FUZZY CONTROL AND MODELING

Analytical Foundations and Applications

Hao Ying
Department of Physiology and Biophysics
Biomedical Engineering Center
The University of Texas Medical Branch, Galveston

IEEE
PRESS

The Institute of Electrical and Electronics Engineers, Inc., New York

This book and other books may be purchased at a discount
from the publisher when ordered in bulk quantities. Contact:

IEEE Press Marketing

Attn: Special Sales
445 Hoes Lane
P.O. Box 1331
Piscataway, NJ 08855-1331
Fax: +1 732 981 9334

For more information about IEEE Press products, visit the

IEEE Online Catalog & Store: http:==www.ieee.org=org=ieeestore.

# 2000 by the Institute of Electrical and Electronics Engineers, Inc.,

3 Park Avenue, 17th Floor, New York, NY 10016-5997.

All rights reserved. No part of this book may be reproduced in any form,
nor may it be stored in a retrieval system or transmitted in any form,
without written permission from the publisher.

Printed in the United States of America.

10 9 8 7 6 5 4 3 2 1

ISBN 0-7803-3497-3
IEEE Order No. PC5729

Library of Congress Cataloging-in-Publication Data

Ying, Hao, 1958±
Fuzzy control and modeling : analytical foundations and applications = Hao Ying.
p. cm. - - (IEEE Press series on biomedical engineering)
Includes bibliographical references and index.
ISBN 0-7803-3497-3
1. Automatic control. 2. Fuzzy systems. 3. Mathematical models. I. Title. II. Series.
TJ211.Y56 2000
629.8- -dc21 00-022760
 IEEE Press
445 Hoes Lane, P.O. Box 1331
Piscataway, NJ 08855-1331

IEEE Press Editorial Board

Robert J. Herrick, Editor in Chief

M. Akay M. Eden M. S. Newman

J. B. Anderson M. E. El-Hawary M. Padgett
P. M. Anderson R. F. Hoyt W. D. Reeve
J. E. Brewer S. V. Kartalopoulos G. Zobrist
D. Kirk

Kenneth Moore, Director of IEEE Press

Catherine Faduska, Senior Acquisitions Editor
Linda C. Matarazzo, Associate Acquisitions Editor
Surendra Bhimani, Production Editor

IEEE Engineering in Medicine and Biology Society, Sponsor

EMB-S Liaison to IEEE Press, Metin Akay

Cover design: William T. Donnelly, WT Design

Technical Reviewers
Anna Bianchi, Polytechnic of Milan
Guanrong Chen, University of Houston
Oscar Gonzalez, Old Dominion University, Norfolk, VA
Reza Langari, Texas A & M University
Frank L. Lewis, University of Texas at Arlington
Liang Wang, CASA, Inc., Los Alamos, NM
John Yen, Texas A & M University

Books of Related Interest from the IEEE Press

ROBUST VISION FOR VISION-BASED CONTROL OF MOTION
Edited by Gregory D. Hager and Markus Vincze
A volume in the SPIE=IEEE Press Imaging Science & Engineering Series
2000 Hardcover 264 pp IEEE Order No. PC5403 ISBN 0-7803-5378-1
EVOLUTIONARY COMPUTATION: Toward a New Philosophy of Machine Intelligence
Second Edition
David B. Fogel
2000 Hardcover 296 pp IEEE Order No. PC5818 ISBN 0-7803-5379-X
UNDERSTANDING NEURAL NETWORKS AND FUZZY LOGIC: Basic Concepts and
Applications
Stamatios V. Kartalopoulos
A volume in the IEEE Press Understanding Science & Technology Series
1996 Softcover 232 pp IEEE Order No. PP5591 ISBN 0-7803-1128-0
THE CONTROL HANDBOOK
Edited by William S. Levine
A CRC Handbook published in cooperation with IEEE Press
1996 Hardcover 1566 pp IEEE Order No. PC5649 ISBN 0-8493-8570-9
 Contents

FOREWORD xvii
PREFACE xix
ACKNOWLEDGMENTS xxiii
LIST OF FIGURES xxv

CHAPTER 1 Basic Fuzzy Mathematics for Fuzzy Control and

Modeling 1
1.1. Introduction 1
1.2. Classical Sets, Fuzzy Sets, and Fuzzy Logic 1
1.2.1. Limitation of Classical Sets 1
1.2.2. Fuzzy Sets 1
1.2.3. Fuzzy Logic Operations 6
1.3. Fuzzi®cation 7
1.4. Fuzzy Rules 8
1.4.1. Mamdani Fuzzy Rules 8
1.4.2. TS Fuzzy Rules 9
1.5. Fuzzy Inference 10
1.6. Defuzzi®cation 11
1.6.1. Generalized Defuzzi®er 12
1.6.2. Centroid Defuzzi®er, Mean of Maximum
Defuzzi®er, and Linear Defuzzi®er 12
1.7. Summary 13
1.8. Notes and References 13
Exercises 13

vii
viii Contents

CHAPTER 2 Introduction to Fuzzy Control and Modeling 15

2.1. Introduction 15
2.2. Why Fuzzy Control 15
2.3. Conventional Modeling 16
2.4. Why Fuzzy Modeling 17
2.5. Two Types of Fuzzy Control and Modeling:
Mamdani Type and TS Type 17
2.6. Typical SISO Mamdani Fuzzy Controllers 17
2.6.1. Fuzzi®cation 18
2.6.2. Fuzzy Rules 21
2.6.3. Fuzzy Inference 23
2.6.4. Defuzzi®cation 24
2.7. Typical MISO Mamdani Fuzzy Controllers 26
2.8. Typical MISO and SISO TS Fuzzy Controllers 28
2.9. Relationship between Fuzzy Control and
Conventional Control 29
2.10. Fuzzy Control vs. Classical Control 30
2.10.1. Advantages of Fuzzy Control 30
2.10.2. Disadvantages of Fuzzy Control 30
2.11. When to Use Fuzzy Control 31
2.11.1. Two Criteria 31
2.11.2. Applicability of Fuzzy Control 32
2.11.3. When to Avoid Fuzzy Control 32
2.12. Analytical Issues in Fuzzy Control 33
2.12.1. Brief Background 33
2.12.2. Signi®cant, Inherent Dif®culties for
Analytical Study of Fuzzy Control 34
2.12.3. Analytical Issues 34
2.13. Fuzzy Modeling 35
2.13.1. Mamdani Fuzzy Model 35
2.13.2. TS Fuzzy Model 36
2.13.3. Relationship between Fuzzy Model
and Fuzzy Controller 36
2.14. Applicability and Limitation of Fuzzy
Modeling 36
2.15. Analytical Issues in Fuzzy Modeling 37
2.16. Summary 37
2.17. Notes and References 38
Exercises 38

CHAPTER 3 Mamdani Fuzzy PID Controllers 41

3.1. Introduction 41
3.2. PID Control 42
3.2.1. Position Form and Incremental Form 42
Contents ix

3.2.2. PI and PD Controllers and Their

Relationship 42
3.3. Different Types of Fuzzy Controllers 43
3.3.1. Linear Fuzzy Controller and Nonlinear
Fuzzy Controller 43
3.3.2. Fuzzy PID Controller, Fuzzy Controller
of PID Type, and Fuzzy Controller of
Non-PID Type 43
3.4. Fuzzy PI=PD Controllers as Linear PI=PD
Controllers 44
3.4.1. Fuzzy PI Controller Con®guration 44
3.4.2. Derivation and Resulting Structures 47
3.5. Fuzzy PI=PD Controllers as Piecewise
Linear PI=PD Controllers 47
3.6. Simplest Fuzzy PI Controller as Nonlinear Variable
Gain PI Controller 51
3.6.1. Derivation and Resulting Structure 51
3.6.2. Characteristics of Gain Variation 52
3.6.3. Performance Enhancement Due to Gain
Variation 53
3.7. Another Simplest Fuzzy PI Controller as Nonlinear
Variable Gain PI Controller 55
3.8. Simulation Comparison between Fuzzy and Linear
PI Controllers 56
3.8.1. System Models and Comparison Conditions 56
3.8.2. Comparison Results for the Linear Models 57
3.8.3. Comparison Results for the Time-Delay
Model and the Nonlinear Model 58
3.8.4. Superior Fuzzy Control Performance
at a Price 64
3.9. Simplest Fuzzy PI Controllers Using Different
Fuzzy Inference Methods 65
3.9.1. Con®gurations of Fuzzy PI Controllers 65
3.9.2. Derivation and Resulting Structures 66
3.9.3. Characteristics of Gain Variation 68
3.9.4. Performance Enhancement by Gain
Variation 70
3.9.5. Unreasonable Gain Variation
Characteristics Produced by the Bounded
Product Inference Method 70
3.9.6. Conclusion on Fuzzy Inference Methods
for Control 73
3.10. Simplest TITO Fuzzy PI Controller as TITO
Nonlinear Variable Gain PI Controller 73
3.10.1. Fuzzy Controller Con®guration 73
3.10.2. Derivation and Resulting Structure 75
3.11. Fuzzy PD Controllers 76
3.12. Fuzzy PID Controllers as Nonlinear PID Controllers
with Variable Gains 77
x Contents

3.13. Summary 78
3.14. Notes and References 78
Exercises 79

CHAPTER 4 Mamdani Fuzzy Controllers of Non-PID Type 81

4.1. Introduction 81
4.2. Multilevel Relay 81
4.3. Linear Fuzzy Rules and Nonlinear Fuzzy Rules 83
4.4. Fuzzy Controller with Linear Fuzzy Rules as Linear
Controller 84
4.4.1. Fuzzy Controller Con®guration 84
4.4.2. Structure Derivation and Explicit Results 86
4.5. Typical Fuzzy Controller with Linear Fuzzy Rules 87
4.5.1. Structure Derivation 88
4.5.2. Resulting Structure 89
4.5.3. Relationship with the Simplest Fuzzy PI
Controller 92
4.6. Fuzzy Controller Using Linear Fuzzy Rules and
Trapezoidal Output Fuzzy Sets 92
4.7. Fuzzy Controller Using Linear Fuzzy Rules and
Three Input Variables 94
4.8. Typical TITO Fuzzy Controller with Linear Fuzzy
Rules 96
4.8.1. Fuzzy Controller Con®guration 96
4.8.2. Derived Structure 97
4.9. Typical Fuzzy Controller with Nonlinear Fuzzy
Rules 98
4.9.1. Fuzzy Controller Con®guration 98
4.9.2. Derivation and Resulting Structure 100
4.9.3. Structure Decomposition and Duality 105
4.9.4. Gain Variation Characteristics 105
4.9.5. Direct Generation of Other Fuzzy
Controllers' Structures 106
4.10. Structure Decomposition of General Fuzzy
Controllers 109
4.10.1. Con®guration of General Fuzzy Controllers 109
4.10.2. Structure Decomposition Theorem 111
4.10.3. Structure of Global Controllers for Linear
Fuzzy Rules 112
4.11. Limit Structure of General Fuzzy Controllers 113
4.11.1. Degree of Nonlinearity for Fuzzy
Controllers with Linear Fuzzy Rules 113
4.11.2. Limit Structure for Fuzzy Controllers
with Linear Rules 114
4.11.3. Limit Structure for General Fuzzy
Controllers with Nonlinear Rules 115
Contents xi

4.12. Structure Decomposition and Limit Structure of

General Fuzzy Models 116
4.13. Summary 117
4.14. Notes and References 117
Exercises 117

CHAPTER 5 TS Fuzzy Controllers with Linear Rule Consequent 119

5.1. Introduction 119
5.2. Why Not Use Nonlinear Rule Consequent 119
5.3. General TS Fuzzy Controllers as Nonlinear Variable
Gain Controllers of PID Type 120
5.3.1. Con®guration of General Fuzzy Controllers 120
5.3.2. Analytical Structure as Nonlinear Controllers
of PID Type 120
5.3.3. General Fuzzy Controllers as Linear
Controllers 121
5.4. Simple TS Fuzzy PI=PD Controllers as Nonlinear
Variable Gain PI=PD Controllers 121
5.4.1. Con®guration of Fuzzy Controller 121
5.4.2. Derivation and Resulting Structure 122
5.4.3. General Characteristics of Variable Gains 124
5.4.4. Three Speci®c Types of Gain Variation
Characteristics 127
5.4.5. Performance Improvement Due to Variable
Gains 128
5.4.6. Design of Gain Variation Characteristics 132
5.4.7. Simulated Control of Tissue Temperature
in Hyperthermia 133
5.5. Typical TS Fuzzy PI=PD Controllers as Nonlinear
Variable Gain PI=PD Controllers 135
5.5.1. Fuzzy Controller Con®guration 135
5.5.2. Derivation and Resulting Structure 137
5.5.3. Analysis of Gain Variation Characteristics 141
5.5.4. Relationship with the Simple TS Fuzzy
Controller 143
5.5.5. Simulated Control of Tissue Temperature in
Hyperthermia 143
5.6. Simpli®ed TS Fuzzy Rule Scheme 145
5.6.1. Disadvantages of TS Fuzzy Rule Scheme 145
5.6.2. Simpli®ed Linear TS Fuzzy Rule Scheme 147
5.6.3. Parameter Reduction as Compared with
Original TS Rule Scheme 148
5.6.4. Simpli®ed Nonlinear TS Fuzzy Rule
Scheme 149
5.6.5. General Analytical Structure of Fuzzy
Controllers with Simpli®ed TS Fuzzy Rules 149
xii Contents

5.7. Simple Fuzzy PI=PD Controllers with Simpli®ed

Linear TS Fuzzy Rule Consequent 150
5.7.1. Con®guration and Explicit Structure
Derivation 150
5.7.2. Gain Variation Characteristics and Their Effect
on Enhancing Control Performance 152
5.7.3. Attaining Desired Gain Variation
Characteristics 154
5.7.4. Other Simple Fuzzy PI=PD Controllers
with Simpli®ed Linear TS Fuzzy Rule
Consequent 155
5.8. Fuzzy PID Controller with Simpli®ed Linear TS
Rule Scheme 157
5.8.1. Con®guration and Explicit Structure
Derivation 157
5.8.2. Simulated Control of Mean Arterial
Pressure 158
5.9. Comparing TS Fuzzy Control with Mamdani Fuzzy
Control 160
5.9.1. Major Features of Mamdani Fuzzy Control 160
5.9.2. Primary Characteristics of TS Fuzzy Control 162
5.9.3. Comparison Conclusions 162
5.10. Summary 163
5.11. Notes and References 163
Exercises 163

CHAPTER 6 Stability Analysis and Design of Mamdani and TS Fuzzy

Control Systems 165
6.1. Introduction 165
6.2. Global Stability, Local Stability, and BIBO
Stability 166
6.2.1. Why Study Local Stability Instead of
Global Stability 166
6.3. Local Stability of Mamdani and TS Fuzzy PID Control
Systems 167
6.3.1. Local Stability Determined by Lyapunov's
Linearization Method 167
6.3.2. System Linearizability Criterion 168
6.4. Local Stability of Mamdani Fuzzy Control Systems
of Non-PID Type 169
6.5. Local Stability of General TS Fuzzy Control
Systems 170
6.5.1. Theoretical Development 170
6.5.2. Numeric Example 171
6.6. BIBO Stability of Mamdani Fuzzy PI=PD Control
Systems 173
Contents xiii

6.6.1. Small Gain Theorem 173

6.6.2. BIBO Stability Conditions 175
6.6.3. Numeric Example 176
6.7. BIBO Stability of TS Fuzzy PI=PD Control
Systems 177
6.7.1. Theoretical Derivation 177
6.7.2. Numeric Example 179
6.8. Design of Mamdani Fuzzy Control Systems 180
6.8.1. Design Principle 180
6.8.2. Justi®cations for Design Principle 181
6.8.3. Design Procedure 182
6.8.4. Design Example 184
6.8.5. System Tuning Guidelines 186
6.8.6. Examples of Designing More Complicated
Fuzzy Control Systems 187
6.9. Design of General TS Fuzzy Control Systems 191
6.9.1. Design Technique 191
6.9.2. Design Examples 192
6.10. General TS Fuzzy Dynamic Systems as Nonlinear ARX
Systems 193
6.11. General TS Fuzzy Filters as Nonlinear FIR=IIR
Filters 195
6.12. Local Stability of General TS Fuzzy Models 195
6.12.1. Local Stability Conditions and Their
Use for Model Quality Check 195
6.12.2. Numeric Example 197
6.13. Design of Perfect Tracking Controllers for General
TS Fuzzy Models 199
6.13.1. Controller Design via Feedback
Linearization Method 199
6.13.2. Stability of Designed Controllers 200
6.13.3. Numeric Examples 202
6.14. Summary 206
6.15. Notes and References 207
Exercises 207

CHAPTER 7 Mamdani and TS Fuzzy Systems as Functional

Approximators 209
7.1. Introduction 209
7.2. Fuzzy Controller and Fuzzy Model as Functional
Approximators 209
7.3. Polynomial Approximation of Continuous
Functions 210
7.4. Suf®cient Approximation Conditions for General
MISO Mamdani Fuzzy Systems 211
xiv Contents

7.4.1. Formulation of General Fuzzy Systems 211

7.4.2. Statement of Approximation Problems 211
7.4.3. Uniform Approximation of Polynomials
by General Fuzzy Systems 211
7.4.4. General Fuzzy Systems as Universal
Approximators 214
7.4.5. Suf®cient Approximation Conditions 215
7.4.6. Numeric Examples 217
7.5. Suf®cient Approximation Conditions for General
MISO TS Fuzzy Systems 219
7.5.1. Suf®cient Approximation Conditions 219
7.5.2. Numeric Example 223
7.6. Necessary Approximation Conditions for General
SISO Mamdani Fuzzy Systems 224
7.6.1. Problem Statement and Assumptions 224
7.6.2. Con®guration of General Fuzzy Systems 225
7.6.3. Lemmas for Establishing Necessary
Conditions 226
7.6.4. Necessary Approximation Conditions 227
7.6.5. Strength and Limitation of SISO Mamdani
Fuzzy Systems as Functional
Approximators 229
7.7. Necessary Approximation Conditions for General
MISO Mamdani Fuzzy Systems 230
7.7.1. Con®guration of General Fuzzy Systems 230
7.7.2. Lemmas for Developing Necessary
Conditions 231
7.7.3. Necessary Approximation Conditions 236
7.7.4. Merits and Pitfalls of MISO Mamdani
Fuzzy Systems as Functional Approximators 238
7.7.5. Numeric Example 238
7.8. Necessary Approximation Conditions for Typical
TS Fuzzy Systems 240
7.8.1. Con®guration of Typical Fuzzy Systems 240
7.8.2. Preparation for Setting Up Necessary
Conditions 242
7.8.3. Necessary Approximation Conditions 247
7.8.4. Advantages and Disadvantages of TS Fuzzy
Systems as Functional Approximators 248
7.9. Comparison of Minimal Approximator Con®guration
Between Mamdani and TS Fuzzy Systems 248
7.9.1. If TS Fuzzy Systems Use Trapezoidal or
Triangular Input Fuzzy Sets 248
7.9.2. If TS Fuzzy Systems Use Other Types
of Input Fuzzy Sets 250
7.9.3. Comparison Results 251
7.10. Conclusions on Mamdani and TS Fuzzy Systems
as Functional Approximators 251
7.11. Summary 251
Contents xv

7.12. Notes and References 252

Exercises 252

CHAPTER 8 Real-Time Fuzzy Control of Biomedical Systems 255

8.1. Introduction 255
8.2. Common Complexity of Biomedical Systems Ideal
for Fuzzy Control 255
8.3. Mamdani Fuzzy PI Control of Mean Arterial Pressure
in Postsurgical Cardiac Patients 256
8.3.1. Hypertension After Cardiac Surgery 256
8.3.2. Patient Model 257
8.3.3. Design of Fuzzy Control Drug Delivery System 257
8.3.4. Clinical Implementation and Fine-Tuning
of Fuzzy Controller 259
8.3.5. Clinical Results 262
8.4. Thermal Treatment of Tissue Lesions 265
8.4.1. Different Kinds of Thermal Therapies 265
8.4.2. Statement of Problems 266
8.4.3. Laser Thermal Therapies 267
8.5. Fuzzy PD Control of Tissue Temperature During
Laser Heating 268
8.5.1. Experimental Setup 268
8.5.2. Design of Mamdani Fuzzy PD Controller 269
8.5.3. Derivation of Fuzzy Controller Structure and
Explicit Results 271
8.5.4. Temperature Control Performance for Laser
Hyperthermia, Coagulation, and Welding 277
8.6. Ultrasound-Guided Fuzzy PD Control of Laser-Tissue
Coagulation 280
8.6.1. Development of Noninvasive Ultrasonic
Sensor 281
8.6.2. Setup for Ultrasound-Guided Fuzzy
Control Experiments 285
8.6.3. Design of the Mamdani Fuzzy PD Controller 286
8.6.4. Control Results of Laser-Tissue Coagulation 287
8.7. Summary 290
8.8. Notes and References 290
BIBLIOGRAPHY 291
INDEX 305
ABOUT THE AUTHOR 309
 Foreword

Close to a quarter of a century has passed since fuzzy control made its debut. During this
period, the literature on fuzzy control and its applications has grown at a geometrical rate.
Taking as an index of growth, the number of papers in the INSPEC database with ``fuzzy
control'' in the title have grown from 38 during 1970±1979 to 214 during 1980±1989 and to
4,356 during 1990±1999, with the data for 1999 not yet complete. And yet fuzzy control has
been, and remains, an object of controversy with someÐespecially within the academic
control systems establishmentÐexpressing the view that anything that can be done with fuzzy
control can be done equally well with conventional methods. This view is re¯ected in the fact
that almost no papers on fuzzy control have been published in the IEEE Transactions on
Automatic Control and this is the backdrop against which the publication of Professor Ying's
monumental work Fuzzy Control and Modeling: Analytical Foundations and Applications
should be viewed.
First, a bit of history. When I wrote my ®rst paper on fuzzy sets in 1965, my expectation
was that the theory of fuzzy sets would ®nd its main applications in the realm of what may be
called humanistic systemsÐsystems exempli®ed by economic systems, societal systems,
biological systems, linguistics, and psychology. It did not take me long, however, to see that
the theory could be applied to mechanistic systems, especially to control. The groundwork for
such applications was laid in my papers ``Toward a Theory of Fuzzy Systems'' (1971); ``A
Rationale for Fuzzy Control'' (1972); ``Outline of a New Approach to the Analysis of
Complex Systems and Decision Processes'' (1973); and ``On the Analysis of Large Scale
Systems'' (1974). These papers, especially my 1973 paper, introduced the basic concepts of a
linguistic variable, fuzzy if-then rule, and fuzzy graph. These concepts have played, and are
continuing to play, key roles in almost all applications of fuzzy set theory (or fuzzy logic),
including fuzzy control.
Although the basic ideas underlying fuzzy control were introduced in these papers, it
was the seminal work of Mamdani and Assilian in 1974±1975, which showed that the ideas
could be used to construct a working model of a fuzzy control system. This was the beginning
of the era of fuzzy control.

xvii
xviii Foreword

What is not fully recognized, however, is that fuzzy control (FC) and conventional crisp
control (CC) are, for the most part, complementary rather than competitive. Thus, FC is rule-
based whereas CC is differential-equation-based; FC is task-oriented whereas CC is set-point-
oriented; and CC is model-based whereas, in the case of FC, what suf®ces is a linguistic, rule-
based description of the model. Today we see more clearly that fundamentally CC is
measurement-based whereas FC is perception-based. In this sense, the role model for FC
is the remarkable human capability to perform a wide variety of tasks without any
measurements and any computations. A canonical example of such tasks is that of driving
a car in city traf®c. Classical control provides no methods for automation of tasks of this type.
Because it is close to human intuition, fuzzy control is easy to learn and easy to apply.
For this reason, there are many applications of fuzzy control in the realm of consumer
products. However, as a system becomes more complex, a limited mastery of fuzzy rule-based
techniques ceases to be adequate. This is when a deep understanding of the theory of fuzzy
control becomes a necessity, and it is this necessity that motivates the analytical theory of
fuzzy control developed in the work of Professor Ying.
Professor Ying's book contains much that is new, important, and detailed. Particularly
noteworthy are the chapters that focus on the Mamdani and Takagi-Sugeno types of
controllers. In these chapters, a novel approach to stability theory is described and a theory
of universal approximation is developed in detail. His linkage of basic theory to real-world
applications is very impressive.
The last chapter in the book deals with a subject in which Professor Ying is a foremost
authority, namely, application of fuzzy control to biomedical systems. Such applications are
likely to grow in importance in the years ahead.
Professor Ying's work should go a long way toward countering the view that fuzzy
control is a collection of applications without a solid theory. The deep theory of fuzzy control
developed by Professor Ying is of great importance both as a theory and as a foundation for
major advances in applications of fuzzy control in industry, biomedicine, and other ®elds. As
the author of Fuzzy Control and Modeling: Analytical Foundations and Applications, he and
the publisher, the IEEE Press, deserve our thanks and congratulations.

Lot® A. Zadeh
Berkeley, CA
February 27, 2000
 Preface

In the past decade, fuzzy system technology Ð especially fuzzy control which is its most
active and victorious component Ð has gained tremendous acceptance in academia and
industry. The worldwide success of countless commercial products and applications has
proved the technology to be not only practical and powerful, but also cost effective. Real-
world systems are nonlinear; accurate modeling is dif®cult, costly, and even impossible in
most cases. Fuzzy control has the unique ability to successfully accomplish control tasks
without knowing the mathematical model of the system, even if it is nonlinear and complex.
Applications are currently being developed in an ad hoc manner requiring signi®cant
trial-and-error effort, however. The fuzzy systems developed are mostly treated as (magic)
black boxes with little analytical understanding and explanation. Thus, there is an urgent need
for developing an analytical theory of fuzzy systems to support and accelerate the growth of
the technology and eliminate the existing misunderstanding and controversy.
The overall objective of this book is to establish comprehensive and uni®ed analytical
foundations for fuzzy control and modeling. My approach is ®rst to establish explicit
relationships between fuzzy controllers/models and their classical counterparts, and then to
utilize the well-developed conventional linear and nonlinear system techniques for analytical
analysis and design of fuzzy systems. The results are uni®ed in an analytical framework and
presented cohesively.

UNIQUENESS OF THE BOOK

This is a unique textbook whose contents are unavailable in any other book. It is the only
book at present that exclusively addresses analytical issues of fuzzy control and modeling by
rigorously connecting fuzzy controllers=models to classical controllers/models. In compar-
ison with other books, the text is unique in the following aspects:

xix
xx Preface

Completely Analytical Approach From the beginning to the end, every topic in the
book is treated analytically. Every fuzzy controller and fuzzy model is studied
mathematically using analytical techniques. All the results are analytical and rigorous.
Connection to Classical Control and System Theories Every result derived and
technique developed is rigorously linked to conventional control and system theories.
The connection is clearly presented and explained in the context of the conventional
theories.
Comprehensive and In-Depth Coverage Major types of fuzzy controllers and models
are covered. For every type, typical con®gurations are systematically examined. The
focus is not only on individual fuzzy systems, but also on their general classes. Many
analytical issues in the analysis and design of fuzzy systems are extensively studied.
Practicality To make the theory practically useful, the system model is assumed to be
unknown throughout the book, except for a portion of the stability analysis. My
approach is to concentrate on fuzzy controllers, as opposed to fuzzy control systems,
and to relate their analytical structures and gain characteristics to their control behavior
in such a way that the linkage holds for general systems. This approach is effective as
evidenced by applications, including a life-critical real-time control application.
Uni®ed and Cohesive Presentation A wide variety of fuzzy controllers are uni®ed in
one analytical framework, which also uni®es the fuzzy models. The common frame-
work makes the presentation consistent and cohesive.
Latest and Long-Term Research Findings Presented in a Textbook Style The book
contents are based on my fuzzy system publications since 1987, including 37 peer-
reviewed journal papers. A signi®cant amount of introductory and background
materials have been added. The materials from the papers are logically integrated
and organized as well as systematically enhanced. Coupled with the above-mentioned
unique features, this book is a self-contained textbook that provides up-to-date
information on some of the most active and fruitful frontiers of analytical research
and development of fuzzy systems.

INTENDED READERSHIP OF THE BOOK

This self-contained textbook is intended for anyone seeking to understand fuzzy control and
modeling in the context of traditional control and modeling. It is also for anyone who is
interested in analytical aspects of fuzzy control and modeling and wants to know precisely
their connections with the classical counterparts.
The book is written for readers who possess a basic knowledge of control and
modeling. Fuzzy mathematics is not a prerequisite nor is highly advanced mathematics;
undergraduate calculus suf®ces. To facilitate the reading and understanding, I provide a brief
introduction or review for every major classical concept, algorithm, and technique before it is
used. Analytical derivation is presented step by step, complete, and easy to follow. Concrete
numeric examples and computer simulation are provided to highlight or con®rm the analytical
work. Graphical representation, including three-dimensional plots, is extensively utilized to
illustrate the theoretical development.
The book can be used as a textbook for engineering senior and graduate students. Since
the book presents state-of-the-art analytical research, a particular topic may be selected for a
research project leading to a Ph.D., M.S., or senior graduation thesis. The text can also be
Preface xxi

used in conjunction with other books as a supplement or reference book to provide analytical
insight and in-depth knowledge on the relevant topics.
This is also a self-study book ideal for engineering professionals in diverse technical
®elds and industries, especially those in the ®elds of control and modeling. The book helps
those people take advantage of their expertise in conventional techniques when using fuzzy
system technology to solve particular problems. It provides a familiar entrance into the ®eld
of fuzzy systems. Given that the analytical theory of fuzzy control and modeling is still in its
early development stage, combining one's expertise with the book could be fruitful in
generating signi®cant theoretical and practical results.

OVERVIEW OF THE BOOK

The book consists of eight chapters that can be divided into four parts. The ®rst part is two
chapters long and contains background material for the rest of the book. Chapter 1 provides a
minimum but adequate amount of fuzzy mathematics and notations for fuzzy control and
modeling. In Chapter 2, both Mamdani and Takagi-Sugeno (TS) fuzzy controllers and models
are introduced. They are mathematically formulated to demonstrate that fuzzy controllers and
models are actually conventional nonlinear variable gain controllers and models, respectively.
The advantages and disadvantages of fuzzy control are pointed out in comparison with
conventional control. This is followed by a discussion of when fuzzy control should and
should not be used, the major analytical issues, and the special technical dif®culties associated
with fuzzy control study. Brief background information is also included regarding the early
attempts in analytical studies. Finally, fuzzy modeling and its strengths, weaknesses, and
usability are discussed.
Chapters 3 to 5 form the second part, which focuses on the analytical structures and
characteristics of a variety of fuzzy controllers and their precise connections to the
conventional controllers. Chapter 3 concentrates on different types of Mamdani fuzzy PID
controllers and compares their performances with the linear PID controller in computer
simulation using various system models. Chapter 4 deals with more complicated and general
classes of Mamdani fuzzy controllers that are of the non-PID type. Structure decomposition
property as well as limit structure are revealed for the fuzzy controllers when the number of
fuzzy rules becomes in®nitely large. Different TS fuzzy controllers with linear rule
consequent are investigated in Chapter 5. A new, simpli®ed TS fuzzy rule scheme is
introduced to reduce the number of design parameters.
The third part, which contains Chapters 6 and 7, provides analytical analysis and design
of different types of fuzzy systems. In Chapter 6, local stability as well as bounded-input
bounded-output (BIBO) stability conditions are established for both Mamdani and TS fuzzy
control systems. System design techniques are developed, including a feedback linearization
scheme for controlling general TS fuzzy models to achieve perfect output tracking control.
Fuzzy systems are also related to nonlinear ARX models and nonlinear FIR=IIR ®lters.
Chapter 7 examines whether fuzzy systems are universal approximators and establishes the
suf®cient and necessary approximation conditions. It also compares Mamdani fuzzy approx-
imators with TS fuzzy approximators in terms of minimal system con®guration and draws
conclusions on the strengths and limitations of the fuzzy approximators as a whole.
Chapter 8, the last part, shows three real-world applications that we have developed
using the analytical work. In one application, a Mamdani fuzzy PI controller is designed,
tuned, and clinically implemented to control mean arterial pressure in real time in postsurgical
cardiac patients in the Cardiac Surgical Intensive Care Unit.
xxii Preface

The bibliography at the end of the book lists the publications cited in the text as well as
other relevant publications that are not cited. Signi®cant effort has been made to ensure the
inclusion of all the publications relevant to the topics and approaches of the book. Never-
theless, given the vast volume of the literature, it is inevitable that the bibliography is still
incomplete.

Hao Ying
Department of Physiology and Biophysics
Biomedical Engineering Center
The University of Texas Medical Branch, Galveston
 Acknowledgments

I extend my appreciation, ®rst to my parents, Meilang Ying and Yiying Zhang. Much of what
I have accomplished can be attributed to their nurturing and love and to their years as my role
models. Their continued encouragement, advice, and help have been invaluable. I am also
very appreciative of my wife Julia Cheng for her understanding and support throughout the
writing of this book for more than two years. I am deeply indebted to my little son Andrew Y.
Ying, who was not able to get my full attention during this time. I am pleased to be able to
devote more time now to my newborn daughter Alice C. Ying.
I wish to thank my Ph.D. advisor, Professor Louis C. Sheppard, for his support and
encouragement. I am also grateful to my master's advisor, Professor Shihuang Shao, for
bringing me into the ®eld of fuzzy control in 1981. My thanks also go to Professor Metin
Akay for inviting me to write this book and for giving me thoughtful advice, as well as to the
IEEE Press editors, Karen L. Hawkins, Linda Matarazzo, and Surendra Bhimani for their
assistance. I am appreciative of my former Ph.D. student Dr. Yongsheng Ding for drawing
part of the ®gures in the ®rst three chapters and compiling part of the bibliography. My
appreciation also extends to the anonymous reviewers for their constructive suggestions and
useful comments.
A signi®cant portion of the results in the book was achieved throught my research
projects partially funded by the Whitaker Foundation and the Texas Higher Education
Coordinating Board. I am very grateful for their support.

Hao Ying
Department of Physiology and Biophysics
Biomedical Engineering Center
The University of Texas Medical Branch, Galveston

xxiii