See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3207690

Advanced PID Control - [Book Review]

Article in IEEE Control Systems Magazine · March 2006

DOI: 10.1109/MCS.2006.1580160 · Source: IEEE Xplore

CITATIONS READS
13 13,222

1 author:

S. Dormido
National Distance Education University
663 PUBLICATIONS 9,955 CITATIONS

SEE PROFILE

All content following this page was uploaded by S. Dormido on 02 July 2014.

The user has requested enhancement of the downloaded file.

 Advanced The goal of [4] was to introduce autotuning and adaptation
PID Control to control practitioners. Autotuning is the capability to auto-
matically adjust the controller parameters on demand from an
BY KARL J. ÅSTRÖM AND operator or an external signal. Adaptation is the continuous
TORE HÄGGLUND adjustment of the parameters to enhance performance. At the
time when [4] was published, several commercial products
with these capabilities were beginning to appear on the mar-
Proportional-integral-derivative ket. Automatic Tuning of PID Controllers discusses many useful
(PID) control is by far the most ideas employed by control practitioners but which are rarely
widely used form of feedback mentioned in standard control texts.
control. Despite the extraordinary The scope of their second book [5] is considerably larger.
ISA, 2006, 461 pp., US$96 progress that has taken place in This book provides a self-contained, well-thought-out, and
ISBN 1-55617-942-1 control theory over the last accessible introduction to a wide range of topics and funda-
20 years, industrial engineers mental principles underlying PID control. The main goal
tend to view much of control theory as esoteric and of little was to provide the technical background for understanding
relevance to industrial problems. In the mid-1980s, the authors PID control in a comprehensive fashion. This goal was
of this text, Åström and Hägglund, turned their attention to achieved with a level of mathematics that is accessible to
the PID controller. At that time, the treatment of PID in most control practitioners.
textbooks was given only a few pages and was restricted to an Based on its table of contents, Advanced PID Control follows
idealized version of its operation. In the academic community, more or less the same structure as its predecessor. However,
there existed the belief that PID control was a research area there are major differences. Traditionally, PID tuning and
with little opportunity. It is fair to say that the authors’ work design have been based on ad hoc techniques that take advan-
changed this situation. Today, many introductory textbooks tage of the special structure of the PID controller. Consequent-
include at least one chapter related to PID control, and there is ly, PID control has not followed the control mainstream,
no doubt that PID control is an active area of research. where robust control theory has provided design methods
In the early 1980s, the tuning formulas proposed by Ziegler based on loop shaping. Advanced PID Control seeks to remedy
and Nichols 40 years earlier represented the state-of-the-art in this situation by applying these ideas to PID control.
industrial PID control. Modern attempts at developing auto-
matic PID tuning were model based, with the exception of CONTENTS OF THE BOOK
Foxboro´s EXACT controller. Introduced in 1984, this con- Chapter 1 provides a brief introduction to the core concepts
troller was based on heuristic logic developed using computer used throughout the book and an outline of the text’s con-
simulation studies [2]. tents. Chapter 2 introduces basic approaches for analyzing the
The relay-tuning method proposed by Åström and behavior of feedback-controlled processes. Time and frequen-
Hägglund in 1984 [3] had its roots in the classical method of cy representations are introduced as a prerequisite to under-
Ziegler and Nichols. That relay-tuning method calculates the standing the dynamics of closed-loop systems. The objective is
controller’s PID parameters using knowledge of only one point to derive simple models that describe the process sufficiently
on the open-loop Nyquist curve. The point of intersection of well in the frequency range that is critical to successful con-
the Nyquist curve with the negative real axis is determined by trol. Two dimensionless parameters, namely, the normalized
increasing the gain of a proportional controller until a sus- time delay and the gain ratio, are introduced to quantify the
tained oscillation of constant amplitude is obtained. A disad- difficulty of controlling a process.
vantage of this method is the fact that the user cannot control A comprehensive study of the PID algorithm is provided
the amplitude of the resulting oscillation. The ingenious idea of in Chapter 3. Taking as a starting point the textbook form of
Åström and Hägglund was the realization that this point can PID, several modifications are discussed that result in a more
be determined by replacing the proportional controller with a practical controller. Implementation issues that must be con-
relay. The relay induces a limit cycle when the loop is closed, sidered in practice are then introduced, including derivative
with the great advantage that the amplitude of the limit cycle filtering, antiwindup mechanisms, and the changing of set-
can be set by the amplitude of the relay characteristic. This point values. The chapter thoroughly examines these issues.
technique, which was patented by Åström and Hägglund, is Alternative PID structures, such as standard noninteracting,
incorporated in many commercial PID controllers. interacting, and parallel structures, as well as conversion from
Advanced PID Control is the most recent of a trilogy of PID one form to another, are treated. The chapter ends by consid-
books written by Åström and Hägglund over the past ering the questions of when PID control is usable and when
20 years. The previous books are Automatic Tuning of PID Con- more sophisticated control is advisable.
trollers [4] and PID Controllers: Theory, Design, and Tuning [5]. Chapter 4 discusses controller design in general. A clear
To understand the contribution of the most recent book, it is overview of the basic design principles is provided. The
helpful to reexamine the first two texts. design objectives are expressed as requirements on load

98 IEEE CONTROL SYSTEMS MAGAZINE » FEBRUARY 2006 1066-033X/06/$20.00©2006IEEE

disturbance response, measurement noise response, setpoint may not capture the robustness issues associated with a Smith
response, and robustness. From a practical point of view, an predictor controller, a new robustness metric—the delay mar-
interesting result for two degree-of-freedom controllers is that gin—is introduced in this chapter.
one can design for robustness and disturbance response inde- Chapter 9 discusses techniques for adaptation and auto-
pendently. This chapter also emphasizes one of the most matic tuning of PID controllers. Automatic tuning of PID con-
important objectives of this book, namely, to demonstrate that trollers can be realized by joining the methods for obtaining
PID control design can be treated using ideas from main- process dynamics presented in Chapter 2 with the methods
stream controller design. for calculating the PID parameters given in Chapters 6 and 7.
Chapter 5 focuses on feedforward design, which is an The authors present several adaptive approaches, including
effective strategy for reducing the effect of measurable distur- gain scheduling, automatic tuning, and continuous adapta-
bances. The ideal feedforward strategy is to use a close tion. The chapter ends with a description of commercial con-
approximation to the inverse process model. The chapter pre- trollers in which adaptive methods have been successfully
sents several alternatives to this approach. The use of a two- employed.
degree-of-freedom controller structure separates the design Before tuning the controller, it is important to conduct a
problems of load disturbance response and setpoint response. loop assessment of the system, including an examination of
Chapter 6 treats special cases of the general PID design sensors and actuators, signal ranges, nonlinearities, noise
problem presented in Chapter 4. Since the complexity of the levels, and disturbances. Chapter 10 provides an overview of
controller is restricted, two alternatives emerge: 1) simplify the methods for commissioning, supervising, and diagnosing
process models in such a way that the design method yields a control loops. The goal is to supervise the performance of the
PID controller or 2) design the controller using a complex system during operation so as to guarantee that specifica-
model and approximate it with a PID controller. This chapter tions are met.
seeks to find a balance between an historical overview and the Up to this point, the book has focused on control of a single
presentation of new methods. A critical review is given of sev- PID loop. Chapter 11 is devoted to the relevant issue of inter-
eral methods, including Ziegler-Nichols, rule-based empirical acting PID loops. Of particular importance, it is shown that
tuning, pole placement, lambda tuning, algebraic design, and controller parameters in one loop can act as a significant load
optimization methods. Strengths, weaknesses, and limitations disturbance to the dynamics of other loops. Basic measures of
of each method are summarized. All of these methods share interaction are presented, such as Bristol’s relative gain array,
the common property that robustness to process variations so as to determine whether the control problem can be solved
must be examined after the design is complete. This philoso- using simple loops. The problem of pairing inputs and out-
phy is not in agreement with the main thrust of recent control puts as well as a design method based on decoupling are pre-
theory, where robustness of the completed design is guaran- sented in this chapter.
teed a priori. The authors present a novel tuning method called Chapter 12 shows how complex control systems can be
MIGO (Ms constrained integral gain optimization), which can built from simple components such as PID controllers, linear
be considered as a translation of robust design principles to filters, gain schedules, and simple nonlinear functions. Vari-
PID design. MIGO seeks to maximize the integral gain (and ous control paradigms are presented, including re petitive
thereby minimize the integrated error) such that loop-gain fre- control, cascade control, midrange and split-range control,
quency response lies outside the classical Ms -circle. ratio control, and control with selectors. Neural and fuzzy
Chapter 7 develops simple tuning rules in the same spirit techniques in the context of PID control are also discussed to
as Ziegler and Nichols. The goal is to obtain rules that can be indicate how they can be interpreted both as rule-based con-
employed for both manual tuning and automatic tuning for a trol and as nonlinear control.
wide range of processes. These tuning rules, dubbed AMIGO Finally, Chapter 13 provides the reader with a discussion
(approximate MIGO), were developed by applying MIGO to a of some implementation issues related to PID control. The
large batch of representative processes and correlating the PID authors follow the historical development starting with pneu-
resulting parameters to simple features of the process dynam- matic and electronic implementation of analog controllers. A
ics. The authors identify key differences between processes detailed presentation of computer implementation issues,
that are delay dominated and those that are lag dominated. such as sampling, prefiltering, and discretization of the PID
The performance of a PI controller can be improved if pre- algorithm, is then provided. Operational aspects and human-
dictive capability is included. Possibilities other than deriva- machine interfaces, such as bumpless transfer at mode switches
tive action may offer improved performance. In this regard, and parameter changes, are also presented.
Chapter 8 examines useful alternatives to dealing with time
delays. Various model-predictive controllers are treated, INTENDED AUDIENCE
including the Dahlin-Higham controller and the minimum- Advanced PID Control is addressed to a broad audience, rang-
variance controller. The treatment of the Smith predictor pro- ing from control practitioners to academic researchers. As a
vides a new perspective on this technique. Since the classical result, the book has several differing objectives. The treatment
robustness metrics, namely, gain margin and phase margin, in the book, however, is well balanced and accessible.

100 IEEE CONTROL SYSTEMS MAGAZINE » FEBRUARY 2006

 It is clear that the authors have taken great care in the text’s [2] E.H. Bristol and T.W. Kraus, “Life with pattern adaptation,” in Proc.
American Control Conf., San Diego, CA, 1984.
presentation to make it accessible to a broad readership. I fore- [3] K.J. Åström and T. Hägglund, “Automatic tuning of simple regulators
see this text being used in several ways: with specifications on phase and amplitude margins,” Automatica, vol. 20,
» As a text for independent study by control practitioners. pp. 645–651, 1984.
[4] K.J. Åström and T. Hägglund, Automatic Tuning of PID Controllers.
The book provides an accessible introduction to the fun- Research Triangle Park, NC: Instrum. Soc. Amer.1988.
damental principles underlying PID control as well as a [5] K.J. Åström and T. Hägglund, PID Controllers: Theory, Design, and Tuning.
wide range of related topics. Research Triangle Park, NC: Instrum. Soc. Amer. 1995.
[6] C.-C. Yu, Autotuning of PID Controllers: Relay Feedback Approach. New
» As one of several texts in an introductory automatic York: Springer-Verlag, 1999.
control course at the graduate level. [7] A. Datta, M.T. Ho, and S.P. Bhattacharyya, Structure and Synthesis of PID
» As part of a training for individuals in industry. I have Controllers. New York: Springer-Verlag, 2000.
[8] W. Quing-Guo, H.C. Chieh, T.J. Hägglund, and K.K. Tan, Eds., Advances
used [5] in this fashion with great success. in PID Control. New York: Springer-Verlag, 2000.
» As a general reference. The text is quite comprehensive [9] J. Quevedo and T. Escobet, Ed., Digital Control 2000: Past, Present and
and would be an invaluable source for researchers and Future of PID Control (IFAC Proceedings Volumes). New York: Pergamon, 2000.
[10] L. Wang and W.R. Cluett, From Plant Data to Process Control: Ideas for
practitioners. Process Identification and PID Design. New York: Taylor & Francis Group, 2000.
[11] A. O’Dwyer, Handbook of PI and PID Controller Tuning Rules. London:
SUMMARY Imperial College Press, 2003.
[12] G.J. Silva, A. Datta and S.P. Bhattacharyya, PID Controllers for Time Delay
Åström and Hägglund´s book is a remarkably clear, accessi- Systems. Cambridge, MA: Birkhäuser, 2003.
ble, and up-to-date text. Several other books have appeared [13] F. Haugen, PID Control of Dynamic Systems. Trondheim: Tapir Forlag,
in the last few years on this subject, although with a more 2004.
[14] Y. Choi and W.K. Chung, PID Trajectory Tracking Control for Mechanical
limited scope [6]–[15]. While the present book has some over- Systems (Lecture Notes in Control and Information Sciences). New York:
lap with these texts in terms of its coverage, Advanced PID Springer-Verlag, 2004.
Control stands on its own as a complete work. The text pro- [15] M.A. Johnson and M.H. Moradi, Ed., PID Control: New Identification and
Design Methods. New York: Springer, 2005.
vides a rigorous yet accessible introduction with a unique
perspective. The strength of the book is its systematic
approach to structuring the PID control problem along the
lines of the major developments in control theory over the
past two decades. This book is a welcome addition to the
existing literature on PID control and will certainly become a
standard reference.
Sebastián Dormido
Essential
MAX PLUS AT WORK
REVIEWER INFORMATION Modeling and Analysis of Synchronized Systems:
Sebastián Dormido graduated from the Universidad Com- A Course on Max-Plus Algebra and Its Applications
plutense of Madrid in 1968 and received the Ph.D. degree in Bernd Heidergott, Geert Jan Olsder,
control engineering from the Universidad del Pais Vasco in
and Jacob van der Woude
1971. From 1968–1975, he was with the Department of Com-
puter Science and Automatic Control of the Universidad Max Plus at Work is the first textbook on max-plus
Complutense and Universidad del Pais Vasco. From algebra, providing a concise and self-contained
1975–1982, he was with the Facultad de Ciencias Físicas of the introduction to the topic.
Princeton Series in Applied Mathematics
Universidad Complutense de Madrid, and, in 1982, he joined
Cloth $50.00 0-691-11763-2
the Facultad de Ciencias of the Universidad Nacional de Edu-
cación a Distancia (UNED). In 1982, he became head of the
Department of Computer Science and Automatic Control. His GENERAL THEORY OF
research interests include process control, predictive control, ALGEBRAIC EQUATIONS
robust control, object-oriented languages for modeling and Etienne Bézout
simulation of hybrid systems, and control education with
special emphasis on remote and virtual labs. He has super- Translated from French by Eric Feron
vised 25 doctoral theses and coauthored more than 180 con- “Bézout’s theorem is central to algebraic geometry.
ference and journal papers. Since 2002, he has served as This book.... is a significant contribution to the field.”
president of the Spanish Association of Automatic Control —Jan Verschelde, University of Illinois at Chicago
(CEA), where he has promoted the relationship between Cloth $49.50 0-691-11432-3 Due April
academia and industry.

REFERENCES PRINCETON 800-777-4726

Read excerpts online
[1] J.G. Ziegler and N.B. Nichols, “Optimum settings for automatic con- University Press www.pup.princeton.edu
trollers,” Trans. ASME, vol. 64, pp. 759–768, 1942.

FEBRUARY 2006 « IEEE CONTROL SYSTEMS MAGAZINE 101

View publication stats