Sunteți pe pagina 1din 12

Asian Journal of Control, Vol. 19, No. 2, pp.

112, March 2017


Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/asjc.1385

SMITH PREDICTOR BASED FRACTIONAL-ORDER-FILTER PID


CONTROLLERS DESIGN FOR LONG TIME DELAY SYSTEMS
Maamar Bettayeb, Rachid Mansouri, Ubaid Al-Saggaf, and Ibrahim Mustafa Mehedi
ABSTRACT
In this paper, an original model-based analytical method is developed to design a fractional order controller combined with a Smith predictor and a modified Smith predictor that yield control systems which are robust to changes in
the process parameters. This method can be applied for integer order systems and for fractional order ones. Based on
the Bodes ideal transfer function, the fractional order controllers are designed via the internal model control principle.
The simulation results demonstrate the successful performance of the proposed method for controlling integer as well
as fractional order linear stable systems with long time delay.
Key Words: Fractional order controller, PID controllers, FO-PID controllers, Smith predictor, long time delay
fractional order system, internal model control, Bodes ideal transfer function.

I. INTRODUCTION
In control system theory, the use of fractional order
controllers can be considered from two different points
of view. The first one is when a fractional-order controller is used to control a fractional-order plant. This
topic began to receive attention in recent years [15].
It is intuitively true, as also argued in [6,7], that the
fractional order models require much more than classical proportionalintegralderivative (PID) controllers
to achieve good closed loop performance. Control of
fractional order systems by fractional order controllers
can also be viewed as generalization of the non-integer
theory. Secondly, the use of fractional order controllers
is justified because they provide superior performance
[810], it is the most popular case and the literature is
now abundant.
In previous work [11,12], we developed a new structure of fractional controllers to control several integer
or fractional order systems with and without time delay.
Manuscript received February 9, 2016; revised May 14, 2016; accepted July 6,
2016.
Maamar Bettayeb is with the Electrical & Computer Engineering Department,
University of Sharjah, United Arab Emirates and Distiguished Adjunct Professor,
Center of Excellence in Intelligent Engineering Systems (CEIES) King Abdulaziz
University, Jeddah, Saudi Arabia (e-mail: maamar@sharjah.ac.ae).
Rachid Mansouri is with the Laboratoire de Commande et Conception des

Systemes
de Production, Mouloud Mammeri University, Tizi Ouzou, Algeria
(e-mail: rachid_mansouri_ummto@yahoo.fr).
Ubaid M. Al-Saggaf and Ibrahim M. Mehedi are with the Center of Excellence in Intelligent Engineering Systems, Electrical and Computer Engineering
Department, King Abdulaziz University, Jeddah, Saudi Arabia (e-mail: usaggaf@kau.edu.sa, imehedi@kau.edu.sa).
This project was funded by the Deanship of Scientific Research (DSR), King
Abdulaziz University, Jeddah, under grant no Gr/34/5. The authors, therefore,
acknowledge with thanks DSR technical and financial support.

The proposed controller is partitioned into two transfer


functions: a fractional order PID (FO-PID)-controller
if the system to be controlled is fractional or an usual
integer PID (IO-PID)-controller if the system is integer
one, cascaded with a fractional integrator if the system is time-delay free or a simple fractional filter if the
system has delay. However, we showed that such controllers will rapidly lose their effectiveness when the time
delay is significant with respect to the process time constant. Other improved control strategies are needed in
this case.
The control of time delay systems has been the subject of intensive research, due to the fact that many industrial processes exhibit an inherent time delay. Time delays
are commonly encountered in process control problems
arising from distance velocity lags, recycle, composition
analysis loops as well as the time required to transport
mass and energy. Processes with long time delays are not
easily controlled using PID controllers due to additional
phase lag caused by time delays that tend to destabilize
the closed-loop system.
The well-known Smith predictor method [13,14] is
successfully applied on stable processes with large time
delays. Unfortunately, the closed-loop performance can
be poor in the case of inevitable mismatch between the
model and the actual process. This mismatched Smith
predictor case has been the focus of much research work.
Several modified Smith predictors were then proposed.
One of these is the structure proposed by Wang et al. [15]
using a simple primary controller and a deliberately mismatched model in order to enhance the performance of
the real system.

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

Asian Journal of Control, Vol. 19, No. 2, pp. 112, March 2017

Due to the advantages of fractional calculus, the


design of fractional order controllers combined with
the Smith predictor (SP-FO-controller) has also become
an attractive research issue. Although the method has
been reported in the literature for integer order systems
[1620], the design of the SP-FO-controllers is still difficult and challenging because of its complex iterative steps
with non-analytical forms. Oddly, the control of fractional order systems with long time delay is not addressed
in the literature.
The design method of the FOF-PID controllers
developed in [11,12] considers only delay free systems
or systems with a small delay. These controllers lose
their effectiveness when the time delay is significant with
respect to the process time constant. In this paper, we
propose to extend this structure to control both fractional and integer order systems with a long dead time.
To do this, the proposed controller is combined with
a Smith predictor scheme. In addition, to improve the
robustness of the FOF-PID controller with respect to
variations of the system delay, the modified Smith predictor scheme [15] is also used. We focus on a simple and
effective analytical design method for the fractional order
controller. Based on Bodes ideal transfer function, the
proposed fractional order controller design can be analytically derived for integer and fractional order models
and can be easily applied to various process models.

II. PRELIMINARY
2.1 Bodes ideal transfer function
Oustaloups research contributions [8,9] with his
well known CRONE controllers have really popularized
fractional control. In the CRONE principle, the closed
loop reference model is given by
f (s) =
1+

1
( )
s
c

R.

as non-integer values. In the Bode diagrams of L(s),


the amplitude curve is a straight line of constant slope
20 dBdec, and the phase curve is a horizontal line at
2 rad.
f (s) plays a central role in the work developed here.
It is used as a reference model to design the controller.
For convenience, it is rewritten as
f (s) =

1
1 + c s+1

0<<1

(3)

When the performance of the closed-loop is specified in the frequency domain, c and are deduced from
the gain crossover frequency c and the phase margin m
by [11,12].
=

m
1
2

and

c =

1
+1
c

(4)

The step response of f (s) (when 0 < < 1) is similar to that of an underdamped second-order system for
which the damping ratio is such that (0 < < 1). Some
other useful formulae characterizing the time response of
f (s) are given in [22] where the authors propose a relationship that approximates the percent overshoot denoted
Mp (%), of the step response as a function of fractional
order . However, often it is rather desired to deduce the
value of giving a chosen overshoot of the step response.
We then used the relationship given in [22] to obtain values of the percent overshoot Mp (%) for different values of
between 0 and 1. We then used the function cftools
of Matlab to interpolate the curve giving as a function of the percent overshoot Mp (%). Equation (5) gives
this relationship:
6.75 105 Mp2 + 0.016 Mp + 0.066

(5)

(1)
2.2 Internal model control

This particular form has very important properties


because its phase margin remains constant independently
of the value of the system gain. Thus, the closed loop
system is robust to process gain variations and the step
response exhibits iso-damping property. Equation (1)
can be obtained if the corresponding open-loop is the
ideal transfer function given by Bode [21] which is
defined as
( )
c
L(s) =
(2)
s
c is the gain crossover frequency of L(s). is the slope
of the magnitude curve, and may assume integer as well

Internal model control (IMC), developed by Morari


and coworkers [2325], has been shown to be a powerful method for controller design. The IMC method
is based on assumed process models and leads to
analytical expressions for the controller settings. The
IMC approach allows model uncertainty and tradeoffs
between performance and robustness to be considered
in a more systematic way. Fig. 1 shows a block diagram
of IMC. The setpoint response of IMC configuration is
given by
CIMC (s) G(s)
y
=
r
1 + CIMC (s)(G(s) Gm (s))

(6)

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

M. Bettayeb et al.: FO-PID Controllers, Smith Predictor, Time Delay Fractional Order System

C(s) =

CIMC (s)
1 CIMC (s)Gm (s)

(13)

2.3 Smith predictor control

Fig. 1. Internal model control structure.

and the load response is given by


1 CIMC (s) Gm (s)
y
=
d
1 + CIMC (s)(G(s) Gm (s))

(7)

If Gm (s) = G(s), (6) and (7) are reduced to


y
= CIMC (s) G(s)
r

(8)

y
= 1 CIMC (s) G(s)
(9)
d
According to the IMC method, CIMC (s) is designed
in two steps:

Step 1. The process model is factored as


+

(s) Gm
(s)
Gm (s) = Gm

(10)

(s)
Gm

is the part of the model inverted by


where
+
the controller and Gm
(s) is the portion that the con+ (s) contains any
troller does not attempt to invert. Gm
time delays and right-half plane zeros and must have
a steady-state gain equal to one.
Step 2. The internal model controller is designed as
CIMC (s) =

1
f (s)
(s)
Gm

(11)

The Smith predictor controller was proposed by


Smith [13,14] for time delay compensation. The block
diagram of the conventional Smith predictive control is
shown in Fig. 3, where G(s) and Gm (s) are stable strictly
proper rational functions characterizing the delay-free
part of the plant and its model, and m are positive
constants standing for the time-delay for respectively the
plant and the model. C(s) is the transfer function of the
controller. The inner loop works to eliminate the actual
delayed output as well as to get the feedback of the predicted output to the controller. This makes it possible to
design the controller C(s) assuming no time-delay in the
control loop. Indeed, the transfer function between the
setpoint r and the output y of the closed-loop system is
Gcl (s) =

C(s)G(s)es
[
] (14)
1 + C(s) Gm (s) + G(s)es Gm (s)em s

In the perfect case where Gm (s) = G(s) and m = ,


(14) reduces to
Gcl (s) =

C(s)Gm (s) s
e
1 + C(s)Gm (s)

(15)

Due to the absence of the time-delay in the characteristic equation of the perfect closed-loop, the controller C(s) can be designed with respect to the time-delay
free part of the model regardless of the troublesome
time-delay. Nevertheless, due to the time delay in the
characteristic equation, (14), the design of C(s) becomes
much more complicated. To resolve the robustness

where f (s) is a low pass filter with a steady-state gain


of one. It typically has the form [23]:
f (s) =

1
(1 + c s)r

(12)

c is the the adjustable parameter which controls


the tradeoff between disturbance rejection performance and robustness of the closed loop system. r is
a positive integer chosen so that the IMC controller
is proper.
If one uses the conventional feedback control
showed in Fig. 2, it can be shown that the two block diagrams of Figs 1 and 2 are identical if controllers C(s) and
CIMC (s) satisfy the relation

Fig. 2. Conventional feedback control.

Fig. 3. Conventional Smith predictor control scheme.

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

Asian Journal of Control, Vol. 19, No. 2, pp. 112, March 2017

CIMC (s) =

D(s)

N(s) 1 + c s+1

(19)

The corresponding conventional feedback controller is then


C(s) =
Fig. 4. Modified Smith predictor structure.

problem of the Smith predictive method, Wang et al. [15]


proposed the modified Smith predictive control scheme
shown in Fig. 4. To design the controller C(s) using the
method proposed in [15], the model delay free part Gm (s)
is partitioned in two transfer functions as follows:
Gm (s) = gm0 (s) gm1 (s)

(16)

where gm0 (s) is a first order transfer function


gm0 (s) =

Km
1 + Tm s

(17)

Km and Tm are respectively the static gain and the time


constant optimized so that gm0 (s) fits at best G(s). This
choice facilitates the design procedure of C(s). gm1 (s) is
given by
gm1 (s) =

1
(1 + Tm s)n1

(18)

n being the order of Gm (s).

III. SMITH PREDICTOR BASED


FRACTIONAL ORDER FILTER PID
CONTROLLER DESIGN PROCEDURE
The conventional feedback control structure shown
in Fig. 2 is considered here where the plant transfer function G(s) is integer or fractional. The proposed design
technique is based on the equivalence between IMC and
conventional feedback. Moreover, the plant being a time
delay system, a Smith predictor is also used. The details
of the controller design for the proposed method is presented in what follows.
Let Gm (s)em s be the model of an integer or fractional order linear time invariant (LTI) system where
Gm (s) = N(s)
. N(s) and D(s) are rational polynomials if
D(s)
Gm (s) is an integer, and irrational if Gm (s) is fractional.
We assume that N(s) and D(s) are coprime and all their
roots are in the left half plane. In addition, Gm (s) is
assumed to be strictly proper.
The closed-loop reference model being that of (3),
according to (11), the internal model controller is
given by

D(s)
1
.
1 + c s+1 em s N(s)

(20)

Equation (20) shows that C(s) is composed of two


terms. The fractional filter
H(s) =

1
1 + c

s+1

(21)

em s

which depends on the time delay m and the two design


parameters c and of the closed reference model. The
Smith predictor is mainly used to implement H(s). The
second portion of C(s) is the inverse of the delay-free part
of the model. We will see in the next section how, using the
integer or fractional integral term contained in H(s), the
D(s)
term N(s)
can be approximated by a PID controller if G(s)
is integer or by a FO-PID controller if G(s) is fractional.
To improve the robustness of the controller with
respect to modeling errors, especially the time delay term,
we use the modified Smith predictor structure. When the
system to be controlled is an integer, according to the
Wangs method [15], C(s) is designed for the first order
model gm0 (s). In this case, the controller C(s) is a simple
fractional integrator cascaded with a PI controller. It is
given by
(
)
1 Tm
1
C(s) =
1
+
(22)
Tm s
c s Km
In the fractional order case, we present the detail of
the method for the so-called one non-integer order plus
time delay (NIOPTD-1) system given by
Gm (s)em s =

K em s
1 + T s

(23)

Using the IMC based controller design method,


C(s) is given by [12]
C(s) =

1
1 + c

s+1

em s

1 + T s
K

(24)

which can be written in the form of a fractional filter


cascaded with a FO-PI controller as
C(s) =

1
c s+1 +

1em s
s

(
)
T
1
1+
K
T s

(25)

On the other hand, if we want to use the modified


Smith predictor structure, to improve the robustness with
respect to uncertainties in the time delay , the delay free

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

M. Bettayeb et al.: FO-PID Controllers, Smith Predictor, Time Delay Fractional Order System

part of model Gm (s) must be approximated as mentioned


in (16), where
gm0 (s) =

K
1+

T
2

(26)

and
gm1 (s) =

1
1+

T
2

(27)

In this case, the controller C(s) is given by


(
)
1
T
1
C(s) =
1+ T
c s+1 2K
s

(28)

3.1 Implementation of H(s)


The fractional controller H(s) in (21) can be
written as
1
c

s+1

1em s
s

The integral operator


D(s)
N(s)

1
s

1
s

3.2 Stability condition of H(s)


Stability of the fractional controller C(s) in (20)
is characterized by the stability of the fractional filter
1
in which the presence of the integral term
c s+1 +1em s
insures a zero steady state error. The stability condition of
the fractional filter is given by the following theorem [26].
Theorem 1. The fractional filter

H(s) =

shows that the fractional controller finally obtained is the


CRONE-controller associated with the Smith predictor.

(29)

is then cascaded with the

term
which is approximated by a PID or a FO-PID
controller. The block diagram of Fig. 5 shows the implementation of the controller C(s).
It can also be implemented by the block diagram of
Fig. 6 since the IMC based method and direct synthesis
method are identical for the delay free system. This figure

Fig. 5. Block diagram of the fractional controller.

Fig. 6. Smith predictor-based fractional order controller.

H(s) =

1
c s+1 + 1 es

(30)

is stable if and only if the time constant c verifies


the condition
( )
+1
2sin
m
2
c >
(31)
[(2 )]+1
To establish the proof, Walton Marshalls method
[27,28] is used. However, the goal is not to find the limit
of the delay m but to find the stability condition of
the filter due to the parameter design c . The proof of
this theorem is based on the analysis of the roots of the
quasi-polynomial p(s, ) = 1 + c s+1 es , see [26] for
more details.

IV. NUMERICAL EXAMPLES


In this section, the proposed methods will be
applied to two typical delayed processes. The two chosen delayed processes present particular difficulties for
the fractional control problem. Indeed, the first process is represented by an integer forth order plus time
delay transfer function and the second one is a typical
NIOPTD-1 system. In all the simulation schemes, the
fractional integral or derivative operators are approximated using Oustaloups approximation with the parameters wl = 104 , wh = 10+2 and the number of cells is
N = 10. In addition, to keep the integrator or differentiator aspects of the fractional operators outside the
approximation frequency range, the fractional integrator
operator s1 is approximated by 1s (s1 ) and the deriva)
(
tive operator s is approximated by s(s1 ). s1 and
( 1 )
s
are respectively the Oustloups approximation of
s1 and s1 .
The simulation scheme is shown if Fig. 7, where
G(s)es is the plant, Gd (s) the disturbance transfer function, SP-FO controller and MSP-FO controller are respectively the standard and the modified

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

Asian Journal of Control, Vol. 19, No. 2, pp. 112, March 2017

Fig. 7. Simulation scheme.

Smith predictor structures combined with a fractional


order controller. r is the setpoint input and d is the
load disturbance.
4.1 High order integer system
The features of the standard and the modified
Smith predictor structures combined with fractional
order controllers are firstly demonstrated on the following high order integer process with long time delay taken
from [29,30].
e10 s
G(s) =
(32)
(1 + s)(1 + 0.5s)(1 + 0.25s)(1 + 0.125s)
To use the Smith predictor structure, the transfer
function (32) is approximated by two models using the
well known Skogestads half rule method [31].

The first model is a first order plus time delay model


given by
G1 (s) =

e10.625 s
1 + 1.25s

(33)

The second one is the second order plus time delay


model
G2 (s) =

e10.25 s
1 + 1.625s + 0.625s2

(34)

The specifications imposed on the closed-loop system are, on one hand, an overshoot of the step response
around 20%. According to (5), the value of which
achieves this objective is = 0.3875 (we took = 0.4).
On the other hand, to have a good trade-off between performance and robustness [32], we have chosen c = 10
(the value of the system time delay). The controllers thus
obtained are:

Based on the first order model (33), the SP-FO-PI


controller is given by
C1 (s) =

(
)
1
1
1.25
1
+
1.25 s
10s0.4

(35)

Fig. 8. Closed-loop step response of the SP-FO-Controller for


the integer order system (32).

Based on the second order model (34), the


SP-FO-PID controller is
C2 (s)=

)
(
1
1
1.625 1+
+0.3846 s
0.4
1.625 s
10s

(36)

Note that, in the simulation tests, the derivative


action of the PID controller is cascaded with the low
1
.
pass filter 1+0.02
s
Note that also, according to the theorem given in
Section 3.2, the limit values of c which guarantees the
stability of the fractional controller are c limit = 3.352
for the SP-FO-PI controller and c limit = 3.188 for the
SP-FO-PID controller. These values are much smaller
than the value adopted, c = 10. A unit step setpoint
input is applied at t = 0 and a negative step disturbance of
magnitude 0.1 is applied at t = 80. The disturbance trans1
fer function is Gd (s) = s+1
. Simulation results are shown
in Figs 8 to 10. Fig. 8 shows the step response of the
closed-loop systems and reference model, Fig. 9 shows
the control signal given by the SP-FO-PI and SP-FO-PID
controllers, and Fig. 10 shows the output response due
to the load disturbance. The performance of each of the
two controllers is summarized in Table I where Mp is
the overshoot of the closed-loop step response, ts (2%)
is the 2% settling time and limit is the maximum time
delay accepted for closed-loop stability. All the results,
except limit , show that both controllers are able to achieve
the objectives.
To improve the performance of the SP-FOController, especially the maximum time delay accepted
for closed-loop stability, the modified Smith predictor is
used. The controller is designed based on the delay free
1
first order model of (33), gm0 (s) =
1.25 . The second
1+

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

M. Bettayeb et al.: FO-PID Controllers, Smith Predictor, Time Delay Fractional Order System

Fig. 9. Control signal of the SP-FO-Controller for the integer


order system (32).

Fig. 11. SP-FO-PI-controller versus modified


SP-FO-PI-controller.
Table II. Performance of the SP-FO-Controller and the
MSP-FO-Controller for the integer order system (32).
Controller
Reference
SP-FO-PI
MSP-FO-PI

Mp (%)

ts (2%)

limit

21,15
21.46
17.64

41.65
42.00
55.42

21.2
23.7

This figure shows that the SP-FO-PI-controller is much


better than the MSP-FO-PI-controller. Nevertheless, the
maximum time delay accepted for closed-loop stability is
greater for the MSP-FO-PI-controller compared to the
SP-FO-PI-controller as shown in Table II.
Fig. 10. Load disturbance of the SP-FO-Controller for the
integer order system (32).

In this second example, consider the fractional


order system given by

Table I. Performance of the SP-FO-Controller


for the integer order system (32).
Controller
Reference
SP-FO-PI
SP-FO-PID

Mp (%)

ts (2%)

limit

21,15
21.46
21.22

41.65
42.00
41.85

21.2
20.8

G(s) =

portion of the model used in the modified Smith predic1


tor structure is gm1 (s) = ( 1.25
)2 . In this case, by using
1+

(28) for = 1, the controller is given by


(
)
1
1
C(s) =
0.625
1
+
0.625 s
10 s0.4

4.2 Fractional order system

(37)

Fig. 11 shows the step response of the reference


model and those of the closed-loop systems for the
SP-FO-PI-controller and for the MSP-FO-PI-controller.

2 e4 s
1 + 3 s0.5

(38)

We then designed two fractional order controllers,


one combined with the usual Smith predictor and the
other with the modified Smith predictor structure. The
parameters of the closed-loop reference model are c =
= 4 and = 0.4 to have an overshoot of the closed-loop
around 20%. The SP-FO-PI Controller is designed based
on the perfect model (38) and the MSP-FO-PI controller
is designed using the method described in Section 2.3
where gm0 (s) = 1+1.52 s0.5 and gm1 (s) = 1+1.51 s0.5 . The
controllers thus obtained are respectively

The FO-PI controller combined with the Smith predictor is

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

Asian Journal of Control, Vol. 19, No. 2, pp. 112, March 2017

Table III. Performance of the SP-FO-PI-Controller and the


MSP-FO-PI-Controller for the fractional order system (38).
Controller
Reference
SP-FO-PI
MSP-FO-PI

Mp (%)

ts (2%)

limit

20.15
20.15
17.04

20.9
20.9
36.15

8.65
10.29

Fig. 12. Closed-loop step response of the SP-FO-PI and


MSP-FO-PI Controllers for the fractional order
system (38).

Fig. 14. SP-FO-PI versus MSP-FO-PI Controllers for 20%


in gain variation.

Fig. 13. Control signal of the SP-FO-PI and MSP-FO-PI


Controllers for the fractional order system (38).

C(s) =

)
(
1
1
1.5
1
+
4 s0.9
3 s0.5

(39)

The FO-PI controller combined with the modified


Smith predictor is
)
(
1
1
(40)
C(s) = 0.9 0.75 1 +
4s
1.5 s0.5

The simulation scheme is that of the Fig. 7 where


2
the disturbance transfer function is Gd (s) = 1+3s
. In this
case too, a unit step setpoint input is applied at t = 0 and
a negative step disturbance of magnitude 0.1 is applied
at t = 50. Figs 12 and 13 show the simulation results
obtained and Table III summarizes the performances
of the two controllers. Since the SP-FO-PI-controller

is designed using the exact model, it gives exactly the


desired performance, Nevertheless, in this case too, the
maximum time delay accepted for closed-loop stability is
larger with the MSP-FO-PI-controller.
To test the robustness of the two control structures
(standard and modified Smith predictors), with respect
to variations of system parameters, we changed each of
these parameters by 20% of their nominal values. Simulation results are shown in Fig. 14 when the value of
the static gain changes, in Fig. 15 when the value of the
time constant changes and in in Fig. 16 when it is the
time delay value that changes. All these results show
that the modified Smith predictor structure is better than
the standard Smith predictor one. Note also that the
iso-damping property of the closed-loop step response
with respect to the time constant is not satisfied.
4.3 Long time delay system
This example treats the control of water distribution in an irrigation main canal pool characterized by
a large time delay [17]. In this section we compare the
performance of the Fractional order controller designed
in [17] and the proposed FOF-PID controller proposed
here to control this system, particularly the robustness of
the two controllers to changes in the static gain and the
time delay of the system. However, the experimental test

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

M. Bettayeb et al.: FO-PID Controllers, Smith Predictor, Time Delay Fractional Order System

Fig. 15. SP-FO-PI versus MSP-FO-PI Controllers for 20%


in time constant.

Fig. 17. Closed-loop step response of the SP-FOF-PIcontroller (curve 1), the SP-PI -controller (curve 2)
and the MSP-FOF-PI-controller (curve 3).
Table IV. Performance of our controller and the
controller proposed in [17].
Controller

Mp (%)

ts (2%)

SP-FOF-PI
SP-PI

6.8
7.3

7540
6800

57.5
76

0.0025
0.0027

Gm (s) =

Fig. 16. SP-FO-PI versus MSP-FO-PI Controllers for 20%


in time delay.

bench being unavailable, the comparison will be done on


the identified second order linear model given by
G(s) =

1.11e550 s
(1 + 1250s)(1 + 50s)

(41)

In [17], the designed controller with Smith predictor is a fractional order PI controller cascaded with an
integer order filter, in order to ensure a zero steady state
error, given by
)
(
)(
1.89
s + 1.52 104
R(s) = 3.71 + 0.2
(42)
s
s
In our case, to obtain a Smith predictor PI controller cascaded with a fractional order filter, the controller design is based on the estimated first order with
time delay model obtained using the Skogestads half
rule method [31].

1.11
e575 s
1 + 1275s

(43)

To obtain, in closed loop, a step response similar to


that obtained in [17] the parameters used to design our
controller are c = 500 s and = 0.05. The resulting
SP-FO-PI controller is then
(
)
1
1
C(s) =
1148.6
1
+
(44)
1275 s
500 s0.05
Simulation results are shown in Fig. 17 and the
performance, in both time and frequency domains,
of the SP-PI -controller proposed in [17] and the
SP-FOF-PI-controller are summarized in Table IV. It
can be seen that the performance of the SP-PI -controller
and the SP-FOF-PI-controller proposed here are met
in both cases. However, the closed-loop step response
obtained using the modified SP-FOF-PI-controller is
very poor.
In Fig. 18, applying the SP-PI -controller (dashed
line) and the proposed SP-FOF-PI-controller (solid line),
the closed-loop step responses are plotted with the
open-loop gain variation from 0.88 to 1.32 ( 20% variation from the nominal value 1.11). It can be seen obviously that the overshoot of the SP-FOF-PI-controller
(solid line) remain constant under gain variations,
i.e., the iso-damping property is exhibited unlike the
SP-PI -controller proposed in [17].

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

Asian Journal of Control, Vol. 19, No. 2, pp. 112, March 2017

system is easily adjusted by only two parameters: to


adjust the overshoot and c to adjust the settling time of
the step response. Note finally that the use of the modified
Smith predictor improves considerably the robustness
of the fractional order controller with respect to variations of system parameters and the maximum time delay
accepted for the closed-loop compared to the standard
Smith predictor scheme.

REFERENCES

Fig. 18. Step response of the Smith predictor based


controllers (42) and (44) with gain variation.
Table V. Performance of our controller and the
controller proposed in [17].

limit

SP-PI

SP-FOF-PI

MSP-FOF-PI

930

870

2900

To show the robustness of the three controllers with


respect to the variation of the time delay, an additional
delay is added to the output of the system. We then
increased its value up to obtain the stability limit; the corresponding value is denoted limit . The obtained results
are summarized in Table V.
These results show that the SP-PI -controller is
more robust than the proposed SP-FOF-PI-controller.
However, it should be noted that the result obtained
using the MSP-FOF-PI-controller is outstanding since
the added delay is five times larger than the system delay.

V. CONCLUSION
A new Smith predictor and a modified Smith predictor based fractional order controller design schemes
for integer or fractional order systems with long time
delay are proposed. Based on the Bodes ideal transfer function and design using the internal model control
method, the proposed fractional order controller has a
very interesting structure for its implementation. Indeed,
the controller obtained is a fractional integrator cascaded
with an integer PID or a FO-PID controller. Two important advantages of the new scheme are that the process
to be controlled can be integer or fractional, and the
fractional integrator parameters are those of the closed
loop reference model. Another important merit of the
proposed method is that the response of the closed loop

1. Hamamci, S., Stabilization using fractional-order


PI and PID controllers, Nonlinear Dyn., Vol. 51,
No. 12, pp. 329343 (2008).
2. Luo, Y. and Y. Q. Chen, Fractional order [proportional derivative] controller for class of fractional
order systems, Automatica, Vol. 45, pp. 24462450
(2009).
3. Narang, A., S. L. Shah, and T. Chen, Tuning of
fractional PI controllers for fractional order system
models with and without time delays, Proc. Amer.
Control Conf., Baltimore, MD, pp. 66746679 (2010).
4. Luo, Y., Y. Q. Chen, C. Y. Wang, and Y. Pi, Tuning fractional order proportional integral controllers
for fractional orders systems, J. Process Control,
Vol. 20, pp. 823831 (2010).
5. Luo, Y., Y. Q. Chen, and Y. Pi, Experimental study
of fractional order proportional derivative controller
synthesis for fractional order systems, Mechatronics, Vol. 21, pp. 204214 (2011).
6. Podlubny, I., Fractional Differential Equations, Academic Press, San Diego (1999).
7. Chen, Y. Q., Ubiquitous fractional order controls,
Proc. Second IFAC Symp. Fractional Deriv. Applicat.,
Porto, Portugal, pp. 481492 (2006).
8. Oustaloup, A., La commande CRONE, Herms Edition, Paris (1991).
9. Oustaloup, A. and B. Mathieu, La commande
CRONE du scalaire au multivariable, Herms Editeur, Paris (1999).
10. Podlubny, I., Fractional-order system and PI D
controllers, IEEE Trans. Autom. Control, Vol. 44,
No. 1, pp. 208214 (1999).
11. Bettayeb, M. and R. Mansouri, IMC-PIDfractional-order-filter controllers design for integer order systems, ISA Trans., Vol. 53, No. 5,
pp. 16201628 (2014).
12. Bettayeb, M. and R. Mansouri, Fractional
IMC-PID-filter controllers design for non integer
order systems, J. Process Control, Vol. 24, No. 4,
pp. 261271 (2014).

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

M. Bettayeb et al.: FO-PID Controllers, Smith Predictor, Time Delay Fractional Order System

13. Smith, O. J. M., Close control of loops with dead


time, Chem. Eng. Process, Vol. 53, pp. 217219
(1957).
14. Smith, O. J. M., A controller to ovecome dead time,
ISA Journal, Vol. 6, pp. 2833 (1959).
15. Wang, Q. G., Q. Bi, and Y. Zhang, Re-design
of smith predictor systems for performance
enhancement, ISA Trans., Vol. 39, pp. 7992
(2000).
16. Feliu-Batlle, V., R. Rivas-Perez, and F. J.
Castillo-Garcia, Fractional order controller robust
to time delay variations for water distribution in
an irrigation main canal pool, Comput. Electron.
Agric., Vol. 69, pp. 185197 (2009).
17. Feliu-Batlle, V., R. Rivas-Perez, F. J. Castillo-Garcia,
and L. Sanchez Rodriguez, Smith predictor
based robust fractional order control: Application
to water distribution in a main irrigation canal
pool, J. Process Control, Vol. 19, pp. 506519
(2009).
18. Castillo-Garcia, F. J., V. Feliu-Batlle, and R.
Rivas-Perez, Time domain tuning of fractional
order controllers combined with a smith predictor
for automation of water distribution in irrigation
main channel pools, Asian J. Control, Vol. 15,
pp. 819833 (2013).
19. Maamri, N., M. Tenoutit, and J. C.
Trigeassou, A comparison of fractional smith
predictors, Proc. Eur. Control Conf., Zurich,
Switzerland, pp. 39913996 (2013).
20. Lino, P. and G. Maione, Loop-shaping and easy
tuning of fractional-order proportional integral controllers for position servo systems, Asian J. Control,
Vol. 15, No. 3, pp. 110 (2013).
21. Bode, H. W., Network Analysis and Feedback Amplifier Design, Van Nostrand, New York (1945).
22. Barbosa, R. S., J. A. T. Machado, and I. M.
Ferreira, Tuning of PID controllers based on Bodes
ideal transfer function, Nonlinear Dyn., Vol. 38,
pp. 305321 (2004).
23. Morari, M. and E. Zafiriou, Robust Process Control,
Prentice-Hall, Englwood Cliffs (1989).
24. Garcia, C. E. and M. Morari, Internal Model Control I. A unifying review and some new results, Ind.
Eng. Chem. Process Des. Dev., Vol. 21, pp. 308321
(1982).
25. Rivera, D. E., M. Morari, and S. Skogestad, Internal model control 4. PID controller design, Ind.
Eng. Chem. Process Des. Dev., Vol. 25, pp. 252265
(1986).
26. Titouche, K., R. Mansouri, M. Bettayeb, and U. M.
Al-Saggaf, Internal Model ControlProportional
Integral DerivativeFractional-Order Filter Controllers Design for Unstable Delay Systems, J. Dyn.

27.

28.

29.

30.

31.

32.

Syst. Meas. Control-Trans. ASME, Vol. 138, No. 2,


p. 021006 (2016).
Walton, K. and J. E. Marshall, Direct method
for TDS stability analysis, IEE Proc. D:
Control Theory Appl., Vol. 134, pp. 101107
(1987).
Silva, G. J., A. Datta, and S. P. Bhattacharyya,
PID Controllers for Time-Delay Systems, Birkhuser,
Boston (2005).
Hgglund, T, A predictive PI controller for processes with long dead times, IEEE Control Syst.
Mag., Vol. 12, No. 1, pp. 5760 (1992).
Vrecko, D., D. Vrancic, D. Jurcic, and S. Strmcnik,
A new modified Smith predictor: The concept,
design and tuning, ISA Trans., Vol. 40, pp. 111121
(2001).
Skogestad, S., Simple analytic rules for model
reduction and PID controller tuning, J. Process
Control, Vol. 13, pp. 291309 (2003).
Shamsuzzoha, M. and S. Skogestad, The setpoint
overshoot method: A simple and fast closed-loop
approach for PID tuning, J. Process Control,
Vol. 20, pp. 12201234 (2010).

Professor Maamar Bettayeb received the


B.S., M.S., and Ph.D. degrees in Electrical Engineering from University of Southern California, Los Angeles, in 1976,
1978 and 1981, respectively. He worked
as a Research Scientist at the BellaireResearchCenter at Shell Oil Development Company, Houston, Texas, USA during 1981/1982.
From 1982 to 1988, He directed the Instrumentation and
Control Laboratory of High Commission for Research
in Algeria. In 1988, He joined the Electrical Engineering Department at King Fahd University of Petroleum
and Minerals, Dhahran, Saudi Arabia. He has been Professor at University of Sharjah UAE since August 2000.
He also held the position of Advisor to the Chancellor
for Graduate Studies and Scientific research for the Years
2004/2006 and Director of Research and StudiesCenter
for the Year 2005/2006 at University of Sharjah. He is
presently the Vice Chancellor for Research and Graduate
Studies at University of Sharjah. He has published over
300 journal and conference papers in the fields of control and signal processing. He has also supervised over
50 M. Sc. and Ph. D. students. His recent research interest are in H optimal control, rational approximation,
signal and image processing, process control, networked
control systems, fractional dynamics and control, nonlinear estimation and filtering, soft computing, wavelets,
renewable energies and engineering education.

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

Asian Journal of Control, Vol. 19, No. 2, pp. 112, March 2017

Professor Rachid Mansouri of Electrical Engineering Department, Mouloud


Mammeri University of Tizi-Ouzou,
Algeria, received the B.S. and M.S.
degrees in Electrical Engineering from
Ecole Nationale Polytechnique dAlger,
Algiers, Algeria in 1990 and 1993, respectively and the Ph.D. degrees in Electrical Engineering
from University of Tizi-Ouzou Algeria in 2008. He
has been Professor at University of of Tizi-Ouzou
since June 2015. His research interests include H
optimization, Electrical machine control, order model
reduction, structured system, and fractional order
systems control.
Professor Ubaid M. Al-Saggaf received
double B.Sc. with highest honors in
Electrical Engineering and mathematics in 1980 from King Fahd University
of Petroleum and Minerals (KFUPM),
Saudi Arabia. He received his M.Sc. in
1983 and the PhD in 1986 in Electrical
Engineering from Stanford University, USA. He joined
King Abdulaziz University in September 2010 where
currently he is the Director of the Center of Excellence in

Intelligent Engineering Systems (CEIES) and a Professor


at the Electrical and Computer Engineering Department.
Prior to that he worked at KFUPM and the Ministry of
Defense. His field of interests and specializations cover
a wide spectrum from theoretical to practical aspects of
engineering including systems, control, signal processing,
optronics and engineering education.
Ibrahim Mustafa Mehedi received the
B.Sc. in Electrical and Electronic Engineering in 2000 from RUET, Bangladesh.
He received the M.Sc. in Aerospace Engineering from UPM, Malaysia in 2005.
Obtaining a Japanese Govt. MEXT scholarship he completed his Ph.D. in Electrical Engineering and Information Systems in 2011 while
he was a Research Assistant of the Global Center of
Excellence (GCOE) of the University of Tokyo and Japan
Aerospace Exploration Agency (JAXA). He joined King
Abdulaziz University, Saudi Arabia in 2011, where currently he is a member of the Center of Excellence in
Intelligent Engineering Systems (CEIES). Prior to that
he worked at KFUPM, and the CocaCola Bottling Plant.
His research interests include space robotics, modern
control system design and implementation.

2016 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd

S-ar putea să vă placă și