinear system in the vicinity of the critical points. Because the linearized equations are more easily solved, it is always desirable to be able to relate the behavior of the actual system to the behavior of the linearized solutions in the vicinity of the operating point. In fact, in our previous work on control systems, we have assumed for nonlinear systems that design of a stable control system based on the linearized equations was adequate to ensure stable operation of the actual system. The basis for this assumption is given by the following theorem of Liapunov (see Letov, 1961). Let the nonlinear equations of a motion be linearized by expansion in deviation variables around a particular critical point. If the linearized solution for the deviation variables is stable, the actual motion will be stable in some vicinity of cou9789x_ch25_553-578.indd 559 8/22/08 3:41:22 PM Confirming Pages 560 PART 7 NONLINEAR CONTROL the critical point. If the linearized solution is neutrally stable (i.e., its characteristic equation has roots on the imaginary axis), no statement can be made about the actual motion. If the linearized solution is unstable, then the actual motion will be unstable. It is necessary to define what is meant by stability and instability of the actual nonlinear motion in the vicinity of the critical point. Although stability in nonlinear systems is a complex subject, for our purposes it will suffice to state that a stable nonlinear motion in the vicinity of a critical point is one for which all phase-plane trajectories in this vicinity travel toward and end at the critical point. An unstable motion is one for which trajectories move away from the critical point. This would mean that while theoretically the state of the system may remain at the critical point indefinitely, any slight disturbance causes the unstable system to move away from the critical point. These conclusions agree with our physical understanding of the pendulum motion, since the steady condition at q ⫽ p is easily destroyed. It is because of Liapunov’s theorem that linear control theory is so successful in control system design. One really hopes to control the system so that it remains permanently in the vicinity of a particular point (i.e., a steady state). However, when serious upsets occur in an automatically controlled plant, moving it far from steady state, it is often necessary to return the plant to manual control until conditions are again close to steady state. This is so because the controllers are designed for satisfactory operation in the linear range only. One of the great drawbacks of linear control theory is the fact that stability of the linearized equations guarantees stability of the nonlinear system only in some vicinity of the particular critical point. No information about the size of this vicinity or about the behavior outside this vicinity is obtained. If the linear vicinity is extremely small, then unknown to the designer who has used linear methods, almost any plant disturbance of practical size may result in control system failure. An example of this behavior will be given later. Limit Cycles The first major difference between linear and nonlinear motions is the possible existence of more than one critical point in the latter type. The second is the possible existence of limit cycles. A limit cycle is defined as a periodic oscillation whose amplitude and frequency depend only on the properties of the system and not on the initial state of the system (provided the initial state lies in a certain nontrivial region of the phase space). In the phase plane, stable limit cycles are recognized as closed curves which are approached asymptotically by all nearby trajectories. Unstable limit cycles are closed curves from which all nearby trajectories diverge. An example of a stable limit cycle is the “steadystate” behavior of a home heating system when controlled by a thermostat. A periodic oscillation in house temperature is always reached, and the amplitude and frequency of the oscillation are independent of the temperature that existed in the house at the time that the furnace was started. Unstable limit cycles can never be realized physically for any system by definition. However, as will be seen later, they divide the phase plane into regions of totally different dynamic behavior and hence are of considerable importance. cou9789x_ch25_553-578.indd 560 8/22/08 3:41:23 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 561 It is important to distinguish between limit cycles and other closed curves that may occur. The linear system t2 d2x dt 2 ⫹x ⫽ 0 has phase-space solution x 2 ⫹ t 2 ( x ) ⫽ C 2 2 (25.8) where x ⫽ dx /dt and the constant C depends on initial conditions. Equation (25.8) defines a family of concentric ellipses in the phase plane. However, these are not limit cycles, because the closed curve which is followed by the system depends on the initial state of the system through the constant C. In the next section, we study some limit cycles occurring in typical control systems. OTHER ASPECTS. We have presented only those aspects of phase-plane analysis that will be of use in the examples to follow. This can be considered only as a brief introduction to the subject, and the interested reader is referred to the references already cited for more information. Among the important subjects that have been omitted are graphical methods for determination of time along a trajectory, various aspects of phase-plane topology, and the mathematical aspects of stability. 25.2 EXAMPLES OF PHASE-PLANE ANALYSIS In this section, we consider two different examples of the use of the phase plane to analyze nonlinear control systems. The first is a simple on/off control system for a stirred-tank heater. The second is the chemical reactor of Chap. 24. In both cases, the systems are second-order and autonomous, so that they are ideal situations for use of the phase plane. On/Off Control of Stirred-tank Heater The use of on/off control offers significant economic advantages over proportional control or other more sophisticated modes of control. The control mechanism is simply a relay that turns on or off depending on the value of the measured variable. The disadvantage is usually that the quality of control is inferior to that realized with proportional control. Consider the stirred-tank heater of Fig. 25–4. Water is being heated to a controlled temperature by mixing with steam. It is assumed for the analysis that the cold water input rate is constant. Heated water overflows into an outlet pipe at the top of the tank, so that no accumulation of mass occurs in the tank. Most of the steam is added, at a fixed flow rate, from the main steam supply. However, this amount of steam is set at a value somewhat less than the amount required to heat the cold water to the desired temperature. An additional amount of steam may be added whenever the solenoid valve is opened. When this additional steam is admitted, the sum of the two steam inputs is enough to heat the water to a temperature somewhat in excess of the desired cou9789x_ch25_553-578.indd 561 8/22/08 3:41:23 PM Confirming Pages 562 PART 7 NONLINEAR CONTROL Main steam supply Temperature sensing Set point Relay Heated water Voltage source Solenoid Control steam Cold water FIGURE 25–4 On/off control of a stirred-tank heater. temperature. A temperature-measuring device such as a thermocouple or RTD transmits the tank temperature to the relay. When this temperature is below the set point, the relay closes, which opens the solenoid valve, thus admitting more steam. Eventually, the additional steam will result in the temperature exceeding the set point, the relay will open, the valve will close, cutting off the additional steam, and the temperature will fall again. It is apparent that an oscillating control will be achieved. In fact, from the discussion in the previous section, we recognize that a limit cycle will occur. We consider now a numerical example of this type of control system. Water at 40⬚ F, at a rate of 100 lb/min, is to be heated to 150⬚ F. The main steam supply is to be set so that it will heat this much water to 125⬚ F, while additional steam, through the controlled solenoid valve, is available to heat the water another 50⬚ F. This means that the steady-state temperatures with the solenoid closed and open, 125 to 175⬚ F, are equally spaced about the set point. Heat losses to the surroundings are negligible. The volume of the tank is 1.6 ft3. The relay control system has a thermocouple for measurement of temperature. This measuring system has a time constant of 30 s. The solenoid valve is very rapid in response. We first analyze this system considering the relay to behave ideally. This means that it opens precisely at the instant the temperature exceeds the set point and closes similarly. Later, we shall correct this to conform more closely to the behavior of actual relays. If the tank is perfectly stirred, it is a first-order system with a time constant of t ⫽ (62) (1.6) ⫽ 1.0 min rV ⫽ w 100 and its transfer function relating changes in the steam input rate to temperature is Gp (s) ⫽ cou9789x_ch25_553-578.indd 562 10 s ⫹1 8/22/08 3:41:24 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 563 where 10⬚F/lb/min of steam flow is the change in steady-state temperature per unit change in steady-state steam flow. The necessary main and controlled steam rates are (using 1000 Btu/lb for latent heat) Qmain ⫽ (125 ⫺ 40 ) (100 ) ⫽ 8 . 5 lb/min 1000 (175 ⫺ 125) (100 ) ⫽ 5.0 lb/min Qcontrolled ⫽ 1000 The amount of steam that would be necessary to maintain the water at a steady-state temperature of 150⬚F is (150 ⫺ 40 ) (100 ) Qs ⫽ ⫽ 11 . 0 lb/min 1000 Hence, in terms of deviation variables, the controller output may be taken as ⫾ 2.5 lb/min of steam. A block diagram may now be constructed for this system, as shown in Fig. 25–5. This diagram uses deviations from 150⬚F as temperature variables, so the set point is taken as zero. The action of the relay is symbolized by the input-output relations, indicating that ⫹2.5 lb/min of steam is admitted when the error is positive and ⫺2.5 lb/min when the error is negative, again in deviation variables. 2.5 0 + − M −2.5 C 10 s+1 B FIGURE 25–5 Block diagram for system of Fig. 25–4. 1 0.5 s + 1 It is convenient to use a dimensionless version of Fig. 25–5. This is provided in Fig. 25–6, where the changes M⬘ ⫽ M 2.5 e⬘ ⫽ e 25 C⬘ ⫽ C 25 B⬘ ⫽ B 25 have been made. 0 + − '= /25 −1 B' = B/25 1 M' = M/2.5 1 s+1 C' = C/25 1 0.5s + 1 FIGURE 25–6 Dimensionless block diagram for system of Fig. 25–4. cou9789x_ch25_553-578.indd 563 8/22/08 3:41:26 PM Confirming Pages 564 PART 7 NONLINEAR CONTROL The usual methods of linear control theory are not applicable to the block diagram of Fig. 25.6. The relay does not obey the principle of superposition in its input-output relation. It is necessary to revert to the differential equations describing the control loop. These are M⬘ ⫽ C⬘ ⫽ dC⬘ ⫹ C⬘ dt (25.9) 1 dB⬘ ⫹ B⬘ 2 dt (25.10) e⬘ ⫽ ⫺B⬘ (25.11) In addition we have 1 M⬘ ⫽ ⫺1 e⬘ ⬎ 0 e⬘ ⬍ 0 (25.12) Combination of Eqs. (25.9) to (25.12) yields ⫺1 1 d 2e ⬘ 3 de ⬘ ⫹ ⫹ e⬘ ⫽ 2 2 dt 2 dt 1 e⬘ ⬎ 0 e⬘ ⬍ 0 (25.13) Equation (25.13) can be rewritten in phase notation as de ⬘ ⫽ e⬘ dt ⫺( 3e⬘ ⫹ 2e ⬘ ⫹ 2 ) d e⬘ ⫽ dt ⫺( 3e⬘ ⫹ 2e ⬘ ⫺ 2 ) e⬘ ⬎ 0 e⬘ ⬍ 0 (25.14) Figure 25–7 is a plot of the phase plane for this system. Note that, in this figure, the e⬘ scale has been expanded by a factor of 10 to magnify the behavior near the origin. It can be seen from Fig. 25–7 that the trajectory approaches the origin. The final state is a limit cycle of zero amplitude and infinite frequency about the origin. In other words, the relay alternately opens and closes at very high frequency, a condition known as chattering. Physically, this condition will never be realized because the dynamics of the solenoid valve and the relay itself would become important. Instead, the final condition will be a limit cycle of high, rather than infinite, frequency and low, rather than zero, amplitude. However, the basic idealization which has led us to this suspect conclusion is in the behavior of the relay. True relays have input-output characteristics more similar to those shown in Fig. 25–8. There is a dead band around the set point, of width 2e⬘0, over which the relay is insensitive to changes in the error signal. Anyone who has made fine adjustments in the setting of a home thermostat has observed this behavior. cou9789x_ch25_553-578.indd 564 8/22/08 3:41:27 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 565 0.6 Solenoid valve closed Solenoid valve open 0.4 0.2 −0.04 −0.02 0.02 0.04 0.06 −0.2 −0.4 −0.6 FIGURE 25–7 Phase-plane trajectory for on-off control system of Fig. 25–4. Output 1 − '0 '0 0 −1 FIGURE 25–8 Characteristics of true relay with dead zone. Consider as an example the case for ε⬘0 ⫽ 0.01. The effect of this dead zone is to change the dividing line between on and off to that shown in Fig. 25–9. The new dividing line has the equation 0 . 01 e⬘ ⫽ ⫺0 . 01 cou9789x_ch25_553-578.indd 565 e⬘ ⬎ 0 e⬘ ⬍ 0 8/22/08 3:41:27 PM Confirming Pages 566 PART 7 NONLINEAR CONTROL 0.6 0.4 0.2 − −0.04 −0.02 o o 0 0.02 0.04 0.06 cl e −0.06 Li m it c y −0.4 −0.6 FIGURE 25–9 Phase plane for system of Fig. 25–4 using relay with characteristics of Fig. 25–8. Now, as shown in Fig. 25–9, all trajectories approach a limit cycle, for which the error amplitude is approximately 0.03. The frequency is finite and is obtained by computing the time around the limit cycle. Although we have not presented here the graphical methods for determining this time, it can always be calculated by noting from the first of Eqs. (25.14) that t ⫽ ∫ dt ⫽ ∫ de ⬘ e⬘ (25.15) Thus, time around the limit cycle can be computed by graphical evaluation of the inte goes to zero. To gral in Eq. (25.15). The only difficulty is near the e⬘ axis, where e⬘ circumvent this, we may use the second of Eqs. (25.14) t ⫽ ⫺∫ d e⬘ 3e⬘ ⫹ 2e ⬘ ⫾ 2 over a small segment of the trajectory as it crosses the e⬘ axis. The result of this graphical calculation is w ⫽ 9.2 rad/min. The frequency thus computed for the error signal is, for obvious physical reasons, the same as the frequency of the controlled signal C⬘. However, the amplitude of C⬘, which is of more direct interest, is not the same as the amplitude of e⬘. It may be found in this case by noting from Eqs. (25.10) and (25.11) that C⬘ ⬇ ⫺ 12 e⬘ ⫺ e ⬘ cou9789x_ch25_553-578.indd 566 8/22/08 3:41:28 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 567 It is therefore clear from Fig. 25–9 that C⬘ attains a maximum value near the switching points where C⬘ ⬇ ⫾0 . 17 Reverting to the original variables, it follows that the water temperature will oscillate with an amplitude of ( 0 . 17 ) ( 25) ⫽ 4 . 25 ⬚F The effect of a small dead zone, 2e0 ⫽ 2(0.01)(25) ⫽ 0.5⬚F, is thus quite significant. In practice, the width of this dead zone is usually an adjustable design parameter. This width is always chosen as a compromise. The wider it is made, the lower will be the limit-cycle frequency, thus saving excessive switching or chatter. However, the limit-cycle amplitude increases with dead-zone width, decreasing the quality of control. Example 25.2. Use of Simulink to generate a phase-plane diagram for a relay with dead zone. Use MATLAB and Simulink to simulate the control system shown in Fig. 25–6 containing a relay with characteristics shown in Fig. 25–8. The Simulink block diagram for the system is shown in Fig. 25–10. Error Error to workspace du/dt Derivative of error Phase-plane trajectory de/dt Derivative to workspace +− Step Relay 1 s+1 Process Relay output Process output 1 0.5 s + 1 Measurement FIGURE 25–10 Simulink block diagram for process in Fig. 25–6. The phase-plane trajectory for Example 25.2 is shown in Fig. 25–11. We can also determine the cycle time by looking at the relay output as shown in Fig. 25–12. cou9789x_ch25_553-578.indd 567 8/22/08 3:41:29 PM Confirming Pages 568 PART 7 NONLINEAR CONTROL 0.4 0.3 0.2 d /dt 0.1 0 −0.1 −0.2 −0.3 −0.4 −0.03 Limit Cycle SOLENOID VALVE CLOSED −0.02 SOLENOID VALVE OPEN −0.01 0 0.01 0.02 0.03 FIGURE 25–11 Phase plane generated using Simulink for Example 25.2. Expanding FIGURE 25–12 Relay ouput from Simulink model. cou9789x_ch25_553-578.indd 568 8/22/08 3:41:29 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 569 The relay cycle time from the graph is 0.7 min/cycle. The frequency is rad 2p rad 1 cycle w ⫽ ⫽ 9.0 min cycle 0 . 7 min This is the same result that we found earlier. We can also determine the maximum and minimum values of C⬘ from the Simulink process output, as shown in Fig. 25–3. Expanding FIGURE 25–13 Minimum and maximum values of C⬘ from Simulink. From the output, we can see that C⬘ 艐 ⫾ 0.173, which is also the same as our earlier result. Simulink provides us with an easy way to study the behavior of these types of systems. Exothermic Chemical Reactor We now wish to consider the phase-plane behavior of the chemical reactor of Chap. 24. This study is based on the paper by Aris and Amundson (1958). For convenience, the dynamic equations are reproduced here: dx A F ⫽ ( x A0 ⫺ x A ) ⫺ ke⫺E /RT x A dt V k ( ∆H ) e⫺E /RT Q (T ) dT F ⫽ ( T0 ⫺ T ) ⫹ xA ⫺ dt V Cp rVC p (24.28) Defining the dimensionless variables t ⫽ cou9789x_ch25_553-578.indd 569 Ft V y ⫽ xA x A0 q ⫽ CP T x A0 ( ∆H ) q0 ⫽ CP T0 x A0 ( ∆H ) 8/22/08 3:41:30 PM Confirming Pages 570 PART 7 NONLINEAR CONTROL these equations become dy ⫽ 1 ⫺ y ⫺ r ( y, q ) dt (25.16) dq ⫽ q 0 ⫺ q ⫹ r ( y,q ) ⫺ q (q ) dt where kVy [⫺ECP /Rx A0 ( ∆H )θ] e F Q (T ) q (θ) ⫽ Frx A0 ( ∆H ) r ( y,θ ) ⫽ As a control heat removal function q (q), Aris and Amundson chose the form q (q ) ⫽ U (q ⫺ q c ) 1 ⫹ K c (q ⫺ q s ) (25.17) where qc is the dimensionless mean temperature of water in the cooling coil. This indicates that the heat removal is always proportional to the difference between the reactor temperature and mean cooling water temperature. In addition, the term in brackets indicates that proportional control on the cooling water flow rate is present. The flow rate is increased by an amount proportional to the difference between the actual reactor temperature q and the desired steady-state temperature qs. This increase in cooling water flow rate is assumed for convenience to cause an approximately proportional increase in heat removal. The constant U is a dimensionless analog of U0A, the overall heat-transfer rate. As a specific numerical example, Aris and Amundson selected the following values for constants: kV ⫽ e 25 F EC p ⫽ 50 Rx A0 ( ∆H ) qs ⫽ 2 q 0 ⫽ q c ⫽ 1 . 75 U ⫽1 Under these conditions, Eqs. (25.16) become dy 50 1/ 2 ⫺1/q ) ⫽ 1 ⫺ y ⫺ ye ( dt dq 50 1/ 2 ⫺1/q ) ⫺ (q ⫺ 1 . 75) 1 ⫹ K c (q ⫺ 2 ) ⫽ 1 . 75 ⫺ q ⫹ ye ( dt It can be seen that there is a critical point of Eqs. (25.18) at y ⫽ 1 2 ⫽ ys (25.18) q ⫽ 2 ⫽ qs and this is the location at which control is desired. This point has the correct steadystate temperature and a 50 percent conversion of reactant. In addition, there may be two cou9789x_ch25_553-578.indd 570 8/22/08 3:41:31 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 571 more critical points of Eq. (25.18) depending on the proportional control constant Kc, as will be discussed below. Since we are primarily interested in control about qs, we make use of Liapunov’s theorem on local stability, presented earlier. Linearizing Eq. (25.18) in deviation variables q ⫺ qs and y ⫺ ys by using Taylor’s series yields d ( y ⫺ ys ) dt d (q ⫺ q s ) ⫽ ⫺2 ( y ⫺ ys ) ⫺ 6 . 25 (q ⫺ q s ) (25.19) K ⫽ y ⫺ ys ⫹ 4 . 25 ⫺ c (q ⫺ q s ) 4 dt 1 where ys ⫽ 2 . As we have seen before, the solution to this linear system is y ⫺ ys ⫽ c1e s1 t ⫹ c2e s2 t q ⫺ q s ⫽ c3e s1 t ⫹ c4 e s2 t where, in this case, s1 and s2 are the roots of [see Eq. (24.6) and the steps following this equation] Kc ⫺ 9 2Kc ⫺ 9 (25.20) s2 ⫹ s⫹ ⫽ 0 4 4 According to the Routh criteria, all coefficients in this characteristic equation must be positive in order that the real parts of the roots s1 and s2 be negative. Hence, we can see immediately from Eq. (25.20) that to achieve a stable node or focus, it is necessary that Kc > 9. Kc = 9+ y Unstable limit cycle 0.5 Stable limit cycle 2 FIGURE 25–14 Stable and unstable limit cycles in an exothermic chemical reactor. However, Aris and Amundson obtained the phase plane for (among other values) a value of Kc slightly greater than 9. This was accomplished by numerical solution of Eqs. (25.18). It was found that in the vicinity of the steady-state point, the situation is as depicted in Fig. 25–14. There are two limit cycles surrounding the stable focus critical point. The inner limit cycle is unstable, and the outer limit cycle is stable, according to the definitions given earlier. It may be seen that any disturbance (or initial condition) which moves the system no further from the critical point than the unstable limit cycle can be controlled. That is, the control system will eventually bring the system back to steady state. However, once the system is forced outside this limit cycle, it will eventually spiral out to the stable limit cycle. Control cannot be restored, and the reactor temperature and cou9789x_ch25_553-578.indd 571 8/22/08 3:41:31 PM Confirming Pages 572 PART 7 NONLINEAR CONTROL concentration oscillate continuously. This example illustrates very well the limitations of linear control theory. All that the linear investigation could reveal is that, for Kc > 9, the system will be stable in some vicinity of the control point. The phase-plane analysis shows that for Kc slightly greater than 9, this vicinity is inside the unstable limit cycle of Fig. 25–14. If Kc is increased further, the two limit cycles disappear and good control can be achieved. This example points out the importance of unstable limit cycles. Although a physical system can never follow an unstable limit cycle, the limit cycle divides the phase plane into distinct dynamic regions for the physical system. Other values of Kc were analyzed by Aris and Amundson. For low values of Kc, there are two other critical points besides the control point. For example, for Kc ⫽ 0.8, there are critical points at y ⫽ 0 . 95 q ⫽ 1 . 77 y ⫽ 0 . 15 q ⫽ 2 . 15 and Linear analysis shows that both these are stable, but for Kc < 9 the control point (y ⫽ 0.5, q ⫽ 2) is not. Phase-plane analysis shows that if the reactor is started at high temperatures, it will come to steady state at the high-temperature critical point, and vice versa. Starting the reactor at the desired control point will be of no avail, as it will leave and go to one of the other steady-state points, depending on the direction of the initial disturbance. For high values of Kc, there is only one critical point, which is at the control point. Phase-plane analysis shows that Kc must exceed approximately 30 before rapid return to steady state at the desired control point, following all disturbances, is achieved. Some phase-plane portraits for this system that were obtained by means of a computer are shown in Figs. 25–15 to 25–17. 1.0 0.8 Kc = 8 y 0.6 0.4 0.2 0 1.75 cou9789x_ch25_553-578.indd 572 2.0 2.25 FIGURE 25–15 Phase-plane portrait of the control of a chemical reactor (limit cycle forms). 8/22/08 3:41:33 PM Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 573 1.0 0.8 Kc = 10 y 0.6 0.4 0.2 0 1.75 2.0 2.25 FIGURE 25–16 Phase-plane portrait of the control of a chemical reactor (no limit cycle forms). 1.0 0.8 Kc = 20 y 0.6 0.4 0.2 0 1.75 2.0 2.25 FIGURE 25–17 Phase-plane portrait of the control of a chemical reactor (no limit cycle forms). This discussion is only a rather brief introduction to the extensive work by Aris and Amundson. The reader is strongly urged to consult the original paper for a more comprehensive treatment of the problem. cou9789x_ch25_553-578.indd 573 8/22/08 3:41:33 PM Confirming Pages 574 PART 7 NONLINEAR CONTROL SUMMARY We have seen that phase-plane analysis can be used for two typical nonlinear control problems. The results of this analysis give extensive information about the control system behavior. PROBLEMS 25.1. For the system shown in Fig. P25–1, use Simulink to plot the trajectory from the initial point e ⫽ 2 and e ⫽ 0. R + M 1 −1 e − M −1 1 10 (s + 1)2 e C FIGURE P25–1 25.2. For the system shown in Fig. P25–2, plot the phase-plane trajectory. Use t ⫽ 1, z ⫽ K ⫽ 1, b ⫽ 0 . 25 . Let R ⫽ 0. R + −b e 1 , 2 1 M 2s 2 b −1 − K +2 C s+1 FIGURE P25–2 25.3. Consider the phase-plane equations x1 ⫽ x 2 x 2 ⫽ ⫺x1 ⫹ 1 x 10 2 ⫺ x12 ⫺ 10 3 x ( 32 ) (a) Determine the type of critical point at x1 ⫽ ⫺1, x2 ⫽ 0. (b) If there are any other critical points, find them. 25.4. The system shown in Fig. P25–4 is to be controlled by an ideal on/off relay. (a) From the block diagram, write the differential equations for phase-plane description of the physical system in the form x1 ⫽ f ( x1 , x 2 ) x 2 ⫽ f ( x1 , x 2 ) where x1 ⫽ c and x 2 ⫽ c. (b) Plot the trajectory that starts at x1 ⫽ 2, x2 ⫽ 0. (c) Determine the values of x1 and x2 where the first switch occurs. R=0 1 + e − M −1 1 s+1 1 s C FIGURE P25–4 cou9789x_ch25_553-578.indd 574 8/22/08 3:41:34 PM Rev. Confirming Pages CHAPTER 25 EXAMPLES OF PHASE-PLANE ANALYSIS 575 25.5. For the control system shown in Fig. P25–5 determine the frequency and amplitude of the limit cycle if one exists. m + R=0 e m N 2 e−0.5s s+1 C − e −2 N FIGURE P25–5 25.6. For the system shown in Fig. P25–6, t ⫽ 1, z ⫽ R ⫽ 0. R + e M −b m 2s 2 b −M − 1 , 2 K ⫽ 1, M ⫽ 1, b ⫽ 0 . 25, and K +2 s+1 C FIGURE P25–6 (a) For K ⫽ 2, does a limit cycle exist? If so, describe it. (b) If a transport lag e⫺s is introduced in the feedback loop, determine if a limit cycle exists for K ⫽ 2. 25.7. The stirred-tank system shown in Fig. P25–7 produces an aqueous solution of salt by use of a solenoid valve that switches from one reagent tank to the other as described below. The reagent tanks contain concentrated solutions of salt. When the measured concentration is above the set point, the control reagent of lower concentration enters the mixing tank at a constant flow rate of 0.01 L/min. When the measured concentration is less than the set point, the control reagent of higher concentration enters the mixing tank at a constant rate of 0.01 L/min. The holdup volume of the tank is 2 L, the transport lag between the tank and measuring element is 1.2 min, and the set point is 2 g salt/L. Control reagent 300 g salt/ L 100 g salt/ L Solenoid valve Relay Set point 0.01 L /min water 1 L /min ci = 0 Concentration-measuring element FIGURE P25–7 cou9789x_ch25_553-578.indd 575 8/26/08 10:11:05 AM Confirming Pages 576 PART 7 NONLINEAR CONTROL (a) Obtain a block diagram, in terms of deviation variables, for this control system. (b) Determine the characteristics of the limit cycle (frequency and amplitude), if one exists. 25.8. For the control system shown in Fig. P25–8, determine if a limit cycle exists for K ⫽ 1, 2, and 3. If a limit cycle exists, describe it in terms of amplitude and frequency. 1 R + e − −1 −1 1 K s(s + 1)2 C FIGURE P25–8 cou9789x_ch25_553-578.indd 576 8/22/08 3:41:36 PM Confirming Pages CHAPTER 25 CAPSULE SUMMARY MATLAB and Simulink can be used to plot phase planes: 1 x2 = dx1/dt 1 s Integrator ++ x1 −5 Gain ++ −5x1 1 x1 x1(0) To Workspace1 0.5 0 −2x2 x2 To Workspace dx2/dt 1 s Integrator1 ++ x2 = dx1/dt (x10 = 0.6, x20 = 0) (x10 = 1, x20 = 0) (x1(0) = −0.6, x2(0) = 0) x2 XY Graph (x1(0) = 1, x2(0) = 1) −0.5 −2 Gain1 −1 0 x2(0) −1.5 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 x1 Limit cycle. A limit cycle is defined as a periodic oscillation whose amplitude and frequency depend only on the properties of the system and not on the initial state of the system (provided the initial state lies in a certain nontrivial region of the phase space). In the phase plane, stable limit cycles are recognized as closed curves that are approached asymptotically by all nearby trajectories. Unstable limit cycles are closed curves from which all nearby trajectories diverge. An example of a stable limit cycle is the “steadystate” behavior of a home heating system when controlled by a thermostat, an on/off controller. 577 cou9789x_ch25_553-578.indd 577 8/22/08 3:41:36 PM Confirming Pages 578 PART 7 NONLINEAR CONTROL Phase-Plane Trajectory for On/Off Phase-Plane Trajectory for Relay Control System True Relay with Dead Zone Control System 0.6 Solenoid valve closed 0.6 Solenoid valve open 0.4 0.4 Output 1 0.2 0.2 0.02 −0.2 −0.4 −0.6 cou9789x_ch25_553-578.indd 578 0.04 − '0 '0 0 0.06 ' −0.06 −0.04 −0.02 '0 0 ' 0.02 0.04 0.06 le − '0 −0.04 −0.02 it c Li m yc −1 −0.4 −0.6 8/22/08 3:41:36 PM Confirming Pages PART VIII COMPUTERS IN PROCESS CONTROL cou9789x_ch26_579-596.indd 579 8/22/08 3:43:58 PM Confirming Pages cou9789x_ch26_579-596.indd 580 8/22/08 3:43:58 PM Confirming Pages CHAPTER 26 MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL I n this chapter, some of the highlights of modern industrial microprocessor-based controllers and distributed control systems are presented. A microprocessor-based controller is essentially a computer programmed to perform the function of a process controller. For our purpose, the term microprocessor is synonymous with computer, and we could refer to a microprocessor-based controller as a computer-based controller. The number of features of these modern controllers is far too great to cover in one chapter. The best way for the reader to acquire some experience with modern controllers is through laboratory and plant use and by attending some of the short courses offered by the major suppliers of the equipment. 26.1 HISTORICAL BACKGROUND During the past fifty years, tremendous development has occurred in process control hardware. The three phases of development are pneumatic control, electronic control, and microprocessor-based control. During the 1940s, the predominant controller was pneumatic, meaning that signals to and from the controller and within the controller mechanism were air-pressure signals that usually varied from 3 to 15 psig. The development of the high-gain operational electronic amplifier during World War II led to the development of the electronic controller and also the analog computer. The electronic controller mimicked the control functions of the pneumatic controller. It also provided some improvements, such as accurate and reproducible control parameter settings and reduction in size of the instruments. In contrast, the pneumatic controller required frequent calibration of the knobs used to set the various controller parameters (Kc,tI,tD). The pneumatic controller had interaction among the control modes and had inherent 581 cou9789x_ch26_579-596.indd 581 8/22/08 3:43:59 PM Confirming Pages 582 PART 8 COMPUTERS IN PROCESS CONTROL lags that became significant at high-frequency operation. There were frequent debates over the pros and cons of pneumatic and electronic controllers. For example, the pneumatic controller was rugged, simple to install, and required little maintenance. Only a source of air pressure was needed to operate the controller. There was initially great concern about the possibility of explosions with the use of electronic controllers, so the instrument cases for these controllers were purged with steady streams of air when used in plants producing flammable substances. The maintenance of electronic controllers also required highly trained technicians. In the 1960s, the chemical industry made its first attempt at computer process control. These control systems used large mainframe computers, for which the control programs had to be written from scratch. The first attempts at computer control were met with mixed reactions. In the 1970s, there appeared on the market the first generation of digital control hardware, which was based on the advances in microprocessorbased technology. This equipment was user-friendly, and all the software accompanied the hardware. The operator did not face the problem of writing computer code to implement the control functions; it was only necessary to learn the instructions needed to configure (set up) the controllers. 26.2 HARDWARE COMPONENTS The hardware requirements for pneumatic, electronic, and microprocessor-based controls are shown in Fig. 26–1. In this figure, all the components are obtained from a manufacturer of control equipment; several of the components are common to the three systems. In Fig. 26–1a, all the signals are pneumatic (3 to 15 psig). The energy needed to operate these pneumatic components is a source of clean, dry air at a pressure of about 20 psig. The pressure can vary from 20 psig by about ⫾ 10 percent without adversely affecting the operation of the instruments. The electronic system shown in Fig. 26–1b requires both electrical and pneumatic power to operate the components. A transducer or converter is needed between the controller and the valve to convert current (4 to 20 mA) to pressure (3 to 15 psig). This is often referred to as an I/P (I ⫽ current, P ⫽ pressure) converter. The components for a microprocessor-based system are shown in Fig. 26–1c. In this case, the control algorithm is software on the computer. The operator interacts with the control system using the computer. The computer can perform many more functions than just the implementation of the control algorithm, as will be discussed later. The chart recorder used on the older pneumatic and electronic systems has been replaced by a monitor on which the transients are shown. Typical controllers are now capable of processing both analog and digital signals. An analog signal is the type that represents a continuous variable that varies over a range of values. A digital signal is a binary signal that can be represented by two states (on, off, or logic 1, logic 0, etc.). Examples of analog signals are the measurement from a temperature transmitter or the signal sent to a valve. Examples of digital signals are the output to a motor, which causes it to be on or off, or the output to an alarm light causing it to be on or off. The focus of this book has been on analog signals that are applicable to continuous control systems. However, there is an important area of control called batch control which frequently must deal with digital (on/off) signals. Batch control, as the name suggests, is the control of processes that are done in a batch cou9789x_ch26_579-596.indd 582 8/22/08 3:43:59 PM Confirming Pages CHAPTER 26 Measured variable MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL 3–15 Transducer psig 3–15 Controller psig Valve 583 To process Recorder (a) Measured variable Transducer 4–20 mA Controller 4–20 mA Transducer 3–15 psig Valve To process Recorder (b) Measured variable Transducer 4–20 4–20 MicroprocessorTransducer mA mA based controller Printer Monitor 3–15 psig Valve To process Keyboard (c) FIGURE 26–1 Controller components for (a) pneumatic control, (b) electronic control, (c) computer or microprocessor-based control. operation. Many examples of batch processing occur in the pharmaceutical industry where small amounts of products of high unit cost are produced. 26.3 TASKS OF A MICROPROCESSOR-BASED CONTROLLER The primary task of a microprocessor-based controller is implementation of a control algorithm; however, the presence of a computer makes it possible to perform a number of other peripheral tasks that are useful in process control and monitoring. Some of these tasks provided in a modern control system are to Implement classical and advanced control algorithms Provide static and dynamic displays on the monitor Provide process and diagnostic alarms Provide mathematical functions Provide data acquisition and storage (archiving) The software to support all these tasks is supplied by the manufacturer of the control equipment. We now look briefly at the nature of each task. cou9789x_ch26_579-596.indd 583 8/22/08 3:43:59 PM Confirming Pages 584 PART 8 COMPUTERS IN PROCESS CONTROL Implementation of Control Algorithms The portion of the software that covers this task is organized into large numbers of blocks that can be connected to solve a specific control problem. A partial listing of the blocks typically provided is as follows: Analog input Analog output Conventional control algorithms (P, PI, PD, PID) Linearization Lead-lag Dead time Self-tuning There are many other blocks that have been omitted from this list because of the limitation of space in this chapter. Also a number of blocks process digital (or logic) signals (on/off) such as comparators, selectors, or timers, which are needed in batch control and automatic plant start-up and shutdown. ANALOG INPUT BLOCK. The analog input block is an analog-to-digital device that converts a continuous signal from a transducer, which is in the form of a current or voltage, to a digital signal that can be used in the microprocessor. ANALOG OUTPUT BLOCK. The analog output block reverses the operation of the analog input block by converting a digital signal, which has been computed in the microprocessor, to a voltage or a current that can be sent out to a transducer in the process in the field. Sometimes this block is called a field output block. CONTROL BLOCK. The control block is a block for which many parameters can be specified. The manufacturer does not give any information on the method of implementing the control algorithms. The computer does not “watch” the process continuously, but rather it “samples” the necessary process signals at discrete time intervals. The sampling period T is one parameter that generally cannot be adjusted in a commercial controller; it is fixed by the developer of the software. Typical values of T in commercial controllers vary from 0.1 to 0.25 s. A controller operating with such a small T can be considered as a continuous controller for many chemical processes with large time constants. Parameters that can be selected are the controller parameters (Kc,tI,tD), limits on set point and controller output, and others. The familiar PID control algorithm from Chap. 9 is a continuous form of the equation p ⫽ K ce ⫹ K ct D de Kc ⫹ dt tI t ∫0 e dt ⫹ ps (9.11) To implement the PID algorithm in the computer (digital) control system, we must discretize the equation, realizing that we are sampling the signals at fixed instants in time. The discrete version of Eq. (9.11) is cou9789x_ch26_579-596.indd 584 8/22/08 3:43:59 PM Confirming Pages CHAPTER 26 MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL T pn ⫽ K c e n ⫹ t I n ∑ ei ⫹ i ⫽1 tD (e n ⫺ e n ⫺1 ) ⫹ ps T 585 (26.1) Equation (26.1) can be interpreted as describing the output from the controller at the nth time interval (at t ⫽ nT). This form of the PID equation is called the position form of the equation. It specifies the exact value of the controller output at any time t (for example, the exact valve position if we are controlling a valve). Another form of this equation is called the velocity form of the equation. It specifies the change in the controller output from one time interval to the next. The velocity form of the PID equation is T tD ∆pn ⫽ K c (e n ⫺ e n ⫺1 ) ⫹ e n ⫹ (e n ⫺ 2e n ⫺1 ⫹ e n ⫺ 2 ) tI T pn ⫽ pn ⫺1 ⫹ ∆pn (26.2) Either the form of Eq. (26.1) or that of (26.2) can be used for the PID controller. Each is used in commercially available control systems. LINEARIZATION BLOCK. The linearization block is used to “straighten out” a non- linear relation. The most common example of the need for this block is in processing a signal from an orifice plate used to measure flow. The signal (pressure) across an orifice plate is proportional to the square of the flow. To obtain a linear relation between flow rate and signal, the signal is sent through a linearization block, which has been configured to extract the square root of the input signal. The linearization block can also be configured to linearize any nonlinear relation that can be plotted on a coordinate system. This aspect of the linearization block can be useful for linearizing the input-output relation to a valve that is nonlinear in behavior. In Chap. 19, an equal-percentage valve was proposed as a device to linearize the relation between flow and valve-top pressure when line loss was large. LEAD-LAG BLOCK. The lead-lag block simulates the lead-lag transfer function K(T1s ⫹ 1)/(T2s ⫹ 1). The parameters K, T1, and T2 can be selected over a wide range of values. If one needs a first-order lag, T1 can be set to zero. We have seen the need for the lead-lag block in feedforward control in Chap. 17. DEAD-TIME BLOCK. The dead-time block simulates dead time (or transport lag) e⫺t d s . For this block, td can be selected over a wide range of values. We have seen the need for this block in the Smith predictor control algorithm of Chap. 17. Figure 26–2 shows a simple flow example using some of the blocks just described. The blocks are connected by computer code at a keyboard during the configuration of the control system. This connection of blocks is called softwiring since it is done through the software. The actual connection between the flow transmitter and the analog input block in the controller, which is made with wires, is called hardwiring. cou9789x_ch26_579-596.indd 585 8/22/08 3:43:59 PM Confirming Pages 586 PART 8 COMPUTERS IN PROCESS CONTROL Linearization (√ ) Control (PID) Analog output Analog input Controller Currentpressure transducer Pressurecurrent transducer Process Delivery Supply Orifice meter FIGURE 26–2 Example of the use of control blocks to control a flow process. SELF-TUNE BLOCK. For years, one of the goals of control engineers has been to develop a device that would automatically tune a controller, online, while the process is operating. Until recently, this goal was reached for some special cases by the application of adaptive control theory, a branch of control that is beyond the scope of this book. In the mid-1980s a commercial device became available that uses the normal transients occurring in a controlled process (caused by set point and load upsets) to update the control parameters of a PID controller. This device is called a self-tuner and is one of the blocks available in the microprocessor-based controller of several hardware manufacturers. When the self-tuner is first applied to a process for which no process identification has been performed, the self-tuner is placed in the pretune phase, during which time the process is subjected to a pulse while it is operated open-loop. The introduction of the pulse and the analysis of the transient are done automatically by the self-tuner. The outcome of the pretune phase of operation is the selection of controller parameters. A conceivable approach to the development of the pretune phase of a self-tuner is to monitor an open-loop step response and apply a tuning method similar to the Cohen-Coon tuning method of Chap. 18. After the pretune phase, the control system is returned to closedloop and the self-tuner continues to monitor the transients and make changes in controller parameters when needed. The self-tuning that occurs during closed-loop operation is based on the characteristics of the transients, such as decay ratio, overshoot, and period of oscillation. The self-tuning algorithm, being proprietary information, is described in only a general manner in the reference manual that accompanies the control equipment. Since many industrial processes are poorly tuned, the general-purpose self-tuner represents an impressive achievement in the application for computer control technology. Displays Software has made the strip chart recorder almost unnecessary. Transients (or trends) produced in a control system can now be displayed on a monitor screen dynamically. cou9789x_ch26_579-596.indd 586 8/22/08 3:44:00 PM Confirming Pages CHAPTER 26 MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL 587 As time progresses, the values of selected variables are displayed as a function of time. The segment of time shown on the screen can be selected to be a few minutes to a few hours to show dynamic detail or long-term trends. Transients that occurred in the past can be stored and displayed again. Many process operators are more comfortable with control instruments that have a faceplate which shows bar graphs or pointers indicating set point, control variable, and output to the valve. In the older instruments, those indicators were obtained by use of mechanical motion or other means. The software for computer-based control can be used to obtain a dynamic display on the screen that mimics the faceplates of older, more traditional instruments. Alarms A part of the control software is devoted to detecting and reporting a problem in the form of an alarm. The alarm takes the form of a visual signal (flashing light), an audible signal (beeping horn), or the actuation of a switch. Examples of the use of switch closures include turning on or off a pump motor or opening or shutting a valve. The alarms are classified as process alarms and diagnostic alarms. The diagnostic alarm detects a malfunction in the control equipment or the loss of communication. For example, if a wire connecting the output of a temperature transmitter to an analog input block breaks, a diagnostic alarm will go off, indicating that the signal to the analog input block is out of range. The manufacturer of the control equipment provides all the software for detecting the problems that trigger diagnostic alarms. The engineer who configures the control blocks selects the variables that are to trigger process alarms and specifies the alarm limits and the type of annunciation (flashing light, beeping horn, etc.) The alarms can be assigned a priority rating. Those variables in a process that are most critical are given the highest priority; less critical variables are given a lower priority. Mathematical Functions Control software provides basic mathematical functions such as summation, subtraction, multiplication, and accumulation (i.e., integration). These functions can be used along with other blocks in the design of a control system. A simple example of these functions is the calculation of mass flow rate of a gas from measurements of velocity, pressure, and temperature. These three measurements are combined according to the following relationship, which is based on the ideal gas law: w ⫽ vAPM RT where w ⫽ mass flow rate, mass/time v ⫽ velocity P ⫽ pressure T ⫽ absolute temperature M ⫽ molecular weight of gas R ⫽ gas constant A ⫽ cross-sectional area for flow cou9789x_ch26_579-596.indd 587 8/22/08 3:44:00 PM Confirming Pages 588 PART 8 COMPUTERS IN PROCESS CONTROL The signal from the math block that represents w can then be sent as the control variable to a control block that controls the mass flow of gas. Data Acquisition and Storage Long-term storage of the transients can be obtained easily with a computer. This task is referred to as archiving. The automatic storage of critical process-control variables on disk or tape can be retrieved later to explain process operating difficulties. The computer can also be used to automatically record or log the type and location of an alarm, the time of a process alarm, the time of acknowledgment of an alarm, and the time it was cleared by operator intervention. This information is useful to supervisors in detecting violation of safety regulations or diagnosing process malfunctions. 26.4 SPECIAL FEATURES OF MICROPROCESSOR-BASED CONTROLLERS In addition to the tasks just described, there are three special features available in modern microprocessor-based controllers that deserve attention. These are limiting, tracking, and anti-reset windup. Each will be discussed separately. Limiting In configuring a control system from basic control blocks, one can select lower and upper limits on controller output and set point. These limits are narrower than the limits inherently present in the hardware. Limits are often placed on a controller output for safety reasons or to protect equipment. For example, if one knows the flow rate of a liquid that causes a tank to overflow, one can set the limit on the output of a controller at a value less than the value that causes overflow. The limits on the controller output are active when the controller is in either automatic or manual mode. An example of a limit on set point is the selection of an upper limit on pressure for a steam-heated sterilizer to prevent damage to the equipment. Tracking A very useful feature of a microprocessor-based controller is tracking. Although tracking is not needed to successfully control a system, its presence is of great convenience to the process operator. Two examples of tracking are set point tracking and controller output tracking. Set point tracking is useful when a controller is transferred from manual to automatic. When a process is started up for the first time, a common procedure is to bring the process on-stream in manual mode. In this case the operator adjusts the output of the controller (which goes to the valve) until the process variable comes to a desired steady state. When the tracking feature is not present in the controller, the set point must be manually adjusted until it equals the process variable before the controller is transferred to cou9789x_ch26_579-596.indd 588 8/22/08 3:44:00 PM Confirming Pages CHAPTER 26 MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL 589 automatic; the process then continues running in a smooth manner. If the operator adjusts the set point to the process variable after switching to automatic, there may be a temporary disturbance in the process variable. The expression for the disturbance is called a bump. With set point tracking, the operator does not need to think about adjusting the set point to the process variable, because it is done automatically. In other words, set point tracking provides “bumpless” transfer when switching from manual to automatic. A second example of tracking can be seen in its use for transferring a cascade system from manual to automatic. (The reader should be familiar with the information on cascade control provided in Chap. 17 to understand this example.) To explain the use of tracking in cascade control, reference to Figs. 17–1b and 17–2b will be made. In starting up this system, the primary controller is placed in a standby condition, and the secondary controller is placed in manual mode. The means for accomplishing this is built into the software of the controller. With the secondary controller in manual mode, its output is adjusted until the temperature of the tank contents T0 is at the desired value. Then, with the control system at steady state and T0 at the desired value, the system is transferred to cascade mode by placing both controllers in automatic. Since the output of the primary controller adjusts the set point of the secondary controller, it is necessary to have the output of the primary controller equal to the jacket temperature Tj when the system is transferred to cascade mode. This goal can be achieved by having the output of the primary controller track the jacket temperature while the secondary controller is used in manual mode to adjust the tank temperature to the desired value. For this example, the set point of the primary controller can also automatically track the tank temperature T0 before the transfer to cascade mode occurs. In this cascade control example, we have seen tracking used for both the set point and the controller output. Anti-Reset Windup A troublesome problem with a controller having integral action (PI or PID) is the possible occurrence of reset windup. When the error to a controller remains large for a long time, the integral action of the controller builds up a large value of output which often approaches the saturation value of the controller output. This accumulation of output is called reset windup. When the process variable returns to the set point, the output of the controller does not immediately return to a value that will hold the process variable at the set point because the controller output has built up (or has been wound up) and must be reduced by the presence of error of opposite sign over some duration of time. Thus the transient for the control variable exhibits a large overshoot that can persist while the output signal is being reduced through integral action being applied to the error of reversed sign. Reset windup typically occurs during the start-up of a process. To gain some insight into the cause of reset windup, consider the start-up of the liquid-level process shown in Fig. 26–3 in which the level in the third tank is to be controlled by a PI controller. The valve is linear and saturates at 0 and at 0.5, as shown in Fig. 26–3b. Upon start-up with the PI controller in automatic mode, the tanks are empty, and the error R ⫺ C is large and positive. The action of the controller on this error will result in a large output M due to proportional action and a rising contribution to M due to the integral action. The output of the controller will be at its saturation value, which is typically about 10 percent above the top of the 4- to 20-mA scale (i.e., 22 mA). cou9789x_ch26_579-596.indd 589 8/22/08 3:44:01 PM Confirming Pages 590 PART 8 COMPUTERS IN PROCESS CONTROL Set point Supply c Flow (a) 0.5 0 0 4 20 Input to value, mA (b) Valve R + PI Process 1 M (s + 1)3 − C (c) FIGURE 26–3 Plant start-up illustrating reset windup (tanks are initially empty): (a) process, (b) linear valve with saturation limits, and (c) block diagram of process. The large saturated value of M will in turn cause the valve to reach its saturation value, which has been taken as 0.5. During the initial phase of the operation, the tanks are being filled at the maximum rate of flow provided by the upper limit of the control valve. During this filling stage of operation, the controller is not exercising any control since the valve is at its limit. As the level rises toward the set point, the large error that existed at start-up gradually diminishes toward zero. If only proportional action were present in the controller, the output of the controller would return quickly to a mid-scale value; however, because of the integral action, the controller output remains high, at its saturation value, long after the process variable first reaches the set point. To reduce the output cou9789x_ch26_579-596.indd 590 8/22/08 3:44:01 PM Confirming Pages CHAPTER 26 MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL 591 M, the integral action must be applied to negative error so that the integration can lower the output to mid-scale. This negative error occurs as a result of the tank level remaining above the set point for some time after the tank level reaches the set point. Other causes of reset windup and some methods to prevent it are discussed by Shinskey (1979). The control system shown in Fig. 26–3c was simulated for a start-up transient with the tanks initially empty; the transient is shown as curve I in Fig. 26–4. The large overshoot in tank level after the level reaches the set point is clearly illustrated. Now that the problem of reset windup has been described, we focus our attention on how to reduce or eliminate it. The development that follows on the use of external feedback to eliminate reset windup is based on the work of Shunta and Klein (1979). c I No external feedback 1.0 Set point II External feedback 0 0 10 t FIGURE 26–4 Start-up transients for system in Fig. 26–3 with and without external feedback. A feature of microprocessor-based controllers is the availability of external feedback in the configuration of a PI or PID controller. The block diagram of a PI controller with external feedback is shown in Fig. 26–5. Controller Set point M Control variable + − M F FIGURE 26–5 Controller with external feedback for use in anti-reset windup. The output of this controller is given by M ( t ) ⫽ Kce ( t ) ⫹ Kc tI 1 ∫0 e ( t ) dt ⫹ t I ∫0 [ F ( t ) ⫺ M ( t )] dt t t (26.3) where M(t) ⫽ controller output e(t) ⫽ error ⫽ set point ⫺ control variable F(t) ⫽ external feedback signal cou9789x_ch26_579-596.indd 591 8/22/08 3:44:01 PM Confirming Pages 592 PART 8 COMPUTERS IN PROCESS CONTROL If the Laplace transform of both sides of Eq. (26.1) is taken, the result is M ( s ) ⫽ Kce ( s ) ⫹ Kce ( s ) 1 ⫹ [ F ( s ) ⫺ M ( s )] st I st I (26.4) If the feedback signal is the controller output F(s) ⫽ M(s), Eq. (26.4) becomes the usual transfer function for a PI controller: 1 M ( s ) ⫽ Kc 1 ⫹ e (s) t Is (26.5) The feedback signal F(t) can be any signal available to the microprocessor-based controller. When F(t) is not equal to M(t), Eq. (26.4) can be solved for M(s) to give M ( s ) ⫽ Kce ( s ) ⫹ F (s) tIs ⫹ 1 (26.6) A controller following this equation provides a signal consisting of proportional action plus first-order tracking of F(t). If F(t) in Eq. (26.3) is taken as the output of the valve (or the output signal of the current-to-pressure transducer that goes to the valve) in our example in Fig. 26–3c, we have the basis for eliminating reset windup. During the filling stage of the tank, the feedback signal F(t) will be constant at the saturation value of the valve output. When the tank level reaches the set point, the error will be zero and the only contribution from the controller output will be the tracked signal represented by the second term on the right side of Eq. (26.6). This value will be less than would be the case if external feedback were not employed. The overall result is that the controller output is less with the external feedback at the time the level first equals the set point, and the overshoot is reduced. The transient using external feedback is also shown in Fig. 26–4 as curve II. Notice that the overshoot is less when external feedback is used. To emphasize the benefit of external feedback for eliminating reset windup, no limits were placed on the output of the controller in the simulation of Fig. 26–3. In practice, there are physical limits on the controller output, and when this is the case, the reduction of overshoot with the use of external feedback may not be so pronounced as shown in Fig. 26–4. 26.5 DISTRIBUTED CONTROL So far we have been concerned in this chapter with the operation of a single controller. Such a controller is referred to as a stand-alone controller because it is not communicating with other controllers, but only with the one control loop of which it is a part. Present-day microcomputer-based control systems have the capability of communicating with other controllers through a network, which is called distributed control. Figure 26–6 shows one version of the communication linkages that are usually present in a distributed control system. Each manufacturer of distributed control systems has a different way of organizing them. cou9789x_ch26_579-596.indd 592 8/22/08 3:44:01 PM Confirming Pages CHAPTER 26 MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL 593 A distributed control system is intended to be used for a large processing facility that involves as many as 50 to 100 loops. Some examples of these types of installations are a refinery, a brewery, and a power plant. In Fig. 26–6, the modules of control equipment that communicate with each other are Control processor (CP) Applications processor (AP) Workstation (WS) Field bus module (FBM) The first three of these modules communicate with one another through a nodebus or “data highway,” as it has been called. The field bus modules serve as devices that interface with transducers and valves in the process. Distributed control systems thus involve small modular controllers mounted in the field, process operator PC workstations in the control room or distributed throughout the plant, and a data highway (perhaps an Ethernet network) connecting the components. The control processor contains the blocks described earlier (analog input, analog output, control, linearization, etc.) that are connected by softwiring to provide the control algorithm required for each loop. Communication between the control processor and the process (a distance away) in the field takes place in the field bus module. Node bus CP CP CP FBM FBM FBM AP WS Printer Monitor Keyboard Process FIGURE 26–6 Typical connections in a distributed control system: CP: central processor, AP: applications processor, WS: workstation, FBM: field bus module. One type of field bus module provides a set of analog inputs and a set of analog outputs that send to and receive from the field continuous signals (4 to 20 mA). Another type of module sends to and receives from the field digital signals that often take the form of switch-contact closures. The application processor is a computer (which runs software) for performing the many tasks described earlier and for managing the communication among modules. The workstation module is connected to a keyboard, a mouse, a monitor, and a printer for use by process operators to interact with the system. At the workstation, the process operator can call up on the screen various displays, change set points and controller parameters, switch from automatic to manual, acknowledge alarms, and perform other tasks needed to operate a control system consisting of many loops. cou9789x_ch26_579-596.indd 593 8/22/08 3:44:02 PM Confirming Pages 594 PART 8 COMPUTERS IN PROCESS CONTROL There are currently a number of types of digital networks (or buses) for plant control systems. In order of increasing complexity, they are sensor buses (used for simple devices and manufacturing control), device buses (which include Profibus and DeviceNet), and field buses (which include Profibus and Foundation Fieldbus) which are used for process control and diagnostics. These types of buses or networks are supported by numerous hardware vendors. “Smart” instrumentation is becoming more and more popular as equipment and control systems are replaced. Smart instrumentation has the capability to provide the control system with additional information and function that “nonsmart” devices do not, such as device ID verification, device diagnostics and status information, secondary process and device variables, remote device configuration and setup. According to the ARC Advisory Group of Dedham, Massachusetts (Fig. 26–7), worldwide about 40 percent of the installed devices are now smart devices D’Aquino and Greene, (May 2003). The rest are pneumatic and electronic. 40–45 million Devices Installed 39% 48% 13% Electronic (4–20 mA) Pneumatic (3–15 psig) Smart FIGURE 26–7 Nature of installed process instrumentation worldwide according to the ARC Advisory Group of Dedham, Massachusetts. The trend is certainly moving toward the installation of computer control systems and smart instrumentation devices. Some of the challenges to be overcome include lack of compatibility with existing plant control systems and wiring, lack of a standard digital protocol, and initial cost. As these systems mature and the obstacles are overcome, we can look forward to exciting advances in process monitoring and control systems’ hardware and software. The net result will be better controlled processes, and better access to process data for monitoring, diagnostics, and reporting capabilities. SUMMARY The computer has greatly changed the nature of industrial process control equipment. The microprocessor has become the heart of control instruments, and software has provided many functions besides the basic control algorithm. When the pneumatic controller cou9789x_ch26_579-596.indd 594 8/22/08 3:44:02 PM Confirming Pages CHAPTER 26 MICROPROCESSOR-BASED CONTROLLERS AND DISTRIBUTED CONTROL 595 was the predominant type, one purchased a controller with very specific attributes (e.g., mode of control, type of measured variable, chart speed). The smart instruments and controllers available today contain not only the conventional control algorithms, but also many other functions such as simulation of basic transfer functions (e.g., leadlag and transport lag), display building, mathematical functions, process and diagnostic alarms, and data acquisition. They also provide logic functions (comparators, timers, counters, etc.) for use in batch control and plant start-up and shutdown, as well as selftuning algorithms. In this chapter, some of the features of modern controllers were discussed (e.g., limiting, tracking, and anti-reset windup). Any controller having integral action can cause reset windup under certain conditions when the error persists for a long time. The result of such a phenomenon is a transient that has large overshoot. Manufacturers of control instruments now offer several methods for reducing reset windup; the one presented in this chapter was use of external feedback. Before computer control appeared, most process loops were served by individual controllers with signals to and from these controllers being collected on a large panel board in a special control room. To obtain communication between the control room and the controllers required much wiring and piping (for pneumatic systems). Today, microprocessor-based control systems have the capability of communicating with other control instruments through networks, called distributed control. A distributed control system can control an entire plant and involve as many as one hundred or more control loops. Since each manufacturer has a different way of organizing a distributed control system, the practicing engineer must obtain the details of a particular system from the manufacturer. Most manufacturers offer a variety of short courses for technicians and engineers on the installation and use of their hardware and software. cou9789x_ch26_579-596.indd 595 8/22/08 3:44:02 PM Confirming Pages CHAPTER 26 CAPSULE SUMMARY DISTRIBUTED CONTROL SYSTEM SCHEMATIC Node bus CP CP CP FBM FBM FBM AP Printer WS Monitor FIGURE 26-6 Typical connections in a distributed control system: CP: central processor, AP: applications processor, WS: workstation, FBM: field bus module. Keyboard Process SMART INSTRUMENTS Smart instruments have the capability to provide the control system with additional information and function that “nonsmart” devices do not, such as device ID verification, device diagnostics and status information, secondary process and device variables, remote device configuration and setup. FORMS OF PID ALGORITHM FOR DIGITAL IMPLEMENTATION Standard continuous form p ⫽ K c e ⫹ K ct D de Kc ⫹ dt tI t ∫0 e dt ⫹ ps (9.11) tD (e n ⫺ e n⫺1 ) ⫹ ps T (26.1) Position form of PID algorithm T pn ⫽ K c e n ⫹ tI n ∑ ei i ⫽1 ⫹ Velocity form of PID algorithm T tD ∆pn ⫽ K c (e n ⫺ e n ⫺1 ) ⫹ e n ⫹ (e n ⫺ 2e n⫺1 ⫹ e n⫺2 ) tI T pn ⫽ pn ⫺1 ⫹ ∆pn (26.2) Either the form of Eq. (26.1) or that of Eq. (26.2) can be used for the PID controller. Each is used in commercially available control systems.. 596 cou9789x_ch26_579-596.indd 596 8/22/08 3:44:02 PM Confirming Pages BIBLIOGRAPHY Aris, R., and N. R. Amundson (1958). Chem. Eng. Sci., 7, 121–155. Anton, H. (1984). Elementary Linear Algebra, 4th ed., New York: Wiley. Bequette, B. W. (2003). Process Control: Modeling, Design and Simulation, Upper Saddle River, N. J.: Prentice-Hall. Bird, R. B., W. E. Stewart, and E. N. Lightfoot (1960). Transport Phenomena, New York: Wiley. Carslaw, H. S., and J. C. Jaeger (1959). Conduction of Heat in Solids, 2d ed., Oxford at the Clarendon Press. Chau, P. C. (2002). Process Control: A First Course with MATLAB, Cambridge UK: Cambridge University Press. Churchill, R. V. (1972). Operational Mathematics. 3d ed., New York: McGraw-Hill. Cohen, G. H., and G. A. Coon (1953). Trans. ASME, 75, 827. Cohen, W. C., and E. F. Johnson (1956). IEC, 48, 1031–34. Cooper, Douglas J. (2005). Practical Process Control using LOOP-PRO Software, Control Station, Inc. Coughanowr, D. R. (1991). Process Systems Analysis and Control, 2d ed., New York: McGraw-Hill. Coughanowr, D. R., and L. B. Koppel (1965). Process Systems Analysis and Control, New York: McGraw-Hill. D’Aquino R. and R. Greene, “Forays in Smart Instrumentation”, Chem. Eng. Progress, 99, 8. Doyle, F. J., III (2000). Process Control Modules: A Software Laboratory for Control Design, Upper Saddle River, N.J.: Prentice-Hall. Evans, W. R. (1948). “Graphical Analysis of Control Systems,” Trans. AIEE, 67, 547–551. Evans, W. R. (1954). Control-System Dynamics, New York: McGraw-Hill. Foxboro Co. (1978). “Principles of Feedforward Control.” Audiovisual tape, No. B0150ME (1 hr, 23 min; 3 parts), Foxboro, Mass. Kuo, B. C. (1987). Automatic Control Systems, 5th ed., Englewood Cliffs, NJ : Prentice Hall. Letov, A. M. (1961). Stability in Nonlinear Control Systems, Princeton, N.J.: Princeton University Press. Lopez, A. M., P. W. Murill, and C. L. Smith (1967). “Controller Tuning Relationships Based on Integral Performance Criteria.”, Instrumentation Technology, 14, No.11, 57. Luyben, W. L., and M. L. Luyben (1997). Essentials of Process Control, New York: McGraw-Hill. Marlin, T. E. (2000). Process Control: Designing Processes and Control Systems for Dynamic Performance, 2d ed., Boston: McGraw-Hill. Marlin, T. E. (2000), Process Control: Designing Processes and Control Systems for Dynamic Performance, 2d ed., New York: McGraw-Hill. 597 cou9789x_bib_597-598.indd 597 8/25/08 2:19:12 PM Confirming Pages 598 BIBILIOGRAPHY McCabe, W. L., J. C. Smith, and P. Harriott (2004). Unit Operations of Chemical Engineering, 7th ed., New York: McGraw-Hill. Mickley, H. S., T. K. Sherwood, and C. E. Reed (1957). Applied Mathematics in Chemical Engineering, 2d ed., New York: McGraw-Hill. Morari, M., and E. Zafiriou (1989). Robust Process Control, Englewood Cliffs, N.J.: Prentice-Hall. Nobbe, L. B. (1961). “Transient Response of a Bubble-Cap Plate Absorber.” M.S. thesis, Purdue University, Indiana. Ogunnaike, B. A., and W. H. Ray (1994). Process Dynamics, Modeling and Control, New York: Oxford University Press. Oldenbourg, R. C., and H. Sartorius (1948). “The Dynamics of Automatic Control,” Trans. ASME, Vol. 77, p. 78. Perry, R. H., and D. W. Green (2007). Chemical Engineers’ Handbook, 8th ed., New York: McGraw-Hill. Philips, C. L., and R. D. Harbor (2000). Feedback Control Systems, 4th ed., Englewood Cliffs, N.J.: Prentice-Hall. Riggs, James B., and M. Nazmul Karim (2007). Chemical and Bio-Process Control, 3d ed., Lubbock, Tex.: Ferret Publishing. Routh, E. J. (1905). Dynamics of a System of Rigid Bodies. Part II. London: Macmillan & Co., Ltd. Seborg, D. E., T. F. Edgar, and D. A. Mellichamp (2004). Process Dynamics and Control, 2d ed., New York: Wiley. Shinskey, F. G. (1979). Process Control Systems, 2d ed., New York: McGraw-Hill. Shunta, J. P. and W. F. Klein (1979). “Microcomputer Digital Control—What it ought to do,” ISA Trans., 18, No. 1, 63. Smith, C. A., and A. B. Corripio (2006). Principles and Practice of Automatic Process Control, 3d ed., New York: Wiley. Smith, O. J. M. (1957). “Closer Control of Loops with Dead Time,” Chem. Eng. Prog., 53, 217. Stephanopoulos, G. (1984). Chemical Process Control. An Introduction to Theory and Practice, Englewood Cliffs, N.J.: Prentice-Hall. Thaler, G. J., and M. P. Pastel (1962). Analysis and Design of Nonlinear Feedback Control Systems, New York: McGraw-Hill. Wilts, C. H. (1960). Principles of Feedback Control, Reading, Mass: Addison-Wesley. Ziegler, J. G., and N. B. Nichols (1942). “Optimum Settings for Automatic Controllers,” Trans. ASME, 64, 759. cou9789x_bib_597-598.indd 598 8/25/08 2:19:13 PM Confirming Pages INDEX A Absorption, dynamics of, 453–57 Adjoint of matrix, 492 Alarm, process and diagnostic, 587 Amplitude ratio, 289 Analog-to-digital converter, 584 Attenuation, 89, 290 Autonomous system, definition, 554 B Batch control, 584 Block diagram, 77, 166–67 chemical reactor, 212–13 standard symbols, 218–19 Bode diagram asymptotic approximations, 302–4 controllers, 311–13 definition, 300 first-order system, 302–5 graphical rules, 307–8 second-order system, 308–11 systems in series, 305–7 transportation lag, 311 Bode stability criterion, 326–27 Bumpless transfer, 398, 589 C Cascade control, 353–60 in valve positioner, 439 Characteristic equation, 255 roots of, 43–44 Chattering, in on-off control, 564 Chemical reactor, 205–8 phase plane of, 569–73 Closed-loop system, 167 Closed-loop transfer functions, 220–23 Cofactor matrix, 492 Cohen-Coon process reaction curve, 397–401 comparison of methods, 401–10 Cohen-Coon tuning, 397–401 Comparator, 166–67 Computer control, 581–94 Control Station, 401–19. See also LOOP PRO Control system response, 228–46 Controller, 192–98 calibration of, 196 cascade, 353–60 digital, 582 feedforward, 361–70 internal model, 378–85 microprocessor-based, 581–94 pneumatic versus electronic, 581–82 ratio, 370–73 Smith Predictor, 373–78 Controller mechanism, 190 Controller modes, choice of, 391–93 motivation for, 197–98 Controller tuning, 391–410 Corner frequency, 302 Criteria of control quality, 393–94 Critical damping, 142 Critical points, analysis of, 556–60 definition of, 554 Cross-controller, 515 Crossover frequency, 326 Custom inputs, 57–58 Cv for valve, 425 D Damping, viscous, 539 Dead zone, in on-off control, 565–69 Decay ratio, 148 Decibel, 303 Derivative action in control, 196 Derivatives, Laplace transform of, 23–25 599 cou9789x_ndx_599-604.indd 599 1/1/70 2:26:24 PM Confirming Pages 600 INDEX Determinant of matrix, 491 Deviation variables, 74 in distributed parameter systems, 458 Differential equations, 26 MATLAB solution, 27, 38 Displays, 587 Distributed control, 592 Distributed-parameter systems, 458–71 E Error, 2–3 External feedback for anti-reset windup, 589–92 F Feedback negative, 167 positive, 167 Feedforward control, 361–70 Foxboro tuning rules, 367 Fieldbus module, 592–95 Figure of merit, 393–404 Filter in internal model control, 381 Final-value theorem, 49 First order plus dead time model, 408–18 First-order lag, 75 First-order system, 71–77 MATLAB/Simulink simulation, 81–84 impulse response, 85–87 in series arrangement, 123–30 interacting, 128–30 noninteracting, 123–27 sinusoidal response, 87–92 step response, 79–84 transfer function, 75 Flow control, 585–86 Focus, 556–57 FOPDT, 408–18 Forcing function, 32 Frequency response, 287–350 Bode diagram, 300–17 Bode stability criterion, 326–27 in control system design, 323–43 of controllers, 311–14 of distributed-parameter systems, 462–63 definition, 300–301 from elliptical phase diagram, 319 experimental determination of, 418–19 gain and phase margins, 327–29 heuristic stability arguments, 299–300, 326–28 Nyquist stability criterion, 326 from pulse test, 415–18 substitution rule, 287–89, 316–17 cou9789x_ndx_599-604.indd 600 of systems, 302–11 in series, 305–7 Ziegler-Nichols settings, 335–37, 394–97 Frequency testing, 418–19 G Gain margin, 328–30 design specifications, 328 Gas absorber, dynamics of, 453–57 Goal Seek Excel, 333 H Heat conduction, dynamics of, 458–63 dynamics, of counterflow, 464–71 steam-jacketed kettle, 443–53 Hysteresis in valves, 438–39 I Impulse function, 22, 54 Initial-value theorem, 51 Instrumentation symbols for P&IDs, 204 Integral action in control, 194–96 Integral of error criteria absolute value of error (IAE), 394 square of error (ISE), 393 time-weighted absolute error (ITAE), 394 Integral, Laplace transform of, 55 Interacting systems, 128–30 in control system, 514–24 in mercury thermometer, 131 Internal model control, 378–85 Inverse of matrix, 491 Inversion of Laplace transforms, 26, 32–42 L Laplace transform, 18–42 of integral, 55 inversion of, 26, 32–42 table, 21–22 use in partial differential equations, 459–61 Lead-lag transfer function, 585 Liapunov, Theorem of, 559–60 Limit cycle, 560–73 in exothermic chemical reactor, 569–73 in on-off control, 560–69 Limiting in controller and valve, 588–92 Linearization, 109–14, 446 in analysis of critical points, 558 Liquid level, 99–105 LOOP PRO simulation, 412–15 Load change, 222, 234–40 Loading, in liquid-level process, 129 LOOP PRO, 401–19 1/1/70 2:26:25 PM Confirming Pages INDEX Lumped-parameter model, of distance-velocity lag, 463–64 for mercury thermometer, 80 M Manometer, 137–47 MATLAB m-file for varying controller gain in Simulink, 232 MATLAB symbolic processing, 23, 26–27 Matrix, 490–95 Matrix differential equation, 479–80 Minor of matrix, 492 Mixing process, 11–12, 14–18, 34–35, 96, 105–06, 117–19, 200, 370–71 Multiloop system, block diagram reduction, 219, 224, 514 Multiple input-multiple output system (MIMO), 512 Multivariable control, 512–29 decoupling, 523 interaction, 512–14 stability, 525–26 Mybode m-file for MATLAB, 304 N Natural frequency, 149 Natural period, 149 Negative feedback, 167 overall transfer function, 222 Node, 557 Nodebus in distributed control, 593 Noninteracting control, 517–24 Nonlinear systems, 533–52 definition of, 533 Nonminimum phase characteristics, 381 Nonminimum phase lag, 463 Nyquist stability criterion, 326 601 Period of oscillation, 149 ultimate, 336 Phase angle, 89, 152, 289 Phase lag, 89 Phase lead, 89 Phase margin, 327–29 design specification, 329 Phase plane, 534–46 Phase space, 534–46 PID equation position form, 585 PID equation velocity form, 585 Poles and zeros, 270, 275 Positive feedback, 167 overall transfer function, 223 Process dynamics, experimental, 410–19 theoretical, 443–71 Process identification, 410–19 Process reaction curve, 398–99 Proportional control, 192 Proportional controller, ideal transfer function, 193 Proportional-derivative control, ideal transfer function, 196–97 Proportional-integral control, ideal transfer function, 194–96 Proportional-integral-derivative control, ideal transfer function, 197 Pulse and doublet testing, 415–18 Pulse function as approximation to unit impulse, 54 response of liquid-level system to, 102–3 Pulse transfer function, 54 Q Qualitative nature of solutions, 43–44 R O Offset, definition, 4, 198 On-off control, 3, 192, 194 of stirred-tank heater, 561–69 Open-loop transfer function, 219 Overall transfer function, from block diagram, 219–24 for positive feedback system, 223 Overdamped response, 141 Overshoot, 147 P Padè approximation to transport lag, 154–55 Partial fractions, 32–42 Pendulum, 543–46 phase plane of, 546 cou9789x_ndx_599-604.indd 601 Ramp function, 78 Ratio control, 370–73 Regulator problem, 167–68 Relay in on-off control, 561–62 Reset windup, 589–92 Resistance, 99 linear, 99 Resonance, 311 Resonant peak, 310–12 Response time, 148–49 Rise time, 148–49 RLOCUS tool MATLAB, 276 Root locus, 269–284 comparison with frequency response, 264 concept, 269 plotting of diagrams with MATLAB, 273 1/1/70 2:26:25 PM Confirming Pages 602 INDEX Roots of equation, MATLAB, 261, 271 Routh test for stability, 258–60 S Saddle point, 557–58 Second order plus dead time model, 408–18 Second-order system, 137–53 simulation with Simulink, 144–47 dynamic parameters τ and ζ, 140–41 impulse response, 150–51 sinusoidal response, 151–53 step response, 141–47 transfer function, 140 Self-tuner, 586 Sensitivity, controller, 179 Servomechanism problem, 167–68 Set point, definition, 1–2 Simulink model of first order system, 83–84 Single input-single output system (SISO), 274, 512 Sinusoidal response with MATLAB, 90–91 SISO tool MATLAB, 274 SOPDT, 408–18 Spring-mass-damper system, 538–43 phase plane of, 541–42 Stability, 252–68 Bode criterion, 326–27 definition, 254–56 in multivariable systems, 525 in nonlinear systems, 560 Routh test, 258–59 Stability of typical roots in characteristic equation, 257 State of system, definition, 553 State variable, 477–78 selection and types, 482–83 State-space methods, 477–529 transfer function matrix, 502–3 transition matrix, 499–500 Steady-state gain, 76 Step function, 19, 78 Step testing, 411–15 Stirred Tank Heater examples, 11–17, 106–9, 231, 235, 238, 242, 243, 245, 296–301, 324 29, 561–69 block diagram for control of, 165–182 closed-loop response of, 235–46 on-off control, 561–69 Substitution rule in frequency response, 287–89 Superposition, 77 Sutro weir, 99 cou9789x_ndx_599-604.indd 602 T Taylor-series expansion, 110, 122, 154–55, 446 Thermometer dynamics, 71–74 Time constant, 74, 80 Tracking in controller and valve, 588–89 Trajectory, definition of, 553 Transducer, 188 Transfer function, 71, 74–77 for distributed-parameter systems, 461, 468 simulation using MATLAB and Simulink, 81–84 Transfer function matrix, 502–3 Transfer lag, 126 Transition matrix, 499–500 Translation of function, 52–54 of transform, 52 Transportation lag simulation with MATLAB and Simulink, 332, 395 as a distributed parameter system, 463–64 Padè approximation, 154–5 transfer function, 154 Tuning rules, 391–410 U Ultimate periodic response, 89 Underdamped response, 141–42 Unity feedback, 229 V Valve, control, 423–42 Cv, 425 characteristics, 427–38 construction, 424–25 equal percentage, 428–30 hysteresis, 438 linear, 428 positioner, 438–39 sizing, 425–26 transfer function, 209–10 Vector, column and row, 490 W Weir, 99 Z Zeros and poles, 270, 275 Ziegler-Nichols settings, 394–97 1/1/70 2:26:25 PM cou9789x_ndx_599-604.indd 603 1/1/70 2:26:25 PM cou9789x_ndx_599-604.indd 604 1/1/70 2:26:25 PM Revised Pages Useful Functions and Laplace Transforms TABLE 2.1 Function Graph Transform 1 u(t) 1 s tu(t) 1 s2 tnu(t) n! s n 1 1 at e cou9789x_fc.indd 1 u(t) 1 s a tneatu(t) n! s a n 1 sin kt u(t) k s2 k 2 5/21/08 6:44:06 PM Revised Pages TABLE 2.1 (Continued) Function Graph Transform cos kt u(t) s s2 k 2 sinh kt u(t) k s2 k 2 cosh kt u(t) s s2 k 2 1 eat sin kt u(t) k (s a)2 k 2 eat cos kt u(t) s a (s a)2 k 2 Area = 1 (t), unit impulse cou9789x_fc.indd 2 1 5/21/08 6:44:09 PM Revised Pages Key Features of Standard Responses of First Order Systems to Common Inputs Step Response of First Order System Ultimate Value Impulse Response of a First Order System 1 1 0.9 0.9 0.8 0.8 0.7 Y*τ/Kp 0.6 Y/Kp Initial “Jump” is to Kp/τ 0.7 Response is 63.2% complete at t=τ 0.5 0.4 0.3 0.6 0.5 Response is 63.2% complete at t=τ 0.4 (Initial “Jump” has decayed to 36.8%) 0.3 0.2 0.2 Initial Slope intersects Ultimate Value at t=τ 0.1 0.1 0 0 1 2 3 0 4 0 1 2 t/τ t/τ Sinusoidal Response of a First Order System 4 Response of First Order System to Ramp Input 1.5 5 phase lag = arctan(−wτ) 1 After an initial transient period, the response is parallel with input. 4.5 ratio=1/√1+(wτ) 2 4 Steady state difference between input and output (after transient) is bτ 3.5 Y/bKp 0.5 Y/AKp 3 0 3 2.5 2 −0.5 1 −1 −1.5 After an initial transient period, the response is periodic with the same frequency 5 10 15 t/τ cou9789x_fc.indd 3 bτ 1.5 input output 0.5 20 0 output lags input by τ 0 1 2 3 4 5 t/τ 5/21/08 6:44:10 PM Revised Pages 2 Standard Form for a Second Order System: d 2Y dY 2 Y X (t ) dt dt 2 Time Constant Damping Coefficient, the magnitude of this parameter determines the nature of the response Sample Second Order System Response to a Unit Impulse Input Sample Second Order System Response to a Unit Step Input 0.7 1.4 underdamped ζ<1 1.2 0.6 1 Output 0.8 0.6 critically damped ζ=1 0.4 critically damped ζ=1 Output underdamped ζ<1 0.5 Input Step overdamped ζ>1 0.3 0.2 overdamped ζ>1 0.1 0.4 0 0.2 −0.1 −0.2 0 0 1 2 3 4 5 6 Time 7 8 9 10 0 2 4 6 8 10 12 Time 14 16 18 20 Terms to Describe an Underdamped (Oscillatory) Second Order Response ) ( Overshoot exp / 1 2 A B ) ( Decay ratio exp 2 / 1 2 (overshoot )2 , radian frequency C A 1 2 n natural frequency 1 Period, T A C Response time limit 1.0 Y (t ) T B 0 cou9789x_fc.indd 4 0 tr Rise time t Response time 5/21/08 6:44:11 PM McGraw-HillÕs CHEMICAL ENGINEERING SERIES Key Features: control classesÉthat this is just another mathematics course disguised as an engineering course ¨ ¨ and Excel¨ have been introduced throughout the book. dynamics and control and not get bogged down in the mathematical complexities of each problem The Solutions to the End-of-Chapter Problems are available to Instructors at the textÕs website: www.mhhe.com/coughanowr-leblanc Electronic Textbook Options This text is offered through CourseSmart for both instructors and students. CourseSmart is an online browser where students can purchase access to this and other McGraw-Hill textbooks in a digital Process Systems Analysis and Control Third Edition half the cost of a traditional text. Purchasing the eTextbook also allows students to take advantage of sales representative or visit www.CourseSmart.com. ISBN 978-0-07-339789-4 MHID 0-07-339789-X www.mhhe.com Third Edition Donald R. Coughanowr Steven E. LeBlanc MD DALIM 976649 7/29/08 CYAN MAG YELO BLACK available for the course material Coughanowr LeBlanc Process Systems Analysis and Control Process Systems Analysis and Control, Third Edition retains the clarity of presentation for which this book is well known. It is an ideal teaching and learning tool for a semester-long undergraduate chemical engineering course in process dynamics and control. It avoids the encyclopedic approach of many other texts on this topic. Computer examples using MATLAB¨ and Simulink¨ have been introduced throughout the book to supplement and enhance standard hand-solved examples. These packages allow the easy construction of block diagrams and quick analysis of control concepts to enable the student to explore Òwhat-ifÓ type problems that would be much more difficult and time consuming by hand. New homework problems have been added to each chapter. The new problems are a mixture of hand-solutions and computational-exercises. One-page capsule summaries have been added to the end of each chapter to help students review and study the most important concepts in each chapter.