International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249 – 8958, Volume-1, Issue-1, October 2011

A Modified Internal Model Control for Unstable

– Time Delayed System
Basil Hamed, Walid Issa

addition, a proposed method for unstable system was

Abstract— A new approach of control design of internal model discussed to cover all design problems.
controller is proposed in this paper. The proposed design method
focuses on modifying the old general structure of IMC and II. MODIFIED IMC
develops a new model structure while saving the same general
concept of using the invertible version of the system in the The general structure of internal model control and its
controller design. The new approach combines the IMC structure design procedure concentrate on getting the model of the
and the traditional structure of a control problem and this process, reusing it as a reference model parallel to the process,
demonstrates an excellent performance and behavior against
and using it for design as shown in Figure 1. The realization
different disturbance inputs and model uncertainty presented in
model parameter mismatch. Beside that a smith predictor is added process here requires a double work, one for the model and
to promote the design to compensate the delayed time systems. another for the controller. Because of the feedback in all cases
Also a proposed stabilizer has mentioned to deal with unstable occupying a position and hardware is implemented for the
systems. system. The proposed controller idea here is to reduce the
amount of hardware used for realization and implementation
Index Terms— IMC, Unstable, Time Delay, Pendulum System, without any additional component. The concept revolves
Smith predictor.
around canceling the parallel reference model and uses the
feedback as usual in the traditional control with some
I. INTRODUCTION
modification on the controller design.
Open-loop unstable processes are difficult to achieve Figure 2 illustrates the new structure of the modified IMC.
equilibrium state. Time delay always exists in the It looks like the traditional control structure but the idea was
measurement loop or control loop, so it is more difficult to focused on the controller design with reserving the concept of
control this kind of process. Using routine control method IMC.
can’t acquire satisfying result [1]. The new proposed IMC structure cancel the repeated
Internal Model Control (IMC) is one of the advanced model appeared in the general IMC structure and presents a
control strategies, which is of good robustness and is easy to new Gc(s) equation.
design and tune. However, the routine IMC is not suit for
unstable process or time delayed systems [2]. Scott A. d(s)
Geddes has designed an IMC for time delay system [3].
U(s) Y(s)
R(s)
However, it is not suitable for large time delay systems. + Gp(s) ++
-
According to Shang, and Wang, the design works well but not
for nonlinear or unstable system [4]. Solving the problem of
unstable system is by using a stabilizer beside IMC (s) -+
meanwhile the controller will become very complex [5, 6]. A
modified IMC was proposed with tuning parameters [7].
d’(s)
However, the tuning problem was raised. The disadvantage of
Kou Yamada [8] is the complexity of the structure and it does
not guarantee the stability when a time delay is added. A Figure 1: General Structure of IMC
proposed design of smith compensator using modified IMC
for an unstable plant with time delay is presented [9]. d(s)
However, this method demands a tuned parameters and an
U(s)
observer beside a PI controller that makes the overall system R(s) Y(s)
+ Gp(s) ++
is complex. In this paper a new approach of IMC was -
proposed to solve the complexity of the old structure beside a
smith predictor was added to compensate the time delay. In

Manuscript received October 29, 2011.

Dr. Basil Hamed is Associate Professor of Electrical Engineering
Department, Islamic University of Gaza, Palestine.
Eng. Walid Issa is a research Assistant at Electrical Engineering Figure 2: Modified IMC
Department, Islamic University of Gaza (IUG) Palestine.
.

56
A Modified Internal Model Control for Unstable – Time Delayed System

The new proposed controller is to cancel the process model second step will require being proper. Let is a
Gp(s) by the term Gp(s)-1 that considered as the inverse of the copy of .
process transfer function and substitute it by different transfer
function Gsc(s) such that:
Step1: Factor the mode into two parts:
-1
Gc(s) = Gp(s) . Gsc(s) (s) = (s). (s) (4)

Where: Gsc(s) is the transfer function of the closed loop (s) contains all Nonminimum Phase Elements in the plant
that will achieve the required criteria as shown in Figure 3.
model, that is all Right- Half-Plane (RHP) zeros and time
E(s) delays. The factor (s), meanwhile, is Minimum Phase and
R(s) Y(s)
+
- invertible.

Then an IMC controller defined as

= (5)
is stable and causal.
Step 2: Augment with such that the final IMC
Figure 3: Controller Closed Loop System controller is now,
= (6)
The output of the system in Figure 3 is

As mention before; the selection of depends on the

This will achieve the specifications required from the original specification of the design.
system to be controlled.
IV. NUMERICAL EXAMPLE
The selection of Gsc(s) is trivial and depends on Y(s)/R(s) that A simple system represents a DC motor with a transfer
can be assumed as a second order system that has the form of: function of:
Y (s ) ωn 2 1.5
= 2 (2) G (s ) = (7)
R (s ) (s + 2ξωn s + ωn 2 ) S + 14S + 40.02
2

Where,
πζ The controller needs to achieve the design speciation: OS% <
−
1−ζ 2
Percent overshoot OS % = e × 100 10% and Ts < 5 sec.
First want to design Y(s) to meet the desired design:
4 4
Settling time Ts = =
ζωn σ
Y (s ) 5.3
π = 2 (8)
Peak time Tp = R (s ) (s + 3s + 5.3)
ωn 1 − ζ 2 Then obtain Gsc(s) from Y(s) such that R(s) is impulse input
5.3
G sc (s ) = (9)
Gsc(s) be can extract from equation (1) to get the form s (s + 3)
Y (s ) And when simulate Gsc(s) as a closed loop system as shown in
G sc (s ) = (3)
R (s ) −Y (s ) Figure 3 it is noted that the system will achieve the
Then have the overall system in Figure 2 and conclude that requirement as shown in Figure 4.
Gc(s) = Gp(s)-1. Gsc(s) will cancel the process behavior but
will add Gsc(s) that guarantees the desired specification to be
achieved.
To have the invertible form of the process faces some
problems. To solve these problems is to use the method that
split the process transfer function to invertible and non
invertible parts then use the invertible part for design.

III. DESIGN PROCEDURE

The IMC design procedure consists of two main steps. The
first step will insure that is stable and causal; the

57
International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-1, Issue-1, October 2011

Figure 4: Gsc(s) Closed Loop response

Figure 6: The Impulse Response for t = 0.3 sec

After applying the proposed controller

Gc(s) = Gp(s)-1 . Gsc(s)
on the system as shown in Figure 5. The response will be the
same response as shown in Figure 4 because the controller
cancels the behavior of the process. Therefore, the controller
can achieve the ideal desired response with systems that can
be completely inverted.

Figure 7: The Impulse Response for t = 1 sec

As known, the smith predictor compensates the time delay

in the systems. After applying smith predictor, the system can
Figure 5: Overall Closed Loop System be dealt as a delay free system.
To be more emphasis, consider the DC motor system
presented in this paper. The system with smith predictor is
V. MODIFIED IMC FOR TIME DELAYED SYSTEMS
shown in Figure 8. Its response is shown in Figure 9 which
For systems with time delay, modified IMC controller will agrees with smith predictor.
not face any problem. Because of the controller design does
not dependent on time delay value. In other words the
controller structure is the same as time delay when varied
because the time delay part is not invertible and will not be
included in design. However, this type of controllers cannot
compensate systems with long time delay. If the motor
transfer functions has a time delay of 0.3 and 1 sec the result
of simulation in Figure 6, and Figure 7 tell us that the time
delay affect the response of the system by shifting it as the Figure 8: The system of DC motor with smith predictor
value of time delay. In addition, the response changed if it was
compared with the ideal one in Figure 4 such that more
overshoot and longer settling time. This result concludes us
to; if the time delay is very long the system will be unstable
and the response also will be unbounded. So it is difficult to
obtain satisfactory performance of control systems with time
delay, which is a well recognized problem in many control
processes. The solution of this problem is represented by
smith predictor.

58
A Modified Internal Model Control for Unstable – Time Delayed System

Figure 9: Result of DC motor system with smith

predictor
Where, K(s) is a stable stabilizing controller. So, the unstable
VI. MODIFIED IMC FOR UNSTABLE SYSTEMS poles of is identical to . Therefore Q(s)
The general rule of IMC based on is that the system to be
is stable. Then obtain that the dotted block is simplified to
controlled must be stable to apply the IMC controller. If the
which is a stable rational system and the controller
system is unstable, it should be stabilized before IMC
controller is applied by any proportional controller or any can be designed easily.
other controllers. This rule is considered as a necessary
condition to apply the modified IMC controller. Therefore, in
all cases, need two controllers to handle unstable systems.
Consider the first order unstable system process with time
delay of the form:

Then choose a proportional controller K to stabilize this Figure 11: Modified IMC for Unstable systems
system as presented in Figure 10. K is intended to stabilize the
delay free unstable model , this simple proportional gain However, is a stable transfer function, it will contain
K will give a stable internal process unstable zeros so the inversion will make a problem. So in this
case another factorizing is recommended as discussed in
section 3 where

And the controller then will consider the term in its

Clearly, is stable if , then choose to make design.

I. Simulation & Results

The simulation will be held on the non-linear pendulum

Then the delayed form will be
system to control its angle and to compare the traditional IMC
with the modified IMC.
The pendulum can be modeled approximately as a linear
second order system [10]:
That can be handled with smith predictor to compensate the
time delay and design a trivial controller.
Y(s)
+
-
The controller transfer function of IMC is
K(s)

a) Figure 10: Stabilizing unstable system

Another solution was proposed to solve the instability. The modified IMC controller is
Consider Figure 11 and let is factorized in another way
such that:

0.8

Where is a stable proper rational function and 0.6

is bi-proper antistable and minimum phase function. 0.4

The term antistable refers to a system with all its poles in the 0.2

open RHP and minimum phase refers to a system with all its 0

zeros in the open LHP. -0.2

And let -0.4

-0.6

-0.8

-1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

59
International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-1, Issue-1, October 2011
Figure 14: A plant/model mismatch of IMC system

Figure 12: Impulse disturbance input response

0.8

0.6

0.4 Figure 15: A plant/model mismatch of modified IMC system

0.2

0
5
x 10
6
-0.2

-0.4 4

-0.6
2
-0.8

-1 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

-2
Figure 13: response of modified IMC
-4

Figure 12 and Figure 13 illustrate the impulse responses of the

-6
pendulum system for the two controllers which show that the 0 2 4 6 8 10 12 14 16 18 20

response of the modified IMC was quicker.

The pendulum system is a nonlinear system and its transfer Figure 16: Response of IMC due to plant/model mismatch
function is a result of the linearization operation so the
mismatch is present in all cases. So we will choose some 1

0.8
parameters and vary their values in the model such that the
0.6
plant and model transfer function are different.
0.4

0.2

-0.2

-0.4

-0.6

-0.8
0 2 4 6 8 10 12 14 16 18 20

Then the controller of IMC takes the form

b)

c) Figure 17: Response of modified IMC due to

plant/model mismatch
The controller of modified IMC takes the form
The system seems that does not have time delay, but in many
cases there is a time delay in almost all systems due to
physical components characteristics and storage elements in
The responses of the two techniques are displayed in Figures the system although it might be very small.
16 and 17. The results are very clear to say that the modified
IMC structure is now the best and overcome the mismatch and Based on this, assume there is a small time delay in the
regulate its output to be zero against the traditional IMC pendulum system beside a mismatch in this delay between the
structure, which behaves unstable, and the controller fails to plant and the model to make the competition worth.
regulate the output.

60
A Modified Internal Model Control for Unstable – Time Delayed System

Figure 18: IMC structure with time delay mismatch Figure 21: Response of a modified IMC time-delayed
system

According to Shamsuzzohal and Lee, they compared the

performance of the PID controller against IMC controller and
the results indicate that the proposed IMC controller provides
fast and smooth set-point response without a loss of
disturbance performance [11].
Figure 19: Modified IMC structure with time delay In the same way, this part compares the results of the
preceding results with the new approach result when applying
In Figure 18 the system has a time delay for the plant t = 2 sec it to the same system.
while its model has t=2.5 sec. Figure 19 has also t=2 sec delay The transfer function of the system is:
for its system.
The time response of each system are shown in Figures 20
and Figure 21 and the responses showed the ability of
modified IMC since it regulates the output and overcome the
perturbation and save the stability. In the other hand, the
traditional IMC lose the control and the response unbounded
to finally yield to instability. The disadvantage of the
modified IMC takes more time response and gets stability but
this is forgiven when compared with the traditional one.
4
x 10
3

Figure 22: Response of both controllers to the proposed

-1
system

-2
Figure 22 exhibit the results of three controllers. The solid
line indicates the traditional IMC, the bold line for modified
-3
0 5 10 15 20 25 30 35 40 45 50 IMC without smith predictor, the dotted line for modified
IMC with smith predictor.
Figure 20: Response of IMC with a mismatch time-delayed Simulation results indicate that the response of the
system modified IMC with smith predictor is superior which
compensate the time delay. The modified without smith
1.2 predictor has a small overshoot but it needs a little smaller
energy to eliminate disturbance than traditional IMC.
1 The traditional IMC suffer from a delay of 3 sec to
compensate the disturbance and overshoot 50%. On the other
0.8
side, the new method without SP has an overshoot of 20% but
0.6
with SP it is 40% but without any delay.

0.4 II. Conclusion

0.2
In this paper, a new approach of control design of internal
model controller was proposed. The proposed design method
0
focuses on modifying the old general structure of IMC and
-0.2
develops a new model structure while saving the same general
0 5 10 15 20 25 30 35 40 45 50
concept of using the invertible version of the system in the
controller design. The new approach combines the IMC
structure and the traditional structure of a control problem and
this demonstrates an excellent performance and behavior

61
International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-1, Issue-1, October 2011
against different disturbance inputs and model uncertainty
presented in model parameter mismatch. Beside that a smith
predictor is added to promote the design to compensate the
delayed time systems. Also a proposed stabilizer has
mentioned to deal with unstable systems.

References
[1] Jeffrey E. Arbogast, Douglas J. Cooper. “Extension of IMC tuning
correlations for non-self regulating processes”. ISA Transactions, 2007,
46(4), 303-311.
[2] Garcia, C. E. and Morari, M., “Internal Model Control - Unifying Review
and Some New Results”, Industrial Engineering Chemical Process
Design and Development, vol. 21,1982.
[3] Scott A. Geddes , Thesis, “Internal Model Control (IMC) of a Fruit
Drying System”, University of Southern Queensland,2006.
[4] Jiliang Shang, Guangguang Wang, “Application Study on Internal Model
Control in Boiler Burning System”, 2010.
[5] Caifen Fu, Wen Tan, “Active control of combustion instability via IMC”,
2008.
[6] JIN Qi-bing FENG Chun-lei LIU Ming-Xin, “Fuzzy IMC for Unstable
Systems with Time Delay” IEEE Pacific-Asia Workshop on
Computational Intelligence and Industrial Application, 2008.
[7] Wen Tan, Horacio J. Marquez, Tongwen Chen, “IMC design for unstable
processes with time delays”,2003, accessed on: April 2011, Online:
http://dsp.vscht.cz/konference_matlab/MATLAB09/prispevky/035_hant
a.pdf .
[8] Kou Yamada, “Modified Internal Model Control for unstable systems”,
Proceedings of the 7th Mediterranean Conference on Control and
Automation (MED99) Haifa, Israel - June 28-30, 1999
[9] Smith, O. J. M., “Closer Control of Loops with Dead Time” Chem. Eng.
Progress, 1975.
[10] Byronic company, “Pendulum system model manual”, 2001, accessed
on: April 2011, online:
http://www.lehigh.edu/inconsy/lab/experiments/PCS_Manual.pdf .
[11] M. Shamsuzzohal, Moonyong Lee, “IMC Based Control System Design
of PID Cascaded Filter”, SICE-ICASE International Joint Conference,
2006.

Dr. Basil Hamed is Associate Professor of Electrical Engineering

Department, Islamic University of Gaza, Palestine, since 1999. He has
Bachelor Degree in Electrical Engineering from New Mexico State
University, NM. USA in the year of 1989, he received Master degree from
University of New Orleans, La. USA in the year of 1992, and earned his PhD
(Fuzzy Control System) from New Mexico State University, NM USA in the
year 1999. He has 15 years of teaching experience and has published many
papers in national and international journals. His fields of interest include
Control Systems, Fuzzy Control, Simulation & Modeling, FPGA, Signal and
Image Processing.

Eng. Walid Issa is a research Assistant August, 2010-Present at Electrical

Engineering Department, Islamic University of Gaza (IUG) Palestine. Since
2007, he has Bachelor Degree in Electrical Engineering from IUG and
received Master degree in 2011 from the same institute. He has worked as a
teaching assistant in 2007 to 2010 in IUG and had an excellent experience of
Microcontrollers, FPGA, Robotics, Power Electronics, Matlab and Labview.

62