4th International Conference on Electrical Engineering (ICEE 2015)

IGEE, Boumerdes, December 13th -15th, 2015

Investigation Of The Effect Of Design Aspects On

Dyanamic ControlOf aBinary Distillation Column
R. BENDIB, H.BENTARZI

Y.ZENNIR

Laboratory of signals and systems

Institute of electrical and electronic engineering
University of M'hamedBougara, Boumerds Algeria
E-mail: bendib_r@yahoo.com, bentarzi_hamid@yahoo.com

Laboratory of automatic
Department of Petrochemical and Process engineering
Skikda university- Algeria
E-mail:youcefzennir@yahoo.com

Abstractthe purpose of this paper is to investigate the

effect of design modification on the dynamic behavior of a
binary distillation column (2x2 MIMO system). The aim is
the use of design modification as a tool to minimize the
interactions between the loops. Two different cases are
studied; the first is the location of feed plate, whereas the
second is the quantity of liquid in the feed plate for DV
control scheme. The dynamic magnitude array criterion is
used to evaluate the degree of interaction between the
loops. The results show that design has a very important
effect on the degree of interactions the fact that can be
used to improve the control aspects for the column. .
KeywordsDistillation column; RGA; interactions; dynamic
control; MIMO

I. INTRODUCTION
Distillation is the most important industrial separation
technology. Considerable attention[10],[9]is devoted to
controlling distillation columns more effectively because:
o They are slow to respond to control action
o As much as 40 per cent of the energy consumed in a
plant is allocated to distillation.
Determining the number of stages required for the desired
degree of separation and the location of the feed tray is merely
the first steps in producing an overall distillation column design
and hence mathematical model. In practice there are several
factors that may affect the design specifications, and it can lead
to some deviations that let the column no longer able to handle
the separation task, that is the product purities which are the
desired outputs for the mathematical model used to define de
distillation column. From control point of view the deviations
in the outputs and be handled by the design of suitable
controllers; however the existing of interactions between the
loops will clearly show this deviation. Several methods are
used to assess the degree of interactions between the loops of
multivariable systems, which are considered in industrial fields
as multi-loop systems that is a set of SISO systems, the fact
that let getting information concerning the magnitude and the
direction of interactions is one of important issues that the
engineer should know before any attempt to choose the
required controller parameters, which lead to a minimum
energy consumption and desired product specification.

II. THE MATHEMATICAL MODELING OF

DISTILLATION COLUM
The dynamic model [1] was developed such that the overall
and the component balances was solved for each tray.
Additional algebraic equations are needed, for the steady state
energy balance, vapour liquid relationships and Francis weir
formula for liquid flow in the reboiler. The derivation of
analytical expressions requires the assumptions of:
Equilibrium stages.
Constant relative volatility.
Constant molarflows.
A. Basic process equations
o Total material balance on stage i
dM dt = L

L +V

V (1)

Material balance for light component on each

)
(
Stage
= L x +V y
L x V y (2)

Algebraic equations :The vapor composition yi is

related to liquid composition xi on the same stage
through the algebraic vapor-liquid equilibrium:

y = ( x ) (1 + (

1)x )

(3)

Where is the relative volatility.

Feed stage i=NF, assuming that the feed is mixed

directly into the liquid at the feed plate.

dM dt = L
(

=L

=V

Lx

(4)
V y + Fx (5)

Lx

(6)
Vy

Dx

(7)

Reboiler : i=1 (Mi=MB, Vi=VB=V)

dM dt = L

2015 IEEE

+V

V +F

Total condenser:i=NT (MNT=MD,LNT=LT)

dM dt = V
(

L +V

(8)

=L

Vy

Bx

between the loops of multivariable system. one of the most

known methods is the RGA ( relative gain array) proposed by
Bristol in 1966 [7]. This method gives a very clear idea about
the level of interactions and gives a good indication about the
best pairing for the chosen loop in multi-loop control
approach; however the main drawback of the RGA is that it
deals with ideal controllers since it considers only steady state
behavior[4]. In the following section we will introduce one
powerful tool used in assessing the interactions which is the
DRMC. This method will be used in the next sections of our
work.

(9)

MD
P
L D,xD
F, zf

A. The Dynamic Relative Magnitude Criterion

The DRMC is a set of plots of magnitude/log frequency for its
elements [4]. The DRMC elements have been arranged into an
array as diagonal and off-diagonal elements and interpreted as
graphical representations, the diagonal elements are like the
RGA [1] relate open loop and closed loop behavior of the fully
controlled system.

MB
B

V
Fig.1 Distillation Column
B. The liquid hold-up

Is the quantity of liquid in the tray column, the liquid holdup

on the tray M n is a function of liquid flow rate leaving the
tray (Ln) [5]. i.e.

Mn

f Ln (10)

Where, f (.) is a function, obtained by including the detailed

tray hydraulic equations, for our case (Benzene, Toluene)
distillation column (Mn)is given by:

S h ow

h d 0 . 703 x n

0 . 589 (1

xn )

(11)

Where h0 w = height of liquid over weir, S: the surface and

(0.703&0.589) are liquid densities of benzene and toluene
respectively.
C. The optimal feed plate location
The feed-stage location is that location which, with a given set
of other operating specifications, will result in the widest
separation between xD (the top product) and xB (the bottom
product) for a given number of stages. Or, if the number of
stages is not specified, the optimum feed location is the one
that requires the lowest number of stages to accomplish a
specified separation between xD and xB. The optimum feed
location can be estimated in the design stage using graphical
method of Mc cab Thiel diagram [where the optimal location is
defined as the intersection of the two operating lines [8].
III. ANALYSIS OF INTERACTIONS IN MULTIVARIABLE
SYSTEMS
The problem associated with control loop interactions can be
stated as follows (following Shinskey [9]):If the steady state or
dynamic gain of a given controlled variable in response to a
given manipulated variable changes when other (initially
open) loops are closed, then interaction exists in the system.
Several methods are used to quantify the degree of interactions

A.1 The construction of DRMC elements

The diagonal elements
The diagonal elements are computed by considering the ratio
of the transfer function where no controller is envisaged to the
system (all loops are open) to the transfer function
where all loops are closed (in automatic) except the loop i

( )=

( )
( )

(12)

( )
( )

The off-diagonal elements

The off-diagonal elements are computed by calculating the
ratio of the transfer function between output (i) and a set point
for input (j) such that
to the transfer function between
the output (i) and its own set point. i.e.

( )=

( )
( )

( )
( )

(13)

A.2 The interpretation of DRMC elements

The DRMC clearly expresses how the individual control loops
respond to their own set-points through the diagonal elements,
and to other set points through the off-diagonal elements.
From the definition of the criterion, the system interaction
caused by the closed control loops, will be very weak for those
pairs of variables with a relative magnitude of unity at loops
resonant frequencies, as the magnitude of the diagonal
elements of the DRMC between controlled variables yi and
manipulated variables ui departs from unity, more interaction
must be expected. The diagonal elements of the DRMC carry

information about how a single loop will respond to changes

in its own set point. However they don't supply any useful
indication about the direction and magnitude of
existingperturbations with other loops. A single loop that does
not see any disturbances in its local performance may see
significant effect from a set point change in another loop.
The off-diagonal elements of the DRMC have been introduced
to give useful information about the direction and magnitude
of the presence of such disturbances propagating between
controlled loops.The off-diagonal elements ijexpress how
much the Jth loop is excited relative to theresponse of the ith
loopwhen a set point is made in the ith loop. The ij for the
range of frequencies where a system works (i.e. the loop
resonant frequencies) should be much smaller than unity for
the rejection of true interaction or disturbance between loops.
For a totally non interacting system dis equal or very Close to
unity, and for only one direction affect in a system od will be
either upper or lower triangular.

The DRMC elements evaluated in the resonant frequencies

indicate that there is a small degree of interaction between the
loops since they are close to unity.
By the same way we will apply the DRMC to evaluate the
degree of interactions but with different feed plate locations the
results are summarized in Fig 4&5. Where the distances
between the DRMC diagonal elements and unity, are
presented for different feed plate locations.
The resonant frequencies where the DRMC elements are
evaluated are included in TABLE 1.
TABLE 1. The resonant frequencies

The feed plate location

r1

r2

Plate N4

0.221

0.06

Plate N5

0.07

0.03

Plate N7

0.3

0.04

Plate N3

0.01

0.01

Plate N1

0.01

0.01

IV. DISCUSSION AND RESULTS

A. The effect of feed plate location on intercations
By the use of graphical method to locate the feed plate - using
the intersection of the two operating lines in Mc-cabe Thiel
diagram- gives that the optimal location is plate No 4.
Application of DRMC to assess the interactions in this case
will lead to the following figures (Fig .2 for DRMC diagonal
elements and Fig.3 for DRMC off-diagonal elements).
The resonant frequencies where the values of DRMC elements
are calculated are:
= 0.221&
= 0.06
/

4
3.5

The firs t diagonal element

10

Magnitude

5
4.5

10

10

1| 2.5

1.5

-1

10

-3

10

10

-2

-1

10
Frequency (rad/min)

10

10

0.5

Magnitude

10

The second diagonal element

10

10

Fig.4 the distance between unity and 1st Diagonal element vs

feed plate location

-1

10

Feed plate location

-3

10

10

-2

-1

10
Frequency (rad/min)

10

10

Fig.2 DRMC Diagonal elements feed plate is stage No=4

35

30

The firs t off-diagonal element

Magnitude

10

10

-5

10

10

-3

10

-2

-1

10
Frequency (rad/min)

10

10

1|

25

20

15

10

10

Magnitude

The second off-diagonal element

10

-2

10

-4

10

10

-3

10

-2

-1

10
Frequency (rad/min)

10

10

Fig.3 DRMC Off-Diagonal elements feed plate is stage No=4

Feed plate location

Fig.5 The distance between unity and 2nd diagonal element
vs.feed plate location

The first off-diagonal element

As it is indicated by the figures (4&5) we deduce that for

the case where the feed plate is plate No=4, we notice that
there is a small degree of interaction, however when we go far
from this location more interaction is expected, as it is noticed
in cases No=1, and No=7.

Magnitude

10

10

-5

10

10

B. The effect of liquid holdup on the feed plate on intercations

for DV control scheme.

In DV configuration (the manipulated variables are distillate

flow D and boilup V [6], Assumingperfectleveland pressure
control
We have,
=

(14)

Hence,
=

1 1
0 1

(15)

And after that we will examine the level of interactions

between the loops for different quantities ofhold up in the feed
tray.
First we will consider no modification on holdup is done,
applying the dynamic relative magnitude criterion and
evaluating the elements in the resonant frequencies
= 0.2

= 0.06

NB. n is the factor by which we augment the liquid holdup

The DRMC diagonal elements and off diagonal elements are
shown in Fig.6 and Fig.7 respectively.

M agnitude

10

10

10

10

M agnitude

10

10

10

-1

10
Frequency (rad/min)

10

10

10

-3

10

-2

-1

10
Frequency (rad/min)

10

The second diagonal element

-1
-3

10

10

-2

-1

10
Frequency (rad/min)

10

10

Fig.6 DRMC Diagonal elements for the case where n=1.

10

Fig.7 DRMC off-Diagonal elements for n=1.

The distillation column under this control scheme is noninteractive compared with the conventional control scheme
(LV) because the diagonal elements are close to unity,
=
1.3 and
= 0.6.

The off-diagonal elements indicate that there is a strong

perturbations that propagate from the top loop to the bottom
loop and this can be depicted from that the fact that the first off
diagonal element is greater than the second,
= 9 and
= 0.009

By the same we will examine the interactions in the following

cases: augmenting the holdup in the feed tray two times, three
times, four times and five times.
Augmenting the holdup in the feed tray in practice can be
realized through a middle vessel, and using a pump we can
keep the level in the feed tray on the required quantity.
More over it should be mentioned that we can not augment the
holdup to any quantity that we want since always the vessel's
volume and the location of the vessel in the plant are
limitations that should be overcome before any realization.
The resonant frequencies where the DRMC elements are
evaluated are shown in TABLE 2.

r1

r2

0.2

0.06

0.73

0.06

0.3

0.0625

0.321

0.071

0.321

0.071

-5

10

-1
-2

10

10

10

The second off-diagonal element

-3

10

-1

10
Frequency (rad/min)

TABLE 2. The resonant frequencies

10
10

-2

The first diagonal element

10

10

Magnitude

The mathematical model of (DV) control scheme [10] can be

deduced from the over all model of the column taking the LV
configuration as a base control scheme and applying the
consistency relations between the control configurations.

-3

The values of DRMC elements for different cases are

depicted in TABLE.3
TABLE 3. The resonant frequencies

The diagonal
elements

n
1
2
3
4
5

The off diagonal

elements

11

22

12

1.3
0.9
1
1
1

0.6
0.7
0.6
0.5
0.5

9
9
10
10
10

21

0.009
0.009
0.01
0.009
0.009

It results from the Table that whenever we augment the holdup

in the feed plate two and three times the degree of interactions
between the loops is reduced until we get a complete
decoupling for the first loop that is a first diagonal element
equal to one in the second case (augmenting the holdup three
times). For the remaining cases (i.e. when we augment the
holdup four and five times) the degree of interactions is small
particularly for the first loop which tends to be independent
from the control action of the second loop controller, but the
second loop will be more interactive than the previous case
(augmenting the holdup three times). The off diagonal
elements show that each time we augment the holdup in the
feed plate the deviation between the values is slightly
augmented, which indicates that the magnitude of the
propagated disturbance is also augmented.
V.

CONCLUSION

In this paper we have investigated the interaction between

the design aspects and dynamic control of a binary distillation
column, by considering two important cases the location of
feed plate and the quantity of liquid holdup on the feed tray.
The results show that with augmenting the liquid holdup in the
feed plate three times we can realize an implicit decoupling
that is the second loop will be totally decoupled from the first
loop which makes the control of both loops easier. In the other
hand with an appropriate choice of feed plate location a
minimum degree of interactions is resulted. Hence a study of
dynamic behavior of the column is recommended before any
decision in the design of the column.
Appendix A. NOTATION
Li and Vi respectively liquid and vapor flow from stage
[Kmole/min]
xi and yi - respectively liquid and vapor composition of
light compnent on stage i[mole fraction].
M i liquid holdup on stage i [ Kmole].
D and B distillate (top) and bottoms product flow rate
[Kmole/min].
L=LT and V=VB
reflux flow and boilup flow
[Kmole/min].

F, zF feed rate [Kmole/min] and feed composition

[Kmole].
qF fraction of liquid in feed.
i. stage n. i=1=bottomn tray. NF= feed
Relative volatility between light and heavy
component.
time constant [min] for liquid flow dynamics on each
stage.
L

REFERENCES
[1] Halvorsen,I.Skogestad,S.(2000) ''DistillationTheory"
Encyclopedia of Separation Science, Academic Press
1117-1134.
[2] Huang,H. Ohshima,M and Hashimoto,L (1994) " Dynamic
interaction and multiloop control system design"
J.Proc.Cont.1994 volume 4,number 1 pages 15-27.
[3] Khelassi A, Bendib.R,A.Benhalla (2012) " control
configurations for optimal control for a binary distillation
column"UKACC International conference on control
Cardiff-UK 793-797. IEEE Xplorer.
[4] King,C,J. (2013) "Separation Process"2nd edition dover
publication .
[5]Luyben,W.(1990)" Process Modeling, Simulation and
Control for Chemical Engineers " second edition McGraw
hill New York.
[6] Luyben,W.(1992) "Practical Distillation Control "Van
Nostrand Reinhold.
[7] Luo,X., Liu.Y,Xu,F. (2014)" Interaction analysis and
decomposition principle for control structure design of
large scale systems"Chiness journal of chemical
engineering 22(1) 59-71.Elsevier
[8] Perry. H.(2007) "Perry's Chemical Engineer' Handbook"
eighth editionMc Graw hill.
[9] Shinskey. F.(1996) "Process-Control Systems" fourth
edition Mc Graw hill.
[10] Skogestad, S. Postlethwaite, I.(2007) "Multivariable Feed
back Control Analysis and Design" John Wiley and sons