IJIREEICE ISSN (Online) 2321 – 2004

ISSN (Print) 2321 – 5526

INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERING
Vol. 4, Issue 6, June 2016

Design of Decoupler and Performance Analysis

of Distillation Column
Palanisamy Boopathi Nevetha1, Maruthai Suresh2
PG Student, Department of CIE, St. Joseph’s College of Engineering, Chennai, India1
Associate Professor, Department of ICE, St. Joseph’s College of Engineering, Chennai, India2

Abstract: This paper deals with the design of decoupler for TITO process to enhance its performance. MIMO process
is where all the inputs and outputs are coupled to each other and therefore it is very difficult to control. In this type of
process the interactions will be very strong and to eliminate such interactions among their loops, it is needed to design a
decoupler. After designing of decoupler then the various control techniques have been done such as PID (proportional-
Integral-derivative), MPC (Model Predictive Controller), MRAC (Model Reference Adaptive Control). The
mathematical model of Wood and Berry distillation column is considered for the analysis. It is one of the most
underestimated fields of chemical engineering and has been around for well over hundred years. Simulation results are
done in MATLAB simulink and comparison analyses of various controllers are carried out. The valuation of diverse
time domain parameters is done to establish that the adaptive control and Model predictive controller has diminutive
overshoot and speedy response as compared to PID controller .The performance evaluation illustrates a good
relationship between PID,MPC and adaptive controller.

Keywords: PID, Auto tuning, Decoupling, Distillation column, MPC, Adaptive.

I. INTRODUCTION

Multi input, multi output (MIMO) systems shows the components PID controller and adaptive Mechanism [19].
process with more than one input and output which require Thus this paper presents the simulated case study of
multiple control loops. These multivariable systems can be distillation column of wood and berry model which is
complex with loop interactions that consequences in designed with decoupler with various controllers such as
variables with unpredicted effects [1]. One of the example PID, MPC, MRAC. Lastly comparing all the performance
of MIMO systems is distillation column. It is the process simulation results like ISE, IAE, ITAE errors is done to
of separation of two or more substances .This process is prove the best controller.
also an example of loop interaction, hence these
interactions is necessary to be avoided as changes in one II. DISTILLATION COLUMN
loop might cause destabilizing changes in another loop[4].
To do so, MIMO systems can be decoupled into individual Process depiction
SISO loops. Decoupling is done using various techniques Distillation column is very often encountered in chemical
that includes pairing of input –output which is made processes for the separation of mixture into its individual
through perceptive guessing or by mathematical methods. components [8]. This column have number of trays and
One such mathematical method is Relative Gain Array each tray has material and heat capacities . A weir on one
(RGA) .To decouple two feedback loops, the decouples side of the tray maintains a liquid level at suitable height
have to be designed and it is done by ideal decoupling on the tray.
method [15]. After the design of decouple suitable
controller is chosen .In recent years many superior
controllers have been discovered. The PI and PID are the
most admired controllers of both in the dominion of the
academic and industrial application [5]. The
implementation of this controllers is easy. It helps to
correct the error among the measured outputs and desired
outputs of the process which improve the transient and
steady state responses as much as possible.
Advanced controller such as MPC is used which is
engender an online feedback control by using the open-
loop optimization [14]. The prediction of the future plant
response with help of a process model is the basic
principle of MPC and it tries to minimize a set horizon
objective function which is the sum of future predicted
errors and control moves. Model Reference adaptive
controller is designed. This controller consists of two Fig-1 schematic diagram of distillation column

Copyright to IJIREEICE DOI 10.17148/IJIREEICE.2016.4610 40

 IJIREEICE ISSN (Online) 2321 – 2004
ISSN (Print) 2321 – 5526

INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERING
Vol. 4, Issue 6, June 2016

Distillation column has 4N+10 variables and control k11 k12 12.8 −18.9
k= =
variables are XD (Distillate composition, mol fraction) , X B k21 k22 6.6 −19.4
(Bottoms composition ),M D (liquid holdup in reflux
drums, mols), M B (Liquid holdup in column base/ reboiler, Therefore
2.009 −1.009
mols) and pressure[14]. ˄=
−1.009 2
Distillation may be carried out either as a batch or as a
Hence the recommended pairing is Y1 and U1 and Y2 and
continuous operation. A distinctive distillation column
U2.
consists of reboiler, condenser, reflux drum etc and has
one feed stream and two of the product streams. The top
V. DECOUPLER AND CONTROLLER DESIGN
distillate is referred as Xd and bottom distillate is referred
as Xb. Distillation column used in different applications in
PID:-
the fields of petroleum industries, chemical industries ,in
Initially PID and PI controller is used and it is not tuned
food processing[7] .
independently rather by decentrailized Relay feedback
method the values of PID and PI is determined [17].
III. PROCESS INTERACTIONS
(PID)
P=0.01; 0.067;
Transfer Function Model (2 x 2 system)
I= -0.23;0.016;
The open loop input-output relationship can be written in
D=0.008; -0.54;
the matrix form

Y1 G11 G12 U1 Decoupling control strategy

= ---(1)
Y2 G21 G22 U2 Here we are using simplified decoupling technique [10]. In
general the best disturbance rejection is not obtained by
The identified MIMO process is Distillation Column perfect decupling but that the best response is obtained by
(Wood and Berry model) some tuning of the decoupler and feedback controller [9].
so decoupling matrix T is taken to decouple a process
12.8e −s −18.9e −3s with a transfer matrix G[16] .Generally some elements of
XD (s)
= 1+16.7s 1+21s R(s)
(2) T is chosen and then feedback controllers are tuned. In
XB (s) 6.6e −7s −19.4e −3s S(s)
simplified decoupling the common choice is that T11=
1+10.9s 1+14.4s
T22= 1 and the off diagonal elements
G21 G21
Process interactions may induce undesirable interactions T12 = − G22- and T21 = − G11
between two or more control loops. Control loop
interactions are due to the presence of a third feedback G12
loop[2]. In order to design the decentralized controller, the T= 1 − G11
. (4)
G21
suitable pairings between manipulated inputs and − 1
G22
controlled outputs are chosen using (RGA) analysis to
weaken the interactions.

IV. RELATIVE GAIN ARRAY

It provides two types of useful information such as

measure of process interactions and Recommendation
about the best pairing of controlled and manipulated
variables. It requires knowledge of steady state gains but
not process dynamics.

λ 1−λ
˄= (λ= λ11) (3)
1−λ λ

where,
1
λ= Fig 2- block diagram of 2x2 decoupler
k12k21
1−
k11k22
λ12 = 1- λ11 = λ21 VI. MODEL PREDICTIVE CONTROLLER

Recommended controller Pairing MPC is the one of the class of advanced control
-It corresponds to the λij which have the largest positive techniques is most often used in the process industries. It
value that is closest to one. –It is suggest not to select the is indeed straight-forward for multivariable systems with
Pairings which correspond to negative [3]. time delays which finds its application in discrete domain
For identified wood and berry model the steady state and also in chemical industries. It has ability to handle
matrix is constraints in a optimal fashion [14]. The basic idea

Copyright to IJIREEICE DOI 10.17148/IJIREEICE.2016.4610 41

 IJIREEICE ISSN (Online) 2321 – 2004
ISSN (Print) 2321 – 5526

INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERING
Vol. 4, Issue 6, June 2016

behind MPC is that the authenticity allows the present VIII. SIMULATION RESULTS AND DISCUSSION
time slot to be optimized, while keeping future time-
horizon. The major principle of MPC is to determine a 1.2
sequence of control moves so that the predicted response
moves to the set point in an optimal manner. The 1
determination of prediction and control Horizon comes
under superposition principle and its calculation based 0.8
upon the current and predicted measurement [14]. It is a

process variable
model based approach. It is found to be giving the better 0.6
response comparatively by the simple build-in process
procedural. 0.4

0.2

-0.2
0 50 100 150 200 250 300 350 400
time
Fig-5 step response of a 2×2 distillation column system
of decoupler and PID controller.(---Xb & ----Xd)

1.4

1.2 Xb
Xd
1
Fig-3 block diagram of MPC
Process Variable

0.8
VII. ADAPTIVE CONTROLLER
0.6
In this day adaptive controller is used as the most
advanced controller in the field of any control field. The 0.4
main two sub components of this controller are PID
controller and athe adaptive Mechanism. Where the PID 0.2
block is fixed , and by keeping the initial plant parameter
the gains have been turned to achieve the total stability. 0

-0.2
0 100 200 300 400 500 600 700 800 900 1000
Time(sec)

Fig- 6 response of distillation column of decoupler and

MPC

1.5
y1
y2

1
Output Response

Fig- 4 The block diagram of MRAC.

0.5

Next comes the Adaptive Mechanism where its output (Ѳ)

gets changed based on error (e) between plant output (Yp)
and reference model output (Ym). How fast it can adapt
(or change its output) depends on parameter called
0
learning rate, gamma. Higher the value of gamma, faster it 0 10 20 30 40 50 60 70 80 90 100
can adapt to any changes in plant. But there are some side Time(sec)

effects also. Controller output (U) is calculated by: U = Uc Fig -7 Response of 2x2 distillation column of decoupler
* theta. and adaptive controller

Copyright to IJIREEICE DOI 10.17148/IJIREEICE.2016.4610 42

 IJIREEICE ISSN (Online) 2321 – 2004
ISSN (Print) 2321 – 5526

INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERING
Vol. 4, Issue 6, June 2016

Table – 1 performance criteria Of PID , MPC and adaptive [14] Sivakumar.R, Shennes Mathew (2013): “Design and Development
of Model Predictive Controller for Binary Distillation Column”
with decoupler.
[15] Branislav T. Jevtovića, , Miroslav R.Mataušekb,(2010) :“PID
controller design of TITO system based on ideal decoupler” Journal
Method Setlting controller IAE ISE ITAE of Process Control Volume 20, Issue 7, August 2010.
Time [16] Pontus Nordfeldt , Tore Hägglund (2006): “Decoupler and PID
145 Loop 1 10.055 4.8790 113.736 controller design of TITO systems”, Journal of Process Control
PID 80 Loop 2 2.1042 1.8109 66.801 Volume 16, Issue 9, October 2006.
[17] L.B.Xiea, , L.S.Shiehb, , F.Pana, , J.S.H. Tsaic, , , J.I. Canelond,
120 Loop 1 5.7002 3.0411 97.310
(2014)“Design of decoupling and tracking controllers for
MPC 130 Loop 2 2.2164 0.8901 64.335 continuous-time transfer function matrices with multiple time
Adaptive 40 Loop 1 3.1516 2.0544 65.331 delays”, Journal of Process Control Volume 24, Issue 1, January
45 Loop 2 1.4419 0.627 52.617 2014.
[18] Juan Garridoa, , , Francisco Vázqueza, Fernando Morillab (2012)
:“Centralized multivariable control by simplified decoupling”-
IX. CONCLUSION Journal of Process Control Volume 22, Issue 6, July 2012.

In this paper, a new control stratergy of combination of

decoupler and adaptive controller was proposed for
decentralized TITO wood and Berry Model. The analysis
evidently explains that set-point in PID control is far from
constrains whereas Mpc control is better than PID.
Compare to other two controls, Adaptive set-point is
closer to constraint. In this paper it was proved that
MRAC tries to bring the process as close as possible to
constraints without infringing them. Here the performance
indices such as IAE(Integral Absolute
Error),ITAE(Integral Time Absolute Error),ISE(Integral
Square Error) based on integral error for a step set point
are considered for comparison as they are generally
accepted as a good measure for system performance.
Hence it is concluded that the performance of MRAC with
decoupler is better to compared with other two controls.

REFERENCES

[1] Francisco Vázquez Fernando Morilla (2015) :“ tuning decentralized

pid controllers for mimo systems with decouplers”-IFAC- jun 2015
[2] Bharathi.M, Selvakumar.C.(2012): “DynamicModeling, Simulation
and Control of MIMO Systems” IJAREEIE- volume 1, issue 2
[3] Manash Kumar Sethi, Tarun Kumar Dan (2012): “Designing of
Decoupler for a 4×4 Distillation Column Processes with PID”
[4] Tao Liu, Weidong Zhang, Furong Gao, (2006) :“ Analytical
decoupling control strategy using a unity feedback control structure
for MIMO processes with time delays”,
[5] Harsha R. Mokadam and Balasaheb M. Patre, Dilip K.
Maghade(2013) :“ Tuning of multivariable PI/PID controllers for
TITO processes using dominant pole placement approach”
[6] Hanuma Naik. R, Ashok Kumar. D.V, Anjaneyulu. K.S.R.(2012) :
“ Controller for Multivariable Processes Based on Interaction
Approach”,
[7] Sutanto hadisupadmo, widodo.R.J, harijono a tjokronegoro1,tatang
hernas soerawijaya, (2013) : “Binary Distillation Column Control
by Decoupling Controller”-conference 21st Telecommunications
forum TELFOR- 2013
[8] Meenakshi1.S, Almusthaliba.A, Vijayageetha.V (2013): “MIMO
Identification and Controller design for Distillation Column”
ijireeice - volume 1, issue 2
[9] Kurt v.t. Waller, (1974):“decoupling in distillation”, AIChE -
Volume 20, Issue3 May 1974
[10] Ramadevi.C, Vijayan.V(2014): “ Design of Decoupled PI
Controller for Quadruple Tank System” -IJSR- Volume 3 Issue 5,
May 2014
[11] Cintia Marangoni, Joel G. Teleken, Leandro, Werle.O. Ricardo A.
F. Machado “Ariovaldo Bolzan5 (2010) :“Multivariable control
with adjustment by decoupling using a distributed action approach
in a distillation column”
[12] Bharathi.M and Selvakumar.C (2012) :“ Interaction reducer for
closed-loop control of multivariable systems”
[13] Truong Nguyen Luan Vu, Moonyong Lee∗ (2010) :“Independent
design of multi-loop PI/PID controllers for interacting multivariable
processes”,

Copyright to IJIREEICE DOI 10.17148/IJIREEICE.2016.4610 43