International Journal of Trend in Research and Development, Volume 3(6), ISSN: 2394-9333

www.ijtrd.com
Modeling and Development of Decentralized PI
Controller for Nonlinear System
1
K.S. Jaibhavani and 2B.Hannuja,
1
Assistant Professor and 2PG Student,
1,2
Electronics and Instrumentation Engineering Department, Valliammai Engineering College, SRM Nagar,
Kattankulathur, TamilNadu, India

Abstract: The main of this project is to control the nonlinear process(MIMO-Multi Input Multi Output) is shown in
system (Quadruple tank process). For better control to achieve Figure1.
modeling of quadruple tank is required. The modeling is
B. Modeling
obtained by linearization technique and Jacobian matrix. The
designing of controller is a challenging task. The system The Modeling of a process is necessary to investigate
response is achieved using Decentralized PI (Proportional how the behavior of a process changes with time under
Integral) controller. The system response for two control influence of changes in the external disturbances and
paring configuration is simulated using MATLAB. manipulated variables and to consequently design an
appropriate controller. In such case a representation of the
Keywords: Quadruple Tank Process, Modeling, Decentalized
process is required in order to study its dynamic behavior. This
PI Controller, Ideal Decoupler, Control Paring Configuration.
representation is usually given in terms of a set of
I. INTRODUCTION mathematical equations whose solution gives the dynamic
behavior of the process. For each tank i=1...4, the
The primary objective of process control is to maintain
mathematical modeling is done by considering mass balance
the process at the desirable operating condition safely and
equation and Bernoulli’s law. The differential equation of
efficiently, while satisfying the environmental and product
quadruple tank is obtained as
quality requirement. In process control various parameters
(level, flow, pressure, temperature) need to be controlled. The dh1 1k1v1 a3 2gh3 a1 2gh1
level is one of the most important parameter required to be = + − 
dt A1 A1 A1
considered, because it indirectly affects remaining parameters.
The main idea of this work is to develop mathematical model
dh2 2k2v2 a4 2gh4 a2 2gh2
and control the process level using Conventional PI Controller. = + − 
The MIMO system response for two control paring is obtained. dt A2 A2 A2
II. DETAILS EXPERIMENTAL dh3 1 − 2 k2v2 a3 2gh3
= − 
A. Quadruple Tank Process dt A3 A3
The Quadruple tank is a laboratory process with four
interconnected tanks and two pumps. The process inputs are u1 dh4 1 − 1 k1v1 a4 2gh4
= − 
and u2 (input voltages to pumps) and the outputs are y1 and y2 dt A4 A4
(voltages from level measurement devices). The target is to
control the level of the lower two tanks with inlet flow rates. Where
The output of each pump is split into two using a three-way
valve. A1,A2,A3,A4-c.s area of the tank1,2,3,4

h1,h2,h3,h4-height of water in tank1,2,3,4

v1,v2-velocity of flow through pump1and2

k1,k2-pump constant

1,2-valve ratio

a1,a2,a3,a4-area of outlet pipe of tank1,2,3,4

g-acceleration due to gravity

By using Jacobian Matrix, the nonlinear differential is

linearized into state space model. The Jacobian matrix is
represented as

Figure1: Schematic Diagram of Quadruple Tank Process δf1/δh1 δf1/δh2 δf1/δh3 δf1/δh4
𝐟
Pump 1 is shared by tank 1 and tank 3, while pump 2   δf2/δh1 δf2/δh2 δf2/δh3 δf2/δh4 
𝐡
is shared by tank 2 and tank 4. Thus each pump output goes to δf3/δh1 δf3/δh2 δf3/δh3 δf3/δh4
two tanks, one lower and another upper diagonal tank and the δf4/δh1 δf4/δh2 δf4/δh3 δf4/δh4
flow to these tanks are controlled by the position of the valve
represented as Ɣ. The schematic diagram of a quadruple tank
IJTRD | Nov-Dec 2016
Available [email protected] 45
 International Journal of Trend in Research and Development, Volume 3(6), ISSN: 2394-9333
www.ijtrd.com
III. RESULTS AND DISCUSSION
δf1/δu1 δf1/δu2
𝐟 δf2/δu1 δf2/δu2  A. Decentralized PI Controller Design
 (𝐡, 𝐮) =
𝐮
δf3/δu1 δf2/δu2 The quadruple process is a MIMO system which has
δf4/δu1 δf2/δu2 multi loops. When a variation occurs in any inputs, the system
becomes nonlinear because of multi loop interactions. In order
The state space model of process is determined using Jacobian to eliminate the process interaction Decentralized controller is
matrix and coefficient of process state space is established used. The interaction is eliminated by using ideal decoupling
from Table1 below. control which is similar to the conventional feedback
controller. The block diagram of decoupling control and
Table 1: Parameters Specification process is shown in Figure3.
Parameters Value
A1=A2=A3=A4 63.585cm/2
a1=a2 1.246cm/2 Decoupling
Operating point of h1 &h2 12.4 &12.7 cm Process Transfer
control Function matrix
Operating point of h3 &h4 1.8 & 1.4 cm
matrix H(s) Gp(s)
k1= k2 3.3cm/3/vs
1 & 2 0.7 &0.6

 0.1232 0 0.2038 0 Figure 3: Block Diagram of Process with Decoupling Control.

A= 0  0.3668 0 0.2310
A. Decoupling Control
0 0  0.2038 0 For a two manipulated variable and two controlled
0 0 0  0.2310 variable, there will be two different configurations available
(n!). When input1 is subjected to change, the output of tank1
0.0367 0 changes and simultaneously tank2 output is changed because
of G21(s) and when input2 is altered, then the output of both
0 0.0316
B= tank1 and 2 are changed because of G12(s). For a 2*2 process
0 0.0211 transfer function, the Decoupling Control H(s) is obtained
0.0157 0 from
1  G12/G11
H(s)=
0.5 0 0 0  G21/G22 1
C=
0 0.5 0 0 The PI Controller value are tabulated
Table 2: PI Controller Gain Value.
0 0
D= Parameter Controller1 Controller2
0 0
Proportional gain 13.1286 18.3467
Integral gain 3.4624 12.8720
The transfer function matrix is determined from state space
model. Below matrix represents the process transfer function The process response for two control paring is
matrix achieved with the same controller values is shown.
G11(s) G12(s)
Gp(s)= = -
G21(s) G22(s) G11(s)
O/p1
0.0184/(s  0.1232) 0.0022/(s  0.1232)(s  0.2038) Gc1
i/p1 G12(s)
0.0018/(s  0.2310)(s  0.3668) 0.0158/(s  0.3668)
From the obtained process transfer function matrix, the
interactive process response is shown in below Figure2. G21(s)

Gc2 G22(s)
i/p2 - O/p2

Figure 4: 1-1/2-2 Control Paring Block Diagram

Figure 2: Interactive Process response of tank1 and tank 2

IJTRD | Nov-Dec 2016

Available [email protected] 46
 International Journal of Trend in Research and Development, Volume 3(6), ISSN: 2394-9333
www.ijtrd.com

Figure 5: 1-1/2-2 Control Paring O/p response of Tank1 Figure 8: 1-2/2-1 Control Paring O/p response of Tank1

Figure 6 1-1/2-2 Control Paring O/p response of Tank2

G11(s)
Gc2 Figure 9: 1-2/2-1 Control Paring O/p response of Tank2

G12(s) CONCLUSIONS
The output response of tank1 and 2 for the quadruple process
is tuned using Decentralized PI Controller. The two control
G21(s) paring configuration response is also shown. We infer that
1. Loop interactions are eliminated.
Gc1 G22(s) 2. The process reaches desired step input value.
Acknowledgments
Figure 7: 1-2/2-1 Control Paring Block Diagram The transient response for two control paring configuration are
tabulated below.

Table 3: Comparison of Two Control Paring Configurations.

Transient
1-1/2-2 control 1-2/2-1
Response
paring Control Paring
Parameters.
Rise time(Sec)
6.06 6.33
for tank1
Rise time(Sec)
4.63 4.39
for tank2
Settling 20.7 20.7
IJTRD | Nov-Dec 2016
Available [email protected] 47
 International Journal of Trend in Research and Development, Volume 3(6), ISSN: 2394-9333
www.ijtrd.com
time(sec) for
tank1
Settling
time(sec) for 12 12.1
tank2
Overshoot(%)
5.27 5.07
for tank1
Overshoot(%)
3.46 4.49
for tank2

From the tabulation we conclude that 1-1/2-2 control paring is

better when compared to 1-2/2-1 control paring.
References
[1] V.D Hajare, B.M.Patre, “Decentralized PID Controller
for TITO Systems Using Characteristic Ratio
Assignment with an Experimental Application ”.
[2] P.Srinivasarao and P.Subbaiah, “Centralized and
Decentralized of Quadruple Tank Process”,
International Journal of Computer Application .
Vol.68, April 2013..
[3] Jignesh Patel and Hasan Vhora, “Development of
Multi Input Multi Output Coupled Process Control
Laboratory Test Setup”,International Journal of
Advanced Research in Engineering and
Technology.Vol.7, pp.97-104,Jan-Feb 2016
[4] D.Angeline Vijula, Anu K, Honey Mol P and Poorna
Priya S “Mathematical Modelling of Quadruple Tank
System”, International Journal of Emerging
Technology and Advanced Engineering. Vol. 3, Dec
2013.
[5] Karl Henrik Johansson,”The Quadruple Tank Process-
A Multivariable Laboratory Process with an
Adjustment Zero”, IEEE Transactions on Control
Systems Technology, Vol. 8, May 2000.
[6] P.Sujatha, Dr.M.Bharathi, Dr.C.SelvaKumar, “Tuning
of Decentralized PI (PID) Controllers for TITO
Process”, International Journal of Scientific &
Engineering Research, Vol. 5, Jan 2014.

IJTRD | Nov-Dec 2016

Available [email protected] 48