You are on page 1of 5

IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 616

OcsIyn and PnaysIs oI OcncIIc PyorIIhm


Dascd LonIrocrs IorIon LIncar LIquId
1ank SysIcm
R.Valarmathi, P.R.Theerthagiri, S.Rakeshkumar
Depatment of Electronics Instrumentation Engineering
School of Electrical Electronics Engineering
SASTRA University, Thanjavur
.S/ruc/- This paper deals with design of various controllers
such as Proportional Integral Derivative (PID) controller,
Internal Model Controller(IMC) and transfer function controller
by using Genetic Algorithm(GA) for a non-linear liquid conical
tank system. Level control of a conical tank is a complex issue
because of the nonlinear nature of the tank. For each stable
operating point, a First Order Process with Time Delay model
was identifed using process reaction curve method. The need for
improved performance of the process has led to the development
of the optimal controllers. GA is an evolutionary algorithm that is
proposed in this aspect. The performance of GA based controllers
is compared with conventional controllers tuned using Ziegler
Nicholas (Z-N) closed loop method and Z-N open loop method.
The comparison is done based on rise time, settling time,
overshoot and Integral Time Absolute Error(ITAE) and it was
found that the GA based IMC controller is better suited for this
process.
AcJw0r05: cncIc g0rIhm, nIcO0 300c L0nIr0,
^0nhnc0r 5J5Icm 0n01 L0nIr0cr.
I. INTRODUCTION
L
ONICAL tanks are extensively used in process industries,
petrochemical industries, hydrometallurgical industries,
food process industries and wastewater treatment industries.
Control of conical tank is a challenging problem because of its
non-linearity ad constantly changing cross section. Hence for
these reasons the conical tank process is taken here. The
innovative technologies ae required for process control
applications. Controlling of nonlinear system in real time is
difcult. For controlling this type of system PID controllers
are used. Conventional PID controllers are simple, robust
provided the system is linear. But the process considered here
has nonlinear characteristics which is represented as piecewise
linearized models .In PID controllers tuning adjustments are
carried out because of the recurrent changes in gain and time
constant parameters of the system.
Open loop and closed loop are used for PID tuning
adjustments. For both open loop and closed loop, Z-N tuning
method is used. Using process reaction curve method, the
trasfer fnction model paameters are found for all the
regions and controller parameters are tuned using Z-N tuning
formula. Z-N method provides a closed loop response with a
quarter-decay ratio [1]. In general PID control is a feedback
control technology and in industries 90% of automatic
controllers are made up of it [2]. A conventional PID
controller has several limitations in handling the nonlinear
process, so various techniques have been proposed to
overcome these limitations [3, 4]. The other techniques are GA
based PID controller, GA based transfer function controller
and GA based IMC controller.
Genetic algorithms ae used in fnding ideal solutions for
non-linea systems [5]. It is also used for tuning PID
controllers. Genetic algorithms used to fnd the best solution
fom the set of possible solutions known as popUlation.
Genetic algorithm composed of two functions namely
encoding scheme and evolution function. The encoding
scheme is used to fnd set of solutions. Primarily a set of
chromosome are installed and they are set to radom value.
From the evolution process subsequent generation are created
fom population ad process is completed afer reaching
predefned value. IMC provides an apparent famework for
control system design and tuning [6, 7]. IMC is based on the
internal model principle. IMC compensates disturbance and
model uncertainty based on its design procedure. IMC
controller depends on the order of the model and control
performance requirements.
The rest of the paper is organized as follows. The
experimental setup is discussed in section II. The modeling is
presented in section III. The controller design is given in
section IV. The simulation and results are presented in section
V. Section VI presents the conclusion of the paper.
II. EXPERIMENTAL SETUP
The experimental setup includes a conical tank, a storage
tak, pump, rotameter, a differential pressure transmiter, a
current to pressure converter (lI converter), a pneumatic
control valve, an interfacing ADAM's module and a personal
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 617
computer (PC).
l - - - - - - - - - - - _,

OPT
I RFFRYO
- pmf


'OT\R
Fig.l Schematic Diagram.
The differential pressure trasmitter is been calibrated to
read the level of 0-38 cm in the conical tank in terms of 4-20
m ad this current output from the DPT is converted into
corresponding voltage This voltage at the input of ADC of
ADAM module is interfaced with computer through the RS-
232 port of the Pc. The output current signal of the DAC is
given to a current to pressure (lIP) converter which is
connected to the pneumatic control valve.
The infow rate is thus adjusted by changing the stem
position of the control valve fom flly open to fully close.
The control signal fom the PC is transmitted to the liP
converter in the form of current signal (4-20) m, which
converts it to corresponding 3-15 psi of compressed air;
fther given as input to the pneumatic control valve. The
pneumatic control valve is actuated by this signal to produces
the required fow rate of water in the conical tank to maintain
its level.
The ADAM's module has 8 analog input ad 4 analog
output channels with the voltage range of IO Volt. The
sapling rate of the module is 18 samples per sec and baud
rate is 9600 bytes per sec with 16-bit resolution. The programs
written in m-fle of MATLAB sofware is then linked via
ADAM's module with the sampling time of 30 seconds. Table
I gives the technical specifcations of experimental setup.
III. MODELING
In process industries system identifcation is usually
caried out by step response based methods. The conical tank
step response, assumption of a suitable model and estimates of
model paaeters ae used to get an operative and exact
mathematical model. The open loop step response is atained
by changing the manual mode output from the controller with
the ideal value of fow through the control valve. The model
validated by process reaction curve method. The FOPTD
model(l) is given by
G
p
(s)
= Kp
l .. Y
(I)
TABLE I
TECHICAL SPECIFICATIONS OF EXPERIMENTAL SETUP
Part Name Details
Material :Stainless Steel
Conical Tank
Top Diameter : 38 cm
Bottom Diameter : 2.54cm,
Height: 38 cm
Reservoir
Material :Stainless Steel ,
Volume : 60 liters
Diferential Pressure
Type Capacitance, Range
Transmitter (DPT)
(2.5 - 250)mbar, Output
(4 - 20)mA
Pump Centrifugal 0.5 HP
Size \ Pneumaticactuated"
Control valve Type: Air to close
Input (3 - 15) psi
Rotanleter Range (0 - 5) Ipm
Air regulator
Size 114" BSP
Range (0 - 2.2 )bar
Iconverter
Input (4 - 20) mA
Output (3 - 15) psi
Pressure gauge Range (0 - 30) psi
8 Analog Input & 4 Analog Output
Data Acquisition Card Sapling rate: 18 saples per sec
(DAC) Baud Rate: 9600 bytes per sec
Resolution: 16 bit
Where K
p
is the process gain, b is the delay time, T is the
time constant. Table 2 gives model parameters of conical tank.
IV. CONTROLLER DESIGN
.. C1J111!1 111C 11111C
GA is an optimization technique that is used to minimize
the user defned objective fnction. For control applications as
consider in this paper, minimizing the steady state error, rise
time, settling time, peak overshoot and performance indices
such as ISE, IAE ad IT AE. The proposed work basically
considered to design a controller that will minimize the IT AE
which is calculated fom the formula (2).
TABLE 2
MODEL PARAMETERS OF CONICAL TAN
Infow
K .
Ed
Level (steady (time
Region rate (elay
(I pm)
(cm) state consta
time )
gain) nt)
l3.1
I 1-2 to 4.5 469 24
17.6
17.6
2 2-3 to 5 309 28
22.6
22.6
3 3-4 to 7.6 503 30
30.2
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 618
Since IT AE only consider the overall error with respect to
time. But the peak overshoot occurs at the initial time of the
response, so error contribution with respect to time is very
low. The controller designed by this technique will provide a
satisfactory response but with high peak overshoot.
In order to reduce the peak overshoot along with IT AE, a
multiobjective fnction was considered as a ftness fnction
for GA. The ftness function was calculated by multiplying the
peak overshoot and IT AE as a penalty if the response has a
peak overshoot. If the response does not have a peak shoot
then the IT AE alone was used as ftness function.
ITAE
[

(2)
o
If percentage overshoot >1 then
fitness functionpercentage overshoot * ITAE
else
fitness functionITAE
end
Fig.2 Agorithm for ftness function
1. O.1.11 1H C1KC111K
The optimal values of the conventional PID controller
parameters k
p
, ki' kd is found using GA. For the PID control
design, it is ensured that controller settings assessed results in
a stable closed loop system. To start up with GA certain
parameters need to be defned. It includes population size,
bounday, ftness fnction ad number of tuning vaiables.
. O.1.111K11K 11lC C1KC111K
For a FOPDT system, Pade approximation is performed
and the transfer fnction of the process is a second order
system. Hence a controller also chosen to be a second order
system as follows,
(4)
Tuning parameters are
a
o
'
a
I
'
a2
,
b
o
'
b
l dHC
b2
.
1. O. 1.11 11 C1KC111K
The method of control, which is based on an accurate
model of the process, leads to the design of a control system
that is stable and robust. A robust control system is one that
maintains satisfactory control in spite of chages in the
dynamics of the process. G'C(S) is proper if n=l, and a good
rule-of-thumb is to choose Tf to be twice as fast as the open
loop response.
The conventional IMC design works well and produces a
response with zero overshoot but with larger rise time and
settling time. In order to have a trade-off between peak
overshoot and settling time GA was used to tune the IMC
controller. The only tunable paraeter in a IMC controller is
'n' as shown in equation(3) The range of n can be fom 0.01
to 2 is considered as constraint for GA optimization.
:
K( n :)
(5)
Fig 3. Schematic of the IMC
V. SIMULATIONS AND RESULTS
The GA based controllers are designed for conical tank.
The controller's performace is compared with the
conventional PID controllers. This comparison is done for all
the three regions of process. The corresponding fgures are
shown in Figure 4, 5 and 6.
The performance of the controllers is compaed in terms of
trasient response as well as IT AE. Table 4 and 5 shows the
performance analysis of controllers.
TABLE 3
PI CONTROLLER GAIN V ALVES FOR ALL REGIONS
REGION
CONTROLLER
Kp Ki Kd
TYPE
ZN-OL 5.2111 0.1086 62.5332
1 ZN-CL 4.32 0.0864 54
GA-PI 2.323 0.02 1.754
ZN-OL 2.6486 0.0473 37.0804
2 ZN-CL 2.4 0.048 30
GA-PI 1.473 0.005 3.65
ZN-OL 2.6474 0.0441 39.711
3 ZN-CL 2.19 0.0365 32.85
GA-PI 1.897 0.005 0.724
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 619
TALE 4
TUNG PATERS VALUES OF TRNSFER FUNCTION
CONTROLLER
REGION
a2 a
l
a
o b2 b
l
b
o
I 0.0081 0.6817 0.7381 2.3285 1.2023 0.3142
2 1.5794 0.5673 0.8479 0.5200 0.8177 0.5752
3 0.9352 0.7087 0.5037 0.4421 0.2986 0.3587
Type of controller Region 3
GA-[MC 1967.20
TALE 6
COMPARISON OF CONTROLLERS PERFORMANCE I TERMS OF ITAE
Type of controller
ZN-OL
ZN-CL
GA-PlD
GA-TF
GA-lMC

. 2
'
I
-
.

1.5
0.5
o
o
---

Region 1 Region 2 Region 3


1324.90 1867.30 1967.20
1685.20 2108.00 2375.70
629.92 1032.70 1742.70
2030.90 7357.40 4706.10
338.20 850.97 906.56
Comparison of Controllers pertormance fr Region-l


---- ---------- ---------- ---------- -----
100 200 300
Time (sec)
400
_GA-IMC
-ZN-CL
-ZN-OL
-GA-
-PID
i
GA-TF
500 600
Fig.4 Comparison of controller's performace for Region- I
Comparison of Controllers pertormance for Region-2
1. 8

I
6 1.
1. 4 `
-
I
- -

-
: 1.2

k
-
I
- - -
C
'Q
I
-
'
1
I


V
-GA-IMC
0.8
E
0.6
- -
I
- - --GA-PID
=
0
<
:
-
<

'

0
:
-GA-TF
0.4 -ZN-CL
o. 2 ---1
o
o 100
-
I
200
-
300
Time (sec)
-
400
-ZN-OL

500
Fig.5 Comparison of controller's performace for Region-2
1.5
0.5
Comparison of conlrollers perfrmance fr Region-3
i
P

------- ---------- ----------- -------------------


I _ I
_

-GA-PID
___ ___________ __________ ___________ ____ -
ZN-OL
-ZN-CL
-GA-TF
-GA-IMC
600
J
J 100 200 300
lime (sec)
400 500
Fig.6 Comparison of controller's performace for Region-3
TALE 7
600
COMPARlSON OF CONTROLLERS PERFORMANCE IN TERMS OF TRANSIENT
RESPONSE
Region
Type of Rise Time Overshoot Settling time
controller (sec) ,) (sec)
ZN-OL 67 29.87 550
ZN-CL 59 30.36 526
1
GA-PI 62 11.34 450
GA-TF 56 7.80 276
GA-lMC 49 5.20 249
ZN-OL 17 44.07 489
ZN-CL 17 43.41 462
2 GA-Pl 31 16.52 363
GA-TF 36 11.82 348
GA-lMC 48 5.48 309
ZN-OL 15 44.54 437
ZN-CL 19 38.95 402
3
GA-Pl 25 27.33 348
GA-TF 37 12.59 333
GA-lMC 54 5.48 329
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 620
N1. CONCLUSION
GA based IMC controller, GA based PID controller and
GA based transfer function controller designed for a non-linear
conical tank system ad it is performance is compared with
conventional PID controllers. The controllers performance
compaison results proves that the GA based IMC controller
has a lesser settling time, rise time, overshoot and IT AE.
Even though the conventional controllers such as PID,
transfer fction and IMC has may empirical formula and
tuning procedures. But there is no unifed procedure for to
tune the controllers for all the process. GA is a optimization
tool that provides a unifed platform for controller tuning as
proposed in this paper. The same work can be extended other
type of nonlinear process such as Spherical tank, Heat
exchanger etc.
REFERENCES
[ I ]. J.G. Ziegler and N.B. Nichols, "Optimum settings for automatic
controllers", Trans.ASM. Vol. 64, pp.759 - 768, 1942.
[2]. Astrom. K, T. Hagglund, "PI Controllers; Theory, Design ad Tuning".
Instrument Societ of America, Research Triangle Park, 1995.
[3]. Herrero.J.M, Blasco.X, Martinez.M and Salcedo.J. V, "Optimal PI
Tuning with Genetic Algorithm for Non Linear Process Models",
15thTrienniai World Congress, 2002.
[4]. I.B. Lee & S.W. Sung, "Limitations and countermeasures of PI
controllers", Industrial & Engineering Chemistry Research, 35, 1996,
pg.2596-2610.
[5]. D. Goldberg, "Genetic algorithms in searching, optimization and
machine learning". Reading, MA: Addison-Wesley, 1989.
[6]. Chipperfeld, A. J., Fleming, P. J., Pohlheim, H. ad Fonseca, C. M, "A
Genetic Algorithm Toolbox for MATLAB". Proc. International
Conference on Systems Engineering, Coventry, U, 6-8 September,
1994.
[7]. W.L. Luyben. "Process modeling, simulation and control for chemical
engineers", Second edition, Tata McGraw Hill, USA, 1990.
[8]. Varsek, T. Urbacic and B. Filipic, "Genetic Algorithms in Controller
Design and Tuning". IEE Trans. Sys.Man and Cyber, Vol. 23, No. 5,
pp.1330-1339, 1993.
[9]. Nitya. S, Sivakumaran. N, Balasubramania. T, and Anantaraman. N,
2008, " !C based controller design for a spherical tank process in real
time", National conference in Advaced Techniques in Instrumentation
Control and Communication Engineering, 2008, pp.173 -178.
[10]. T O.Mahony, C J Downing and K Fatla, "Genetic Algorithm for PI
Parameter Optimization: Minimizing Error Criteria". Process Control
and Instrumentation 2000,26-28 July 2000, Universit of Stracthclyde,
pg. 148-153, July 2000.
[ I I ]. Marlin, T. E., "Process Control: Designing Processes and Control
Systems for Dynamic Performance". McGraw Hill, New York, 2000.
[12]. Luyben, W. L., "Process Modeling, Simulation ad Control for Chemical
Engineers". McGraw-Hill, New York, 1990.
[13]. Krishnaswamy, P. R., Chan, B. E. M., and Rangaiah, G. P., "Closed-loop
tuning of process control systems". Cher. Eng. Sci. 42, 2173-2182,
1987.
[14]. Garcia and Morari(1982) Garcia, CE. and MMorari (1982). "Internal
Model Control-I . A uniting review and some new results", Ind. Eng.
Cher. Process Des. & Dev., 21, pp.308-323.
[15]. Garcia and Morari(l985) Garcia, CE. and MMorari (1985). "Internal
Model Control-2. Design procedure for multivariable systems", Ind. Eng.
Cher. Process Des. & Dev., 24, pp.472-484.
[16]. Garcia and Morari(1985b) Garcia, CE. ad M.Morari (1985). "Internal
Model Control-3. Multivariable control law computation and tuning
guidelines", Ind. Eng. Chem. Process Des. & Dev., 24, pp.484-494.
ISBN: 978-81-909042-2-3 2012 IEEE

You might also like