Académique Documents
Professionnel Documents
Culture Documents
Hao Ying
Department of Physiology and Biophysics
Biomedical Engineering Center
The University of Texas Medical Branch, Galveston
IEEE
PRESS
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.
10 9 8 7 6 5 4 3 2 1
ISBN 0-7803-3497-3
IEEE Order No. PC5729
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
FOREWORD xvii
PREFACE xix
ACKNOWLEDGMENTS xxiii
LIST OF FIGURES xxv
vii
viii Contents
3.13. Summary 78
3.14. Notes and References 78
Exercises 79
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.
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.
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.
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