Model Based PID Controller

Design for Process Industry

Mohammad Shamsuzzoha
“Model Based PID Controller Design for Process Industry” presents a
clear, multidimensional, representation of IMC-PID control for both students
and specialists working in the area of IMC based PID control. It mainly
focuses on the theory and application of the IMC based PID control.
This book includes several chapters which cover the broad range of the
tropics. The book mainly focused on the theory of IMC based approach for
the PID controller tuning.

i
 Preface

This book is devoted to PID controller’s theory and its application. The PID
controllers are probably the most widely used industrial controller in the
process industries. It remains an important control tool for three reasons: past
record of success, wide availability and simplicity in use. Its stability analysis
is extremely easy to carry out and the design trade-off between performance
and robustness is clearly understood.
It incorporates recent development of the IMC-PID control technology in the
industrial practice. The emphasis has been given to find the best possible
approach to get the simple and optimal solution of the industrial users. This
book includes several chapters which cover simple PID to PID with filter
controller tuning. The book focused on the model based controller tuning for
both simple PID and PID with filter control structure.
The book overviews in brief are: Chapter 1 presents a detail overview of
PID controllers tuning method that can be applied to stable, integrating, and
unstable plants. Chapter 2, Determine the optimum IMC filter which gives
the best performance of the resulting PID controller for several representative
process models. This filter is used to design a PID controller for disturbance
rejection and the corresponding analytical IMC-PID tuning rule. Provide the
guidelines for selection of closed-loop time constant  for several process

models to cover a wide range of   . Conduct a robust study by inserting a

perturbation uncertainty in each parameter simultaneously (for the worst case
model). Several illustrative examples are included to demonstrate the
superiority of the proposed tuning methods. Chapter 3, This study is mainly
focused for a simple and efficient method for the design of a PID filter

ii
controller with enhanced performance. A closed-loop time constant   
guideline is recommended for a wide range of time-delay/time-constant
ratios. A simulation study was performed to illustrate the superiority of the
proposed method for both nominal and perturbed processes. Chapter 4, An
analytical method for the design of a PID controller cascaded with a second-
order lead-lag filter is developed for various types of time-delay process. The
proposed PID●filter controller is designed based on the IMC-PID principle
and gives a better response than the conventional PID controllers for
obtaining the desired, closed-loop response. Several examples are provided
for comparing the results with the conventional PID controllers. In the
proposed PID●filter structure, the resulting control system becomes
equivalent to controlling a fast dynamic process by integral control, which
dramatically improves the performance. Some discussions with the 
guideline for particular robustness levels are also provided. Chapter 5, A
simple analytical method is proposed for the design of a PID·filter controller,
in order to achieve enhanced performance for first order unstable and
integrating delay processes. A closed-loop time constant,  , guideline was
recommended for a wide range of time-delay/time-constant ratios. A
simulation study was performed for both unstable and integrating delay
processes to show the superior performance of the proposed method for both
nominal and perturbed processes.
I would like to thank Professor Moonyong Lee, for providing me the
opportunity to work in this exciting field. His keen insights and clear
guidance gave me great encouragement to carry out this research.

Dr. Mohammad Shamsuzzoha

iii
Youngnam University, South Korea

Presently author is affiliated with ADNOC Refining Research

Center, ADNOC Refining Co., Post Box: 3593; Abu Dhabi, UAE

Present address of author:

Dr. Mohammad Shamsuzzoha,
Senior Process Simulation Engineer, ADNOC Refining Research
Center, ADNOC Refining Co., Post Box: 3593; Abu Dhabi, UAE
Email: smzoha@gmail.com

iv
 ABSTRACT

The PID controllers are probably the most widely used industrial controller
in the process industries. It remains an important control tool for three
reasons: past record of success, wide availability and simplicity in use. It is
well known that the IMC is a powerful framework for control system design
and implementation, and it has sound theoretical foundation. Its stability
analysis is extremely easy to carry out and the design trade-off between
performance and robustness is clearly understood. It has attracted the
attention of industrial users because there is only one user-defined tuning
parameter, which is directly related to the closed-loop time constant or
equivalently. The motivation behind present study is the lack of unified
framework in the literature for the IMC-PID controller design which gives
the enhanced and robust control performance.
Therefore, the present study is devoted to develop a unified framework for
IMC based PID controller design and analysis which is divided in several
sections.
In section one the focus is to develop the IMC-PID controller design for
improved disturbance rejection of time delayed processes. An optimal IMC
filter structure is proposed for the several representatives stable and unstable
process to design a PID controller that gives an improved disturbance
rejection response. The simulation studies of several process models show
that the proposed design method provides better disturbance rejection for lag
time dominant processes, when the various controllers are all tuned to have
the same degree of robustness according to the measure of maximum
sensitivity. A closed-loop time constant  guideline is proposed for several

process models to cover a wide range of   ratios. Section two is devoted

v
for PID filter controller design for first order time delay processes for
enhanced performance. An analytical tuning method for a PID controller
cascaded with a lead/lag filter is proposed for FOPDT processes based on the
IMC design principle. The simulation study shows that the proposed design
method provides better disturbance rejection than the conventional PID
design methods. A guideline of a single tuning parameter of closed-loop time
constant    is provided for several different robustness levels. Section three
contains analytical design of PID controller cascaded with a lead-lag filter for
time-delay processes. The proposed design method is based on to obtain a
desired, closed-loop response for setpoint tracking. The process dead time is
approximated by using the appropriate Pade expansion to convert the ideal
feedback controller to the PID●filter structure with little loss of accuracy. The
resulting PID●filter controller efficiently compensates for the dominant
process poles and zeros and drastically improves the closed-loop
performance. Section four contains the analytical design of enhanced
PID·filter controller for integrating and first order unstable processes with
time delay. An analytical design for a PID controller cascaded with a first
order lead/lag filter is proposed for integrating and first order unstable
processes with time delay. In the simulation study, the controllers were tuned
to have the same degree of robustness by measuring the maximum
sensitivity, Ms . For the selection of the closed-loop time constant,  , a
guideline is also provided over a broad range of time-delay/time-constant
ratios. The simulation results obtained for the suggested method were
compared with those obtained for other recently published design methods to
illustrate the superiority of the proposed method.

CONTENTS

vi
 1. Introduction .....................................................................................1
1.1. Motivation .......................................................................................1
1.2. Review of PID controller.................................................................5
1.2.1 Analytical techniques.............................................................5
1.2.2 IMC control strategy..............................................................5
1.3. Outline of the thesis.........................................................................6
1.4 References.........................................................................................9

2. IMC-PID Controller Design for Improved Disturbance Rejection

of Time Delayed Processes ………………………………………15

2.1. Introduction ....................................................................................15

2.2. IMC-PID approach for PID controller design................................18
2.3. IMC-PID tuning rules for typical process models..........................20
2.3.1 First-order plus dead time process (FOPDT)........................20
2.3.2 Delayed integrating process (DIP)........................................21
2.3.3 Second-order plus dead time process (SOPDT)...................24
2.3.4 First-order delayed integrating process (FODIP).................24
2.3.5 First-order delayed unstable process (FODUP)....................25
2.3.6 Second-order delayed unstable process (SODUP)...............26
2.4. Performance and robustness measure...............................................26
2.4.1 Integral error criteria.............................................................26
2.4.2 Overshoot..............................................................................26
2.4.3 Maximum sensitivity  Ms  to modeling error........................26
2.4.4 Total variation (TV)..............................................................27
2.4.5 Set-point and derivative weighting.......................................27
2.5. Simulation Results................................................................................28
2.5.1 Example 2.1. FOPDT............................................................28

vii
 2.5.2 Example 2.2. DIP (Distillation column model)....................33
2.5.3 Example 2.3. SOPDT............................................................38
2.5.4 Example 2.4. FODIP (Reboiler level model)........................43
2.5.5 Example 2.5. FODUP...........................................................47
2.5.6 Example 2.6. SODUP...........................................................51
2.6. Discussions...........................................................................................55
2.6.1 Effect of  on the tuning parameters..................................55

2.6.2  guideline for PID parameter tuning................................58

2.6.3 Beneficial range of the proposed method.............................61
2.6.4 Optimum filter structure for IMC-PID design ....................63
2.7. Conclusions...........................................................................................66
2.8. Literature cited......................................................................................67

3. An Enhanced Performance PID Filter Controller for First Order

Time Delay Processes………………………………………………..69

3.1. Introduction ..........................................................................................69

3.2. IMC controller design procedure..........................................................71
3.3. PID filter controller design for FOPDT Process...................................73
3.4. Robust stability.....................................................................................76
3.5. Simulation study...................................................................................79
3.5.1 Example 3.1: Lag time dominant process.............................80
3.5.2 Example 3.2: Equal lag time and dead time process............83
3.5.3 Example 3.3: Dead time dominant process...........................85
3.5.4 Example 3.4: Polymerization process...................................87
3.6. Conclusions...........................................................................................90
3.7. Literature cited......................................................................................91

viii
4. Analytical Design of PID Controller Cascaded with a Lead-Lag
Filter for Time-Delay Processes ……………………………………93

4.1. Introduction ..........................................................................................93

4.2. IMC-PID approach for PID●filter controller design.............................95
4.2.1 Tuning rule for first-order plus dead time process................96
4.2.2 Tuning rule for second-order plus dead time process...........97
4.2.3 Tuning rule for SOPDT process with overshoot response....99
4.2.4 SOPDT and FOPDT processes with inverse response.........99
4.2.5 First-order delayed integrating processes............................100
4.3. Simulation studies................................................................................101
4.3.1 Example 4.1. FOPDT process..............................................101
4.3.2 Example 4.2. underdamped SOPDT process.......................102
4.3.3 Example 4.3. Overdamped SOPDT process........................106
4.3.4 Example 4.4. SOPDT model with a strong lead term..........107
4.3.5 Example 4.5. SOPDT process with inverse response..........110
4.3.6 Example 4.6. First-order delayed integrating process.........110
4.4. Discussion............................................................................................114
4.5. Conclusions..........................................................................................118
4.6. References............................................................................................119

5. Analytical Design of Enhanced PID·Filter Controller for Integrating

and First Order Unstable Processes with Time Delay…………….121

5.1. Introduction .........................................................................................121

5.2. Design procedure.................................................................................124
5.3. Proposed tuning rule............................................................................126
5.3.1 First-order delay unstable process (FODUP).......................126
5.3.2 Delayed integrating process (DIP).......................................128

ix
5.4. Simulation studies................................................................................129
5.4.1 Performance and robustness measure..................................130
5.4.1.1 Integral error criteria................................................130
5.4.1.2 Overshoot.................................................................130
5.4.1.3 Maximum sensitivity  Ms  to modeling error...........130
5.4.1.4 Total variation (TV).................................................131
5.4.2 Example 5.1. Lag time dominant FODUP...........................131
5.4.3 Example 5.2. Dead time dominant FODUP.........................134
5.4.4 Example 5.3. DIP process....................................................137
5.4.5 Example 5.4. Distillation column model.............................140
5.4.6 Example 5.5. Comparison with modified Smith predictor.. 143
5.4.7 Closed-loop time constant  guideline..............................146
5.5. Conclusions..........................................................................................147
5.6. References............................................................................................148
6. Conclusions............................................................................................151

x
 List of Tables

Table 2.1. IMC-PID Controller Tuning Rules…………………………….23

Table 2. 2. PID Controller Setting for Example 2.1………………………30
Table 2. 3. Robust Analysis for Example 2.1 …………………………….31
Table 2. 4. PID Controller Setting for Example 2.2 ……………………...35
Table 2. 5. Robust Analysis for Example 2.2…………………………......35
Table 2. 6. PID Controller Setting for Example 2.3………………………41
Table 2. 7. Robust Analysis for Example 2.3…………………………......42
Table 2. 8. PID Controller Setting for Example 2.4………………………45
Table 2.9. Robust Analysis for Example 2.4 …………………………......45
Table 2. 10. PID Controller Setting for Example 2. 5…………………….49
Table 2. 11. Robust Analysis for Example 2.5……………………………50
Table 2. 12. PID Controller Setting for Example 2. 6……………………53
Table 2.13. Robust Analysis for Example 2. 6……………………………54
Table 3.1 PID controller parameters and performance matrix…………...82
Table 3.2 Robustness analysis for example 3.1………………………......82
Table 3.3 PID controller parameters and performance matrix………..…84
Table 3.4 PID controller parameters and performance matrix ………......86
Table 4.1.Design rules of the proposed PID●filter controller……………..98
Table 4.2. Controller parameters and resulting performance……………..104
Table 4.3. Controller parameters and resulting performance indices…….105
Table 4.4. Controller parameters and resulting performance indices…….108
Table 4.5. Controller parameters and resulting performance indices…….109
Table 4.6. Controller parameters and resulting performance indices…….112
Table 4.7. Controller parameters and resulting performance indices…….113

xi
Table 5.1. PID controller setting for Example 5.1………………………..132
Table 5.2. PID controller setting for Example 5.2………………………..137
Table 5.3. PID controller setting for Example 5.3. ………………………138
Table 5.4. PID controller setting for Example 5.4.. ………………………143

xii
 List of Figures

Figure 1.1 PID controller representations…………………………………2

Figure 2.1. Block diagram of IMC and classical feedback control………19
Figure 2.2. Simulation results for unit step disturbance …………………32
Figure 2.3. Simulation results for unit step set-point change……………..33
Figure 2.4. Simulation results for unit step disturbance…………… …….34
Figure 2.5. Simulation results for unit step set-point change……………..37
Figure 2.6. Simulation results for unit step disturbance…………………..39
Figure 2.7. Simulation results for unit step set-point change …………….40
Figure 2. 8. Simulation results for unit step disturbance………………….44
Figure 2.9. Simulation results for unit step set-point change …………….47
Figure 2. 10. Simulation results for unit step disturbance………………...48
Figure 2.11. Simulation results for unit step set-point change……………51
Figure 2.12. Simulation results for unit step disturbance…………………52
Figure 2.13. Simulation results for unit step set-point change……………55
Figure 2. 14. Proportional gain  K c  setting for different  ……………..56
Figure 2. 15. Integral time constant   I  setting for different  …………57
Figure 2. 16. Derivative time constant   D  setting for different  ……..58
Figure 2. 17.  guidelines for FOPDT…………………………………...59
Figure 2. 18.  guidelines for DIP……………………………………….59
Figure 2. 19.  guidelines for SOPDT……………………………………60
Figure 2. 20.  guidelines for FODIP……………………………………60
Figure 2. 21.  guidelines for FODUP…………………………………...61

xiii
Figure 2. 22. Performance of the proposed filter vs. the conventional filter..62
Figure 2. 23. Plot of  vs. IAE for different tuning rules for FOPDT……65
Figure 2. 24. Plot of Ms vs. IAE for different tuning rules for FOPDT…..66
Figure 3.1 Block diagram of IMC and classical feedback control………..72
Figure 3.2 Simulation results for example 3.1…………………………….83
Figure 3.3 Simulation results for example 3.2…………………………….85
Figure 3.4 Simulation results for example 3.3…………………………….87
Figure 3.5 Simulation results of the polymerization process……………..88
Figure 3.6 Simulation results for model mismatch……………………….89
Figure 3.7  guideline for the proposed tuning method…………………90
Figure 4.1 Block diagram of IMC and classical feedback control………..95
Figure 4.2 Simulation result of proposed tuning method for example 4.1..102
Figure 4.3 Simulation result of proposed tuning method for example 4.2..103
Figure 4.4 Simulation result of proposed tuning method for example 4.3..106
Figure 4.5 Simulation result of proposed tuning method for example 4.4..107
Figure 4.6 Simulation result of proposed tuning method for example 4.5..110
Figure 4.7 Simulation result of proposed tuning method for example 4.6..111
Figure 4.8 Closed-loop responses by various controllers………………..116
Figure 4.9 Comparison of the ITAE generated by various tuning rules….117
Figure 4.10  guide lines for FOPDT and SOPDT………………………118
Figure 5.1. Block diagram of IMC and classical feedback control……….124
Figure 5.2. Response of the nominal system for Example 5.1……………133
Figure 5.3. Responses of the model mismatch system for Example 5.1…134
Figure 5.4. Response of the nominal system for Example 5.2……………136
Figure 5.5. Responses of the model mismatch system for Example 5.2…136
Figure 5.6. Responses of the model mismatch system for Example 5.2…139
Figure 5.7. Responses of the model mismatch system for Example 5.3…140

xiv
Figure 5.8. Response of the nominal system for Example 5.4…………...142
Figure 5.9. Responses of the model mismatch system for Example 5.4…142
Figure 5.10. Simplified structure for the modified Smith predictor……...143
Figure 5.11. Response of the nominal system for Example 5.5………….145
Figure 5.12. Responses of the model mismatch system for Example 5.5...146
Figure 5.13.  guidelines for FODUP……………………………………147

xv
xvi
1. INTRODUCTION

1.1 Motivations

PID control is a name commonly given to three-term control. The mnemonic

PID refers to the first letters of the names of the individual terms that make
up the standard three-term controller. These are P for the proportional term, I
for the integral term and D for the derivative term in the controller.
Three-term or PID controllers are probably the most widely used industrial
controller. Even complex industrial control systems may comprise a control
network whose main control building block is a PID control module. The
three-term PID controller has had a long history of use and has survived the
changes of technology from the analogue era into the digital computer
control system age quite satisfactorily.
It was the first (only) controller to be mass produced for the high-volume
market that existed in the process industries.
The introduction of the Laplace transform to study the performance of
feedback control systems supported its technological success in the
engineering community. The theoretical basis for analyzing the performance
of PID control is considerably aided by the simple representation of an
Integrator by the Laplace transform,  1 s  and a Differentiator using [s].
Conceptually, the PID controller is quite sophisticated and three different
representations can be given. First, there is a symbolic representation, Figure
1.1(a), where each of the three terms can be selected to achieve different
control actions. Secondly, there is a time domain operator form, Figure 1.1(

1
b), and finally, there is a Laplace transform version of the PID controller
Figure 1.1(c). This gives the controller an s-domain operator interpretation
and allows the link between the time domain and the frequency domain to
enter the discussion of PID controller performance.

e uc
I

D
(a)

kP

e(t) uc(t)
kI∫t

kDd ∕dt
(b)

kP

e(s) uc(s)
kI ∕s

kDs
(c)

Fig. 1.1 PID controller representations

PID control remains an important control tool for three reasons: past record
of success, wide availability and simplicity in use. These reasons reinforce
one another, thereby ensuring that the more general framework of digital
control with higher order controllers has not really been able to displace PID
control. It is really only when the process situation demands a more
sophisticated controller or a more involved controller solution to control a
complex process that the control engineer uses more advanced techniques.

2
Even in the case where the complexity of the process demands a multi-loop
or multivariable control solution, a network based on PID control building
blocks is often used.
A time delay may be defined as the time interval between the start of an
event at one point in a system and its resulting action at another point in the
system. Delays are also known as transport lags or dead times; they arise in
physical, chemical, biological and economic systems, as well as in the
process of measurement and computation. Methods for the compensation of
time delayed processes may be broadly divided into parameter optimized (or
PID based) controllers, in which the controller parameters are adapted to the
controller structure, and structurally optimized controllers, in which the
controller structure and parameters are adapted optimally to the structure and
parameters of the process model [1, 2]. The PID controller and its variations
(P, PI or PD) is the most commonly used controller in process control
applications, for the compensation of both delayed and non-delayed
processes. Koivo and Tanttu [3], for example, suggest that there are perhaps
5-10% of control loops that cannot be controlled by single input, single
output (SISO) PI or PID controllers; in particular, these controllers perform
well for processes with benign dynamics and modest performance
requirements [4, 5]. PID controllers have some robustness to incorrect
process model order assumptions and limited process parameter changes. The
controller is also easy to understand, with tuning rules that have been
validated in a wide variety of practical cases. It has been stated that 98% of
control loops in the pulp and paper industries are controlled by SISO PI
controllers [6] and that, in process control applications, more than 95% of the
controllers are of PID type [5]. However, Ender [7] states that, in his testing
of thousands of control loops in hundreds of plants, it has been found that

3
more than 30% of installed controllers are operating in manual mode and
65% of loops operating in automatic mode produce less variance in manual
than in automatic (i.e. the automatic controllers are poorly tuned). A details
review of the PID controller work is available in [8]. Other reviews based on
the details elements of the topics treated are in [3, 9-16].
The PID controller may be implemented in continuous or discrete time, in a
number of controller structures [17]. The ideal continuous time PID
controller is expressed in Laplace form as follows:
 
GPID  K c 1 
 Is
1
 D s 

 1.1
with K c = proportional gain,  I = integral time constant and  D = derivative

time constant. If  I   and  D  0 (i.e. P control), then the closed loop

measured value will always be less than the desired value for processes
without an integrator term, as a positive error is necessary to keep the
measured value constant, and less than the desired value. The introduction of
integral action facilitates the achievement of equality between the measured
value and the desired value, as a constant error produces an increasing
controller output. The introduction of derivative action means that changes in
the desired value may be anticipated, and thus an appropriate correction may
be added prior to the actual change. Thus, in simplified terms, the PID
controller allows contributions from present, past and future controller
inputs.
In many cases, the designs of PID controllers for delayed processes are based
on methods that were originally used for the controller design of delay-free
processes. However, PID controllers are not well suited for the control of
dominant delay processes [18]. It has been suggested that the PID
implementation is recommended for the control of processes of low to

4
medium order, with small delays, when controller parameter setting must be
done using tuning rules and when controller synthesis may be performed a
number of times [1].

1.2 Review of PID controller

1.2.1 Analytical techniques

Controller parameters may be determined using analytical techniques. Some
methods minimize an appropriate performance index; Harris and Mellichamp
[19], for instance, outline a methodology to tune a PI or PID controller to met
multiple closed loop criteria. These criteria are subsumed into a single
performance index that is an arbitrary function of relevant frequency domain
parameters; the method reflects the important point that there is no one set of
tuning values that provide the optimum response in all respects. Other such
methods to determine compensators for delayed SISO processes have also
been described, both in continuous time [20-37] and discrete time [1, 2, 38-
44]. Compensators for delayed MIMO processes have also been proposed in
continuous time [45-47] and discrete time [2, 48].
Alternatively, a direct synthesis strategy may be used to determine the
controller parameters. Such strategies may be defined in the time domain,
possibly by using pole placement [4, 49-59] or in the frequency domain,
possibly by specifying a desired gain and/or phase margin [60-66].
Robust methods, based on the IMC design procedure, may be used to design
analytically an appropriate PID controller for a FOPDT process model both
with delay uncertainty and with general parameter uncertainty [67-68].
1.2.2 IMC control strategy

5
The idea inherent in IMC has been around in one form or another for several
decades. The Smith predictor [69] contains the reference model idea of the
IMC. The IMC structure was formally introduced by [70]. It is a powerful
control design strategy for linear system [71-73]. It uses the process model as
the internal model to predict the process output. When the model is perfect,
the IMC system becomes an open-loop system and controller design and
stability analysis issue become trivial. When a model mismatch exists, by
appropriately modifying the difference, robustness can be obtained. The IMC
enables the transient response and robustness to be addressed independently.
Single–loop control and most of the existing advanced controller such as the
linear quadratic optimal controller and Smith predictor can equivalently be
put into the general IMC form [70, 15]. The advantages of IMC are exploited
in many industrial applications [70].
It is well known that the IMC is a powerful framework for control system
design and implementation [67], and it has sound theoretical foundation. Its
stability analysis is extremely easy to carry out and the design trade-off
between performance and robustness is clearly understood. It has attracted
the attention of industrial users because there is only one user-defined tuning
parameter, which is directly related to the closed-loop time constant or
equivalently, the closed loop bandwidth. On the other hand, the vast majority
of controllers being used in industry are of the PID type due to its simplicity
and popularity [5].

1.3 Outline of the Book

It is, however, noted that by far the most widely used controllers in the
process industries are PID controllers, so it is worth exploring the

6
relationship between IMC and PID in order to gain insight into the tuning of
this simpler controller, its performance and limitations.
It is essential to emphasize that there is lack of unified framework of the
IMC-PID controller designed for the several class of the process model
which gives the enhanced and robust control performance.
Therefore, this study mainly focused to develop a unified framework for IMC
based PID controller design method for several class of process model. For
the clarity and ease of understanding the present study is divided in the
following chapters.
Chapter 2:
Determine the optimum IMC filter which gives the best performance of the
resulting PID controller for several representative process models. This filter
is used to design a PID controller for disturbance rejection and the
corresponding analytical IMC-PID tuning rule. Provide the guidelines for
selection of closed-loop time constant  for several process models to cover

a wide range of   . Conduct a robust study by inserting a perturbation

uncertainty in each parameter simultaneously (for the worst case model).
Several illustrative examples are included to demonstrate the superiority of
the proposed tuning methods.

Chapter 3:
This study conducted for a simple and efficient method for the design of a
PID filter controller with enhanced performance. A closed-loop time constant
   guideline is recommended for a wide range of time-delay/time-constant
ratios. A simulation study was performed to illustrate the superiority of the
proposed method for both nominal and perturbed processes.
Chapter 4:

7
An analytical method for the design of a PID controller cascaded with a
second-order lead-lag filter is developed for various types of time-delay
process. The proposed PID●filter controller is designed based on the IMC-
PID principle and gives a better response than the conventional PID
controllers for obtaining the desired, closed-loop response. Several examples
are provided for comparing the results with the conventional PID controllers.
In the proposed PID●filter structure, the resulting control system becomes
equivalent to controlling a fast dynamic process by integral control, which
dramatically improves the performance. Some discussions with the 
guideline for particular robustness levels are also provided.
Chapter 5:
A simple analytical method is proposed for the design of a PID·filter
controller, in order to achieve enhanced performance for first order unstable
and integrating delay processes. A closed-loop time constant,  , guideline
was recommended for a wide range of time-delay/time-constant ratios. A
simulation study was performed for both unstable and integrating delay
processes to show the superior performance of the proposed method for both
nominal and perturbed processes.

8
1.4 References

1. Isermann, R., Digital control systems Volume 1. Fundamentals,

deterministic control, Springer-Verlag, 1989.
2. Isermann, R., Digital control systems Volume 2. Stochastic control,
multivariable control, adaptive control, applications, Springer-Verlag,
1991.
3. Koivo, H.N. and Tanttu, J.T. (1991). Proc. IFAC Intelligent Tuning
and Adaptive Control Symposium, Singapore, 75.
4. Hwang, S.-H. (1993). Chemical Engineering Communications, 124,
131.
5. Astrom, K.J. and Hagglund, T. (1995). PID controllers: theory, design
and tuning, Instrument Society of America.
6. Bialkowski, W.L. (1996). The Control Handbook. Editor: W.S. Levine,
CRC/IEEE Press, Boca Raton, Florida, 1219.
7. Ender, D.B. (1993). Control Engineering, September, 180.
8. O’Dwyer (2000). Technical Report 2000-02,
http://www.docsee.kst.dit.ie/aodweb, April.
9. Bueno, S.S., De Keyser, R.M.C. and Favier, G. (1991). Journal A,
32(1), 28.
10. Astrom, K.J., Hagglund, T., Hang, C.C. and Ho, W.K. (1993). Control
Engineering Practice, 1, 699.
11. Astrom, K.J., (1996). Proc. 13th World Congress of IFAC, San
Franscisco, CA, USA, Plenary Volume, 1.
12. Chen, G. (1996). International Journal of Intelligent Control and
Systems, 1(2), 235.
13. Gorez, R. (1997). Journal A, 38, 3.

9
14. Seborg, D.E., Edgar, T.F. and Shah, S.L. (1986). AIChE Journal, 32,
881.
15. Fisher, D.G. (1991). The Canadian Journal of Chemical Engineering,
69, 5.
16. Lelic, M. and Gajic, Z. (2000). Preprints, Proc. PID ’00: IFAC
Workshop on digital control, Terrassa, Spain, 73.
17. Astrom, K.J. and Wittenmark, B. (1984). Computer controlled
systems: theory and design, Prentice-Hall International Inc..
18. Hagglund, T. and Astrom, K.J. (1991). Automatica, 27, 599.
19. Harris, S.L. and Mellichamp, D.A. (1985). AIChE Journal, 31, 484.
20. Lee, J., Cho, W. and Edgar, T.F. (1990). AIChE Journal, 36, 1891.
21. Nishikawa, Y., Sannomiya, N., Ohta, T. and Tanaka, H. (1984).
Automatica, 20, 321.
22. Patwardhan, A.A., Karim, M.N. and Shah, R. (1987). AIChE Journal,
33, 1735.
23. Zevros, C., Belanger, P.R. and Dumont, G.A. (1988). Automatica, 24,
165.
24. Schei, T.S. (1994). Automatica, 30, 1983.
25. Di Ruscio, D. (1992). Modeling, Identification and Control, 13, 189.
26. Nakano, E. and Jutan, A. (1994). ISA Transactions, 33, 353.
27. Abbas, A. and Sawyer, P.E. (1995). Computers and Chemical
Engineering, 19, 241.
28. Wang, L., Barnes, T.J.D. and Cluett, W.R. (1995). IEE Proceedings -
Control Theory and Applications, 142, 265.
29. Wang, F.-S., Yeh, C.-L. and Wu, Y.-C. (1996). Transactions of the
Institute of Measurement and Control, 18, 183.

10
30. Ham, T.W. and Kim, Y.H. (1998). Industrial Engineering Chemistry
Research, 37, 482.
31. Hizal, N.A. (1997). Turkish Journal of Engineering and Environmental
Sciences, 21, 83.
32. Astrom, K.J., Panagopoulos, H. and Hagglund, T. (1998). Automatica,
34, 585.
33. Ruano, A.E.B., Fleming, P.J. and Jones, D.I. (1992). IEE Proceedings,
Part D, 139, 279.
34. Ruano, A.E.B., Lima, J.M.G., Mamat, R. and Fleming, P.J. (1996).
Journal of Systems Engineering, 6, 166.
35. Wu, C.-J. and Huang, C.-H. (1997). Journal of the Franklin Institute,
334B, 547.
36. Leva, A. and Colombo, A.M. (1999). IEE Proceedings – Control
Theory and Applications, 146(2), 137.
37. Liu, G.P. and Daley, S. (1999). Control Engineering Practice, 7, 821.
38. Cameron, F. and Seborg, D.E. (1983). International Journal of
Control, 38, 401.
39. Ralston, P.A.S., Watson, K.R., Patwardhan, A.A. and Deshpande, P.B.
(1985). Industrial and Engineering Chemistry Process Design and
Development, 24, 1132.
40. Bortolotto, G., Desages, A. and Romagnoli, J.A. (1989). Chemical
Engineering Communications, 86, 17.
41. Vega, P., Prada, C. and Aleixandre, V. (1991). IEE Proceedings - Part
D, 138, 303.
42. Yang, Y., Jia, C., Chen, J. and Lu, Y. (1991). Computers in Industry,
16, 81.

11
43. Miura, N., Imaeda, M., Hashimoto, K., Wood, R.K., Hattori, H. and
Onishi, M. (1998). Journal of Chemical Engineering of Japan, 31, 626.
44. Kelly, J.D. (1998). The Canadian Journal of Chemical Engineering,
76, 967.
45. Wang, F.-S. and Wu, T.-Y. (1995). Transactions of the Institute of
Measurement and Control, 17, 27.
46. Puleston, P.F. and Mantz, R.J. (1995). Industrial and Engineering
Chemistry Research, 34, 2993.
47. Ham, T.W. and Kim, Y.H. (1998). Journal of Chemical Engineering of
Japan, 31(6), 941.
48. Woldai, A., Al-Gobaisi, D.M.K., Dunn, R.W., Kurdali, A. and Rao,
G.P. (1996). Control Engineering Practice, 4(5), 721.
49. Schuster, T. (1989). Periodica Polytechnica Series - Electrical
Engineering, 33, 263.
50. Aguirre, L.A. (1992). Electronics Letters, 28, 2269.
51. Zhao, Z.-Y., Tomikuza, M. and Iskra, S. (1993). IEEE Transactions on
Systems, Man and Cybernetics, 23, 1392.
52. Hwang, S.-H. and Shiu, S.-J. (1994). International Journal of Control,
60, 265.
53. Sung, S.W., Lee, I.-B. and Lee, J. (1995). Industrial Engineering
Chemistry Research, 34, 4127.
54. Prokop, R. and Meszaros, A. (1996). Journal of Electrical
Engineering, 42(11-12), 287.
55. Daley, S. and Liu, G.P. (1999). IEE Computing and Control
Engineering Journal, April, 51.
56. Jung, C.L., Song, H.K. and Hyun, J.C. (1999). The Canadian Journal
of Chemical Engineering, 77, 180.

12
57. Jung, C.L., Song, H.K. and Hyun, J.C. (1999). Journal of Process
Control, 9(3), 265.
58. Atherton, D.P. (1999). IEE Computing and Control Engineering
Journal, April, 44.
59. Majhi, S. and Atherton, D.P. (1999). IEE Proceedings – Control
Theory and Applications, 146(5), 415.
60. Edgar, T.F., Heeb, R. and Hougen, J.O. (1981). Computers and
Chemical Engineering, 5, 225.
61. Kim, Y.H. (1995). Journal of Chemical Engineering of Japan, 28, 118.
62. Shafiei, Z. and Shenton, A.F. (1997). Automatica, 33, 2223.
63. Natarajan, K. and Gilbert, A.F. (1997). The Canadian Journal of
Chemical Engineering, 75, 765.
64. Wang, Q.-G., Hang, C.-C. and Bi, Q. (1997). Transactions of the
Institute of Chemical Engineers, 75(A), 64.
65. Wang, Q.-G., Hang, C.-C., Zhu, S.-A. and Bi, Q. (1999). Journal of
Process Control, 9(4), 291.
66. Fung, H.-W., Wang, Q.-G. and Lee, T.-H. (1998). Automatica, 34,
1145.
67. Morari, M. and Zafiriou, E. (1989). Robust process control, Prentice-
Hall Inc..
68. Horn, I.G., Arulandu, J.R., Gombas, C.J., VanAntwerp, J.G. and
Braatz, R.D. (1996). Industrial Engineering ChemistryResearch, 35,
3437.
69. Smith, C.L., Corripio, A.B. and Martin, J. (1975). Instrumentation
Technology, December, 39.
70. Garcia, C.E and M. Morari (1982). Industrial and Engineering
Chemistry: Process Design and Development 21, 208-323.

13
71. Rivera, D.E., Morari, M. and Skogestad, S. (1986). Industrial and
Engineering Chemistry Process Design and Development, 25, 252.
72. Fruehauf, P.S., Chien, I.-L. and Lauritsen, M.D. (1993). ISA
Transactions, 33, 43.
73. Lee, Y., Park, S., Lee, M. and Brosilow, C. (1998). AIChE Journal, 44,
106.

14
2. IMC-PID CONTROLLER DESIGN FOR IMPROVED
DISTURBANCE REJECTION OF TIME DELAYED
PROCESSES

2.1. Introduction

The proportional integral derivative (PID) control algorithm is widely used in

process industries because of its simplicity, robustness and successful
practical application. Although advanced control techniques can provide
significant improvements, a well designed PID controller has proved to be
satisfactory for a large number of industrial control loops. The well-known
IMC-PID tuning rules have the advantage of only using a single tuning
parameter to achieve a clear trade-off between closed-loop performance and
robustness to model inaccuracies. The IMC-PID controller provides good set-
point tracking but has a sluggish disturbance response, especially for the
process with a small time-delay/time-constant ratio.1-9 However, as
disturbance rejection is much more important than set-point tracking for
many process control applications, a controller design that emphasizes
disturbance rejection rather than set-point tracking is an important design
problem that has been the focus of renewed research recently.

The IMC-PID tuning methods by Rivera et al.1, Morari and Zafiriou6, Horn et
al.3, Lee et al.4 and Lee et al.9, and the direct synthesis method by Smith et
al.10 (DS) and Chen and Seborg5 (DS-d) are examples of two typical tuning
methods based on achieving a desired closed-loop response. These methods
obtain the PID controller parameters by computing the controller which gives

15
the desired closed-loop response. Although this controller is often more
complicated than a PID controller, the controller form can be reduced to that
of either a PID controller or a PID controller cascaded with a first- or second-
order lag by some clever approximations of the dead time in the process
model.

Several workers (Morari and Zafiriou6, Lee et al.4, Lee et al.9, Chien and
Fruehauf 2, Horn et al.3, Chen and Seborg5, Skogestad8) have reported that the
suppressing load disturbance is poor when the process dynamics are
significantly slower than the desired closed-loop dynamics.

In fact, regarding the disturbance rejection for lag time dominant processes,
the well-known old design method by Ziegler and Nichols 11 (ZN) shows
better performance than the IMC-PID design methods based on the IMC

filter f  1   s  1 . Horn et al.3 proposed a new type of IMC filter which

r

includes a lead term to cancel out the process dominant poles. Based on this
filter, they developed an IMC-PID tuning rule which leads to the structure of
a PID controller with a second-order lead-lag filter. The performance of the
resulting controller showed a clear advantage over those based on the
conventional IMC filter. Chen and Seborg5 proposed a direct synthesis design
method to improve disturbance rejection for several popular process models.
To avoid excessive overshoot in the set-point response, they utilized a set-
point weighting factor. In order to improve the set-point performance by
including a set-point filter, Lee et al.4 proposed an IMC-PID controller based
on both the filter suggested by Horn et al. 3 and a two-degree-of-freedom
(2DOF) control structure. Lee et al.9 extended the tuning method to unstable
processes such as first- and second-order delayed unstable process (FODUP
and SODUP) models and for the set-point performance 2DOF control

16
structure proposed. Skogestad8 proposed a model reduction technique to
reduce the higher order process model to a lower order model and also
developed the SIMC-PID rule for improved disturbance rejection in several
lag time dominant processes.

It is clear that in the IMC-PID approach, the performance of the PID

controller is mainly determined by the IMC filter structure. In most previous
works of IMC-PID design, the IMC filter structure has been designed just as
simple as to satisfy a necessary performance of the IMC controller. For
example, the order of the lead term in the IMC filter is designed small
enough to cancel out the dominant process poles and the lag term is simply
set to make the IMC controller proper. However, the performance of the
resulting PID controller is not only based on the IMC controller performance
but also based on how closely the PID controller approximates the ideal
controller equivalent to the IMC controller, which mainly depends on the
structure of the IMC filter. Therefore, for the IMC-PID design, the optimum
IMC filter structure has to be selected considering the performance of the
resulting PID controller rather than that of the IMC controller.

In the present study we have the following objectives:

1. Determine the optimum IMC filter which gives the best performance of
the resulting PID controller (Skogestad8: “best” could, for example, mean it
minimizes the integrated absolute error (IAE) with a specified value of the
sensitivity peak, Ms ) for several representative process models. This filter is
used to design a PID controller for disturbance rejection and the
corresponding analytical IMC-PID tuning rule.

2. Provide the guidelines for selection of closed-loop time constant  for

several process models to cover a wide range of   .

17
3. Conduct a robust study by inserting a perturbation uncertainty in each
parameter simultaneously (for the worst case model). Several illustrative
examples are included to demonstrate the superiority of the proposed tuning
methods.

2.2. IMC-PID Approach for PID Controller Design

Figure 2.1-a shows the block diagrams of IMC control, where GP is the

process, G P the process model, q the IMC controller. In the IMC control
structure, the controlled variable is related as:

GP q  %q
1 G 
fR R   P  GD d (2.1)
 %
1  q GP  GP   %
1  q GP  GP  

% ), the set-point and disturbance responses

For the nominal case (i.e., GP  GP

are simplified to:

C %
 GP qf R (2.2)
R
C %q  G
 1  G (2.3)
d 
P  D

In the classical feedback control structure shown in Figure (2.1-b), the set-
point and disturbance responses are represented by:
C GcGP f R
 (2.4)
R 1  GcGP
C GD
 (2.5)
d 1  GcGP

where Gc denotes the equivalent feedback controller.

18
According to the IMC parameterization (Morari and Zafiriou6), the process
% is decomposed into two parts:
model GP

%P P
G (2.6)
P M A

where PM and PA are the portions of the model inverted and not inverted,

respectively, by the controller ( PA is usually a non-minimum phase and

contains dead times and/or right half plane zeros); PA  0  = 1 .

d
GD
IMC GD d
setpoint controller
filter Process
R C setpoint Controller Process
fR +- q Gp ++ filter
R C
fR +- Gc Gp ++
Process model
~
Gp -+

Fig. 2.1 (a) The IMC structure Fig. 2.1 (b) Classical feedback control structure

Figure 2.1. Block diagram of IMC and classical feedback control

The IMC controller is designed by:

q  PM -1 f (2.7)
The ideal feedback controller that is equivalent to the IMC controller can be
% , and the IMC controller, q :
expressed in terms of the internal model, GP

q
Gc  (2.8)
%q
1 G P

Since the resulting controller does not have a standard PID controller form,
the remaining issue is to design the PID controller that approximates the
equivalent feedback controller most closely. Lee et al. 4 proposed an efficient

19
method for converting the ideal feedback controller Gc to a standard PID

controller. Since Gc has an integral term, it can be expressed as:

g  s
Gc  (2.9)
s
Expanding Gc in Maclaurin series in s gives:

1 g ''  0  2 
Gc   g  0   g '  0  s  s  ...  (2.10)

s 2 

The first three terms of the above expansion can be interpreted as the
standard PID controller which is given by:
 1 
Gc  K c 1   Ds  (2.11)
 Is 
where
Kc  g '  0 (2.12a)
 I  g '  0 g  0 (2.12b)

 D  g ''  0  2 g '  0  (2.12c)

2.3. IMC-PID Tuning Rules for Typical Process Models

This section proposes the tuning rules for several typical time-delayed
process models.

2.3.1. First-Order Plus Dead Time Process (FOPDT)

The most commonly used approximate model for chemical processes is the
FOPDT model as given below:
Ke  s
GP  GD  (2.13)
 s 1

20
where K is the gain,  the time constant and  the time delay. The

optimum IMC filter structure is found as f    s  1   s  1

2 3
. The

resulting IMC controller becomes q    s  1   s  1 K   s  1 . Therefore,

2 3

the ideal feedback controller which is equivalent to the IMC controller is:
  s  1   s  1
2

Gc  (2.14)
K   s  1  e  s   s  1 
3 2
 
The analytical PID formula can be obtained from eq 2.12 as:
I
KC  (2.15a)
K  3  2    

3    2   
2 2 2

    2    2
I
 3  2   
(2.15b)

    3 3 2
  2 
D 
 2    2 6
 3  2     3   2  2   
2 2 2

I  3  2   
(2.15c)
The value of the extra degree of freedom  is selected so that it cancels out
the open-loop pole at s  1  that causes the sluggish response to the load

disturbance. Thus,  is chosen so that the term  1  Gq  has a zero at the pole

of GD . That is, we want  1  Gq  s 1   0 and 1    s  1 e   s  1  s 1   0 . The

2  s 3

value of  after some simplification becomes

21
   
3

12


   1   1   e    
     
 
 

(2.16)

2.3.2. Delayed Integrating Process (DIP)

Ke  s
GP  GD  (2.17)
s

The delayed integrating process (DIP) can be modeled by considering the

integrator as a stable pole near zero. This is necessary because it is not
possible to apply the aforementioned IMC procedure for DIP, since the term
of  disappears at s  0 . Usually, the controller based on the model with a
stable pole near zero can give a more robust closed-loop response than that
based on the model with an integrator or unstable pole near zero as suggested
by Lee et al.9 Therefore, DIP can be approximated to FOPDT as shown
below:
Ke  s Ke  s  Ke  s
GP  GD    (2.18)
s s  1/  s 1

where  is an arbitrary constant with a sufficiently large value.

Accordingly, the optimum filter structure for DIP is same as that for the

FOPDT model, i.e., f    s  1   s  1

2 3
.

22
Process KC I D 

Ke  s I  3 2

2
 2    2      6      
3 3 2 2
 
 
3

12

 2      
2
 K  3  2        2   2
3    2    2
22    1   1   e    
s  3  2     3  2    
2       
I  3  2     
  4  
2  
4

12 1   e 1  1    2   1   e   2 1
Ke  s
I
  1   2  1  
 6 2

2
2
 1   2   3  3 1
 4  6  2

 2 
2
1 
  1  
  2  1 
   2  

  1s  1   2 s  1               6  2     

K  4  1   
2
 4  1    1 2 1 2 1 2
 4  1   
2


1 2
 
4

I  4  1     2   2 2  1   e  2  1   2 1
  2  
 
4
 2  
4

 2 1   e    1     1   e   1
 3  3 1 
2 

       
 4  6    2  1 
2    
Ke  s I
     1  
 6 2

2
2
 1   2         1   2   
 4  1   

6 2    1   2
2

2      4 
s   s  1  K  4  1     4  1    I

 4  1     2   2 1   e    1  1
   

Ke  s 
I
 3 2

2
 2    2      6  
3 3 2
  2  
 
3

12

 s 1
K  3  2        2   2
 3  2   
 2   2    3  2   

 3 2  
2
2
 2    2      1   e 
   


 1

I  3  2      
 4   6  
3 3 2
  2  
   
4 12

 a  2  a         6   2  2    
2 2

 6       1 e   1

2 2
Ke  s I 2

2
 2    2  4  2     
  s  1  as  1  2 I  4  2        
K  4  2      a    2     
 4  2   
Table 2.1. IMC-PID Controller Tuning Rules

23
Therefore, the resulting IMC controller becomes

q    s  1   s  1 K   s  1
2 3
and the ideal feedback controller is,

  s  1   s  1
2

GC  2 . The resulting PID tuning rules are listed in

K   s  1  e s   s  1 
3
 
Table 2.1.

2.3.3. Second-Order Plus Dead Time Process (SOPDT)

Consider a stable SOPDT system:

Ke  s
GP  GD  (2.19)
  1s  1   2 s  1
The optimum IMC filter structure is found as f    2 s  1s  1   s  1
2 4
. The
2

IMC controller becomes q   1s  1   2 s  1  2 s  1s  1 K   s  1  4
and the
ideal feedback controller equivalent to the IMC controller is
 1s  1   2 s  1   2 s  1s  1
2

GC  . The resulting PID tuning rules are listed in

K   s  1  e  s   2 s 2  1s  1 
4

 
Table 2.1.
2.3.4. First-Order Delayed Integrating Process (FODIP)

Consider the following FODIP system:

Ke  s
GP  GD  (2.20)
s   s  1

The above process can be approximated as the SOPDT model and it

becomes:
Ke  s  Ke  s
GP  GD   (2.21)
s   s  1   s  1   s  1
where  is an arbitrary constant with a sufficiently large value. Thus, the
optimum IMC filter is same as that for the SOPDT,

24
 f    2 s 2  1s  1   s  1
4
. Therefore, the resulting IMC controller becomes

 
q    s  1   s  1  2 s 2  1s  1 K   s  1 . The resulting PID tuning rules
4

are listed in Table 2.1.

2.3.5. First-Order Delayed Unstable Process (FODUP)

One of the most popular unstable processes with time delay is the FODUP:
Ke  s
GP  GD  (2.22)
 s 1

The optimum IMC filter is found to be f    s  1   s  1

2 3
. Therefore,

the IMC controller becomes q    s  1   s  1 K   s  1 and the ideal

2 3

  s  1   s  1
2

feedback controller is GC  2 . The resulting PID tuning

K   s  1  e s   s  1 
3
 
rules are given in Table 2.1.

2.3.6. Second-Order Delayed Unstable Process (SODUP)

The process model is:

Ke s
GP  GD  (2.23)
  s  1  as  1
The optimum IMC filter is found to be f    s  1   s  1
2 4
. The IMC

controller is q    s  1  as  1   s  1 K   s  1 . The resulting PID tuning

2 4

rules are shown in Table 2.1.

2.4. Performance and Robustness Measure

25
In this study, the performance and robustness of the control system are
evaluated by the following indices.

2.4.1. Integral Error Criteria

Three popular performance indices based on integral error are used to

evaluate the performance: Integral of the Absolute Error (IAE), Integral of
the Squared Error (ISE), Integral of the Time-weighted Absolute Error
(ITAE).

IAE   e  t  dt
0
(2.24)

 e t
2
ISE  dt (2.25)
0


ITAE   t e  t  dt
0
(2.26)

where the error signal e  t  is the difference between the set-point and the
measurement. The ISE criterion penalizes larger errors, whereas the ITAE
criterion penalizes long-term errors. The IAE criterion tends to produce
controller settings that are between those for the ITAE and ISE criteria.

2.4.2. Overshoot

Overshoot is a measure of how much the response exceeds the ultimate value
following a step change in set-point and/or disturbance.

2.4.3. Maximum Sensitivity  Ms  to Modeling Error

To evaluate the robustness of a control system, the maximum sensitivity, Ms

, which is defined by Ms  max 1/[1  G pGc (iw)] , is used. Since Ms is the

inverse of the shortest distance from the Nyquist curve of the loop transfer

26
function to the critical point  1, 0  , a small value indicates that the stability
margin of the control system is large. Typical values of Ms are in the range
of 1.2 ~ 2.0 (Åström et al.12). For fair comparison, throughout all our
simulation examples all the controllers compared were designed to have the
same robustness level in terms of the maximum sensitivity, Ms .

2.4.4. Total Variation (TV)

To evaluate the manipulated input usage we compute TV of the input u  t  ,

which is sum of all its moves up and down. If we discretize the input signal


as a sequence [u1 , u2 , u3 ...., ui ...], then TV   ui 1  ui should be as small as
i 1

possible. TV is a good measure of the smoothness of a signal (Skogestad8).

2.4.5. Set-Point and Derivative Weighting

The conventional form of the PID controller that is used for the simulation in
this study is given as:
 1 
GPID  K c 1   D s  (2.27)
  I s 

A more widely accepted control structure that includes set-point weighting

and derivative weighting is given by Åström and Hägglund13. The PID
controller after set-point and derivative weighting becomes:
 1
t
d cr  t   y  t   
u  t   K c  br  t   y  t      r  t   y  t   d   D   (2.28)
 I 0 dt 
 

where b and c are additional parameters. The integral term must be based on
error feedback to ensure the desired steady state. The controller given by eq
2.28 has a structure with two degrees of freedom. The set-point weighting
coefficient b is bounded by 0  b  1 and the derivative weighting coefficient

27
c is also bounded by 0  c  1 . The overshoot for set-point changes
decreases with decreasing b .
The controllers obtained for different values of b and c respond to
disturbance and measurement noise in the same way as a conventional PID
controller, i.e., different values of b and c do not change the closed-loop
response for disturbances (Chen and Seborg5). Therefore, the same PID
tuning rules developed for this study are also applicable for the modified PID
controller in eq 2.28. However, the set-point response does depend on the
values of b and c . In this study, the coefficient c was fixed as c  1 for all
simulation examples while the set-point filter

 
f R   b I s  1  I D s 2   I s  1 was used with 0  b  1 .

2.5. Simulation Results

This section demonstrates the simulation study for the six different kinds of
process model, which are widely used in process industries and have also
been studied by other researchers. In every simulation study, different
performance and robustness matrices have been calculated and compared
with other existing methods.

2.5.1. Example 2.1. FOPDT

The following FOPDT model (Chen and Seborg5) has been considered.
100e 1s
GP  GD  (2.29)
100 s  1
A unit step disturbance is acting at the plant input and the corresponding
simulation result is shown in Figure 2.2. All the tuning methods except the
ZN method11 were adjusted to have the same robustness level as Ms  1.94
by varying the  value. The performance matrix for the disturbance rejection

28
is listed in Table 2.2. The proposed method was compared with other existing
IMC-PID methods (Horn et al.3; Lee et al. 4; Rivera et al.1), the direct
synthesis method (Chen and Seborg5) and the ZN method11. The performance
matrix containing IAE, ISE and ITAE shows that the proposed method
performs better than the other IMC-PID methods. IAE, ISE, ITAE and
overshoot values are minimized for the ZN method. However, the ZN
method gives the Ms value of 2.29. The proposed method has a better
performance indices than the ZN method when Ms  2.29 .
The resulting output response when the unit step changes are introduced into
the set-point is shown in Figure 2.3. It is clear that under the 1DOF control
structure, any controller with good disturbance rejection essentially
accompanies an excessive overshoot in the set-point response. To avoid this
water-bed effect, a 2DOF control structure is used. The corresponding
response is shown in the same figure. For the 1DOF case, the output response
for the proposed method has a larger overshoot than all the other methods
except the ZN method, but the settling time is faster for the proposed method
than all the other methods. The overshoot for the proposed controller can be
eliminated without affecting the disturbance response by setting b  0.4 in
the 2DOF controller, as suggested by Åström and Hägglund 13 and Chen and
Seborg5. Based on the comparison of the output response and the value of the
performance matrices listed in Table 2.2, it can be concluded that the
proposed controller shows the best performance.
The robustness of the controller is evaluated by inserting a perturbation
uncertainty of 20% in all three parameters simultaneously to obtain the worst

case model mismatch, i.e., GP  GD  120e1.2s  80s  1 as an actual process.

The simulation results for the proposed and other tuning rules are given in
Table 2.3 for both the set-point and the disturbance

29
 Table 2. 2. PID Controller Setting for Example 2.1     0.01 (FOPDT)

Tuning  Kc I D Ms set-point disturbance

methods
TV IAE Overshoot TV IAE Overshoot

Proposed 1.51 0.827 3.489 0.356 1.94 1.67 3.08 1.45 1.96 4.30 1.26
(b=1.0)
Proposed 1.51 0.827 3.489 0.356 1.94 0.79 2.37 1.03 1.96 4.30 1.26
(b=0.4)
DS-d5 1.2 0.828 4.051 0.353 1.94 1.57 3.06 1.40 1.88 4.89 1.27

Lee et al.4 1.32 0.810 3.928 0.307 1.94 1.54 3.09 1.44 1.87 4.85 1.31

Horn et 1.68 15.146 100.5 0.497 1.94 1.62 3.56 1.31 1.69 6.64 1.47
al.3
ZN11 - 0.948 1.99 0.498 2.29 3.25 3.58 1.67 3.04 3.22 1.17

IMC1 0.85 0.744 100.5 0.498 1.94 1.18 2.11 1.0 1.58 84.47 1.29

 1  1  cs  ds 2
Horn et al.3, Gc  K c 1   Ds  where a  100.2127; b  21.2687; c  4.2936, and d  0
 Is  1  as  bs
2

30
 Table 2. 3. Robust Analysis for Example 2.1 (FOPDT)

Tuning methods set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 6.36 5.47 2.12 7.33 6.31 1.95

(b=1.0)

DS-d5 6.20 5.23 2.06 7.19 6.28 1.96

Lee et al.4 5.93 5.61 2.09 7.00 6.80 2.1

Horn et al.3 4.44 5.15 1.84 5.91 7.76 2.23

ZN11 24.21 12.89 2.56 24.13 12.50 1.83

IMC1 4.58 3.26 1.43 6.15 84.59 1.92

31
rejection. The error integral values for the disturbance rejection of the
proposed method and Chen & Seborg5 (DS-d) appear to be almost similar
whereas DS-d has slight better performance. For the servo response, the IMC
has clear advantage and the proposed, DS-d, Lee et al.4 and Horn et al.3 give
almost similar robust performance.
1.5

1.2

0.9
Process Response

Proposed
0.6 Lee et al. (1998)
DS-d
Horn et al.
ZN
IMC
0.3

-0.3
0 4 8 12 16
Time

Figure 2.2. Simulation results of PID controllers for unit step disturbance

32
 1.7
Proposed (b=1)
Lee et al. (1998)
DS-d
1.5 Horn et al.
ZN
Proposed (b=0.4)

1.2
Process Response

0.9

0.6

0.3

0
0 4 8 12 16
Time

Figure 2.3. Simulation results of PID controllers for unit step set-point change

2.5.2. Example 2.2. DIP (Distillation Column Model)

The distillation column model studied by Chien and Fruehauf 2 and Chen &
Seborg5 (DS-d) was considered. The distillation column separates a small
amount of a low-boiling material from the final product. The bottom level of
the distillation column is controlled by adjusting the steam flow rate. The
process model for the level control system is represented as the following
DIP model which can be approximated by the FOPDT model as:
0.2e 7.4 s 20 Ke 7.4 s
GP  GD   (2.30)
s 100 s  1
The method proposed by Chen and Seborg5, Lee et al. 9
and Chien and
Fruehauf2 was used to design the PID controllers, as shown in Figure 2.4 and
Table 2.4.   11.3 was chosen for the proposed method,   9.15 for Chen
and Seborg5,   11.0 for Lee et al.9 and   15.28 for Chien and Fruehauf2,

33
resulting in Ms  1.90 . Figure 2.4 shows the output response, where the
proposed tuning rule results in the least settling. Chien and Fruehauf’s
method2 has the slowest response and requires the highest settling time. The
performance matrix in Table 2.4 shows that the proposed method has the
lowest error integral value while Chien and Fruehauf’s 2 method has the
highest value at an equal robustness level. Based on Table 2.4 and Figure 2.4,
it is clear that the proposed method performs better than the other
conventional methods.

2
Proposed
DS-d
1.8 Lee et al. (2000)
Chien and Fruehauf

1.4
Process Response

0.6

0.2

-0.2
0 30 60 90 120 140
Time

Figure 2.4. Simulation results of PID controllers for unit step disturbance

34
 Table 2. 4. PID Controller Setting for Example 2.2 (Distillation Column Model)

Tuning
methods
 Kc I D Ms set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 11.3 0.531 24.533 2.467 1.90 1.128 24.04 1.49 1.99 49.19 1.95
(b=1.0)

Proposed 11.3 0.531 24.533 2.467 1.90 0.58 18.41 1.07 1.99 49.19 1.95
(b=0.4)

DS-d5 9.15 0.543 31.15 2.558 1.90 1.0 23.28 1.39 1.84 57.47 1.93

Lee et al.9 11 0.536 35.137 2.286 1.90 0.95 23.46 1.38 1.78 65.35 1.98

Chien and 15.28 0.526 37.96 3.339 1.90 0.99 23.68 1.27 1.81 71.88 1.86
Fruehauf2

35
Table 2. 5. Robust Analysis for Example 2.2 (Distillation Column Model)

set-point disturbance
Tuning methods

TV IAE Overshoot TV IAE Overshoot

Proposed 1.89 27.81 1.88 3.27 49.86 2.54

(b=1.0)

DS-d5 1.80 25.38 1.77 3.09 57.47 2.52

Lee et al.9 1.74 25.85 1.75 3.06 65.34 2.57

Chien and 1.62 25.14 1.58 2.90 71.69 2.43

Fruehauf2

36
The simulation results for the unit set-point are shown in Figure 2.5. The
overshoot for the proposed method is large but the settling time is minimized
for the 1DOF controller. The overshoot can be minimized by using the 2DOF
controller with b  0.4 . The results in Table 2.4 and Figure 2.4 show the
superior performance of the proposed method.

1.5
Proposed (b=1.0)
DS-d
Lee et al. (2000)
Proposed (b=0.4)
Chien and Fruehauf
1.2

0.9
Process Response

0.6

0.3

0
0 30 60 90 120 140
Time

Figure 2.5. Simulation results of PID controllers for unit step set-point change

The robustness of the controller is evaluated by inserting a perturbation

uncertainty of 20% in all two parameters simultaneously. The worst plant-

model mismatch case after 20% perturbation is GP  GD  0.24e 8.88s s .

The simulation results for all tuning rules are given in Table 2.5. It is clear
that the proposed method has better performance for disturbance rejection
followed by DS-d and Lee et al. 9 The set-point response of the DS-d, Lee et

37
al.9 and Chien and Fruehauf2 are almost similar and superior to the proposed
method.

2.5.3. Example 2.3. SOPDT

Consider the SOPDT model described by Chen and Seborg5:

2e  1 s
GP  GD  (2.31)
 10 s  1  5s  1
The proposed, DS-d (Chen and Seborg5), ZN, SIMC (Skogestad8) and DS
(Smith et al.10, Seborg et al.7) methods were used to design the PID
controller. The DS and IMC-PID (Rivera et al. 1) gives the exactly same
tuning formula for the SOPDT process. The parameters of the PID controller
settings for the DS and ZN were taken from Chen and Seborg5. All the other
methods were adjusted to have the equal Ms value of 1.87 for a fair
comparison. Figure 2.6 shows the output response for all tuning methods
mentioned above. The proposed method has a similar overshoot to the DS-d
method, while the ZN tuning method gives the highest peak. The settling
time is the lowest for the proposed method, whereas the DS method shows
the slowest response. Apart from the DS-d method, all the other tuning
methods have either higher overshoot or slower response for disturbance
rejection.
The simulation results for the unit set-point are shown in Figure 2.7. The
overshoot for the ZN method is largest followed by the proposed method but
the settling time is lowest for the proposed method among all the 1DOF
controllers. The overshoot can be minimized by using the 2DOF controller
with b  0.4 . Based on the performance shown in the above figures and the
performance matrix in Table 2.6, the proposed method has the best
performance.

38
To evaluate the robust performance, the worst plant-model mismatch case

was considered as GP  GD  2.4e

1.2 s
 8s  1  4s  1 by inserting a
perturbation uncertainty of 20% in all three parameters simultaneously. The
simulation results for all tuning rules are given in Table 2.7. When compared
to the other methods, the error integral values for the proposed method are
the best for both set-point and disturbance rejection and the overshoot for the
proposed method is similar to those by the DS-d and DS methods. The ZN
method has a higher overshoot than the other methods.
Proposed
0.2
DS-d
SIMC
DS
ZN

0.15
Process Response

0.1

0.05

-0.05
0 5 10 15 20 25 30 35 40 45 50
Time

Figure 2.6. Simulation results of PID controllers for unit step disturbance

39
 Proposed (b=1.0)
1.6 DS-d
SIMC
DS
Proposed (b=0.4)
1.4 ZN

1.2
Process Response

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30 35 40 45 50
Time

Figure 2.7. Simulation results of PID controllers for unit step set-point change

40
 Table 2. 6. PID Controller Setting for Example 2.3 (SOPDT)

Tuning
methods
 Kc I D Ms set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 1.6 6.415 6.859 1.9798 1.87 11.59 5.66 1.41 1.83 1.06 0.14
(b=1.0)
Proposed 1.6 6.415 6.859 1.9798 1.87 4.86 4.72 1.0 1.83 1.06 0.14
(b=0.4)
DS-d5 2.4 6.384 7.604 2.0977 1.87 10.82 5.67 1.34 1.71 1.19 0.14

SIMC8 0.43 3.496 5.72 5.0 1.87 5.514 10.08 1.36 1.47 2.50 0.16

DS10 0.5 5.0 15.0 3.33 1.91 7.18 6.53 1.12 1.42 3.0 0.15

ZN11 - 4.72 5.83 1.46 2.26 12.75 8.73 1.65 2.71 1.79 0.21

41
 Table 2. 7. Robust Analysis for Example 2.3 (SOPDT)

Tuning methods set-point disturbance

TV IAE Overshoot TV IAE Overshoot
Proposed 41.88 5.11 1.46 6.41 1.09 0.17
(b=1.0)

DS-d5 47.93 5.14 1.38 7.40 1.21 0.17

SIMC8 50.36 8.71 1.37 14.19 2.31 0.18

DS10 82.83 6.34 1.22 16.50 3.00 0.17

ZN11 14.90 6.00 1.68 3.09 1.29 0.24

42
2.5.4. Example 2.4. FODIP (Reboiler Level Model)

Consider a level control problem proposed by Chen and Seborg5. It is an

approximate model for a liquid level in the reboiler of a steam heated
distillation column, which is to be controlled by adjusting the control valve
on the steam line. The process model is given by:
1.6  0.5s  1
GP  GD  (2.32)
s  3s  1

This kind of “inverse response time constant” (negative numerator time

constant) can be approximated as a time delay such as   0inv s  1  e

inv
 0

This is reasonable since an inverse response has a deteriorating effect on

control similar to that of a time delay (Skogestad8).
Therefore, the above model can be approximated as:
1.6e 0.5 s
GP  GD  (2.33)
s  3s  1

The above process model can be treated as FODIP and the tuning parameters
can be estimated by the analytical rule proposed in Table 2.1.
Figure 2.8 shows the output response of the proposed tuning method and its
comparison with the DS-d, IMC and ZN methods. The PID controller
settings for all the other methods were taken from Chen and Seborg5. The
value   0.935 was selected for the proposed method, which gives
Ms  1.94 . From the figure, the proposed output response has a small
overshoot and a fast settling time followed by the DS-d and IMC methods
while the ZN method has a very aggressive response with significant
overshoot and oscillation that takes a long time to settle. The performance
values are also listed in Table 2.8 and indicate the clear advantage of the
proposed method over other tuning rules.

43
The output responses for the unit set-point are shown in Figure 2.9. The
overshoot by the proposed method with the 1DOF structure is somewhat
large but shows a fast settling time before reaching its final value. The
overshoot can be drastically minimized by using the 2DOF controller. It is
apparent from Figure 2.8 and Table 2.8 that the proposed method is superior
over the other tuning methods for the disturbance rejection.
The robustness of the controller is evaluated considering the worst case under
a 20% uncertainty in all three parameters as

GP  GD  1.92  0.6s  1 s  2.4s  1 . The simulation results for all tuning rules

are given in Table 2.9. The error integral and overshoot values for the
proposed method prove to be the best. The overshoot and IAE of the ZN
method is the highest among the other tuning methods.
0.5
Proposed
DS-d
IMC
ZN

0
Process Response

-0.5

-1

-1.5
0 5 10 15 20 25 30
Time

Figure 2. 8. Simulation results of PID controllers for unit step disturbance

44
 Table 2. 8. PID Controller Setting for Example 2.4 (Level Control Problem)
Tuning
methods
 Kc I D Ms set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 0.935 -1.456 4.195 1.250 1.94 4.70 3.087 1.40 3.43 2.96 -0.66
(b=1.0)
Proposed 0.935 -1.456 4.195 1.250 1.94 1.49 2.536 1.01 3.43 2.96 -0.66
(b=0.4)
DS-d5 1.6 -1.25 5.3 1.45 1.93 3.70 3.42 1.32 3.14 4.32 -0.72

IMC1 1.25 -1.22 6.0 1.5 1.95 3.58 3.505 1.28 3.10 5.00 -0.74

ZN11 - -0.752 3.84 0.961 2.77 2.89 7.401 1.79 4.03 9.65 -1.42

45
 Table 2.9. Robust Analysis for Example 2.4 (Level Control Problem)

Tuning methods set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 17.89 2.318 1.26 13.2 3.13 -0.59

(b=1.0) 3

DS-d5 13.85 2.74 1.23 11.9 4.47 -0.67

5

IMC1 14.70 2.85 1.21 12.9 5.14 -0.68

3

ZN11 2.59 4.49 1.70 3.71 6.47 -1.49

46
 1.8
Proposed (b=1.0)
DS-d
1.6 IMC
Proposed (b=0.4)
ZN
1.4

1.2

1
Process Response

0.8

0.6

0.4

0.2

-0.2
0 5 10 15 20 25 30
Time

Figure 2.9. Simulation results of PID controllers for unit step set-point change

2.5.5. Example 2.5. FODUP

The following FODUP (Huang & Chen14 and Lee et al.9) was considered:
1e 0.4 s
GP  GD  (2.34)
s 1
The closed-loop time constant   0.63 is selected for the proposed tuning
method for Ms  3.08 . Setting   0.5 results in the same value of
Ms  3.08 for Lee et al.’s 9 method, thus providing a fair comparison. The
controller setting parameters for the other existing method were taken from
Lee et al.9. Figure 2.10 shows the output response of all tuning methods, with
the proposed method showing the clear advantage. In Table 2.10 the

47
performance and robustness matrices are listed for all tuning rules, and the
proposed method shows a clear advantage over the other methods.
The response for the unit set-point of the 1DOF controller is shown in Figure
2.11 where every 1DOF controller show a significant overshoot. By using the
2DOF controller, the overshoot can be minimized, as shown in the response
of the proposed method by setting b  0.1 . The performance shown in
Figures 2.10 and 2.11 and Table 2.10 demonstrates the clear advantage of the
proposed method over the other methods.
For the robustness study, the controller is evaluated by inserting a
perturbation uncertainty of 20% to all three parameters and finding the actual
0.48 s
process as GP  GD  1.2e (1.2s  1) . The simulation results of the model

mismatch for the all tuning rules are given in Table 2.11. In the model
mismatch case, as seen from the performance matrix in Table 2.11, the
proposed method gives a superior performance over all the other methods
both for set-point and disturbance rejection.
Proposed
Lee et al. (2000)
0.7 Rotstein and Lewin
Huang and Chen

0.6

0.5
Process Response

0.4

0.3

0.2

0.1

0 2 4 6 8 10 12 14
Time

Figure 2. 10. Simulation results of PID controllers for unit step disturbance

48
 Table 2. 10. PID Controller Setting for Example 2. 5 (FODUP)

Tuning methods  Kc I D Ms set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 0.63 2.573 2.042 0.207 3.08 9.58 2.12 2.06 3.71 0.92 0.65
(b=1.0)
Proposed 0.63 2.573 2.042 0.207 3.08 2.44 1.30 1.07 3.71 0.92 0.65
(b=0.1)
Lee et al.9 0.5 2.634 2.519 0.154 3.03 9.13 2.09 2.21 3.54 0.96 0.71

De Paor and O - 1.459 2.667 0.25 4.92 11.01 8.74 2.53 7.02 5.96 1.13
Malley16
Rotstein and - 2.250 5.760 0.20 2.48 6.32 3.74 1.82 3.03 2.56 0.72
Lewin 17
Huang and - 2.636 5.673 0.118 3.21 9.14 3.28 2.19 3.62 2.15 0.76
Chen14

49
 Table 2. 11. Robust Analysis for Example 2.5 (FODUP)

Tuning methods set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 14.90 2.11 2.45 5.63 0.86 0.76

(b=1.0)
Lee et al.9 17.49 2.51 2.58 6.49 1.03 0.82

De Paor and 7.38 5.62 2.31 4.79 3.95 1.08

O Malley16
Rotstein and 7.97 3.48 2.05 3.69 2.56 0.83
Lewin17
Huang and 18.29 3.24 2.49 6.90 2.15 0.86
Chen14

50
 Proposed (b=1.0)
Lee et al. (2000)
Rotstein and Lewin
2 Huang and Chen
Proposed (b=0.1)

1.5
Process Response

0.5

0
0 2 4 6 8 10 12 14
Time

Figure 2.11. Simulation results of PID controllers for unit step set-point change

2.5.6. Example 2.6. SODUP

The following unstable process was considered for the present study (Huang
& Chen14, and Lee et al.9):
1e 0.5 s
GP  GD  (2.35)
 5s  1  2 s  1  0.5s  1
The above model was approximated to the SODUP model by Huang &
Chen14, and also used by Lee et al.9
1e 0.939 s
GP  GD  (2.36)
 5s  1  2.07 s  1
The value   1.2 was used for the Lee et al. method 9 which has Ms  4.35 .
For the proposed tuning rule, the value   0.938 was adjusted to give
Ms  4.35 for the fair comparison with Lee et al. 9. The other methods by
Huang & Chen14 and Huang & Lin15 were also included in the simulation and
the controller parameters were obtained from the Lee et al. 9. Figure 2.12

51
shows the output response of the proposed tuning method compared to the
other existing tuning rules. The proposed tuning method has a fast settling
time compared to all the other existing methods. Lee et al.’s 9 have a smaller
peak but it is slower and more oscillatory. In Table 2.12 the controller setting
parameters and performance indices are given, where the proposed method
shows clear advantage over the others.
Figure 2.13 compares the output responses for the unit set-point change. It is
clear that the 2DOF controller can improve the set-point response by
eliminating the overshoot. Table 2.13 shows performance index values, with
20% uncertainty in gain and dead time. The proposed method shows better
response both for set-point and disturbance rejection when compared with
Lee et al. method9. Due to different Ms bases in the nominal case, it is
difficult to get the fair comparison, Huang & Lin 15 has more robust
performance followed by Huang & Chen14 as performance indices appear in
Table 2.13.
0.3
Proposed
Lee et al. (2000)
Huang and Chen
0.25 Huang and Lin

0.2

0.15
Process Response

0.1

0.05

-0.05

-0.1
0 5 10 15 20 25
Time

Figure 2.12. Simulation results of PID controllers for unit step disturbance

52
 Table 2. 12. PID Controller Setting for Example 2. 6 (SODUP)

Tuning methods  Kc I D Ms set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 0.938 7.017 5.624 1.497 4.35 36.44 5.35 1.89 5.21 0.85 0.20
(b=1.0)
Proposed 0.938 7.017 5.624 1.497 4.35 12.27 3.49 1.05 5.21 0.85 0.20
(b=0.3)
Lee et al.9 1.20 7.144 6.696 1.655 4.34 36.45 5.19 1.71 5.12 0.95 0.19

Huang and - 6.186 7.17 1.472 3.63 26.04 5.57 1.85 4.25 1.16 0.23
Chen 14
Huang and Lin15 - 3.954 4.958 2.074 2.18 11.99 8.99 1.86 3.16 2.19 0.29

Poulin and - 3.050 7.557 2.07 1.86 8.71 11.0 1.88 3.00 3.81 0.40
Pomerleau18

Table 2.13. Robust Analysis for Example 2. 6 (SODUP)

53
Tuning methods set-point disturbance

TV IAE Overshoot TV IAE Overshoot

Proposed 167.03 13.54 2.04 23.56 1.95 0.22

(b=1.0)

Lee et al.9 248.09 15.19 1.84 34.50 2.26 0.21

Huang and Chen14 65.05 7.915 1.96 10.48 1.44 0.25

Huang and Lin15 13.34 7.303 1.82 3.43 1.78 0.31

Poulin and 8.25 8.69 1.79 2.90 3.12 0.41

Pomerleau18

54
 Proposed (b=1)
Lee et al. (2000)
1.8
Huang and Chen
Huang and Lin
Proposed (b=0.3)
1.6

1.4

1.2
Process Response

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25
Time

Figure 2.13. Simulation results of PID controllers for unit step set-point change

2.6. Discussions

2.6.1. Effect of  on the Tuning Parameters

The proposed IMC based PID tuning method has a single tuning parameter,

 , that is related to the closed-loop performance as well as the robustness of

the control system. It is important to analyze the effect of  on the PID

parameters, K c ,  I and  D . Consider a FOPDT model given by:

1e  s
GP  GD  (2.37)
1s  1
The PID parameters were calculated using the proposed method for different

closed loop time constant  values for each case of    0.25, 0.5, 0.75 and
1.0 .

55
Figure 2.14 shows the K c variation with  for different values of the  

ratio. As the   ratio decreases, the effect of  on K c is severe. From the

above figure, it is clear that K c becomes less sensitive to  with increasing

  . As the value of  increases sufficiently, K c variation is significantly

reduced.
25

Lamda=0.25
Kc vs   0.25
Lamda=0.5
Kc vs   0.50
Kc    0.75
vs Lamda=0.75
Kc    1.0
vs Lamda=1

20

15

10

0
0 0.3 0.6 0.9 1.2

Figure 2. 14. Proportional gain  K c  setting for different  values

Figure 2.15 shows the variation of  I with  . The  I value increases initially

with increasing  for different   ratios, but as the   ratio increases

from 0.25 to 1.0, the variation of  I comparatively decreases, as clearly

indicated by the    1.0 line. The trend of  I increasing with  reverses

56
after a specific  value for each   ratio. The above mention   ratios,

 I starts decreasing with increasing  after some significantly bigger value

of  . It is important to note that even for large values of  ,  I has positive

value.
1.5

1.3

1.1

0.9

0.7

   0.25

계열1
0.5
   0.50

계열2
   0.75
계열3
   1.0
계열4

0.3
0 0.3 0.6 0.9 1.2

Figure 2. 15. Integral time constant   I  setting for different  values

Figure 2.16 shows the variation of  D with  for different   ratios. The

value of  D decreases with increasing  at different   ratios maintaining

to a positive value.

57
 0.5


계열1   0.25

계열2   0.50
   0.75
계열3

0.4    1.0
계열4

0.3

0.2

0.1

0
0 0.3 0.6 0.9 1.2

Figure 2. 16. Derivative time constant   D  setting for different  values

2.6.2.  guideline for PID parameter tuning

Since the closed-loop time constant  is the only user-defined tuning

parameter in the proposed tuning rule, it is important to have some 
guidelines to provide the best performance with a given robustness level.
Figure 2.17 shows the plot of   vs.   ratios for different Ms values for
the FOPDT model. Figure 2.18 shows the  guideline plot for the DIP
model, where  can be calculated for the desired  value at different Ms
values. Figure 2.19 shows the  guidelines for the SOPDT model. It is
necessary to describe the model reduction technique proposed by Skogestad 8.
Using Skogestad’s8 half rule, we can convert a SOPDT model into a FOPDT
model. For the   value of FOPDT converted, we obtain the   value
from the plot in Figure 2.19 for the SOPDT model. Although this model

58
reduction technique introduces some modeling error, it is within the
acceptable limit. Figures 2.20 and 2.21 show the  guideline plots for the
FODIP and FODUP models, respectively.
1

0.8

0.6

0.4

0.2

Ms=1.6
Ms=1.7
Ms=1.8
Ms=1.9
Ms=2.0
0
0 0.5 1 1.5 2

Figure 2. 17.  guidelines for FOPDT

25

20

15

10

Ms=1.6
Ms=1.8
Ms=2.0
0
0 2 4 6 8 10 12 14

Figure 2. 18.  guidelines for DIP

59
0.6

0.5

0.4

0.3

0.2

Ms=1.8
Ms=1.9
Ms=2.0
0.1
0 0.6 1.2 1.8 2.4 3

Figure 2. 19.  guidelines for SOPDT

0.5

0.4

0.3

0.2

0.1

Ms=1.6
Ms=1.8
Ms=2.0
0
0 0.5 1 1.5 2 2.5

Figure 2. 20.  guidelines for FODIP

60
 1.2

0.8

0.6

0.4

0.2

Ms=3.0
Ms=3.5
Ms=4.0
0
0 0.1 0.2 0.3 0.4 0.5

Figure 2. 21.  guidelines for FODUP

2.6.3. Beneficial range of the proposed method

The load performance by the proposed PID controller is superior as a lag

time dominates, but the superiority to that based on the conventional filter
diminishes as a dead time dominates. In the case of a dead time dominant
process (i.e.,    1 ), the filter time constant should be chosen as
     for stability. Therefore, the process pole at 1/  is not a
dominant pole in the closed-loop system. Instead, the pole at 1/ 

determines the overall dynamics. Thus, introducing the lead term   s  1 into
the filter to compensate the process pole at 1/  has little impact on the
speed of the disturbance rejection response. Furthermore, the lead term
generally increases the complexity of the IMC controller, which in turn
degrades the performance of the resulting controller by causing a large
discrepancy between the ideal and the PID controllers. It is also important to

61
note that as the order of the filter increases, the power of the denominator

term   s  1 also increases, which causes an unnecessarily slow output

response. As a result, in the case of a dead time dominant process a
conventional filter without any lead term could have an advantage. Figure
2.22 compares the IAE values of the load responses by the PID controllers
based on the proposed filter and the conventional filter for the process model,
GP  GD  10e  s (1s  1) . The plot clearly indicates that the IAE gap

between the proposed and conventional filter decreases as the   ratio

increases.
The robustness of the controller based conventional filter and proposed study
for dead time dominant process has similar performance for perturbation
uncertainty in process parameters
40
Proposed Filter
Conventional Filter

35

30

25
IA E

20

15

10

0
0 0.6 1.2 1.8 2.4 3

Figure 2. 22. Performance of the proposed filter vs. the conventional filter

62
2.6.4. Optimum filter structure for IMC-PID design

One common problem with the conventional IMC-PID approaches is that the
IMC filter is usually selected based on the resulting IMC performance while
the ultimate goal of the IMC filter design is to obtain the best PID controller.
In the conventional approach for the filter design, it is assumed that the best
IMC controller results in the best PID controller. However, since all the
IMC-PID approaches utilize some kind of model reduction techniques to
convert the IMC controller to the PID controller, an approximation error
necessarily occurs. Therefore, if the IMC filter structure causes a significant
error in conversion to the PID controller, although it gives the best IMC
performance, the resulting PID controller could have poor control
performance. The performance of the resulting PID controller depends on
both the conversion error and the dead time approximation error, which is
also directly related to the filter structure and the process model. Therefore,
there exists an optimum filter structure for each specific process model that
gives the best PID performance. For a given filter structure, as  decreases
the discrepancy between the ideal and the PID controller increases while the
nominal IMC performance improves. This indicates that an optimum 
value also exists which balances these two effects to give the best
performance. Therefore, the best filter structure as defined in this paper is
that which gives the best PID performance for the optimum  value.

To find the optimum filter structure, we evaluate the IMC filters with the

  s  1   s  1
r rn
structure of for the first order models and

 s    s  1
r 2r n
2
2
 1s  1 for the second order models, where r and n are

63
varied from 0 to 2, respectively. Our investigation shows that a high order
filter structure generally gives a better PID performance than a low order
filter structure. For example, for an FOPDT model, it is found that the high
order filter, f  s    s  1 2  s  1 3 , provides the best disturbance rejection in
terms of IAE. Based on the optimum filter structures, we derived the PID
controller tuning rules for several representative process models, which are
listed in Table 2.1.

Figure 2.23 shows the variation of IAE with  for several tuning methods
for the FOPDT model studied in the earlier sections (in example 2.1). The
tuning rules proposed by Horn et al.3 and Lee et al.4 are based on the same

filter f  s     s  1   s  1 . Horn et al.3 use a 1/1 Pade approximation for the

2

dead time for calculating  and the PID parameters also. Lee et al. 4 obtained
the PID parameters using a Maclaurin series approximation. Since both two
methods use the same IMC filter structure, the IMC controllers of Lee et al.4
and Horn et al.3 coincide with each other as seen in Figure 2.23. Due to the
approximation error in e  s when calculating PID, the performance of Lee et
al.4 method shows a clear advantage over Horn et al. 3 method. It is clear from
this figure that down to some optimum  value, the ideal (or IMC) and the
PID controllers have no significant difference in performance, and after some
minimum IAE point the gap rises sharply towards unstable limits. The
smallest IAE value can be achieved by the proposed tuning method while the
Horn et al.3 tuning method shows the worst performance. It is also apparent
that for the case of model mismatch where a large  value is required, the
proposed method provides the best performance.
It is also worth while to visualize the performance and robustness of the

controller design. The M s and IAE are well known indices for robustness and

64
performance, respectively. Figure 2.24 shows the plot of Ms vs. minimum
IAE for the different tuning methods for the FOPDT model used in example
2.1. The figure clearly illustrates that for a constant Ms value, the PID
controller by the proposed method always produces a lower IAE value than
those by the other tuning rules.

3.5

3
IA E

2.5

2
Proposed IMC
Proposed IMC- PID
Lee et al. IMC
Lee et al. IMC- PID
Horn et al. IMC
Horn et al. PID
DS-d Ideal
DS-d PID
1.5
0.3 0.6 0.9 1.2 1.5

Figure 2. 23. Plot of  vs. IAE for different tuning rules for FOPDT

65
 5
Proposed
Lee et al.
DS-d
Horn et al.

4.5

3.5
Ms

2.5

1.5
1.5 3 4.5 6 IAE 7.5 9

Figure 2. 24. Plot of Ms vs. IAE for different tuning rules for FOPDT

2.7. Conclusions

Optimum IMC filter structures were proposed for several representative

process models to improve the disturbance rejection performance of the PID
controller. Based on the proposed filter structures, tuning rules for the PID
controller were derived using the generalized IMC-PID method. The
simulation results demonstrate the superiority of the proposed method when
the various controllers are all tuned to have the same degree of robustness in
terms of maximum sensitivity. The proposed method becomes more
beneficial as the process is lag time dominant. The robustness analysis was
conducted by inserting a 20% perturbation in each of the process parameters
in the worst direction, with the results demonstrating the robustness of the
proposed method against the parameter uncertainty. The closed-loop time

66
constant  guidelines were also proposed for several process models over a

wide range of   ratios.

2.8 Literature Cited

(1) Rivera, D. E.; Morari, M.; Skogestad, S. Internal Model Control. 4. PID
Controller Design. Ind. Eng. Chem. Process Des. Dev. 1986, 25, 252.
(2) Chien, I.-L.; Fruehauf, P. S. Consider IMC Tuning to Improve Controller
Performance. Chem. Eng. Prog. 1990, 86, 33.
(3) Horn, I. G.; Arulandu, J. R.; Christopher, J. G.; VanAntwerp, J. G.;
Braatz, R. D. Improved Filter Design in Internal Model Control. Ind. Eng.
Chem. Res. 1996, 35, 3437.
(4) Lee, Y.; Park, S.; Lee, M; Brosilow, C. PID Controller Tuning for
Desired Closed-Loop Responses for SI/SO Systems. AIChE J. 1998, 44,
106-115.
(5) Chen, D.; Seborg, D. E. PI/PID Controller Design Based on Direct
Synthesis and Disturbance Rejection. Ind. Eng. Chem. Res. 2002, 41,
4807-4822.
(6) Morari, M.; Zafiriou, E. Robust Process Control; Prentice-Hall:
Englewood Cliffs, NJ. 1989
(7) Seborg, D. E.; Edgar, T. F.; Mellichamp, D. A. Process Dynamics and
Control; John Wiley & Sons; Second Edition, New York, 2004.
(8) Skogestad, S. Simple Analytic Rules for Model Reduction and PID
Controller Tuning. J. Process Control. 2003, 13, 291-309.
(9) Lee, Y.; Lee, J.; Park, S. PID Controller Tuning for Integrating and
Unstable Processes with Time Delay. Chem. Eng. Sci. 2000, 55, 3481-
3493.

67
(10) Smith, C. L.; Corripio, A. B.; Martin, J. Controller Tuning from
Simple Process Models. Instrum. Technol. 1975, 22 (12), 39.
(11) Ziegler, J. G.; Nichols, N. B. Optimum Settings for Automatic
Controllers. Trans. ASME 1942, 64, 759-768.
(12) Åström, K. J.; Panagopoulos, H.; Hägglund, T. Design of PI
Controllers Based on Non-Convex Optimization. Automatica 1998, 34,
585-601.
(13) Åström, K. J.; Hägglund, T. PID Controllers: Theory, Design, and
Tuning, 2nd ed,; Instrument Society of America: Research Triangle Park,
NC, 1995.
(14) Huang, H. P.; Chen, C. C. Control-System Synthesis for Open-loop
Unstable Process with Time Delay. IEE Process-Control Theory and
Application 1997, 144, 334.
(15) Huang, C. T.; Lin, Y. S. Tuning PID Controller for Open-loop
Unstable Processes with Time Delay. Chem. Eng. Communications 1995,
133, 11.
(16) De Paor, A. M. Controllers of Ziegler Nichols Type for Unstable
Process with Time Delay. International Journal of Control 1989, 49,
1273.
(17) Rotstein, G. E.; Lewin, D. R. Control of an Unstable Batch Chemical
Reactor. Computers in Chem. Eng. 1992, 16 (1), 27.
(18) Poulin, ED.; Pomerleau, A. PID Tuning for Integrating and Unstable
Processes. IEE Process Control Theory and Application 1996. 143(5),
429.

68
3. AN ENHANCED PERFORMANCE PID FILTER
CONTROLLER FOR FIRST ORDER TIME DELAY
PROCESSES

3.1 Introduction

Proportional integral derivative (PID) controllers have been the most popular
and widely used controllers in the process industries because of their
simplicity, robustness and wide ranges of applicability with near-optimal
performance. However, it has been noticed that many PID controllers are
often poorly tuned and a certain amount of effort has been made to
systematically resolve this problem.

The effectiveness of the internal model control (IMC) design principle has
made it attractive in the process industries, where many attempts have been
made to exploit the IMC principle to design PID controllers for both stable
and unstable processes (Morari and Zafiriou, 1989). The IMC-PID tuning
rules have the advantage of using only a single tuning parameter to achieve a
clear trade-off between the closed-loop performance and robustness. The PID
tuning methods proposed by Rivera et al. (1986), Morari and Zafiriou (1989),
Horn et al. (1996), and Lee et al. (1998) are typical examples of the IMC-
PID tuning method. The direct synthesis (DS) method proposed by Smith et
al. (1975) and the direct synthesis for the disturbance (DS-d) method
proposed by Chen and Seborg (2002) can also be categorized into the same
class as the IMC-PID methods, in that they obtain the PID controller

69
parameters by computing the ideal feedback controller which gives a
predefined desired closed-loop response. Although the ideal controller is
often more complicated than the PID controller for time delayed processes,
the controller form can be reduced to that of either a PID controller or a PID
controller cascaded with a low order filter by performing appropriate
approximations of the dead time in the process model.
The control performance can be significantly enhanced by cascading the PID
controller with a lead/lag filter, as given by Eq. (3.1).

 1  1  as
Gc  K c  1 
  I s
 Ds 
 1  bs
(3.1)

where Kc ,  I and  D are the proportional gain, integral time constant, and
derivative time constant of the PID controller, respectively, and a and b are
the filter parameters.
The structure of the PID controller cascaded with a filter was also suggested
by Rivera et al. (1986), Morari and Zafiriou (1989), Horn et al. (1996), Lee
et al. (1998) and Dwyer (2003). The PID filter controller in Eq. (3.1) can
easily be implemented in modern control hardware.
It is essential to emphasize that the PID controller designed according to the
IMC principle provides excellent set-point tracking, but has a sluggish
disturbance response, especially for processes with a small time-delay/time-
constant ratio (Morari and Zafiriou, 1989; Chien and Fruehauf, 1990; Horn et
al., 1996; Lee et al., 1998; Chen and Seborg, 2002; Skogestad, 2003). Since
disturbance rejection is much more important than set-point tracking for
many process control applications, a controller design that emphasizes the
former rather than the latter is an important design goal that has recently been
the focus of renewed research.

70
 In the present study, a simple and efficient method is proposed for the
design of a PID filter controller with enhanced performance. A closed-loop
time constant    guideline is recommended for a wide range of time-
delay/time-constant ratios. A simulation study was performed to illustrate the
superiority of the proposed method for both nominal and perturbed
processes.

3.2. IMC Controller Design Procedure

Figures 3.1-(a) and (b) show the block diagrams of the IMC control and
equivalent classical feedback control structures, respectively, where GP is the
process, G P the process model, q the IMC controller, f R the set-point filter,
and Gc the equivalent feedback controller.
For the nominal case (i.e., GP  G%P ), the set-point and disturbance responses
in the IMC control structure can be simplified as:

y  GP qr  (1  G%P q)GD d (3.2)

According to the IMC parameterization (Morari and Zafiriou, 1989), the

process model G%P is factored into two parts:

%p p
G (3.3)
P m A

where pm is the portion of the model inverted by the controller, pA is the

portion of the model not inverted by the controller and pA  0   1 . The

noninvertible part usually includes the dead time and/or right half plane zeros
and is chosen to be all-pass.
To obtain a good response for processes with poles near zero, the IMC
controller q should be designed to satisfy the following conditions.

71
 Disturbance d
Disturbance
GD
IMC GD d
Setpoint controller
filter Process
r y
Setpoint Controller Process
fR +- q Gp ++ filter
r y
fR +- Gc Gp ++
Process model
~
Gp -+

Fig. 3.1 Block diagram of IMC and classical feedback control systems:
(a) The IMC structure; (b) Feedback control structure

1. If the process Gp has poles near zero at z1 , z2 , L , z m , lim

x  then q should have
zeros at z1 , z2 ,L , zm .
2. If the process GD has poles near zero, zd1 , zd2 ,L , zdm , then 1  GP q should have
zeros at zd1 , zd2 ,L , zdm .
Since the IMC controller q is designed as q  pm1 f , the first condition is
satisfied automatically. The second condition can be fulfilled by designing
the IMC filter f as

im1  i s i  1
f 
( s  1) r (3.4)

where  is an adjustable parameter which controls the tradeoff between the

performance and robustness; r is selected to be large enough to make the
IMC controller (semi-)proper; i are determined by Eq. (3.5) to cancel the
poles near zero in GD .

pA ( im1 i s i  1)
1  GP q s  z  1 0 (3.5)
d 1 ,L zdm
( s  1) r sz d 1 ,L , zdm

Then, the IMC controller comes to be

( im1  i s i  1)
q  pm1 (3.6)
( s  1) r

Thus, the closed-loop response is

72
 pA  im1 i s i  1  pA  im1  i s i  1 
y
  s  1
r
r  1 
   s  1
r
 GD d
 (3.7)
 

From the above design procedure, one can achieve a stable closed-loop
response by using the IMC controller.
3.3. PID filter Controller Design for FOPDT Process
The ideal feedback controller that is equivalent to the IMC controller can be
expressed in terms of the internal model G P and the IMC controller q :

q
Gc  (3.8)
1  G P q

Substituting Eqs. (3.3) and (3.6) into Eq. (3.8) gives the ideal feedback
controller:

( im1  i s i  1)
pm1
( s  1) r
Gc  (3.9)
pA  im1  i s i  1
1
  s  1
r

Let us consider the first order plus dead time (FOPDT) process, which is
most widely utilized in the chemical process industries, as a representative
model.

Ke  s
GP  GD 
 s 1
(3.10)

where K is the gain,  the time constant, and  the time delay. The IMC
filter structure is

 s 1
f  (3.11)
  s  1 2

It is noticed that the IMC filter form in Eq. (3.11) was also utilized by Lee et
al. (1998) and Horn et al. (1996). The resulting IMC controller becomes

73
   s  1   s  1
q (3.12)
K   s  1
2

Therefore, the ideal feedback controller is obtained as

  s  1   s  1
Gc 
K   s  1  e s   s  1 
2 (3.13)
 

Since the ideal feedback controller in Eq. (3.13) does not have the PID filter
controller form, the remaining issue is how to design the PID filter controller
that approximates the ideal feedback controller most closely.
Approximating the dead time e s with a 2/2 Pade expansion

  s  2s2 
1   
2 12
e  s
 
(3.14)
  s  2 s2 
1   
 2 12 

results in Gc as

  s  2s2 
  s  1 1      s  1
 2 12 
Gc 
 2  s  2s2    s  2s2  (3.15)
K   s  1  1       s  1  1   
  2 12   2 12  

It is important to note that the 2/2 Pade approximation is precise enough to

convert the ideal feedback controller into a finite dimensional feedback
controller with barely any loss of accuracy. Expanding and rearranging Eq.
(3.15) gives
  s  2 s2 
1   
2 12    s  1   s  1
Gc  
K  2      s    2  2  2    2 2  
       2          (3.16)
  2   12 6 2  2  12  3 
1   2      s   2     
s  s
 2      

 
 

74
As seen in Eq. (3.16), the resulting controller has the form of the PID
controller cascaded with a high order filter. The analytical PID formula can
be obtained as
  
KC  ; I  ; D  (3.17)
2 K  2      2 6

The value of the extra degree of freedom  is selected so that it cancels out
the open-loop pole at s  1  that causes a sluggish response to load

1    s  1 e s   s  1 2 
disturbances. From Eq. (3.5), this requires   s 1   0 . Thus, the

value of  is obtained as

   2 
   1  1   e   (3.18)
    

Furthermore, it is obvious from Eq. (3.5) that the remaining part of the
denominator in Eq. (3.16) contains the factor   s  1 . Therefore, the filter
parameter b in Eq. (3.1) can be obtained by taking the first derivative of Eq.
(3.19) below

    2  2  2    2 2 
2   
 2       12 6

2  2  12 
     s3
1
 2     
s
 2     
s 
 2      (3.19)
 cs 2

 bs  1 
  s  1

and substituting s  0 as

b
  2     2   (3.20)
 2     

The filter parameter a in Eq. (3.1) can be easily obtained from Eq. (3.16) as
a (3.21)
Since the high order cs 2 term has little impact on the overall control
performance in the control relevant frequency range, the remaining part of

75
the fraction in Eq. (3.16) can be successfully approximated to a simple first
order lead/lag filter as (1  as) /(1  bs ) . Our simulation result (although not
shown in this paper) also confirms the validity of this model reduction.
The lead term   s  1 in the closed-loop transfer function of Eq. (3.7) causes
excessive overshoot in the set-point response, which can be eradicated by
adding the set-point filter f R as:

 s  1
fR  (3.22)
 s 1

where 0    1 . The extreme case with   0 has no lead term in the set-point
filter which would cause a slow servo response. On the other hand,   1
means that there is no set-point filter.  can be adjusted online to obtain the
desired speed of the set-point response. The proposed study is also applicable
to the process with negligible dead time while it is mainly focused on the first
order time delay process.

3.4. Robust Stability

The well-known robust stability theorem can be utilized to analyze the robust
stability of the proposed controller.
Robust Stability Theorem (Morari and Zafirou, 1989): Let us assume that all
plants Gp in the family 

 Gp  i   G
%  i  
 l m   
p
  Gp :
%  i 
G
(3.23)
 p 

have the same number of RHP poles and that a particular controller Gc

stabilizes the nominal plant G%p . Then, the system is robustly stable with the

controller Gc if and only if the complementary sensitivity function % for the

nominal plant G%p satisfies the following bound:

76
 %l m  sup %l m     1 (3.24)
 

Since % G%p q  G%p %

pm1 f for the IMC controller, the resulting Eq. (3.24)
becomes:

m   1  
%% 1
G p pm f l m (3.25) Thus,

the above theorem can be interpreted as %%

l m  1 %=1 G 1
, which
p pm f

guarantees robust stability when the multiplicative model error is bounded by

 m  s   l mm .

% s   m  s  1 (3.26)


where  m  s  defines the process multiplicative uncertainty bound. i.e.,


 m  s   Gp  G%p G%p .  This uncertainty bound can be utilized to represent the

model reduction error, process input actuator uncertainty, and process output
sensor uncertainty, etc., which are very frequent in the actual process plants.
For the FOPDT process, the complementary sensitivity function % s  can be
obtained as

% s  
  s  1 e s
(3.27)
  s  1 2

Substituting Eq. (3.27) and  into Eq. (3.26) yields the robust stability
constraint required for tuning the adjustable parameters 

    2  
 1   1   e    s  1
       1
 (3.28)
  s  1 2
m  s 


77
Substituting s  i into Eq. (3.28) results in

 2 
 2        2 
2

  1  1    e   1
      1
 
 2 2  1

 m    (3.29)


It is possible for uncertainty to occur in any of the three process parameters

i.e.,  ,  , and K . Consequently, we have to consider the uncertainty in the
different parameters separately. Let us consider the FOPDT process having
the uncertainty in all three parameters as

 K  K  e    s
Gp  (3.30)
  s  1   s  1

It is most common practice that the FOPDT model approximated from the
high order process in the real process plant. Due to this for the time constant
uncertainty it is assumed that the small time constant  is neglected/missing
in developing the nominal model as considered in Eq. (3.30) (Seborg et al.,
2004). Then the process multiplicative uncertainty bound becomes
 K   s
1  K  e
m  s     1 (3.31)
  s  1

Substituting the above result into Eq. (3.29), we obtain the robust stability
constraint as follows:
 2 
 2        2 
2

  1  1    e   1
   
  
1
,   0
    1
2 2
 K   i
1  K  e
(3.32)
  1
  i  1

The above robust stability constraint is very useful to adjust  where there is
uncertainty in the process parameters. The robust stability constraint in Eq.

78
(3.32) can also be used to determine the maximum allowable values of
uncertainty in K   and  or various combinations of them for which

robust stability can be guaranteed. For example, a plot of %   l mm    vs. 

can be constructed for a small value of any parametric uncertainty and/or

combination of different uncertainties.

3.5. Simulation Study

This section deals with the simulation study conducted for three
representative FOPDT processes: the lag time dominant process, the equal
dead time and lag time process, and the dead time dominant process.
To evaluate the robustness of a control system, the maximum sensitivity, Ms,

which is defined by Ms  max 1/[1  Gp Gc (iw)] , is used. Since the Ms is the inverse
of the shortest distance from the Nyquist curve of the loop transfer function
to the critical point  1, 0  , a small Ms value indicates that the stability
margin of the control system is large. The Ms is a well-known robustness
measure and is used by many researchers (Skogestad and Postlethwaite,
1996; Åström et al., 1998; Chen and Seborg, 2002; Skogestad, 2003;).
Typical values of Ms are in the range of 1.2  2.0 (Åström et al., 1998; Seborg
et al., 2004). To ensure a fair comparison, it is widely accepted for the
model-based controllers (DS-d, DS, and IMC) to tune by adjusting  so that
the Ms values become the same values. Therefore, throughout all our
simulation examples, all of the controllers compared were designed to have
the same robustness level in terms of the maximum sensitivity, Ms .
To evaluate the closed-loop performance, two performance indices were
considered in the case of both a step set-point change and a step load
disturbance, viz., the integral of the time-weighted absolute error (ITAE)

79
 

defined by ITAE   t e  t  dt, and the overshoot which acts as a measure of how
0

much the response exceeds the ultimate value following a step change in the
set-point and/or disturbance.
In this paper, the simulation study has been conducted using the PID
controller in the form of Eq. (3.1). However, for real implementation, the
 1  D s  1  as
“parallel form” of the PID controller, G  s   K c 1    , which is widely
  I s 0.1 D s  1  1  bs

used in the real processes, can be applied to approximately the same

performance.
To evaluate the usage of manipulated input values, we compute TV of the

input u  t  , which is the sum of all of its movement of up and down. If we



discretize the input signal as a sequence [u1 , u2 , u3 ...., ui ...], then TV   ui 1  ui

i 1

should be as small as possible. TV is a good measure of the smoothness of a

signal (Skogestad and Postlethwaite, 1996; Chen and Seborg, 2002;
Skogestad, 2003).

3.5.1 Example 3.1: Lag time dominant process     0.01

Consider the following FOPDT process (Chen and Seborg, 2002; Seborg et
al., 2004):
100e 1s
Gp  GD 
100s  1
(3.33)

The proposed PID filter controller is compared with other controllers based
on existing methods, such as the DS-d method, and those proposed by Rivera
et al. (1986), Horn et al. (1996), Lee et al. (1998) and Lee et al. (1998) with
a conventional filter. The controller parameters, including the performance
and robustness matrix, are listed in Table 3.1. In order to ensure a fair
comparison, all of the controllers compared are tuned to have Ms  1.94 by

80
 adjusting  . Figure 3.2 compares the set-point and load responses obtained
using the proposed method, the DS-d method, and the methods proposed by
Lee et al. (1998) and Horn et al. (1996). The 2DOF controller using the set-
point filter was used in the DS-d method and the methods proposed by Lee et
al. (1998) and Horn et al. (1996) to obtain an enhanced set-point response. It
is important to note that the set-point filter used for the set-point response has
a clear benefit when the process is lag time dominant. In this case, it is
observed that 0.4    0 gives smooth and robust control performances. In the
proposed controller,  in the set-point filter is selected as   0.45. The
closed-loop response for both the set-point tracking and disturbance rejection
signifies that the proposed method provides a superior response for the same
robustness.
The robust performance is evaluated by inserting a perturbation uncertainty
of 20% in all three parameters in the worst direction simultaneously and
finding the actual process as Gp  GD  120e1.2 s  80s  1 . The simulation results for
the model mismatch for various methods are given in Table 3.2. The
performance and robustness indices obviously demonstrate that the proposed
method has more robust performance than the others.

Table 3.1 PID controller parameters and performance matrix for example 3.1     0.01
Set-point
Tuning methods  Kc I D Disturbance
ITAE Overshoot ITAE Overshoot
M
Proposed method 1.131 0.124 0.50 0.167 1.96 0.007 14.55 1.206
N
Lee et al. (1998) 1.330 0.806 3.947 0.3068 8.46 0.0 19.77 1.314
O
DS-d 1.202 0.826 4.059 0.353 3.13 0.015 20.43 1.273

Horn et al. (1996)P 1.689 15.038 100.50 0.497 12.45 0.0 31.18 1.478
Rivera et al. (1986)Q 0.408 0.714 100.50 0.4975 3.86 0.025 3785.0 1.411

Lee et al. (1998)R 0.248 0.805 100.41 0.399 3.15 0 .018 3354.0 1.273

 1  1  as
81 1.45s  1
M Gc  Kc  1   D s  , where a  3.222 , b  0.139; f R 
  Is  1  bs 3.22s  1
1 2.03s  1
N fR  ; O  f R 
3.66s  1 1.43s 2  4.06  1
 1  1  as 1
P   Gc  Kc 1   s   D s  1  bs  cs2 , where a  4.311 , b  100.2 , c  21.34; f R 
 I  4.31s  1
 1  1
Q Gc  K c 1   s   D s  1  bs , where b  0.145
 I 
R The Lee et al. (1998) method based on the conventional IMC filter form of f  1   s  1
* 1DOF controller is used only for the methods of Rivera et al. (1986)Q and Lee et al. (1998)R
Table 3.2 Robustness analysis for example 3.1

Tuning methods Set-point Disturbance

Proposed ITAE Overshoot ITAE Overshoot
Proposed methodM 13.36 0.3418 33.36 1.8798
N
Lee et al. (1998) 14.40 0.0387 76.26 1.7399
DS-dO 40.41 0.2817 98.25 1.7011
Horn et al. (1996)P 16.32 0.0091 62.27 2.2096
Rivera et al. (1986)Q 28.73 0.5354 3766.0 2.1152
Lee et al. (1998)R 21.21 0.5797 3338.0 1.9450

82
 1 (a)

0.8
Process variable

0.6

0.4 Proposed method

Lee et al. (1998)
0.2 DS-d
Horn et al. (1996)
0
0 3 6 Time [min] 9 12 15

1.5
(b) Proposed method
Lee et al. (1998)
DS-d
Process variable

1 Horn et al. (1996)

0.5

0
0 3 6 9 12 15
Time [min]

Fig. 3.2 Simulation results for example 3.1

3.5.2 Example 3.2: Equal lag time and dead time process     1

Consider the process model described by Chen and Seborg (2002) as follows

1e 1s
Gp  GD  (3.34)
1s  1

The proposed PID filter controller is compared with the DS-d controller and
the controllers designed by Lee et al. (1998), Horn et al. (1996), Rivera et al.
(1986) and Lee et al. (1998) with a conventional filter. The controller
parameter values are listed in Table 3.3 along with the performance matrix,
where Ms  1.84 is selected for all controller designs. Unit step changes are
introduced both in the set-point and in the disturbance for the simulation. The
simulation results in Figure 3.3 indicate that both the disturbance and the set-
point responses are faster in the proposed controller. The 2DOF controller
structure is used for each design method except Rivera et al. (1986), and Lee

83
et al. (1998) with a conventional filter.   0 is selected for the proposed
controller. It is clear from Figure 3.3 and Table 3.3 that the proposed
controller exhibits better performance for both the set-point and disturbance
response.

Table 3.3 PID controller parameters and performance matrix for example 3.2
   1

Set-point Disturbance
Tuning methods  Kc I D ITAE Overshoot ITAE Overshoot
Proposed methodM 0.499 0.458 0.5 0.166 2.28 0.0034 2.66 0.626
Lee et al. (1998)N 0.596 1.042 1.304 0.270 2.83 0.0025 3.31 0.622
DS-dO 0.771 1.055 1.444 0.313 2.71 0.0006 3.85 0.633
P
Horn et al. (1996) 0.73 1.010 1.50 0.333 3.59 0.0002 4.12 0.672
Q
Rivera et al. (1986) 0.503 0.998 1.50 0.333 2.33 0.1481 4.26 0.670
Lee et al. (1998)R 0.309 1.055 1.382 0.289 1.95 0.1273 3.52 0.634

1
M a  0.907, b  0.102; fR 
0.908s  1
1 0.722s  1
N fR  ; O fR 
0.94s  1 0.452s 2  1.44s  1
1
P a  0.975, b  1.179, c  0.179; f R 
0.976s  1
Q b  0.167
* 1DOF controller is used only for the methods of Rivera et al. (1986)Q and Lee et al. (1998)R

84
 1 (a)

P ro c e s s v a ria b le
0.8

0.6

0.4 Proposed method

Lee et al. (1998)
0.2 DS-d
Horn et al. (1996)
0
0 2 4 6 8
Time [min]
0.7
(b) Proposed method
Lee et al. (1998)
P ro c e s s v a ria b le

0.5 DS-d
Horn et al. (1996)

0.3

0.1

-0.1
0 2 4 6 8
Time [min]

Fig. 3.3 Simulation results for example 3.2

3.5.3 Example 3.3: Dead time dominant process     5

Consider the process with a long dead time studied by Luyben (2001) and
Chen and Seborg (2002)

1e 5 s
Gp  GD  (3.35)
1s  1

The proposed and aforementioned design methods are compared. The

controller settings with the performance matrices are given in Table 3.4. All
of the controllers are designed to have Ms  1.74. Since in the case of a dead
time dominant process, the 1DOF controller is sufficient to achieve
satisfactory control performance, no set-point filter is used for any design
method.
The set-point and load responses are shown in Figure 3.4. From this figure, it
is apparent that the proposed controller and the one designed by Lee et al.
(1998) with the conventional filter provide similar responses, while the DS-d
and Horn et al. (1996) methods exhibit sluggish responses and take a long
time to settle the response.

85
The proposed controller has excellent performance when the lag time
dominates, but its performance becomes similar to that of the methods based
on the conventional filter when the dead time dominates. When    1 , the
filter time constant should be chosen as      for the sake of closed-loop
stability. Therefore, the process pole at 1/  is not a dominant pole in the
closed-loop system. Instead, the pole at -1/λ determines the overall dynamics.
Thus, introducing the lead term   s  1 into the IMC filter to compensate the
process pole at 1/  has little impact on the disturbance response.
Furthermore, the lead term usually increases the complexity of the IMC
controller, which in turn degrades the performance of the resulting PID
controller by causing a larger discrepancy between the ideal feedback
controller and thus the PID controller.
It is also important to note that as the order of the filter increases, the power
of the denominator term   s  1 also increases, which can cause an
unnecessarily slow output response. As a result, in the case of a dead time
dominant process, the PID controller based on the IMC filter that includes no
lead term offers better performance.

Table 3.4 PID controller parameters and performance matrix for example 3.3     5
Set-point
Tuning methods  Kc I D Disturbance
ITAE Overshoot ITAE Overshoot
Proposed methodM 1.408 0.366 2.5 0.833 29.95 0.045 68.34 0.989
N
Lee et al. (1998) 1.423 0.408 2.799 0.721 30.79 0.064 67.67 0.992
DS-dO 2.706 0.316 2.555 0.053 37.57 0.011 85.88 0.984
P
Horn et al. (1996) 2.648 0.430 3.5 0.714 41.76 0.050 87.01 0.993
Q
Rivera et al. (1986) 3.117 0.431 3.5 0.714 40.86 0.043 87.06 0.992
R
Lee et al. (1998) 1.798 0.417 2.838 0.759 30.66 0.064 66.77 0.990

M a  0.998, b  0.689
P a  2.164, b  3.155, c  2.156 , Q b  0.96

86
 1 (a)
P ro ce ss va ria b le

0.8

0.6

Proposed method
0.4 DS-d
Horn et al. (1996)
0.2 Lee et al. (1998)R

0
0 5 10 15 20 25 30
Time [min]

1 (b)
Proposed method
0.8
P ro ce ss va ria b le

DS-d
Horn et al. (1996)
0.6
Lee et al. (1998)R

0.4

0.2

-0.2
0 5 10 15 20 25 30
Time [min]

Fig. 3.4 Simulation results for example 3.3

3.5.4 Example 3.4: Polymerization process

An important viscosity loop in a polymerization process was identified by

Chien et al. (2002) as follows:

3e10 s
Gp  GD 
100s  1
(3.36)

The above-mentioned process has a large open-loop time constant of 100 min
and a dead time of 10 min, which is also quite noteworthy. Chien et al.
(2002) designed the PI controller with the modified Smith Predictor (SP) by
approximating the above process in the form of an integrating model with a
long dead time. Figure 3.5 compares the nominal responses by the proposed
PID filter controller and that by the modified SP. In the proposed controller,
  8.0 is selected and the resulting tuning parameters are obtained as

87
 K c  0.6446,  I  5.0,  D  1.6667, a  23.4146 and b  0.9781. The simulation was
conducted by inserting the step set-point change at t  0 followed by a load
step change of -1.0 at t  90 .
The proposed controller uses a simple feedback control structure without any
dead time compensator. Nevertheless, the proposed PID filter controller
provides a superior performance, as shown in Figure 3.5. The disturbance
rejection afforded by the proposed controller has a smaller settling time,
whereas the modified SP controller described by Chien et al. (2002) shows a
sluggish and required long settling time.
1

0.9

0.8

0.7
P ro c e s s v a ria b le

0.6

0.5

0.4

0.3

0.2

Proposed method
0.1
Chien et al. (2002)

0
0 30 60 90 120 150 180
Time [min]

Fig. 3.5 Simulation results of the polymerization process

As regards the set-point response, the modified SP controller has an initially

fast response, because of the elimination of the dead time, but afterwards it
becomes slow. On the other hand, the speed of the response for the proposed
controller is uniform and the settling time is similar to that by the modified
SP.
It is important to note that the SP control configuration has a clear advantage
of eliminating the time delay from the characteristic equation, which is very
effective to set-point tracking performance. However, this advantage is lost if

88
the process model is inaccurate. In order to evaluate the robustness against
model uncertainty, a simulation study was conducted for the worst case of
model mismatch by assuming that the process has a 20% mismatch in the
three process parameters in the worst direction, as follows

3.6e 12 s
Gp  GD 
80 s  1
(3.37)

The closed-loop responses are presented in Figure 3.6. Notice that the
proposed method and the modified SP method described by Chien et al.
(2002) have similar disturbance rejection responses for the model mismatch
case. However, the set-point response afforded by the modified SP controller
shows severe oscillation, while the proposed controller gives a more robust
response.
1.5
Proposed method
Chien et al. (2002)

1
P ro c e s s v a ria b le

0.5

0
0 100 200 300 400 500 600
Time [min]

Fig. 3.6 Simulation results of the polymerization process with model mismatch

In the proposed tuning rule, the closed-loop time constant  controls the
tradeoff between the robustness and performance of the control system. As 
decreases, the closed-loop response becomes faster and can become unstable.

89
On the other hand, as  increases, the closed-loop response becomes stable
but sluggish. A good tradeoff is obtained by choosing  to give an Ms value
in the range of 1.2  2.0 (Åström et al., 1998; Seborg et al., 2004). The 
guideline for several robustness levels is plotted in Figure 3.7.

10
Ms=1.4
Ms=1.5
Ms=1.6
Ms=1.8
1 Ms=1.9

0.1

0.01
0.01 0.1 1 10

Fig. 3.7  guideline for the proposed tuning method

3.6 Conclusions

A simple analytical design method for a PID controller cascaded with a

lead/lag filter was proposed based on the IMC principle in order to improve
its disturbance rejection performance. The proposed method also includes a
set-point filter to enhance the set-point response like the 2DOF controller
suggested by Lee et al. (1998), Horn et al. (1996) and Chen and Seborg
(2002). FOPDT processes with three representative different   ratios were
used for the simulation study. The proposed PID filter controller consistently
provides superior performance over the whole range of the   ratio, while

90
the other controllers based on the IMC-PID design methods take their
advantage only in a limited range of the   ratio. In particular, the proposed
controller shows excellent performance when the lag time dominates. The
proposed controller was also compared with the more sophisticated
controller, such as the modified Smith Predictor, in the case of the viscosity
loop in a polymerization process. The result shows that the proposed
controller gives satisfactory performance without the external dead time
compensator. A guideline of closed-loop time constant  was also proposed
for a wide range of   ratio.

3.7 Literature Cited

(19) Åström, K. J., H. Panagopoulos and T. Hägglund; “Design of PI

Controllers Based on Non-Convex Optimization,” Automatica, 34,
585–601 (1998)
(20) Chen, D. and D. E. Seborg; “PI/PID Controller Design Based on
Direct Synthesis and Disturbance Rejection,” Ind. Eng. Chem. Res.,
41, 4807–4822 (2002)
(21) Chien, I. L. and P. S. Fruehauf; “Consider IMC Tuning to Improve
Controller Performance,” Chem. Eng. Prog., 86, 33–41 (1990)
(22) Chien, I. L., S. C. Peng and J. H. Liu; “Simple Control Method for
Integrating Processes with Long Dead Time,” J. Process Control, 12,
391–404 (2002)
(23) Dwyer, A. O.; Handbook of PI and PID Controller Tuning Rules;
Imperial College Press, London, U.K. (2003)
(24) Horn, I. G., J. R. Arulandu, J. G. Christopher, J. G. Van Antwerp and
R. D. Braatz; “Improved Filter Design in Internal Model Control,” Ind.
Eng. Chem. Res., 35, 3437–3441 (1996)

91
(25) Lee, Y., S. Park, M. Lee and C. Brosilow; “PID Controller Tuning for
Desired Closed-Loop Responses for SI/SO Systems,” AIChE J., 44,
106–115 (1998)
(26) Luyben, W. L.; “Effect of Derivative Algorithm and Tuning Selection
on the PID Control of Dead Time Processes,” Ind. Eng. Chem. Res.,
40, 3605–3611 (2001)
(27) Morari, M. and E. Zafiriou; Robust Process Control, Prentice-Hall,
Englewood Cliffs, U.S.A. (1989)
(28) Rivera, D. E., M. Morari and S. Skogestad; “Internal Model Control. 4.
PID Controller Design,” Ind. Eng. Chem. Process Des. Dev., 25, 252–
265 (1986)
(29) Seborg, D. E., T. F. Edgar and D. A. Mellichamp; Process Dynamics
and Control, 2nd ed., John Wiley & Sons, New York, U.S.A. (2004)
(30) Skogestad, S.; “Simple Analytic Rules for Model Reduction and PID
Controller Tuning,” J. Process Control, 13, 291–309 (2003)
(31) Skogestad, S. and I. Postlethwaite; Multivariable Feedback Control;
Analysis and Design, John Wiley & Sons, New York, U.S.A. (1996)
(32) Smith, C. L., A. B. Corripio and J. Martin; “Controller Tuning from
Simple Process Models,” Instrum. Technol.,22, 39–44 (1975)

92
4. ANALYTICAL DESIGN OF PID CONTROLLER
CASCADED WITH A LEAD-LAG FILTER FOR TIME-
DELAY PROCESSES

4. 4.1. Introduction

Due to their efficient and robust performance with a simple algorithm,

proportional-integral-derivative (PID) controllers have gained wide
acceptance in most industrial applications. Furthermore, recent developments
of modern control systems have enabled the PID controller to be combined
with various simple control algorithms in a quick and easy manner to
enhance the control performance. To cascade a PID controller with a lead-lag
filter is a typical example of taking advantage of the facility by using the
modern control systems. However, in spite of its significant potential to
improve control performance, the PID●filter structure has not entered wide
use in the process industries. The only popular trial so far is to add a first- or
second-order lag filter in series with the PID controller for filtering high
frequency noises. This limited application is mainly due to the lack of
systematic design methods for the PID●filter control system.
Many previous studies have investigated the conventional PID controller
design since Ziegler and Nichols (1942) published their classical method. The
internal model control (IMC)-based approach has gained widespread
acceptance for the design of the PID controller in process industries because
of its simplicity, robustness, and successful practical applications (Rivera et

93
al., 1986; Morari and Zafiriou, 1989; Chien and Fruehauf, 1990; Horn et al.,
1996; Lee et al., 1998; Lee et al., 2006). It also has a practical advantage in
that a clear tradeoff between closed-loop performance and robustness is
achieved with a single tuning parameter. In the IMC-PID methods, the PID
controller parameters are obtained by first computing the ideal feedback
controller that gives a desired closed-loop response. It is well known that
most processes in the chemical industries can be represented by a first- or
second-order plus dead time model, either with or without single zero. In the
time-delay processes, the ideal feedback controllers are more complicated
than the standard PID controllers are. The controller form can then be
reduced to that of a PID controller by clever approximations of the process
dead time. However, all these methods concern the design of the conventional
PID controller, while the control system with the PID controller often only
shows a sluggish response and overshoots to the time-delay processes where
lag time dominates.
In this study, an analytical method for the design of a PID controller
cascaded with a second-order lead-lag filter is developed for various types of
time-delay process. The proposed PID●filter controller is designed based on
the IMC-PID principle and gives a better response than the conventional PID
controllers reported in earlier studies for obtaining the desired, closed-loop
response. Several examples are provided for comparing the results with the
conventional PID controllers. In the proposed PID●filter structure, the
resulting control system becomes equivalent to controlling a fast dynamic
process by integral control, which dramatically improves the performance.
Some discussions with the  guideline for particular robustness levels are
also provided.

94
4.2. IMC-PID approach for PID●filter controller design

Figs. 4.1-a and b show the block diagrams of IMC control and equivalent
classical feedback control structures, respectively, where GP is the process,
% the process model,
G q the IMC controller, and Gc the equivalent
P

feedback controller. In the IMC control structure, the controlled variable is

related as:

GP q 1  G%P q 
C R  GD d (4.1)
%
1  q GP  G
P 1  q GP  G

%
P   

% ), the set-point and disturbance responses
For the nominal case (i.e., GP  GP

are simplified as:

C %
R
 GP q (4.2)
C
 1  G%P q  GD (4.3)
d 

According to the IMC parameterization (Morari, 1989), the process model

% is factored into two parts:
GP

%P P
G (4.4)
P M A

where PM is the portion of the model inverted by the controller, PA is the

portion of the model not inverted by the controller and PA  0   1 . The

noninvertible part usually includes dead time and/or right half plane zeros
and is chosen to be all-pass.
Disturbance
Disturbance d GD
d
IMC GD
controller Process Controller Process
R C
+- q Gp ++ R C
setpoint +- Gc Gp ++
setpoint Output
Process model
~
Gp -+

Fig. (4.1-a) IMC structure Fig. (4.1-b) Classical feedback control structure

95
 Fig. 4.1 Block diagram of IMC and classical feedback control

The IMC controller is designed by:

q  PM -1 f (4.5)
where the IMC filter f is usually set as:
1
f (4.6)
( s  1)r
The ideal feedback controller equivalent to the IMC controller can be
% , and the IMC controller, q :
expressed in terms of the internal model, GP

q PM1
Gc   (4.7)
1  G%P q ( s  1) r  PA

Suppose that the time-delay process can be represented by first- or second-

order dynamics. The ideal feedback controller in Eq. (4.7) can be converted
into the PID controller cascaded with a second-order lead-lag filter by using
the following appropriate Pade approximations of the dead time:
 1  1  cs  ds 2
Gc  K c 1   Ds  (4.8)
 Is  1  as  bs
2

where K c ,  I and  D are the proportional gain, integral time constant, and

derivative time constant of the PID controller, respectively, and a, b, c and

d are the filter parameters. The second-order lead-lag filter is easily
implemented using the modern control hardware. The PID●filter controller in
Eq. (4.8) is an extension to the modified PID controller structure developed
by Rivera et al. (1986), Lee et al. (1998), Horn et al. (1996) and Dwyer
(2003).

4.2-1. Tuning rule for first-order plus dead time process

96
The most commonly used approximate model for chemical processes is the
first-order plus dead time (FOPDT) model as given below:
Ke s
GP  (4.9)
 s 1

The process model GP is factored into two parts: PM  K   s  1 and PA  e s . The

IMC controller is designed as q  PM f    s  1 K   s  1 for the desired closed-loop

-1

response, e /   s  1 . From Eq. (4.7), the ideal feedback controller is given

 s

 s 1
by Gc  . Approximating the dead time e  s with a 2/2 Pade
K   s  1  e  s 

expansion as 
e s  1   s / 2   2 s 2 /12   1   s / 2   s /12 ,
2 2
we obtain the following
tuning rule of the PID●filter controller after some simplification:

     2
KC  , I  , D  , a  , b , c , d  0 (4.10)
2K      2 6 2    12     

Note that the 2/2 Pade approximation is precise enough to convert the ideal
feedback controller into a PID●filter controller with little loss of accuracy.

4.2-2. Tuning rule for second-order plus dead time process

Consider a general second-order plus dead time (SOPDT) process system,
Ke  s
Gp  (4.11)
  2 s 2  2 s  1 
where  is the damping factor. The process model is decomposed into

 
PM  K  2 s 2  2 s  1 and PA  e  s . The resulting IMC controller is then given as

 
q  PM -1 f   2 s 2  2 s  1 K   s  1 for the desired closed-loop response, e  s /   s  1 .

 s  2 s  1
2 2

From Eq. (4.7), the ideal feedback controller is given by Gc 
K  s  1  e  s 
.

97
Substituting a 2/2 Pade approximation into the dead time term, we obtain
the tuning rule of the PID●filter controller given in Table 4.1.

98
Table 4.1. Design rules of the proposed PID●filter controller for various process models

Case Process model KC I D a b c d

A Ke  s      2  -
 s 1 2K      2 6 2    12     
    2
 s
B Ke  2
2
  s  2 s  1
2 2
 2K      2 6 2    12     

C

K   a s  1 e  s
s  2 s  1

2K     
      a   2   a    2 12   a 2   a 2   a 2 12  2 2
  
2 2
2 6   
D   a s  1
  s  2 s  1
Ke  s 
2 K      2 a 
      a   2   a    2 12   a 2   a 2   a 2 12  2 2
     2 a       2 a 
2 2
2 6
E K   a s  1 e  s        a   2   a    2 12   a 2   a 2   a 2 12  -
2 K      2 a       2 a 
 s 1 2 6      2 a 
F Ke  s 1 -    3  2 6  -  -
K    
s   s  1 3   
G K   a s  1 e  s 1 -      a    2   a 3   a 2 6  -
   6 2 a 3 
s   s  1 K      2 a 
     2 a 
a
3  3 
    2 a 


99
4.2-3. Tuning rule for SOPDT process with overshoot response
Consider a general SOPDT process with a lead term,
K   a s  1 e  s
Gp  (4.12)
 2
s 2  2 s  1

The negative zero in the model causes an excessive overshoot in the open-
loop response. Since the process model is factored into


PM  K   a s  1  2 s 2  2 s  1  and PA  e  s , the IMC controller is given as

q    2 s 2  2 s  1 K   s  1   a s  1 for the desired closed-loop response

e  s /   s  1 , while the equivalent ideal feedback controller is represented as

Gc 
 1   s / 2   s /12    s  2 s  1
2 2 2 2

K  1   s / 2   2 s /12    s  1   s  1    s  1  1   s / 2  
2
s /12  
2 2 by replacing the
a a

dead time with a 2/2 Pade approximation. Neglecting all terms in the
denominator higher than third-order by replacing

 1 s / 2  s 2 2
/12    a s  1   s  1 with  1   s / 2    a s  1   s  1 , we obtain the

proposed PID●filter structure. The truncation of the higher order terms has no
impact on the performance of the resulting controller. The tuning formula
derived is given in Table 4.1.
4.2-4. SOPDT and FOPDT processes with inverse response
Consider a SOPDT model with a positive zero,

Gp 
  a s  1 Ke  s
(4.13)
 2
s 2  2 s  1 
The noninvertible portion becomes PA    a s  1 e   a s  1 by applying the
 s

all-pass form for the non-minimum part   a s  1 . Accordingly, the invertible

portion is PM  K   a s  1   s  2 s  1 . For the desired closed-loop response,

2 2

100
  a s  1 e s   s  1   a s  1 , the ideal feedback controller is given as

Gc 
 s  2 s  1
2 2

. The PID●filter structure can be obtained in

K   a s  1   s  1    a s  1 e  s 

the same manner as follows in the SOPDT model with a lead term. The
resulting tuning rule is given in Table 4.l.
The same procedure can be used to obtain the tuning rule for a FOPDT
process with inverse response as,

K   a s  1 e  s
GP  (4.14)
 s 1

The resulting tuning rule is also given in Table 4.l.

4.2-5. First-order delayed integrating processes
Consider a first-order delayed integrating process,
Ke  s
Gp  (4.15)
s   s  1

G p is factored into PM  K s   s  1 and PA  e  s . For the desired closed-loop

response, e  s /   s  1 , the ideal controller can be converted to the PID●filter

controller by approximating the dead time with a 2/1 Pade approximation and
the resulting tuning rule is given in Table 4.1.
For the following first-order delayed integrating process with inverse
response,
K   a s  1 e  s
Gp  (4.16)
s   s  1

where G p is decomposed as PM  K   a s  1 s   s  1 , and PA    a s  1 e s   a s  1 .

The desired closed-loop response is set as   a s  1 e   s  1   a s  1 . The

 s

PID●filter controller can be derived from the ideal feedback controller by

101
using a 2/1 Pade approximation. The resulting tuning rule is given in Table
4.1. Note that the proposed IMC-PID approach essentially leads to a PD
structure (with no integral action) for integrating processes.

4.3. Simulation Studies

To evaluate the robustness of a control system, the maximum sensitivity, M s

, which is defined by M s  max 1/[1  G p Gc (iw)] , is used. Since M s is the

inverse of the shortest distance from the Nyquist curve of the loop transfer

function to the critical point  1, 0  , a small value indicates that the stability

margin of a control system is large. Typical values of M s are in the range of

1.2 2.0 as recommended by Åström and Hägglund (1998). For fair
comparison under the same robustness level, throughout all our simulation
examples all the controllers compared were designed to have the same Ms
value of Ms  1.60 . To evaluate the closed-loop performance, two
performance indices were considered to a step set-point change and a step
load disturbance: the integral of the time-weighted absolute error (ITAE),

defined by ITAE   t e  t  dt , and the overshoot as a measure of how

0

much the response exceeds the ultimate value following a step change in set-
point and/or disturbance. Robust performance was also evaluated in all
examples by perturbing +20% uncertainties in the gain and the dead time.

4.3.1. Example 4.1. Consider the following FOPDT process studied by Lee
et al. (1998):

1e 3s
GP  (4.17)
10s  1

102
Conventional PID controllers based on the methods of Lee et al. (1998) and
Rivera et al. (1986) are used for comparison. For the proposed method, Lee
et al. (1998)’s method and Rivera et al. (1986)’s method,  values of 2.031,
1.774, and 2.592, respectively, are selected so that every controller satisfies
Ms  1.60 . The closed-loop responses of the three controllers for the nominal
case are shown in Fig. 4.2. The values of controller parameters and resulting
performance indices are listed in Table 4.2.

Fig. 4.2 Simulation result of proposed tuning method for example 4.1

The simulation results indicate that the proposed PID●filter controller

provides a fast and smooth set-point response without any loss of disturbance
performance. The robust performance of the proposed controller is evident in
the results of the model mismatch case in Table 4.2. The proposed controller
still gives the smallest ITAE in the set-point response.

4.3.2. Example 4.2. Consider the following underdamped SOPDT process

model,

103
 1e 4 s
Gp  (4.18)
 
100 s 2  10 s  1

where   0.5 . The proposed PID●filter controller is compared with the

conventional PID controllers based on the methods of Lee et al. (1998) and

Chien & Fruehauf (1990). All controllers are tuned to have M s  1.60 by
adjusting  . The resulting parameter settings and performance indices are
listed in Table 4.3. Fig. 4.3 compares the closed-loop responses achieved by
the three controllers. The results show that the proposed controller exhibits
significant advantage over the other two controllers not only in terms of
servo performance but also in load performance. The proposed controller
also shows superior robust performance in the model-mismatch case.

Fig. 4.3 Simulation result of proposed tuning method for example 4.2

104
Table 4.2. Controller parameters and resulting performance indices for the FOPDT process

set-point disturbance
Method Kc I D a b c nominal 20% mismatch nominal 20%
case mismatch
ITAE overshoot ITAE overshoot ITAE overshoot ITAE overshoot
Proposed 0.298 1.5 0.50 0.605 0.302 10 14.09 0 27.33 0.25 80.06 0.32 75.86 0.41
  2.031
Lee et al. 2.292 10.9 0.856 - - - 27.30 0.04 36.86 0.27 75.02 0.32 71.17 0.40
  1.775
Rivera et 2.057 11.5 1.304 0.695 - - 39.31 0.03 47.78 0.24 91.58 0.35 86.38 0.44
al.   2.592

105
Table 4.3. Controller parameters and resulting performance indices for the underdamped SOPDT process
set-point disturbance
Method K c I  D a b c d nominal 20% mismatch nominal 20%
mismatch
ITAE overshoot ITAE overshoot ITAE overshoot ITAE overshoot
Proposed 0.298 2.0 0.666 0.807 0.538 10 100 24.8 0.0 49.6 0.24 271.9 0.34 275.5 0.39
  2.707
Lee et al. 1.268 10.43 9.893 - - - - 392.7 0.28 411.3 0.35 322.8 0.43 326.6 0.50
  2.1105
Chien 1.264 10.0 10.0 - - - - 424.7 0.30 441.6 0.37 333.2 0.43 337.5 0.50
and
Fruehauf
  3.91

106
4.3.3. Example 4.3. Consider the overdamped SOPDT process studied by
Seborg et al. (2004),

2e 1s
Gp  (4.19)
 10s  1  5s  1
The proposed controller was compared with the PID controllers based on the
methods of Lee et al. (1998) and Chien & Fruehauf (1990). All controllers

are designed to meet M s  1.60 . The performance indices in Table 4.4 show
the superiority of the proposed PID●filter controller in both model mismatch
and nominal cases. Fig. 4.4 shows the closed-loop responses of the three
controllers in the nominal case. As seen from the figure, the proposed
controller has no overshoot and negligible settling time compared with the
other two controllers.

Fig. 4.4 Simulation result of proposed tuning method for example 4.3

107
4.3.4. Example 4.4. Simulation study was carried out for the following
SOPDT model with a strong lead term:

1 15s  1 e 6 s
Gp  (4.20)
 10 s  1  10 s  1
The proposed PID●filter controller is compared with the conventional PID
controllers based on the methods of Lee et al. (1998) and Chien & Fruehauf

(1990). All controllers are tuned to have the same robustness as M s  1.60 .
The controller parameter settings and the resulting values of the performance
indices are listed in Table 4.5. The closed-loop responses of the three
controllers are compared in Fig. 4.5. The results show that the proposed
controller gives a superior performance both for set-point and load responses
over the other two controllers. The proposed controller is also robust in the
face of model uncertainty.

Fig. 4.5 Simulation result of proposed tuning method for example 4.4

108
Table 4.4. Controller parameters and resulting performance indices for the overdamped SOPDT process

set-point disturbance
Method Kc I D a b c d nominal 20% nominal 20%
mismatch mismatch
ITAE overshoot ITAE overshoot ITAE overshoot ITAE overshoot
Proposed 0.149 0.5 0.166 0.202 0.034 15 50 1.55 0.0 3.08 0.25 56.88 0.17 55.95 0.17
  0.6766
Lee et al. 3.723 15.11 3.418 - - - - 73.79 0.12 71.36 0.12 69.57 0.20 68.2 0.20
  0.515
Smith 3.793 15.0 3.333 - - - - 71.56 0.12 69.06 0.13 67.08 0.20 65.77 0.20
  0.9773
Chien and 3.793 15.0 3.333 - - - - 71.56 0.12 69.06 0.13 67.08 0.20 65.77 0.20
Fruehauf
  0.9773

109
Table 4.5. Controller parameters and resulting performance indices for the second order delayed overshoot process
set-point disturbance
Method Kc I  a b c d nominal 20% nominal 20%
D

mismatch mismatch
ITAE overshoot ITAE overshoot ITAE overshoot ITAE overshoot

Proposed 0.310 3.0 1.0 15.82 12.41 20 100 53.37 0.0 121.3 0.34 188.7 0.62 185.6 0.95
  3.666
Lee et al. 0.513 6.43 4.874 - - - - 184.6 0.07 184.4 0.15 320.5 0.62 340.5 0.95
  6.544
Chien and 0.50 5.0 5.0 - - - - 245.3 0.16 286.7 0.27 323.3 0.62 397.7 0.95
Fruehauf
  3.999

110
4.3.5. Example 4.5. Consider the following SOPDT process with inverse
response,
1 5s  1 e 5 s
Gp  (4.21)
 100s 2

 10 s  1

The proposed PID●filter controller is compared with the conventional PID

controllers based on the methods of Lee et al. (1998) and Chien & Fruehauf

(1990) while all controllers are tuned to meet M s  1.60 . The controller
parameter settings and the resulting values of the performance indices are
listed in Table 4.6. The closed-loop responses of the three controllers are
shown in Fig. 4.6. The superior performance achieved by the proposed
controller is readily apparent from the figure.

Fig. 4.6 Simulation result of proposed tuning method for example 4.5

4.3.6. Example 4.6. Set-point response was studied for the following first-
order delayed integrating process,

111
 1e 5 s
Gp  (4.22)
s  10 s  1

As shown in Fig. 4.7 and Table 4.7, the proposed controller shows superior
performance over the other two controllers.

Fig. 4.7 Simulation result of proposed tuning method for example 4.6

112
Table 4.6. Controller parameters and resulting performance indices for the second order delayed inverse process

set-point disturbance
Method Kc I D a b c d nominal 20% nominal 20%
mismatch mismatch
ITAE overshoot ITAE overshoot ITAE overshoot ITAE overshoot

Proposed 0.106 2.50 0.833 2.6 4.9 10 100 309.5 0.0 384.0 0.15 1181 0.95 1876.0 1.26
  8.44
Lee et al. 0.448 10.02 9.628 - - - - 666.7 0.20 1153. 0.46 1208 0.94 2000.0 1.25
  7.381 0
Chien and 0.449 11.01 10.09 - - - - 509.2 0.11 781.5 0.34 1031 0.93 1487.0 1.24
Fruehauf
  14.512

113
Table 4.7. Controller parameters and resulting performance indices for the first order delayed integrating process

set-point
Method Kc I D a c
nominal 20% mismatch
ITAE overshoot ITAE overshoot

Proposed 0.122 - 1.666 0.134 10 54.63 0.0 83.39 0.18

  3.156

Lee et al. 0.094 - 11.185 - - 314.40 0.0 258.6 0.0

  5.5455
Chien and 0.110 45.598 7.806 - - 1054.0 0.36 960.7 0.42
Fruehauf
  15.299

114
4.4. Discussion

Consider a lag time dominant FOPDT process where   and   .

Using a 2/2 Pade approximation for the dead time, the ideal feedback

controller to give the desired closed-loop response, e /   s  1 , can be

 s

converted to the PID●filter controller as:

Gc 
 s /12   s / 2  1
2 2


  s  1
K     s    2
  (4.23)
 s2  s  1
12      2    

The loop transfer function then becomes:

1 1
L( s)  GcG p  
       2
s 2
  (4.24 )
 s  s  1
 12      2      

under the assumption that the process dead time e  s is effectively

compensated for by the lead term  1   s / 2   s /12  in the lead-lag filter. It is

2 2

clear that the dominant pole in the process transfer function is compensated
for by the PID controller. Consequently, the resulting control system is

  2  
equivalent to controlling the fast process 1  s2  s  1 with the
12      2    

integral controller. Note that as  increases, the PID●filter controller

approaches the conventional PID controller.

For a lag time dominant SOPDT process, the ideal feedback controller is
converted to the PID●filter controller as:

Gc 
 s /12   s / 2  1
2 2


 s  2 s  1
2 2

K     s    2
  (4.25)
 s2  s  1
12      2    

115
The loop transfer function is exactly the same as Eq. (4.24) by the FOPDT
model.

Since the resulting filter has the damping factor  f  3 4      , it is

always underdamped. The maximum value the filter can have is 0.866 when
  . As the  value decreases, the underdamping of the filter increases
which can cause a severe robustness problem.

Fig. 4.8 compares the closed-loop responses by various controllers for the
previously studied process G p  2e  10s  1  5s  1 . The approximated loop
1s

transfer function by Eq. (4.24) is (for   0.6766 ):

1 1
L( s )  (4.26)

1.6766 s 0.0336 s 2  0.2018s  1 
In the figure, the PID●filter (1/1Pade) is derived using the 1/1 Pade
approximation and consists of the PID controller cascaded with a first-order
lead-lag filter. As seen in the figure, the response by the proposed PID ●filter
controller follows the response by the ideal controller very closely while that
by the PID●filter(1/1Pade) controller is quite different. This indicates that the
proposed PID●filter controller based on the 2/2 Pade expansion approximates
the ideal controller almost perfectly while the 1/1 Pade expansion does not.
Furthermore, the response based on the approximated loop transfer function
by Eq. (4.26) shows quite a similar trend with that by the PID●filter
controller, which indicates that the lead term of the filter plays an important
role in improving closed-loop performance by compensating for the dead
time effect.

116
 1.2

Ideal
PID.filter (Proposed)
0.8 Approximated by Eq. (26)
Process Response

PID.filter(1/1 Pade)
PID (Lee et al.)
0.6

0.4

0.2

0
0 10 20 30 40 50
Time

Fig. 4.8 Closed-loop responses by various controllers

The proposed PID●filter controller has a clear advantage as the lag time
dominates. Fig. 4.9 compares the ITAE values of the set-point responses for
various dead time to lag time ratios for the FOPDT model used in Example
4.1 ( by changing  while fixing  ).  is chosen so that Ms  1.6 for every
case. As seen from the figure, the proposed PID●filter controller gives the
smallest ITAE value among all other controllers over the lag time dominant
range. As   increases (i.e., the dead time dominates), the benefit gained by
the proposed PID●filter controller is diminished. In Fig. 4.9, the conventional
PID controller based on the methods of Lee et al. (1998) gives the smallest
ITAE value over the other controllers when   goes beyond 2.

117
 1000

100
ITAE

Proposed
Lee et al.
10 Rivera et al.

1
0 0.4 0.8 1.2 1.6 2 2.4

Fig. 4.9 Comparison of the ITAE generated by various tuning rules for Ms  1.60

 guideline. In the proposed tuning rule, the closed-loop time constant 

controls the tradeoff between robustness and performance of the control
system. As  decreases, the closed-loop response becomes faster and can
become unstable. On the other hand, as  increases, the closed-loop
response becomes sluggish and stable. A good tradeoff is obtained by
choosing  to give a Ms value in the range of 1.2~2.0. The  guideline plot
for several robustness levels is shown in Fig. 4.10. As seen from the figure,
the resulting plot is almost a straight line and the desired  value can also be
obtained from a linear relation by  /    ( /  ) , where  = 1.360,
0.677,0.372, and 0.337 for Ms =1.4, 1.6, 1.8, and 2.0, respectively. For small
  (typically less than 0.2), a detuning process might be considered to

118
account for the constraint on the manipulated variable. Since the  value for
a particular Ms is independent on the damping factor in the process model,
the plot in Fig. 4.10 can also be used for  setting of the SOPDT model.

3.5
Ms=1.4
3
Ms=1.6
2.5 Ms=1.8
Ms=2.0
2

1.5

0.5

0
0 1 2

Fig. 4.10  guide lines for FOPDT and SOPDT for different Ms values

4.5. Conclusions

An analytical design method for a PID controller in series with a second-

order lead-lag filter is proposed for various types of time-delay process. By
using the appropriate Pade expansion to the process dead time, the ideal
feedback controller can be converted into a PID●filter structure with little
loss of accuracy. The resulting PID●filter controller efficiently compensates
for the dominant process poles and zeros and drastically improves the closed-

119
loop performance. Simulation results demonstrate the superior performance
of the proposed PID●filter controller over the conventional PID controllers.

4.6. References

(1) Åström, K. J., Panagopoulos. H., & Hägglund, T., (1998), Design of PI
controllers based on non-convex optimization, Automatica, 34, (5), 585-
601.
(2) Chien, I. L., Fruehauf, (1990), Consider IMC tuning to improve
controller performance, Chem. Eng. Prog., 86 (10), 33.
(3) Dwyer, Aidan O., (2003), Handbook of PI and PID controller tuning
rules; Imperial College Press, London.
(4) Horn, I. G., Arulandu, J. R., Christopher, J. G., VanAntwerp, J. G., &
Braatz, R. D., (1996), Improved filter design in internal model control,
Ind. Eng. Chem. Res., 35, 3437.
(5) Lee, Y., Park, S., & Lee, M., (2006), Consider the generalized IMC-PID
method for PID controller tuning of time-delay processes, Hydrocarbon
Processing, pp. 87-91.
(6) Lee, Y., Park, S., Lee, M., & Brosilow, C., (1998), PID controller tuning
for desired closed-loop responses for SI/SO systems, AIChE Journal, 44,
No.1, pp. 106-115.
(7) Morari, M., and E. Zafiriou, (1989), Robust Process Control, Prentice
Hall, Englewood Cliffs, NJ.
(8) Rivera, D. E., Morari, M., & Skogestad, S., (1986), Internal model
control, 4. PID controller design, Ind. Eng. Proc. Des. Deu.. 25, 252.
(9) Seborg, D. E., Edgar, T. F., & Mellichamp, D. A., (2004), Process
dynamics and control, John Wiley & Sons, Second edition, New York.

120
(10) Skogestad, S., Postlethwaite, I., (1996), Multivariable feedback control:
analysis and design; John Wiley & Sons, New York.
(11) Smith, C. L., Corripio, A. B., & Martin, J. Jr., (1975), Controller tuning
from simple process models, Instrum. Technol., 22, 12. 39.
(12) Ziegler, J. G., and Nichols, N. B., (1942), Optimum settings for
automatic controller, Trans. ASME, 64, 759.

121
5. ANALYTICAL DESIGN OF ENHANCED PID·FILTER
CONTROLLER FOR INTEGRATING AND FIRST ORDER
UNSTABLE PROCESSES WITH TIME DELAY

5.1. Introduction

The proportional-integral-derivative (PID) control algorithm has three-term

functionality enabling the treatment of both transient and steady-state
responses; it provides a generic and efficient solution to real world control
problems. The wide application of PID control has stimulated and sustained
research and development in order to “get the best out of PID’’, and the
search is on to find the next key technology or methodology for PID tuning.
Many of the important chemical processing units in industrial and chemical
practice are open-loop unstable processes that are known to be difficult to
control, especially when there exists a time delay, such as in the case of
continuous stirred tank reactors, polymerization reactors and bioreactors
which are inherently open loop unstable by design. Furthermore, many of
these processes are usually run batch-wise, and as a result of possible
formulation changes, may operate with significant batch-to-batch variability.
Clearly, the tuning of controllers to stabilize these processes and to impart
adequate disturbance rejection is critical. Moreover, integrating processes are
very frequently encountered in process industries and many researchers have
suggested that for the purpose of designing a controller, considerable
numbers of chemical processes could be modeled using an integrating
process with time delay. Consequently, there has been much interest in the

122
literature in the tuning of industrially standard PID controllers for open-loop
unstable systems as well as for integrating processes.
The effectiveness of the internal model control (IMC) design principle has
made this method attractive in the process industry, which has led to much
effort being made to exploit the IMC principle to design equivalent feedback
controllers for unstable processes (Morari and Zafiriou, 1989). The IMC
based PID tuning rules have the advantage of using only one tuning
parameter to achieve a clear trade-off between closed-loop performance and
robustness.
It is well known that the IMC structure is very powerful for controlling stable
processes with time delay and cannot be directly used for unstable processes
because of the internal instability (Morari and Zafiriou, 1989). For this
reason, some modified IMC methods of two-degree-of-freedom (2DOF)
control were developed for controlling unstable processes with time delay,
such as those proposed by Lee et al. (2000), Yang et al. (2002), Wang and
Cai (2002), Tan et al. (2003), and Liu et al. (2005). In addition, 2DOF control
methods based on the Smith-Predictor (SP) were proposed by Majhi and
Atherton (2000), Kwak et al. (1999), and Zhange et al. (2004) to achieve a
smooth nominal setpoint response without overshoot for first order unstable
processes with time delay. It is a notable merit of the modified IMC methods
and the modified SP methods that the nominal setpoint response tends to be
faster without overshoot for unstable processes. In fact, the common
characteristic of the abovementioned modified IMC and SP methods is the
use of a nominal process model in their control structures, which is
responsible for their good performance in this respect. It should be noted that
most existing 2DOF control methods are restricted to the unstable processes
in the form of a first order rational part plus time delay, which in fact, cannot

123
represent a variety of industrial and chemical unstable processes well
enough. Besides, there usually exist unmodeled dynamics that inevitably tend
to deteriorate the control system performance.
The delay integrating process has a clear advantage in the identification test,
because the model contains only two parameters and is simple to use for
identification. Some of the well accepted PID controller tuning methods for
delay integrating processes are those proposed by Chien and Fruehauf
(1990), Lubyen (1996), and Chen and Seborg (2002).
Modern control hardware provides the microprocessor implementation for a
flexible combination of conventional control algorithms to achieve enhanced
control performance. The PID controller cascaded with a first order lead/lag
filter is a typical example. The main reason for using the PID·filter controller
is that it provides better performance without tribulation. Earlier, many
authors (Morari and Zafiriou, 1989; Lee et al., 2000; Horn et al., 1996;
Luyben, 1996) proposed the use of the PID controller cascaded with a first or
second order filter, as described in Eq. (5.1) below.
 1  1  as
Gc  K c 1   Ds  (5.1)
 Is  1  bs

where K c ,  I and  D are the proportional gain, integral time constant, and

derivative time constant of the PID controller, respectively, and a and b are
the filter parameters.
It should be emphasized that the design principle of the aforementioned
tuning methods for unstable and integrating delay processes is complicated
and that the modified IMC structure is difficult to implement in a real process
plant in the presence of model uncertainty.
In this paper, a simple analytical method is proposed for the design of a
PID·filter controller, in order to achieve enhanced performance for first order

124
unstable and integrating delay processes. A closed-loop time constant,  ,
guideline was recommended for a wide range of time-delay/time-constant
ratios. A simulation study was performed for both unstable and integrating
delay processes to show the superior performance of the proposed method for
both nominal and perturbed processes.

5.2. Design procedure

The IMC controller (Figure 5.1a) is a competent method for control system
design (Morari and Zafiriou, 1989). Nevertheless, for unstable processes the
IMC structure cannot be implemented because any bounded input, d , will

produce unbound output, y , if G p is unstable. As discussed in Morari and

Zafiriou (1989), the IMC approach to designing a controller for an unstable
% if the following conditions are satisfied for
process is possible for G p  G p

the internal stability of the closed-loop system:

Disturbance d Disturbance
GD
d
IMC GD
Setpoint controller
filter Process Setpoint Controller Process
r y filter
fR +- q Gp ++
r y
fR +- Gc Gp ++
Process model
~
Gp -+

Figure 5.1 (a) The IMC structure Figure 5.1 (b) Classical feedback control structure

Figure 5.1. Block diagram of IMC and classical feedback control

(i) q is stable.
(ii) G p q is stable.

(iii)  1  G p q  GD is stable.
These three conditions result in the well known standard interpolation
conditions (Morari and Zafiriou, 1989):

125
• If the process model, G p , has unstable poles, up1 , up2 ,L , upm , then q

should have zeros at up1 , up2 ,L , upm .

• If the process model, GD , has unstable poles, dup1 , dup2 , L , dupm , then

1  G p q should have zeros at dup1 , dup2 , L , dupm .

Since the IMC controller, q , is designed as q  pm f

1 1
in which pm

includes the inverse of the model portion, the controller satisfies the first
condition. The second condition could be satisfied through the design of the
IMC filter, f . For this, the filter is designed as
 im1  i s i  1
f  (5.2)
( s  1) r

where m is the number of poles to be canceled;  i are determined by Eq.

(5.3) to cancel the unstable poles in GD ; r is selected large enough to make

the IMC controller proper.
pm1 ( im1  i s i  1)
1  Gpq  1 0 (5.3)
s  dup1 , dup2 ,L , dupm ( s  1)r s  dup , dup
1 2 ,L , dupm

Then, the IMC controller comes to be

(im1  i s i  1)
1
q p m (5.4)
( s  1) r
Thus, the resulting setpoint and disturbance rejection is obtained as
y p A  im1  i s i  1
 Gp q  (5.5)
  s  1
r
r

y  p A   im1  i s i  1 
 (1  G p q )GD   1   GD (5.6)
   s  1 
r
d
 

126
  
The numerator expression  i 1  i s  1 in Eq. (5.5) causes an unreasonable
m i

overshoot in the servo response, which can be eliminated by adding the

setpoint filter f R as
1
fR 
   i si  1
m
i 1
(5.7)

From Figure 5.1, a feedback controller Gc which is equivalent to the IMC

controller q is represented by
q
Gc  (5.8)
%q
1 G P

The resulting ideal feedback controller is obtained as

(  im1  i s i  1)
pm1
( s  1) r
Gc  (5.9)
p A  im1  i s i  1
1
  s  1
r

Although the resulting controller in Eq. (5.9) does not have a PID controller
structure, we can design a PID controller cascaded with a first order filter that
resembles the equivalent feedback controller very closely. This will be
discussed in the next section.
5.3. Proposed tuning rule
The first order delay unstable process (FODUP) is the representative model,
which is commonly utilized for many unstable processes in the chemical
process industry. Consequently, this section describes the design of the
tuning rule for the FODUP and extends it to the delay integrating process
(DIP).
5.3.1. First-order delay unstable process (FODUP)
Ke  s
GP  GD  (5.10)
 s 1

127
where K is the gain,  the time constant and  the time delay. The IMC
filter structure exploited here is given as
 s 1
f  (5.11)
  s  1
3

The resulting IMC controller can be obtained as follows

q
  s  1   s  1
(5.12)
K   s  1
3

The IMC controller in Eq. (5.12) is proper and the ideal feedback controller
which is equivalent to the IMC controller is

Gc 
  s  1   s  1
K   s  1  e  s   s  1 
3 (5.13)
 

Approximating the dead time e  s with a 1/2 Pade expansion gives

 6  2 s 
e s  (5.14)
 6  4 s   s  2 2

Substituting Eq. (5.14) for the dead time in Eq. (5.13) results in

Gc 
 6  4 s   s    s  1   s  1
2 2

(5.15)
K   s  1  6  4 s   s     s  1  6  2 s  
3 2 2
 

It is important to note that the 1/2 Pade approximation is precise enough to

convert the ideal feedback controller into a PID cascaded first order filter
with barely any loss of accuracy, while retaining the desired controller
structure. Simplifying and rearranging Eq. (5.15), we obtain
Gc 
 6  4 s   s 
2 2
  s  1   s  1
 K  6  18  6  s  
1 
 
2   2  12  18 2
s
 
3 2  12 2  6 3 2 3 2 2  4 3 3
s 
s 
 3 2 
s4 
(5.16)

 6  18   6   6  18  6   6  18  6   6  18   6  

It can be recognized from Eq. (5.16) that the resulting controller has the form
of the PID controller cascaded with a high order filter. The analytical PID
tuning formula can be obtained from Eq. (5.16) as

128
 4
KC   (5.17a)
K  6  18  6 

 I  2 3 (5.17b)
D  4 (5.17c)
It is obvious from Eq. (5.3) that the denominator in Eq. (5.16) contains the
factor   s  1 . Therefore, the filter parameter b in Eq. (5.1) can be obtained
by taking the first derivative of Eq. (5.16) as described below

1
 2   2
 12  18 2  s   3 2
 12 2  6 3  s   3 
2
2 2
 4 3 s 3

 3 2
s4
 6  18  6   6  18  6   6  18  6   6  18  6  (5.18)
 
ds 3  cs 2  bs  1 
  s  1

and substituting s  0 gives

b
 2   2
 12  18 2 
 (5.19)
 6  18  6 
Since the high order ds  cs
3 2
  term has little impact on the overall control

performance in the control relevant frequency range, the remaining part of

the fraction in Eq. (5.16) can be successfully approximated to a simple first
order lead/lag filter (1  as ) /(1  bs ) where a   .

The value of  is designed to remove the open-loop unstable pole at s  1 

. This method chooses  such that the term 1  G p q  has a zero at the pole

of GD , i.e., 1  G p q  s 1   0 . Therefore, the designed value of  is

obtained from
   3 
    1   e   1 (5.20)
   

5.3.2. Delayed integrating process (DIP)

129
 Ke  s
G p  GD  (5.21)
s
The DIP can be modeled by considering the integrator as an unstable pole
near zero. This is mandatory since it is not practicable to implement the
aforementioned IMC design procedure for the DIP, because the term,  ,
vanishes at s  0 . As a result, the DIP can be approximated to the FODUP as
follows:
Ke  s Ke  s  Ke  s
G p  GD    (5.22)
s s  1/   s 1

where  is an arbitrary constant with a sufficiently large value.

Accordingly, the optimum IMC filter structure for the DIP is identical to that

for the FODUP model, i.e., f    s  1   s  1

3
.

Therefore, the resulting IMC controller becomes

q    s  1   s  1 K   s  1
3
and the subsequent PID tuning rules are

obtained as
4
KC   (5.23a)
K  6  18  6 

 I  2 3 (5.23b)

D  4 (5.23c)
a  (5.23d)

b
 2   2
 12  18 2 
 (5.23e)
 6  18  6 
   3 
   1   e   1 (5.23f)
   

5.4. Simulation Results

This section is devoted to the simulation study which is classified as follows

130
1. Example 5.1. Deals with a lag time dominant (    0.4 ) first order
delay unstable process. This is the most popular process model and has
been included in performance comparisons by many researchers.
2. Example 5.2. Shows the performance superiority of the proposed method

in the dead time dominant     1.2  first order delay unstable process.
3. Example 5.3. Performance comparison for the delay integrating process.
4. Example 5.4. Application of the proposed method to the distillation
column model, which is widely used in the literature.
5. Example 5.5. Performance comparison of the proposed PID·filter
controller with the modified Smith predictor (SP) controller.
In the simulation study, the performance and robustness of the control system
were evaluated using the following indices to ensure a fair comparison.

5.4.1. Performance and robustness measure

5.4.1.1 Integral error criteria

To evaluate the closed-loop performance, the integral of the time-weighted

absolute error (ITAE) criterion was considered in the case of both a step
setpoint change and a step load disturbance. The ITAE is defined as

ITAE   t e  t  dt
0
(5.24)

5.4.1.2 Overshoot

Overshoot is a measure of how much the response exceeds the ultimate

value following a step change in the setpoint and/or disturbance.

5.4.1.3. Maximum Sensitivity  Ms  to Modeling Error

131
To evaluate the robustness of a control system, the maximum sensitivity, Ms

, which is defined as Ms  max 1/[1  G pGc (iw)] , was used. Since Ms is the
inverse of the shortest distance from the Nyquist curve of the loop transfer
function to the critical point  1, 0  , a small Ms value indicates that the
stability margin of the control system is large. Ms is a well known
robustness measure and has been used by many researchers (Chin and
Seborg, 2002; Skogestad and Postlethwaite, 1996). To ensure a fair
comparison, it is widely accepted for the model-based controllers (DS-d and
IMC) to be tuned by adjusting  so that the Ms values are the same.
Therefore, throughout all of our simulation examples, all of the controllers
compared were designed to have the same robustness level in terms of the
maximum sensitivity, Ms .

5.4.1.4. Total Variation (TV)

To evaluate the manipulated input usage, we compute the total variation ( TV

) of the input u  t  which is the sum of all its moves up and down. If we

discretize the input signal as a sequence [u1 , u2 , u3 ...., ui ...], then


TV   ui 1  ui should be as small as possible. The TV is a good measure of
i 1

the smoothness of a signal (Skogestad and Postlethwaite, 1996).

5.4.2. Example 5.1. Lag time dominant FODUP     0.4 
An extensively published FODUP model (Hung and Chen, 1997; Lee et al.,
2000; Tan et al., 2003, Liu et al., 2005, Majhi & Atherton, 2000) was
considered for the performance comparison
1e0.4 s
G p  GD  (5.25)
1s  1

132
For the above FODUP model, the recently published paper of Liu et al.
(2005) demonstrated the superiority of their method over those of Tan et al.
(2003) and Majhi and Atherton (2000). In this simulation study, the proposed
method was compared with those of Liu et al. (2005) and Lee et al. (2000).
The design of the disturbance rejection is identical for both Liu et al.’s (2005)
and Lee et al.’s (2000) methods. However, for the setpoint response, Liu et
al. (2005) used a modified IMC structure, while Lee et al. (2000) applied a
setpoint filter. For the methods of both Liu et al. (2005) and Lee et al. (2000),
  0.5 was used in the simulation, which results in Ms  3.03 . To obtain a

fair comparison,  was also adjusted in the proposed method    0.20  to

obtain Ms  3.03 . The controller parameters, including the performance and
robustness matrix, are listed in Table 5.1.

Table 5.1. PID controller setting and performance matrix for Example 5.1
tuning  Kc I D Ms set-point disturbance
methods ITAE overshoot ITAE overshoot
Proposed 0.2 0.46 0.26 0.10 3.0 0.61 1.01 0.75 0.61
Liu et al. 0.5 2.63 2.51 0.15 3.0 0.40 1.0 1.51 0.69
Lee et al. 0.5 2.63 2.51 0.15 3.0 1.27 1.0 1.51 0.69

- Proposed method: a  1.5779 , b  0.1053 ; f R  1  1.5779s  1

- Liu et al.: K c  2 , C  s    s  1  0.4s  1

- Lee et al.: f R  1  2.3566s  1

Figure 5.2 shows the comparison of the proposed method with those of Liu et
al (2005) and Lee et al. (2000), performed by introducing a unit step change
in both the setpoint and load disturbance. For the servo response, the setpoint
filter is used for both the proposed method and that of Lee et al. (2000),
whereas a three controller element structure is used for the method of Liu et

133
al (2005). As is apparent from Figure 5.2 and Table 5.1, the proposed method
results in an improved load disturbance response. Since the design of the
disturbance rejection is identical for both Liu et al.’s (2005) and Lee et al.’s
(2000) methods, the same PID tuning setting and consequently an identical
disturbance rejection response is obtained in both cases. For the servo
response, the method of Liu et al. (2005) seems better, but the settling times
of Liu et al.’s (2005) method and the proposed method are comparable, while
Lee et al.’s method (2000) shows the slowest response with a long settling
time.
1

0.8
Process Variable

0.6

0.4

0.2 Proposed
Liu et al.
Lee et al.
(a) 0
0 1 2 3 4 5 6
Time
0.7
Proposed
Liu et al.
Lee et al.
0.5
Process Variable

0.3

0.1

-0.1

(b) 0 1 2 3 4 5 6
Time

Figure 5.2. Response of the nominal system for Example 5.1

It is important to note that the well known modified IMC structure has the
theoretical advantage of eliminating the time delay from the characteristic

134
equation. Unfortunately, this advantage is lost if the process model is
inaccurate. Besides, real process plants usually incorporate unmodeled
dynamics that inevitably tend to deteriorate the control system performance
severely. The robustness of the controller was investigated by inserting a
perturbation uncertainty of 10% in all three parameters simultaneously

towards the worst case model mismatch, i.e., G p  GD  1.1e0.44s  0.9s  1 .

The simulation results are presented in Figure 5.3 for both the set-point and
the disturbance rejection. It is obvious from Figure 5.3 that the proposed
controller tuning method has an excellent setpoint and load response, while
the modified IMC controller corresponding to Liu et al.’s (2005) method has
the worst setpoint response for the model mismatch. The better setpoint
response for the nominal case afforded by the SP controller is achieved by
sacrificing the robustness of the closed-loop system. For the disturbance
rejection, the methods of Liu et al. (2005) and Lee et al. (2000) are identical
and perfectly overlapped.

135
 Process Variable 1

0.8

0.6

0.4
Proposed
0.2 Liu et al.
Lee et al.
(a) 0
0 1 2 3 4 5 6
Time

Proposed
0.8 Liu et al.
Lee et al.
0.6
Process Variable

0.4

0.2

-0.2
(b)
0 1 2 3 4 5 6
Time

Figure 5.3. Responses of the model mismatch system for Example 5.1

5.4.3. Example 5.2. Dead time dominant FODUP     1.2 

Consider an unstable dead time dominant process (Hung and Chen, 1997;
Tan et al., 2003) as follows:
1e 1.2 s
GP  GD  (4.26)
 1s  1
Tan et al. (2003) previously demonstrated the superiority of their method
over other methods including that of Hung and Chen (1997). The  value
for both the proposed method and Lee et al.’s (2000) method was adjusted
for Ms  10.71 . The controller setting parameters are listed in Table 5.2. To
test the performance of the control system, the load disturbance has a step
change of magnitude 0.1, and a setpoint with a magnitude of 1 is added. The

136
simulation results are provided in Figure 5.4 for both the setpoint tracking
and the disturbance rejection. It is clear from Figure 5.4 and Table 5.2 that
the proposed method results in an improved load disturbance response. The
proposed method shows superiority for the load disturbance over the other
controllers. The setpoint response given by the method of Tan et al. (2003) is
the best among all of the methods, whereas Lee et al.’s (2000) method has
the slowest response with a long time to reach the steady state. It is important
to note that Tan et al.’s (2003) method has a modified IMC structure using
three individual controllers. In the proposed method, the servo response is
initially slow, but the settling times for both the proposed method and Tan et
al.’s (2003) method are similar. The modified IMC structure proposed by
Tan et al. (2003) has the merit of providing a nominal setpoint response, but
it loses when the process has unmodeled dynamics.
The robustness of the controller is evaluated by inserting a perturbation
uncertainty of 5% in all three parameters simultaneously to obtain the worst

case model mismatch, i.e., G p  GD  1.05e1.26s  0.95s  1 as an actual

process, whereas the controller settings are those calculated for the process
with the nominal model. Figure 5.5 shows both the setpoint and disturbance
rejection responses for model mismatch. The controller settings of the
proposed method provide the most robust performance for both the servo and
regulatory problems. The methods of Tan et al. (2003) and Lee et al. (2000)
give an unstable oscillatory response for both the setpoint and disturbance
rejection, as is apparent from Figure 5.5.

137
 1

0.8
Process Variable

0.6

0.4
Proposed
0.2 Tan et al.
Lee et al.
(a)
0
0 5 10 15 20 25
Time
0.7
Proposed
Tan et al.
0.55
Lee et al.
Process Variable

0.35

0.15

(b)
-0.05
0 5 10 15 20 25
Time

Figure 5.4. Response of the nominal system for Example 5.2

138
 Proposed
1.5 Tan et al.
Process Variable Lee et al.

0.5

(a) 0
0 5 10 15 20 25
Time
1

0.5
Process Variable

-0.5
Proposed
Tan et al.
(b) Lee et al.
-1
0 5 10 15 20 25
Time

Figure 5.5. Responses of the model mismatch system for Example 5.2

Table 5.2 . PID controller setting and performance matrix for Example 5.2

tuning  Kc I D Ms Set-point disturbance

methods ITAE overshoot ITAE overshoot
Proposed 1.1 0.03 0.8 0.3 10.7 13.6 1.0 15.4 0.54
Tan et al. K 0  2 , K1   s  1  2s  1 , 7.1 1.0 17.9 0.66
K 2  1.1 0.49 s  1
Lee et al. 2.9 1.17 51.0 0.57 10.7 34.6 1.0 33.2 0.61

- Proposed method: a  29.7476 , b  0.2709 ; f R  1  29.7476s  1

- Lee et al.: f R  1  50.4356s  1

5.4.4. Example 5.3: DIP process

The following DIP model was considered which was previously studied by
other researchers (Luyben, 1996; Visioli, 2001; Chidambaram and Padma
Sree, 2003).

139
 0.0506e 6 s
GP  GD  (5.27)
s

Chidambaram & Padma Sree previously demonstrated the superiority of their

method over that of Visioli (2001). In the simulation study, we compared the
proposed method, with those of Lee et al. (2000), Luyben (1996), and
Chidambaram and Padma Sree (2003). For both the proposed method and
that of Lee et al. (2000),  was adjusted to obtain Ms  2.79 . The design
method of Chidambaram and Padma Sree’s (2003) and Luyben’s (1996)
methods is not based on the  tuning method and we therefore used their
respective values without adjusting the Ms value.
The proposed controller was designed by considering the DIP as
GP  5.06e6 s  100 s  1 . Figure 5.6 shows the closed-loop output response for a
unit step change in both the setpoint and load disturbance. The controller
setting parameters and performance matrix for both the setpoint and load
disturbance are listed in Table 5.3. To eliminate the overshoot in the setpoint
response, a setpoint filter is used in the proposed method and that of Lee et
al. (2000), while in the methods of Chidambaram and Padma Sree (2003) and
Luyben(1996), a set-point weighting type filter.

f R    I s  1   I D s 2   I s  1 , where 0    1 , is used. In Chidambaram and

Padma Sree’s paper, the recommended value of   0.4 , while in the Luyben
method (1996)   0.7 is employed. On the basis of the comparison of the
output response and the values of the performance matrices listed in Table
5.3, it is apparent that the proposed controller shows the best performance.
Table 5. 3. PID controller setting and performance matrix for Example 5.3
tuning methods  Kc I D Ms Set-point disturbance
ITAE overshoot ITAE overshoot
Proposed 2.1 1.09 4.0 1.5 2.79 101.4 1.0 60.2 0.31
Lee et al. 4.8 3.6 19.22 2.24 2.79 179.5 1.0 102.1 0.34
Chidambaram - 4.06 27.0 2.7 3.81 239.0 1.0 178.9 0.32

140
 and
Padma Sree
Luyben - 2.56 56.32 3.56 2.24 302.2 1.0 1053 0.37

- Proposed method: a  13.1974 , b  1.4414 ; f R  1  13.1974s  1

- Lee et al.: f R  1  16.7065s  1
- Chidambaram and Padma Sree: f R  (10.80s+1) 72.9s +27s  1
2
 
-Luyben: Gc  K c  1  1  I s   D s   1  bL s  where bL  0.382 , f R  (39.42s+1)  200.55s 2 +56.32s  1

To confirm the robust performance of the proposed method, it was assumed

that there are 20% parameter perturbations in K and  simultaneously

towards the worst case model mismatch, i.e., G p  GD  0.0506e 7.2 s s . As

shown in Figure 5.7, both the proposed method and that of Luyben (2003)
are robust to parameter perturbation. Note that the robust performance of
Luyben’s (2003) method is achieved at the expense of the sluggish nominal
response. The proposed method has better performance indices in the case of
both the nominal and model mismatch when it is tuned to have the same Ms
as Luyben’s method (2003).

141
 1

0.8
Process Variable

0.6

0.4
Proposed
Lee et al.
0.2 Chidambaram and Padma Sree
Luyben
(a) 0
0 20 40 60 80 100
Time
0.38
Proposed
Lee et al.
0.28 Chidambaram and Padma Sree
Process Variable

Luyben

0.18

0.08

(b)
-0.02
0 20 40 60 80 100
Time

Figure 5.6. Response of the nominal system for Example 5.3

142
 1.5

1.2
Process Variable

0.9

0.6

Proposed
0.3 Lee et al.
Chidambaram and Padma Sree
(a) Luyben
0
0 20 40 60 80 100
Time
0.5
Proposed
Lee et al.
Chidambaram and Padma Sree
Luyben
0.3
Process Variable

0.1

-0.1

(b)

0 20 40 60 80 100
Time

Figure 5.7. Responses of the model mismatch system for Example 5.3

5.4.5. Example 5.4. Distillation column model

Distillation remains the most widely used separation technique in the
petrochemical and chemical process industries for the separation of fluid
mixtures. The operation of the distillation column is extremely critical,
because of the purity requirements of the products. The distillation column
separates a small amount of a low-boiling material from the final product.
The bottom level of the distillation column is controlled by adjusting the
steam flow rate. The process model for the level control system is
represented by the delay integrating process. The distillation column model
studied by Chien and Fruehauf (1990), and Chen and Seborg (2002) was

143
considered for the present study, which can be approximated to the FODUP
model as follows:
0.2e 7.4 s
G p  GD  (5.28)
s
The methods proposed by Chen and Seborg (2002) and Lee et al. (2000) were
used to design the PID controller, as shown in Figure 5.8. The λ value was
selected for each method to give Ms  1.90 . The controller settings are listed
in Table 5.4. The proposed controller was designed by considering the DIP as

20e 7.4 s
GP  .
100 s  1

Figure 5.8 shows the output response, where the proposed tuning rule results
in the least settling time for both the servo and disturbance rejection,
followed by that of Chen and Seborg (2002). Lee et al.’s (2000) method has
the slowest response and requires the longest settling time for both the
setpoint and disturbance rejection. A set-point weighting type filter is used
for the method of Chen and Seborg (2002) to reduce the overshoot in the
setpoint response. On the basis of Figure 5.8 and the performance indices
listed in Table 5.4, it is evident that the proposed method performs better
than the other conventional methods for both the servo and regulatory
problems.
The robustness of the controller is also evaluated by inserting a perturbation
uncertainty of 75% in the gain and 20% in the dead time simultaneously

towards the worst case model mismatch, as follows G p  GD  0.35e 8.88s s .

The simulation results for the plant-model mismatch are given in Figure 5.9
for the both servo and regulatory problems. It should be mentioned that the
controller settings are those calculated for the process with nominal process
parameters. The responses indicate that the proposed method has less

144
oscillatory response as well as the minimum settling time for both the
setpoint and disturbance rejection. The method of Chen and Seborg (2002)
shows more oscillation, followed by that of Lee et al. (2000). It seems that
the proposed method gives good performance, even for severe process
uncertainties.
1
P rocess V ariable

0.8

0.6

0.4

Proposed
0.2
Chen and Seborg
Lee et al.
(a) 0
0 20 40 60 80 100 120
Time
2
Proposed
Chen and Seborg
Lee et al.
P rocess V ariable

1.5

0.5

(b)
0
0 20 40 60 80 100 120
Time

Figure 5.8. Response of the nominal system for Example 5.4

145
 1

0.8
Process Variable

0.6

0.4
Proposed
0.2 Chen and Seborg
Lee et al.
(a) 0
0 30 60 90 120
Time
3.5
Proposed
Chen and Seborg
2.5 Lee et al.
Process Variable

1.5

0.5

-0.5

(b)
-1.5
0 30 60 90 120
Time

Figure 5.9. Responses of the model mismatch system for Example 5.4

Table 5.4. PID controller setting and performance matrix for Example 5.4

tuning  Kc I D Set-point disturbance

methods
ITAE overshoot ITAE overshoot
Proposed 5.56 0.095 4.93 1.85 452.7 1.0 1470. 1.9
0
Chen and 9.15 0.543 31.15 2.55 604.9 1.0 1794. 1.9
Seborg 0
Lee et al. 11.0 0.536 35.13 2.28 737.0 1.0 2292. 1.9
0

- Proposed method: a  26.659 , b  3.0952 ; f R  1  26.659s  1

- Chen and Seborg: f R  1 79.691s  31.15s  1
2
 
- Lee et al.: f R  1  32.6734s  1

5.4.6. Example 5.5. Comparison with the modified Smith predictor

146
The proposed controller is compared with the modified Smith predictor
(Zhang et al., 2004) given in Figure 5.10, which has a more complicated
structure with three controllers.

GD

ys y
Gcs +
+ Gp ++

Gcd

Hs + -

Figure 5.10. Simplified structure for the modified Smith predictor

For the purpose of comparison, we consider the process described below

which was studied by Zhang et al. (2004), Majhi and Atherton (2000), and
Kwak et al. (1999).
e 0.5 s
G p  GD  (5.29)
1s  1
The proposed method was compared with that of Zhang et al. (2004) because
their methods were shown to be superior to the previously reported SP
methods, such as those of Majhi and Atherton (2000) and Kwak et al. (1999).
The controllers given by Zhang et al. (2004) are
s 1
Ccs  s   (5.30a)
0.5s  1

e 0.5 s
Hs  s  (5.30b)
0.5s  1
 1 
Ccd  s   2.6483 1   0.2185s  (5.30c)
 2.4669 

In the proposed controller,   0.235 is selected and the resulting tuning

parameters are obtained as K c  0.3701 ,  I  0.3333 ,  D  0.125 , a  2.1056

and b  0.955 . The simulation was conducted by inserting a unit step change

147
in both the setpoint and load disturbance. For the servo response, the setpoint

filter f R  1  2.1056s  1 is used for the proposed method.

Figure 5.11 shows a comparison of the nominal response obtained by the
proposed PID·filter controller and that obtained by Zhang et al.’s (2004)
modified SP controller. Note that the proposed controller uses a simple
feedback control structure without any dead time compensator. Nevertheless,
the proposed controller provides superior performance, as shown in Figure
5.11. The disturbance rejection afforded by the proposed controller has a
smaller settling time, whereas the modified SP controller described by Zhang
et al. (2004) shows an aggressive response with significant overshoot and
oscillation that requires a long time to settle.

1
Process Variable

0.8

0.6

0.4

0.2 Proposed
Zhang et al.
(a) 0
0 2 4 6 8
Time
0.88
Proposed
0.7 Zhang et al.
Process Variable

0.5

0.3

0.1

(b)
-0.1
0 2 4 6 8
Time

Figure 5.11. Response of the nominal system for Example 5.5

148
As regards the servo response, the modified SP controller has an initially fast
response, because of the elimination of the dead time. The proposed method
has an initially slow response, but the settling time is similar to that afforded
by the modified SP controller.
It is important to note that the SP control configuration has the clear
advantage of eliminating the time delay from the characteristic equation,
which is very effective in improving the setpoint tracking performance.
However, this advantage is lost if the process model is inaccurate. In order to
evaluate the robustness against model uncertainty, a simulation study was
conducted for the worst case model mismatch by assuming that the process
has a 5% mismatch in the three process parameters in the worst direction, as
follows
1.05e 0.525 s
G p  GD  (5.31)
0.95s  1
The closed-loop responses are presented in Figure 5.12. Notice that the
modified SP method described by Zhang et al. (2004) gives a severe
oscillatory response on the verge of instability for both the servo and
regulatory problems, whereas the proposed controller gives a more robust
response. In practice, the robustness is as important as the nominal
performance. One key in designing the controller is the tradeoff between its
robustness and nominal performance. As shown in Figure 5.12, the proposed
method provides not only better nominal performance but also excellent
robustness, while using a simple feedback control structure.

149
 1

0.8
Process Variable

0.6

0.4

Proposed
0.2
Zhang et al.
(a)
0
0 1 2 3 4 5 6 7 8
Time
1
Proposed
Zhang et al.
0.7
Process Variable

0.4

0.1

-0.2

(b)
-0.5
0 1 2 3 4 5 6 7 8
Time

Figure 5.12. Responses of the model mismatch system for Example 5.5

5.4.7. Closed-loop time constant  guideline

The closed-loop time constant  is a user-defined tuning parameter in the
proposed tuning rule. It is directly related to the performance and robustness
of the proposed tuning method and, therefore, it is important to have a
guideline for setting the  value in order to provide both a fast and robust

response for a given   ratio.

Figure 5.13 shows the plot of   versus   for the FODUP model. It is
important to notice that the desirable Ms value to give robust control
performance in an unstable process tends to gradually increase as the dead
time increases. In Figure 5.13, for instance, the Ms values corresponding to

150
the recommended  values are approximately Ms  3.0 for    0.5 ;

Ms  4.0 for 0.5     0.6 ; Ms  5.0 for 0.6     0.8 .

Furthermore, on the basis of an extensive simulation study, it is
recommended that the starting value of  be equal to the process time delay.
If the resulting closed loop performance or robust stability is not acceptable,
then the  value should be monotonously increased or decreased until the
desirable trade-off between the nominal and robust performances is achieved.

2.5

1.5

0.5

0
0 0.3 0.6 0.9 1.2 1.5

Figure 5.13.  guidelines for FODUP

5.5. Conclusions

151
A simple analytical design method for a PID∙filter controller was proposed
based on the IMC principle for the FODUP and DIP processes. The proposed
PID∙filter controller can easily be implemented on the modern control
hardware. The proposed method affords an excellent improvement in both
the setpoint and disturbance rejection for the FODUP and DIP processes.
Several representative processes frequently used in many previous studies
were considered in the simulation. The simulation was conducted by tuning
the various controllers to have the same degree of robustness in terms of
Ms value in order to provide a fair comparison. The proposed controller
consistently provided superior performance over the whole range of the  
ratio. The robustness study was conducted by inserting a perturbation
uncertainty in all parameters simultaneously to obtain the worst case model
mismatch, and the proposed method was found to be superior to the other
methods. The proposed controller was also compared with more
sophisticated controllers such as the modified Smith predictor. The result
showed that the proposed controller gives satisfactory performance in both
the nominal and model mismatch cases, without any external dead time
compensator. The closed-loop time constant,  , guideline was also proposed
for a wide range of   ratios.
5.6. References
Chen, D., Seborg, D.E., 2002. PI/PID controller design based on direct
synthesis and disturbance rejection, Industrial and Engineering Chemistry
Research 41, 4807-4822.
Chidambaram, M., Padma Sree, R., 2003, A simple method of tuning PID
controllers for integrator/dead-time processes. Computers and Chemical
Engineering 27, 211-215.

152
Chien, I.L., Fruehauf, P.S., 1990. Consider IMC tuning to improve controller
performance Chemical Engineering Progress 86, 33-41.
Horn, I.G., Arulandu, J.R., Christopher, J.G., VanAntwerp, J.G., Braatz, R.
D., 1996. Improved filter design in internal model control, Industrial and
Engineering Chemistry Research 35, 3437-3441.
Hung, H.P., Chen, C.C., 1997, Control-system synthesis for open-loop
unstable process with time-delay, IEE Proceedings 144, 334–346.
Kwak, H.J., Sung, S.W., Lee, I., Park, J.Y., 1999. Modified Smith predictor
with a new structure unstable processes. Industrial and Engineering
Chemistry Research 38, 405-411.
Lee, Y., Lee, J., Park, S., 2000. PID controller tuning for integrating and
unstable processes with time delay, Chemical Engineering Science 55,
3481-3493.
Liu, T., Zhang, W., Gu, D., 2005. Analytical design of two-degree-of-
freedom control scheme for open-loop unstable process with time delay,
Journal of Process Control, 15, 559–572.
Luyben, W.L., 1996. Design proportional-integral-derivative controllers for
integrating/ deadtime processes, Industrial and Engineering Chemistry
Research 35, 3480-3483.
Majhi, S., Atherton, D.P., 2000. Obtaining controller parameters for a new
smith predictor using autotuning, Automatica 36, 1651–1658.
Morari, M., Zafiriou, E., 1989. Robust Process Control, Prentice-Hall:
Englewood Cliffs, NJ,.
(33) Skogestad, S., Postlethwaite I., 1996, Multivariable Feedback Control;
Analysis and Design; John Wiley & Sons; New York.
Tan, W., Marquez, H.J., Chen, T., 2003. IMC design for unstable processes
with time delays. Journal of Process Control 13, 203–213.

153
Wang, Y.G., Cai, W.J., 2002. Advanced proportional-integral-derivative
tuning for integrating and unstable processes with gain and phase margin
specifications, Industrial and Engineering Chemistry Research 41, 2910–
2914.
Visioli, A., 2001. Optimal tuning of PID controllers for integral and unstable
processes. IEE Proceeding Control Theory and Application 148, 180.
Yang, X.P., Wang, Q.G., Hang, C.C., Lin, C., 2002. IMC-based control
system design for unstable processes, Industrial and Engineering
Chemistry Research 41 17, 4288–4294.
Zhang, W.D., Gu, D., Wang, W., Xu, X., 2004. Quantitative performance
design of a modified smith predictor for unstable processes with time
delay. Industrial and Engineering Chemistry Research 43, 56–62.

154
6. CONCLUSIONS

The PID controller is most popular and widely used industrial controller in
the process industries. The well-known IMC-PID tuning rules have the
advantage of only using a single tuning parameter to achieve a clear trade-off
between closed-loop performance and robustness to model inaccuracies.
Stability analysis of the IMC-PID controller is extremely easy to carry out
the design, and trade-off between performance and robustness is clearly
understood. On the basis of the exhaustive present study conducted in the
above chapters have the following key conclusions:
(i) The brief literature survey clearly indicate the need of the unified
framework of the IMC-PID controller design for the several class of the
process model which gives the enhanced and robust control performance. (ii)
Optimum IMC filter structures were proposed for several representative
process models to improve the disturbance rejection performance of the PID
controller. Based on the proposed filter structures, tuning rules for the PID
controller were derived using the generalized IMC-PID method. The
proposed method becomes more beneficial as the process is lag time
dominant. The closed-loop time constant  guidelines were also proposed

for several process models over a wide range of   ratios. (iii) A simple
analytical design method for a PID controller cascaded with a lead/lag filter
was proposed based on the IMC principle in order to improve its disturbance

155
rejection performance. The proposed method also includes a set-point filter
to enhance the set-point response and FOPDT processes with three
representative different   ratios were used for the simulation study. The
proposed PID filter controller consistently provides superior performance
over the whole range of the   ratio and the proposed controller shows
excellent performance when the lag time dominates. The proposed controller
was also compared with the more sophisticated controller, such as the
modified Smith Predictor, in the case of the viscosity loop in a
polymerization process. The result shows that the proposed controller gives
satisfactory performance without the external dead time compensator. A
guideline of closed-loop time constant  was also proposed for a wide range
of   ratio. (iv) An analytical design method for a PID controller in series
with a second-order lead-lag filter is proposed for various types of time-delay
process. By using the appropriate Pade expansion to the process dead time,
the ideal feedback controller can be converted into a PID●filter structure with
little loss of accuracy. The resulting PID●filter controller efficiently
compensates for the dominant process poles and zeros and drastically
improves the closed-loop performance. Simulation results demonstrate the
superior performance of the proposed PID●filter controller over the
conventional PID controllers. (v) A simple analytical design method for a
PID∙filter controller was proposed based on the IMC principle for the
FODUP and DIP processes. The proposed method affords an excellent
improvement in both the setpoint and disturbance rejection for the FODUP
and DIP processes. The simulation was conducted by tuning the various
controllers to have the same degree of robustness in terms of Ms value in
order to provide a fair comparison. The proposed controller consistently
provided superior performance over the whole range of the   ratio. The

156
robustness study was conducted by inserting a perturbation uncertainty in all
parameters simultaneously to obtain the worst case model mismatch, and the
proposed method was found to be superior to the other methods. The
proposed controller was also compared with more sophisticated controllers
such as the modified Smith predictor. The result showed that the proposed
controller gives satisfactory performance in both the nominal and model
mismatch cases, without any external dead time compensator. The closed-
loop time constant,  , guideline was also proposed for a wide range of  
ratios.

157