1

complete substitute gasoline or as a feedstock for chemical

Multivariable control in a distillation column system

improvement by static decoupling
Victor medinaa, Jos D Prezb, Victor Pugliesec
alcohol [1].
AbstractThese instructions give you guidelines for preparing
papers for IEEE Transactions and Journals. Use this document
as a template if you are using Microsoft Word 6.0 or later.
Otherwise, use this document as an instruction set. The electronic
file of your paper will be formatted further at IEEE. Paper titles
should be written in uppercase and lowercase letters, not al
uppercase. Avoid writing long formulas with subscripts in the
title; short formulas that identify the elements are fine (e.g., "Nd
FeB"). Do not write (Invited) in the title. Full names of
authors are preferred in the author field, but are not required.
Put a space between authors initials. Define all symbols used in
the abstract. Do not cite references in the abstract. Do not delete
the blank line immediately above the abstract; it sets the footnote
at the bottom of this column.
Index TermsEnter key words or phrases in alphabetical
order, separated by commas. For a list of suggested keywords,
send a blank e-mail to keywords@ieee.org or visit
http://www.ieee.org/organizations/pubs/ani_prod/keywrd98.txt

I.

INTRODUCTION

Distillation is one of the most important separation operations,

mainly due to allowing separation of both ideal and non-ideal
mixtures to a large scale [1] This process is used to separate
two or more components of a solution liquid, using the
distribution of substances between the gas phase and liquid
according to differences in their relative volatilities or their
boiling points.
Extractive distillation is one of the main techniques used to
separate binary mixtures azeotropic. In this operation a third
component also known as agent is added separation or solvent,
which is characterized by being non-volatile, miscible, high
point no additional boil and form azeotropes. The solvent
affects the activity coefficient phase liquid component, so that
the mixture can be efficiently separated into the pure products
[2].one of the main applications of the extractive distillation is
in the separation of the mixture azeotropic ethanol-water. This
separation has a great industrial interest, because of potential
renewable energy, ethanol being used as an additive or

a.

Mechanical Engineering department, Universidad

Barranquilla Colombia vjmedina@uninorte.edu.co

del

Norte

b.

Mechanical Engineering department, Universidad del

Barranquilla Colombia. jperezd@uninorte.edu.co

Norte,

c.

Mechanical Engineering department, Universidad del

Barranquilla Colombia. vpugliese@uninorte.edu.co

Norte,

In the extractive column the solvent is added in a stage near

the top of the column and usually obtained as the top product
of more volatile component with a high purity, in this case
ethanol, and as bottoms product component is obtained
together with the less volatile solvent; This stream is fed to
another column called recovery column wherein the second
component is obtained at the top and the solvent in the
background which is again fed back into the first column.
Among the solvents studied for such separation are ethylene
glycol [3], glycerin [4], tetraethylene [1],most of these studies
focus on the feasibility of steady state operation and
Sensitivity analysis of the main variables that affect the
process.
In the distillation process as well as in many other chemical or
physical separation processes, different variables influence the
operation of the process variables such as temperature,
pressure, level and flow among others. To know and
manipulate these variables is essential for making a Detailed
operation both in steady state and dynamic analysis; for this
initially required automation and instrumentation of the spine
of the plant and then perform a control study aimed to
establish the different control loops, determining the Variable
couples manipulated - controlled variable that will maintain
the process in an operating point even specific to the influence
of different disturbances.
The study of control strategies in distillation columns has been
conducted over the Last 60 years [5] both by industry and
academia, despite this, there is general agreement among the
authors to choose the best configuration control [6]The goal of
many of these studies is to derive a procedure systematic and
comprehensive strategies to select control in distillation
columns, the problem is that there are plenty of distillation
columns and therefore different types of control structures.
Some techniques has been succesfull in the multivariable
control of the distillation collumns, this is the case of
Ogunnaike and Ray [7] in the 80s, whom developed a multidelay
compensator which, in the absence of model error, removes all
time delays from the closed loop characteristic equation. The
compensator, which can handle state, output, or input delays, was
tested through simulation on a number of test problems, giving
good results.

One of the early work in this regard was the doctoral thesis of
Vincent Grassi in 1991, showing their results in the book
Luyben 1992 [2]); it presents a methodology for the design of
control systems in distillation columns raising different
schemes or strategies and analyzing them through lax
simulations.

Recently the work of marangoni [8] presents an

approach which uses a multivariable control
strategy with distributed action in order to
minimize operation transients in distillation
columns when a disturbance in the temperature
feed is introduced. [9]

When the controllers are in automatic, the interaction between

loops can affect the performance of the control system. Second
part of this work is to use the technique of decoupling
described by [Smith], which will be done for static
decoupling. Decoupling can improve the control system if it is
applied carefully. The relative gain matrix provides an
indication of when decoupling can be beneficial. Static
decoupling is accomplished by leaving out the dynamic terms

The Purpose of this article is simulate the pilot-scale

distillation column, and show how using a decoupling
technique of interacting loops improves the performance of
control systems.

When the decouplers are implemented, the relationship

between the controlled and manipulated variables is

y ( s )=G p ( s ) D ( s ) x ( s )
1

II.

METHODOLOGY

To obtain a multivariable control process is important start

with the characterization of the system, where the transfer
functions between each manipulated and controlled variable
gain is realized, these variables are mated together to obtain
the necessary data to tune controllers.
Consider an open loop multivariable system with 3 inputs and
y i ,i=1,2,3 are
3 outputs, where x i , i=1,2,3 and
the mentioned inputs and outputs, respectively. G(s) is process
transfer functions matrix, expressed by

g11 ( s ) g12 ( s ) g 13 ( s )
G ( s )= g21 ( s ) g22 ( s ) g 23 ( s )
g31 ( s ) g32 ( s ) g 33 ( s )

In order to pair controlled and manipulated variables, we use a

interaction measure define as,

ij =

K ij
'
ij

K
where ij is the relative gain for the pair y i and x j
'
. K ij is the open loop steady state gain and K ij is the
close loop steady state gain. When the effect of interaction in a
multivariable control system, the relative gain for each gain is
closed to unity. [Smith]
For a n x n system, to obtain the relative gain matrix we
use the transpose of the inverse of the steady-state gain matrix
and multiply it with the original matrix. [Bristol 1966]

ij =( B ji )1 K ij
Values closest to 1, indicate the combination of controlled and
manipulated variables, which minimize the effect of
interaction.
Tuning controller are performed using the equations
recommended by Daling. []

If the inverse G p
is obtain, then the elements of the
decoupler matrix are calculated as

D ij ( s )=

B ij ( s )
B jj ( s )

The improvement by using decouplers was analyzed for a case

of study.
III.

CASE OF STUDY

The case of study was taken from [7]. The distillation column
for a binary ethanol-water system used in this study, have 19
trays, of 12 inch diameter each one and are made of copper.
The outputs are the overhead ethanol mole fraction
( y 1 (t ) ) , the side stream ethanol mole fraction ( y 2 ( t ) ) ,
and the tray number 19 temperature, which correspond to
bottom position, ( y 3 ( t ) ) .
The inputs are the reflux flow rate
product flow rate

( x3(t ))

( x2 ( t ) )

( x1 ( t ) )

, the side stream

and the reboiler stream pressure

. The feed flow rate and temperature are variable,

being the principal disturbances,

respectively.

d 1 ( t ) and d 2 ( t ) ,

Origin: Ogunnaike and Ray [7]

TABLE 2. SYSTEMS CONSTRAINTS

Upper
Constraints

Lower
constraints
Reflux
Rate

[ ]

Overhead
Mole
Fraction
Ethanol

Side Stream Flow

0.046
gpm

Side
Stream
Mole
Fraction
Ethanol

Rate

Reboiler Steam

20
psig

Temperat
ure on
tray

0.00694 gpm

0.1 gpm

Reboiler
Steam

15.6 psig

34 psi

] [

k:

[ ]

The open loop gain matrix-ordered:

[ ][

0.7

0.52

This allows get the transpose of the inverse of the gain matrix
steady state and multiplying each term of the resulting matrix
by the corresponding term in the matrix original. The terms
obtained are relative gains.[9]
That is, to obtain the relative gain matrix

92c

[ ]

Then, from open loop gains, the relative gains are determined.

k 11 k 12 k 13
0.66 0.61 0.0049
K= k 21 k 22 k 23 = 1.11 2.36 0.012
k 31 k 32 k 33 34.68 46.2 0.87

Controlled
Variables
0.18
gpm

Stream
Flow Rate

and the outputs are open-loop gains

Table 1. USUAL SYSTEM STEADY-STATE VALUES

Reflux Flow Rate

Side

ij
0.0011
( 26.32 s+ 1 ) e2.66 s
0.14 e12 s
0.0049 es
( 7.85 s+1 ) ( 14.63
s+1 )
9.06 s+1
k =0.66 6.2 s+ 1
k 12=0.61
(s)
x 111
1.2 s
3.44 s
10.5 s
0.012 e
0.0032 ( 19.62 s +1 ) e
d1 ( s )
k 13=0.0049
0.53 e
+
x2 ( s )
7.09 s+1
k 21=1.11 6.9 s +1
k 22=2.36
( 7.29 s +1 )( 8.94
s+ 1 )
d2 ( s )
0.22 s
x
3 (s)
0.6 s
2.6 s
0.87 e
k 23=0.012
11.54 e
0.32 e
9.3 s+1
k 31=34.68
k
=46.2
k 33=0.87
7.01 s+ 1 32
7.76 s +1

Table 1 shows the steady state values for controlled and

manipulated variables. Table 2 shows constrains for
manipulated variable in order to operate in safe conditions.

Manipulated
Variables

0.245 gpm

The gains present in the transfer functions relating the inputs

The column model takes the following form

0.61 e3.5 s
8.64 s+ 1
2.36 e3 s
5 s+1
46.2e9.4 s
10.9 s+1

0.068 gpm

Origin: Ogunnaike and Ray [7]

Fig 1. Ethanol distillation column process

0.66 e2.6 s
6.7 s+1
y1 ( s)
1.11 e6.5 s
=
y2 ( s)
3.25 s+1
y3 ( s )
34.68 e9.2 s
8.15 s+1

Flow

K
(1)'K .

1.9618 0.6648 0.2970

= 0.6699 1.8921 0.2222
0.2919 0.2273 1.5192

The best relationships between inputs and outputs are given by

11=1.9618

Reset
time

6,7

9,3

Reflux Flow Rate / Overhead Ethanol

Mole Fraction

22=1.8921

Side Stream Product Flow / Side Stream

Ethanol Mole Fraction

33=1.5192

Reboiler Stream Pressure / Tray Number

19 Temperature

each controlled variable was simulated. First with the

condition SISO and later with MIMO condition. the figure
shows the performance comparison between MIMO condition
and SISO condition. In each case the response of the
controlled variable takes much more time of stabilization in
MIMO condition than in SISO condition. This is the most
important aspect to improve with the technique of decoupling.

Fig 2 Overhead ethanol mole fraction control loop performance

Tuning of Controllers
The controlled variables are concentrations and temperature.
The concentration signal has important noise levels, so it is
recommended to use PI controllers. In the case of the
temperature controller [9] it is recommended that when
downtime is less than a quarter of the time constant of the
process the PI controller is used.
Tuning controllers used was as follows in Table 3. for
multivariate control.

Fig 3. Side stream ethanol mole fraction control loop performance

TABLE 3 TUNING OF CONTROLLERS
Controlle
r1

Gain

5,8/2

Controll
er 2

-1.1/2

Controller 3

145.7/4

Fig 4. Tray #19 temperature control loop performance

Fig 6. Side stream ethanol mole fraction decoupled performance

Decoupling technique
Figure 6 shows a decrease in stabilization time when
decoupling technique is used.
The Figures 5 and 7 shows that the stabilization time doesnt
decreases when the technique of decoupling is used .this is
mainly due to the fact that when the relative gain is greater
than 1 means that there is an important relationship between
variables, then applying the technique of decoupling is no
greater potential for improvement

Fig 7. Tray #19 temperature decoupled performance

IV. CONCLUSION
A conclusion section is not required. Although a conclusion
may review the main points of the paper, do not replicate the
abstract as the conclusion. A conclusion might elaborate on the
importance of the work or suggest applications and extensions.

Fig 5. overhead ethanol mole fraction decoupled performance

V. BIBLIOGRAFA
[1] T. Ravagnani y p. joao, Three phase fluidized
distillation, the canadian journal of chemical
engineering, pp. 881-890, 2001.
[2] v. Grassi y W. Luyben, Practical distillation control, New
York: Van Nostrand Reinhold, 1992.

6
[3] I. Gil y A. Uyazn, Estudio de la destilacin extractiva
con etilenglicol para deshidratar etanol azeotrpico., de
XXII Colombian Congress of Chemical Engineering,
bucaramanga, 2003.
[4] A. Uyazn, I. Gil, J. Aguilar y G. Rodrguez, Produccin
de alcohol carburante por destilacin extractiva:
Simulacin del proceso con glicerol, Ingeniera e
Investigacin., vol. 26, n 1, pp. 45-50, 2006.
[5] S. Skogestad, The dos and donts of distillation column
control, de Norwegian University of Science and
Technology, Trondheim, Norway, 1998.
[6] S. hurowitz, J. Anderson y M. Duvall, An Annalysis of
controlability statistics for distillation configuration
selection, de AIChe Annual Meeting, Dallas, Texas,

1999.
[7] B. Ogunnaike y L. Lemaire, Advance Multivariable
control of a pilot plant Distillation column, AIChE
JOURNAL, vol. 29, n 4, pp. 632-640, 1983.
[8] c. Marangoni, J. Teleken y W. Leandro, Multivariable
control with adjustment by decoupling using a distributed
action, 7th IFAC International Symposium on Advanced
Control of Chemical Processes,, vol. 7, n 1, pp. 857-862,
2009 .
[9] C. Smith y A. Corripio, Control Automtico de Procesos,
Mxico: Limusa Wiley, 2014.