Chapter 3

Macro-modelling of distillation columns

at unsteady state

Distillation in a tray column is a counter-current stage-wise mass transfer operation. In

distillation columns, the liquid flows down through the downcomer on to the tray below,
crosses the tray and falls over the outlet weir into the downcomer of the next tray. The
vapour enters through the holes in the tray from the vapour space of the tray below and
flows upward through the pool of liquid which flows cross-current to the vapour flow.
Because of the outlet weir, there is always a positive height of liquid maintained on the tray.
All the mass and heat transfer takes place in the tray, while the vapour rises up through the
liquid. The two phase mixture on the tray may be of the liquid continuous (i.e. vapour in the
form of bubbles) or of the vapour continuous (i.e. liquid in droplets and vapour in jets) type,
depending upon the flow-regimes of the liquid and vapour flow. In the macro-modelling or
lumped-modelling of a distillation column, spatial variation of any variable like composition,
temperature, fluid velocity or pressure within a tray is not taken into account. These variables
are lumped into a single variable for a particular tray. The above kind of modelling
encompasses the following types of equations
* Overall material balance and component material balance equations.
* Energy balance equations.
* Equilibrium relations between the phases.
* Hydrodynamic relations describing the liquid flow over the weir and the liquid and

vapour pressure drop.

3.1 Development of a rigorous model

A rigorous model of a distillation tray as shown schematically in figure (3.1) has been
developed and the mathematical equations representing the operation of this tray has been
described in this section. As shown in figure (3.1), the liquid from the (j-1)th tray flows into
the downcomer and subsequently to the j-th tray. The vapour from the (j+l)th tray flows
upwards through the holes in the j-th tray and comes into contact with the liquid there. The
time taken by the liquid to travel from one tray to another is Ti and that by the vapour is t2.
The j-th tray in the figure has been divided into three distinct logical zones. The liquid from
the (j-1)th tray comes down to zone 1 of the j-th tray and it is assumed that there is no
contact with any vapour in this region. Therefore, it can be presumed that the liquid
composition in this zone will be same as that in the liquid flowing down from the (j-1)th tray.
In the second zone on the tray, where the liquid and the vapour forms a dispersed phase, all
the mass and heat transfer take place. The zone above this dispersed phase which constitutes
the vapour escaping the vapour-liquid mixture on the tray is the third zone. This zone fills
the rest of the physical volume of the whole section between j-th and (j-1)th tray. Even
though the model developed here, is termed above as a rigorous model, many assumptions
have been made to represent mathematically the distillation process on a tray.

3.1.1 Major assumptions

* The model developed below is macroscopic in nature i.e. the spatial variations of
the independent variables have been ignored and they have been lumped as single
average variables.
* The second zone of the tray consists of perfectly mixed liquid of uniform
concentration and the mass fraction of the vapour in this liquid pool has been
considered to have negligible effect on the balance equations.

* Zone three of the tray consists of perfectly mixed vapour of uniform concentration.

29
 L
j-1
3 vi
1
1
1
2
j-th tray
V +1
I
I'
j+1

Figure 3.1 Schematic diagram of a model tray

* The liquid and vapour streams are in thermodynamic equilibrium and any departure
from the equilibrium condition can be represented by a single parameter called
efficiency which in turn will be a function of the liquid and vapour flow rates, tray
geometry and other factors affecting mass transfer.
The above assumptions are common in both steady-state and dynamic modelling of a
distillation column. The following additional assumptions were made for the dynamic
modelling :
* The response of the mass-transfer relations and hydrodynamic relations are at least
an order of magnitude faster than that of the mass and energy conservation equations.

Therefore they can be described by algebraic equations, whereas the mass and energy
balance equations have to be described by ordinary differential equations.

30
3.1.2 Model equations

Equations describing the dynamic behaviour of a distillation column

Component material balance

d (1) +Ai (2)x (2) +m(3 )

) 2Li (t-T
Li_1(t-T 1 4_ 1) -Lb?) -Viy ji+
dt" fi j j _ (3.1)

Vi+1(t-T2)),J+14(t-T2)+fji

Overall material balance

d 1 2 3 (3.2)
— (Mj( )+Mj( ) +Mj( ))=1,j_1(t-Ti)-Li-Vj+ Vi+1(t-T2)+F1
dt

Enthalpy balance :

d (1) (2) 3
— (E. +E.1 +E(
dt 1 .1
)) 171 L (2) h (2) I
I-1 J-1,1 j-1.1 t-T
L
- iEx
fi (2) 11 (2) -
VjEy
ji
(3.3)
F
+V+I Ey j+1,i (t-s 2)+h,J +Q
j+14H I

P) 119)
(3.4)
E (1) = M M E X
.1

(2) (2)%--% (2). (2) (3.5)

E. =M. 2, x..I n.II

(3.6)
Ej(3)=i43)E yigifii

Equilibrium relation

(3.7)
y; = kir15(,2)

31
Summation equation

(3.8)

Efficiency equation

E - Yfi-Y1+1,i
(3.9)
Yfi -Yj+1,4

Pressure drop relation

v
j+1
VV LrN (3.10)
P -Pj = ARV, rj+1,PpPj.i,Ppi-i,

Holdup relations

.(1) L (1), (3.11)

M = M.(1)(L.
r )

A42) = mi(2) (4 pjL42)) (3.12)

(3) (3), 2) V (3.13)

M . = Mj . (Mj V 0 %eft.

Thermodynamic properties

= (3.14)

hji = hji(Tj,Pj,xj) (3.15)

(3.16)
Hfi = Hfi(TrPrV

L
pi = pi (Trxii)
L (3.17)

V
P1 = p (TpPpYji)
V (3.18)

In the above equations and all other equations which will be developed later in this chapter,

the following conventions are followed regarding the use of subscripts and superscripts, unless
otherwise stated .

The superscripts (1,2 or 3) refer to the particular zone of the tray and the superscripts

32
 (L or V) refer to the liquid or the vapour phase respectively.

The first subscript or the single subscript refers to the tray number, while the second
subscript refers to the component number.

3.1.3 Pressure drop relations

The relationships for the pressure drop across a distillation tray have been developed and
discussed in great detail by a number of authors (Holland and Liapis (1983)); Lockett (1986);
Smith (1963); Van Winkle (1967)). The relevant equations, which have been used in the
present model will be presented in this section to give the exact nature of equation (3.10).
The total pressure drop of the vapour when it flows through the tray holes and the liquid pool
on the tray is the sum total of the dry hole pressure drop, ho and the pressure drop through
the aerated liquid, hLi.

P —P
j+1
gc _ h ±h
(3.19)
L g
Pj

The dry tray pressure drop is given by the orifice equation,

2 y
6, uh p (3.20)
hof —= — —

g Co p

where, ho = dry hole pressure drop of the vapour across the j-th tray

uh = linear velocity of the vapour through the holes

Ah = total hole area

Co = discharge coefficient , values of which are presented in the form of charts as a

33
 function of hole diameter, tray thickness and active area (Smith (1963)). This chart
has been put into the form of a linear equation. This is presented in appendix A.
The pressure drop through the aerated liquid, hLi , is given by

JiLl = 13(h0w +h w) (3.21)

•Cdhere h0 = equivalent clear liquid height over the weir, the expression for which is given

in the section for the calculation of the liquid holdup on a tray.

= height of the outlet weir.
13 = aeration factor which is given by (Van Winkle (1967)),

V)0.177709
p
(3 = 1.0 -0.28423127(unir—

cohere un is the vapour velocity based upon net tray area.

3.1.4 Holdup relations

The liquid holdup on zone 1 of the tray is given by,

1) L,1
= A d xZ xp m (3.22)

• here A d = downcomer cross-sectional area,

= height of the liquid in downcomer.

The liquid height in the downcomer can be calculated from the pressure balance of the liquid,

Z — (P1—P-1-1)gc +h +h +h
J
(3.23)
ticj wi
Pi-1 g

where hde = head loss resulting from the flow of the liquid under the downcomer and this is

34
 given by (Holland and Liapis (1983)),

h„=-0.057
A dc

Q = volumetric flow rate of the liquid under the downcomer.

Ade = clearance area between the downcomer and the floor of the tray.
The liquid holdup in zone 2 of the tray is given by,

(3.24)
M (2) = A„x(hw+how)xp.,42i

Ohere Aac = active area of the tray.

ic and h.„, have been defined previously.

The liquid height over the outlet weir, ho„., is calculated from the Francis weir formula which
is given by (Lockett (1986)),

ho„ = pg)--1/3(152)213 (3.25)

2 W)

where how = liquid height over the weir.

Q = volumetric flow rate of the liquid.

W = outlet weir length.
The volumetric vapour holdup in the zone 3 of the tray is the total volumetric space between
two successive trays, Ej , minus the liquid volumetric holdup on the tray. The molar vapour
holdup is, therefore, computed from,

Aff2)
MP) = [ xpv (3.26)
P. i

The above correlations have been used to express the pressure drop and the holdup of the
liquid on a sieve tray in terms of the state variables like liquid and vapour flow rate,

35
compositions, temperature and pressure. However, the formulation of the model is in no way
dependent upon the nature of these correlations. Any other correlations (for other kinds of

trays for example) can easily be used with the model.

3.1.5 Efficiency

In normal distillation calculation, each tray is assumed to be an equilibrium stage, where the
liquid phase and the vapour phase is considered to be in thermodynamic equilibrium with
each other. However, this never happens in reality. The non-equilibrium behaviour of a tray
is represented by a factor called the efficiency of the tray. A tray is 100% efficient, when
it is an equilibrium tray. The lower the efficiency, the further the tray is away from its
equilibrium condition. There are different kinds of definitions for the efficiency that have
been suggested in the literature (Hausen (1953); Holland (1963); Murphree (1925); Standart
(1965)). Murphree plate efficiency, even though it is not the most theoretically sound, has
been most popular with the distillation models for its ease of implementation. The calculation
of the efficiency from its fundamental principles is a very complex process and is dependent
upon the physical properties of the liquid-vapour mixture, tray geometry, flow-regimes and
the mixing pattern of the liquid and vapour. However, different researchers have tried to
establish empirical correlations between the tray efficiency and the vapour velocities and the
liquid holdup on a tray. In the dynamic behaviour of a column, the vapour flow rate through
the column changes, if there is a change in terms of the feed flow rate or the composition.
The equations correlating the efficiency with the vapour velocities through the holes have
been incorporated into the distillation model to account for the above-said change in the
efficiency. The data obtained from the literature have been fitted into equation form to be

used by the model. Details of this have been described in chapter 4.

36
3.2 Simplified model

The rigorous model equations (3.1 to 3.18) have been simplified and rearranged based on the
following arguments from physical and computational point of view.

The vapour holdup terms used in the balance equations do not contribute significantly unless
the column is operated at high pressure. Even though the model developed here is of general
nature, the vapour holdup term was neglected because most of the distillation systems operate
in pressure regimes where this is insignificant. Inclusion of this term for the high pressure
columns does not make any fundamental difference to the solution method except an increase
in computer memory and time requirement.
It is assumed that the liquid composition and the temperature on a tray are uniform. Thus
their values in zone one of the tray are the same as those in zone two.
It is further assumed that there is no delay in the liquid and vapour flow from any tray to the
next tray. Based upon this assumption, T i and T2 in the material and energy balance equations
are neglected.
The formulation of the model equations was also guided by the simplicity and ease of use in
numerical computation. The model equations can be written in different forms with different
variables. For example, instead of treating the liquid height in the downcomer as one of the
dependent variables (Gallun and Holland (1982)), this and other similar kinds of variables
have been substituted in the set of equations to be solved simultaneously to minimise the
memory requirement.

3.2.1 Simplified model equations

The following are the simplified versions of the rigorous equations presented earlier. In
addition to the equations for the general trays, the equations for the condenser (j=1) and the
reboiler (j=N) are also described. The pressure drop equation in the condenser is taken from
Holland and Liapis (1983). The reflux accumulator is assumed to be a horizontal cylinder.

37
 -

The liquid holdup at the bottom of the column is provided in the form of the liquid collected
in the base of the column.

Component material balance

d (3.27)
— (MAE) = j=2,N i=1,M
dt

d (3.28)
— = j=1 i=1,M
dt

= j=N i=1,111 (3.29)

dt n r

Overall material balance :

—(M) = L -V +V + -F j=2,N-1 (3.30)

dt J-1

d (3.31)
— (M) =-L -V j=1
dt 1+1

d (3.32)
— (M) = L -L -V j=N
dt I

Enthalpy balance :

d (Ei)
y xiihii-ViE y
(3.33)

+1,i+1EYJ+1411pu+hiF +91 j=2,N-1

d (Ei) = Vi+ ,E
I x j,- ViE j=1 (3.34)

(Ei) = Li_,E j=N (3.35)

38


Ei = MiE xfilzji j = 1,N (3.36)

Equilibrium relation and summation equation :

y = Eigkfixii +yj + 1,1 (1 -Eji) j=2,N-1 i=1,M (3.37)

E = 1 j=2,N-1 (3.38)

For total condenser

yii = xii j= 1 i= 1,M (3.39)

L = RxV j=1 (3.40)

For partial condenser

11 = A1 x11 j = 1 i=1,M (3.41)

E = 1 j=1
(3.42)

Pressure drop relations

Pj+i -P1 = piL(hod +1/

41) 1- j =2,N-1 (3.43)
gc

P 1--P
Ppi - pi = v2 p V)2[ j + I j =1 (3.44)
( V20 2 ref

Holdup relations

= [fl ac (hw +ho,)+AdZi]xp Lmi j=2,N-1 (3.45)

2Z.1 (3.46)
M=L 11x( Da" )2 +(Z, — —Dacc)x(ZixDacc-Zi2)112xp;
acc 2 \ Dacc ) 2 2

(3.47)
J l"mi
M = A txZ.xp j=N

39
In all the above equations wherever a summation term appears the summation is done over
i=1 to M.

3.3 Control equations

There are many variations available for the control schemes of a distillation column (Luyben
(1973)). One of the popular schemes has been used in the present model. This section

describes each of the control loops and its mathematical formulation. The superscript 0 in
this section refers to the steady state values of the corresponding manipulated variables and
the superscript 'set' refers to the preset values of the corresponding controlled variables.

3.3.1 Pressure control

When the condenser in the column is a partial condenser or a total condenser with the liquid
at its bubble point, one of the popular control schemes for the control of the pressure is to
manipulate the coolant flow to the condenser and hence the condenser duty. The
Proportional-Integral (PI) control of such a scheme can be represented by,

= (2:) +1Cci[(PseL/31) + 1.f(P'-Pi)dt1 (3.48)

In the case of a total condenser with subcooled liquid the pressure of the condenser is
assumed to be constant at a certain preset value. The governing equation for pressure control

thus becomes,

131-Pset =0 (3.49)

40
3.3.2 Control of condenser temperature

In the case of a total condenser with subcooled liquid the temperature of the liquid stream is
controlled by changing the coolant flow rate and thereby the condenser heat duty. The PI-

control scheme for this can be expressed by the equation,

Q, = Q c43 +1C,21(T-T+ f s
( Tcl-Tt1 (3 .5 0)
Ti
2

3.3.3 Control of liquid level in condenser and reboiler

The liquid level in the condenser is controlled by manipulating the valve in the distillate
stream and that in the reboiler is done by manipulating the flow rate of the liquid outlet
stream from it. The equations governing the control schemes for the liquid levels in the

condenser and the reboiler are given by,

V1 = 11 + I C c [(Zisoet -Z 1) + -I-- f (Zr -Z ddt1 (3.51)

3 T 1,

and

LN = LN° +1C1gNset -Zd+ -1— f (ZNset -Zddt1 (3.52)

T
14

respectively.

3.3.4 Control of distillate composition

The composition of the distillate is controlled by changing the reflux stream. The

manipulation of the ref lux stream is done according to the logic of the controller used.

41
In the case of a PI controller this is represented by the equation,

L1 = Li°+1C,[(xd
-
5
set
-xd)+ 1 f(xd
set
-xd)dt (3.53)
T r,

3.3.5 Control of bottom stream composition

The composition of the bottom stream is controlled by manipulating the steam flow rate to
the reboiler and hence the reboiler heat duty. The PI controller equation for this scheme is
given by,

Qr = Qr0 +K,ef(xb -xd+—

set 1f set
(xb -xb)dt (3.54)
T r6

3.4 Arrangement of model equations

The model equations described above form a set of differential and algebraic equations.
These equations are to be solved for the distillation variables at different time steps. The
model equations are arranged stage-wise in the same way as Naphthali and Sandholm (1971)
did for their steady state model. Benefits from this kind of arrangement can be obtained by
exploiting the block triangular structure of the Jacobian matrix.

The set of differential-algebraic equations can be written in the vector form,

F(X,t) = 0 (3.55)

, here
F = [FI,F2,

X = [X,,X2, , X]T

F; = [Cp,••,qm,OpEitDpgil,••,Q.NA,PplijiT

42
 Co is the component material balance equation

Oj is the overall material balance equation

• is the enthalpy balance equation

Di is the equation for the definition of the term E in the energy balance
equation.
is the equation defining the equilibrium relation.

Sj is the summation equation

Pi is the pressure drop equation

Hj is the holdup relation

X1. = vim,
M. Lppppp,
V. T. E

This system of differential-algebraic equations are solved simultaneously using the BDF
method proposed by Gear (1971b). This is discussed in detail in chapter 4.

43