1284 Chem. Eng. Technol. 2007, 30, No.

9, 1284–1291

Noori Sotudeh1 Research Article

Bahram Hashemi Shahraki1

1
Petroleum University of
Technology, Ahwaz, Iran.
A Method for the Design of Divided Wall
Columns
A divided wall column has been modeled as a Petlyuk column with no heat trans-
fer across the column wall. The feed to the column has been generalized as a mix-
ture of saturated liquid and vapor (0 <q <1). The pinch compositions at connec-
tion points, the Fenske equation and the Gilliland correlations, which have been
commonly used by previous investigators to design divided wall columns, have
not been applied. In contrast, a shortcut method based on Underwood’s equa-
tions has been introduced. Moreover, it is shown that the split of the internal re-
flux over both sides of the middle wall of the column is bounded and a method
for choosing the proper value of the split ratio is suggested. This is a novel ap-
proach not attempted before by investigators.

Keywords: Columns - Distillation, Columns - Divided wall, Energy savings, Modeling

Received: May 05, 2007; accepted: May 27, 2007
DOI: 10.1002/ceat.200700175

1 Introduction
Distillation is the most reliable process for the separation of
the components of a multi-component liquid. However, when
separation and high purities are desired, a single distillation
column is not sufficient. For a liquid stream with n compo-
nents, a cascade consisting of n–1 columns with their related
accessories such as condensers and reboilers, are required. This
arrangement performs the necessary separation but with the
consumption of too much materials and energy. By process in-
tegration, the components of a multi-component mixture
could be distilled and separated at high purities with the con-
sumption of smaller amounts of utilities and energy. Process
integration also helps to reduce the cost of the process equip-
ment.
One way of integrating distillation columns is to employ
either thermally coupled distillation columns, which are
known Petlyuk columns [1] or applying divided wall columns,
DWCs, designed by Kaibel [2], Figs. 1a) and 1b). The special
advantage of a dividing wall column is the achievement of a
large energy saving together with the capital savings from the
use of a single shell for the distillation. Capital savings also re-
sult from the use of a single reboiler and condenser compared
with a conventional arrangement. Theoretical studies have
shown that it can save, on average, ca. 30 % of energy costs
compared to conventional schemes [3]. From the point of
mathematical modeling, a DWC is practically identical to a Pe-
tlyuk column, if the heat transfer across the column wall is ne-
glected or the wall is insulated, as shown by Lestak et al. [4].
Apart from the work of Cerda and Westerberg [5], Trianta-
fyllou and Smith [6] and Muralikrishna et al. [7], all of the
work published on such models has used done minimum va-
por flow calculations without consideration of the number of
plates and internal reflux values. Hence, none of these models
can provide either a basis for preliminary optimization or an
initial solution for a rigorous simulation.
Glinos and Malone [8] presented the correct algorithm for
analyzing a Petlyuk column. Their analysis was developed for a
ternary mixture. The authors base their results on estimating
the location of the pinch point in the connection points. One
step in their development assumes that the composition of the
A
A
F
F
B B

C
C
(a) (b)
–
Correspondence: N. Sotudeh (noorisotudeh@gmail.com), Ahwaz Pet- Figure 1. The schematic diagram of the columns: (a) DWC, and
roleum University of Technology (PUT), 61991-27917 Khouzestan, Iran. (b) Petlyuk column.

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com

Chem. Eng. Technol. 2007, 30, No. 9, 1284–1291 Modeling 1285

liquid return stream from the main column is

D2 = A+ B
at the pinch composition. However, this as-
sumption is not always true. {x A, D , xB , D , xC , D = 0}
2 2 2

Fidkowski and Krolikowski [9] presented a

method to find the minimum energy require- V2 L2
ment for a Petlyuk column. They restricted N3
their analysis to only a saturated liquid feed
and the development is very complex algebrai- D1 Section 2

cally. Because the analysis is limited to saturat-

ed liquids and pinch point conditions, their re-
sults are not generalized to all feed qualities N4
V1 L1
and more than three-component mixtures.
Such an extension would not be easy due to V2 L2
W2 S = A+B+C
the complexity of the algebra involved in their
derivation. The authors discovered that the F Section 1 {x A, S , xB , S , xC , S }
minimum reflux occurs over a range of middle z A , z B , zC
component split fractions. However, they did D3
not come to a decision on which split was best. V3 L3
Smith et al. and Muralikrishna et al. used V 1 L1 N5
Underwood’s equations [10] to find the mini-
mum reflux ratio and the Fenske equation to
find the minimum number of trays at the total Section 3
reflux conditions. Then, by using these values W1
in the Gilliland correlation, they estimated the N6
required number of theoretical trays for a given V3 L3
operating reflux ratio. The current authors be- W3 = C + B
lieve that the application of the Fenske equa-
tion for the estimation of the minimum num- { x AW = 0, x BW , x CW }
3 3 3

ber of trays of a divided wall column is not

correct since the situation in this column is dif- Figure 2. More detailed view of a Petlyuk column (Used as a model for a DWC).
ferent from that of a conventional one. The
difference arises from the fact that the compo-
sition of the liquid stream returning from the main column is
not equal to the composition of the vapor entering the main
column at the connection point, while Fenske equation has
been derived based on the equalities of these compositions.
The present work attempts to introduce a method for the
design of divided wall columns based on the Underwood equa-
tions. In this method, the minimum vapor flow and minimum
reflux ratio in the column are determined. Following this, by
choosing an operating reflux ratio in the range of 1.2–1.5 Rmin;
the total number of trays in the tower and the side stream loca-
tion are calculated1). The methods used for locating the feed
tray and splitting of the internal liquid reflux over both sides
of the middle wall of the column, are also presented. An illus-
trative example is also given to compare this design method
with the work of previous investigators.
2 Material Balance Equations for a DWC
By referring to Fig. 2, the material balance equations for the
current model are:
A balance: FzA ˆ D2 xAD2 ‡ SxAS ‡ W3 xAW3 (1)
– sections. The choice of the highest value of the minimum va-
1) List of symbols at the end of the paper.
B balance: FzB ˆ D2 xBD2 ‡ SxBS ‡ W3 xBW3 (2)
C balance: FzC ˆ D2 xCD2 ‡ SxCS ‡ W3 xCW3 (3)
Also
X
xi;D2 ˆ 1 …4†
X
xi;S ˆ 1 …5† , with i = A, B, C
X
xi;W3 ˆ 1 …6†
where [xij, i = A, B, C, and j = D2, S, W3] are the mole percen-
tages of the components in the product streams. From the
above equations, it is seen that there are twelve unknowns,
while the known parameters are the feed rate and its composi-
tions. Therefore, in order to solve the above equations, six of
the unknowns must be specified. The specified parameters are
x
suggested to be xAW3 ; xCW3 ; xAD2 ; xCD2 ; xBS and AS .
xCS
3 Minimum Vapor Flow Calculations
As seen from Fig. 2, the column in the current model has been
divided into three different sections, 1, 2 and 3. The minimum
vapor flow and pinch condition could occur in any of these
por flow would ensure that the required separation of the

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com

1286 N. Sotudeh et al. Chem. Eng. Technol. 2007, 30, No. 9, 1284–1291

components across each section of the column could be Vmin 1 is calculated based on the accepted roots and is cho-
achieved simultaneously. Any smaller value of the minimum sen to be:
vapor flow might lead to inconsistencies, as described by Yaws
[11]. Vmin 1 = Max{Vmin 1 (h1), Vmin 1 (h2)} (15)

Both Vmin 1 (h1) and Vmin 1 (h2) change with b. Fig. 3 shows
3.1 Minimum Vapor Flow in Column Section 1 their variations with b and help one to find the maximum
point at a given b value. The value of b at which Vmin 1 (h1) =
Consider that a saturated feed of compositions zA, zB, and zC, Vmin 1 (h2) is called the preferred split and from work by
and quality q, (0 < q < 1) is fed to section 1. The relative vola- Stichlmair, is denoted by bp [12]. bp is given by:
tilities of A, B and C with respect to C are assumed to be con-    
stant and equal to aA, aB and aC, respectively. Therefore, the aA :FzA aA :FzA
a h
fractional recovery of component i in the top product can be b P ˆ aA :Fz 1   aA h2
 (16†
B B aB :FzB
defined as:
aB h1 aB h2
xi;D1 D1
ri;T ˆ …7† , i = A, B, C From Fig. 3, it is easily found that for b > bp, Vmin 1 (h2) is
zi F
maximized but at b < bp, Vmin 1 (h1) is minimized. The other
where D1 is the net stream flowing to the top of section 1. This flow streams inside section 1 are calculated by writing the ma-
permits one to present the top component compositions in terial balance equations. These streams are as follows:
terms of their recoveries, as specified below:
L min1 ˆ V min1 D1 (17†
rA;T ˆ 1 ; rB;T ˆ b ; rCT ˆ 0
Fz Fz
xA;D1 ˆ A ; xB;D1 ˆ b B ; xC;D1 ˆ 0 …8†
D1 D1 V min1 ˆ V min1 …1 q†F (18†
D1 ˆ zA F ‡ b zB F
 
Similarly for the net stream flowing to the bottom of section aB ZB F ac Zc F
V min1 ˆ …1 b† ‡ (19†
1, one can write: aB h ac h

xi;W1 W1
ri;B ˆ …9† L min1 ˆ L min1 ‡ qF (20†
zi F

rA;B ˆ 0 ; rB;B ˆ 1 b ; rCB ˆ 1 Vmin 1

Fz Fz Vmin,1 Based on
xA;w1 ˆ 0; xB;w1 ˆ …1 b† B ; xC;w1 ˆ C (10†
W1 W1 θ1 Vmin,1 Based on
W1 ˆ zB F…1 b† ‡ FzC θ2

By using Underwood’s equations at the minimum reflux

conditions:
X ai zi
…1 q† ˆ …11† i ˆ fA; B; Cg
ai h

X ai xiD D1
V min;1 ˆ 1
…12† i ˆ fA; B; Cg βp β
ai h
Figure 3. Graph of Vmin 1 vs. B.
and combining Eqs. (8) and (12), one obtains:

aA zA F a z F
V min;1 ˆ ‡ B B b (13†
aA h aB h 3.2 Minimum Vapor Flow in Column Section 2

Solving Eq. (11) for its roots, results in three values being
found for h. Inspection of these values shows that one of them
is negative and should be rejected. The remaining two roots
are in the following ranges:
aA > h1 > aB > h1 > aC (14)
The net stream from the top of section 1 is fed to section 2,
i.e., F2 = D1. The top stream product of section 2 is D2. The
compositions of this stream, xi;D2 , have been found earlier
through the process of solving the material balance equations,
Eqs. (1–6). Therefore, at the minimum reflux condition, one
can write:

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com

Chem. Eng. Technol. 2007, 30, No. 9, 1284–1291 Modeling 1287

where F3 ˆ W1 ˆ …1 b†zB F ‡ zC F
Vmin
Vmin
V min, 2
X ai xiW W3
V min;3 ˆ 3

V min,3 + (1 − q) F ai h″
V min, 2
…26† f i ˆ B; C g

The use of the results of Eq. (10) and knowl-

V min,3 + (1 − q) F
edge that:
L min;3 L min;3 L min;1
q″ ˆ ˆ (27),
W1 W1

βp β βp β permits one to write Eq. (25) as:

βb2
βb
aB …1 b†zB F a z F
b) a) V min;1 ˆ ‡ C C …28†
aB h″ aC h″
Figure 4. Variation of (a) Vmin 2, and (b) V min;3 ‡ …1 q†F with b.
Solving equation (28) for h″, gives two roots,
X ai xi;D F2 with the acceptable one occurring when aB > h″ > aC. From
F2 …1 q′† ˆ 1
…21†
ai h′ comparison of Eq. (28) with Eq. (19), it is seen that when b is
where i ˆ fA; Bg; and F2 ˆ D1 ˆ zA F ‡ bzB F greater than bp and Vmin 1 = Vmin 1 (h2), the acceptable root of
Eq. (28) is h″ = h2. At this condition, applying h″ = h2 in Eq.
P ai xi;D2 D2 (26) would give a constant value for V min3 in the whole range
and V min2 ˆ …22† ; i ˆ fA; B g
ai h′ from b = bp–1. But for b < bp, since Vmin 1 = Vmin 1 (h1), the
The use of the results of Eq. (8) and knowledge that: acceptable root of Eq. (28) is not h1 because it is not in the de-
fined range. At this condition, Eq. (28) will give a new value
L min2 L min2 L min1 for h″ which can change with b. This leads to a graph of V min3
q′ ˆ ˆ …23† vs. b as shown in Figs. 4a) and 4b).
D1 D1

permits one to rewrite Eq. (21) as:

4 Comparing the Minimum Vapor Flow
V min;1
a z F a z F
ˆ A A ‡ B B b …24† Streams in Column Sections 1, 2 and 3
aA h′ aB h′ and Choosing the Minimum Vapor Flow
Two values are obtained for h′, by solving Eq. (24).The ac-
of the Tower
ceptable value occurs when h′ lies in the range of aA > h′ > aB.
From inspection of the equations, the minimum vapor flow
From comparison of Eqs. (13) and (24), it is seen that for
from the top of the Petlyuk column should be chosen as:
b ≤ bp, where Vmin 1 = Vmin 1 (h1), h′ would be equal to h1.
The introduction of this result into Eq. (22) gives a constant
V min2 ˆ V min petlyuk
value for Vmin 2 in the entire range from b = 0–bp. However,
when b is greater than bp and Vmin 1 = Vmin 1 (h2), h′ is not ˆ MaxfV min;2 ; V min;3 ‡ …1 q†Fg …29†
equal to h2, since h2 is not in the defined range given for h′.
Therefore, a new value for h′ is obtained at this condition. In This result was first presented by Halvorsen et.al. [13]. The
this case, Vmin 2 no longer remains constant. These conditions authors noted that the selection of a proper value for Vmin pe-
are shown in Figs. 4a) and 4b). tlyuk was dependent on the value of the split of B, i.e., on the
value of b. The authors’ findings are summarized in Tabs. (1)
and (2).
3.3 Minimum Vapor Flow in Column Section 3
Table 1. Conditions for choosing Vmin petlyuk when h′ = h1.
The net stream from the bottom of section 1 is fed to section
3, i.e., F3 = W1. The bottom stream product of section 3 is W3, Range of b b < bb1 b b1 ≤ b ≤ b p b > bp
the compositions of this stream, xi;w3 , have been found earlier
Vmin petlyuk V min;3 ‡ …1 q†F Vmin 2 Vmin 2
through the process of solving the material balance equations,
Eqs. (1–6). Therefore, at the minimum reflux condition, one
can write: Table 2. Conditions for choosing Vmin petlyuk when h″ = h2.
X ai xi;w F3
F3 …1 q″† ˆ 1
…25† Range of b b < bp bp ≤ b ≤ b b2 b > bb2
ai h″
Vmin petlyuk V min;3 ‡ …1 q†F V min;3 ‡ …1 q†F Vmin 2

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com

1288 N. Sotudeh et al. Chem. Eng. Technol. 2007, 30, No. 9, 1284–1291

The balanced splits b b1 and b b2 , which are seen in Tabs. (1) 0 P .P 1

ai xi;D ai xi;D
and (2), are obtained for the conditions when Vmin 2 = V min;3 2 2

Log@P .P A
ai u ′2 ai u ′1

‡…1 q†F or: ai xi;L ai xi;L

ai u ′2 u ′1
N3 ˆ  i …36†
a

X ai xi;D D2 X ai xiW W3 u′2

2
ˆ 3
‡ …1 q†F …30† Log
u′1
ai h′ ai h″

Calculations show that: and

1 1
0P .P 1
aC zC aC h1 aC h″ ai xi;L ai xi;L
b b1 ˆ 1 ‡ (31)
Log@P .P A
1 1 ai w ′2 ai w ′1
aB zB
ai xi;s ai xi;s
aB h1 aB h″
ai w ′2 ai w ′1
N4 ˆ   …37†
where h″ is found from Eq. (30) by setting h′ = h1, and: w′2
Log
w′1
1 1

b b2 ˆ
aA zA aA h2 aA h′
…32† In the above equations, xi;D2 and xi,S are known values. u′1 ;
aB zB 1 1
u′2 ; w′1 and w′2 are found from Underwoodþrimes definition
aB h2 aB h′
equation as follows:
where h′ is found from Eq. (30) by setting h″ = h2. X ai xi;D D2
From the above discussion and noting on the results given V2 ˆ 2
…38†
ai u′
in Tabs. (1) and (2), it seen that there exists a flat optimal re-
gion for minimum vapor flow in the tower in the range be- f i ˆ A; B g where aA > u′1 > aB > u′2 > aC
tween the preferred split and the balanced splits, which will
give the desired separations by consuming minimum amounts
X ai xiW W2
of energy. In order to decide which value of b should be chosen V2 ˆ 2
…39†
in this range, it is advised that the calculations of the number ai w′
of trays be performed at both terminals, b = bb and b = bp. f i ˆ A; B g where w′1 > aA > w′2 > aB
Hence, the calculation with the lower tray number will reveal
the proper value of b. The unknown parameters in Eqs. (36) and (37) are N3, N4
and xi,L. In the present case xi,L is the composition of the liquid
entering section 1. Since xB,L = 1–xA,L, there are two equations
5 Number of Trays and three unknowns, N3, N4 and xA,L. It is clear that by in-
creasing xA,L in Eq. (36), N3 decreases while N4 in Eq. (37) is
After finding Vmin petlyuk, the minimum reflux ratio of the increased, Fig. 5a). These trends cause that the total number of
tower is found from: trays of section 2, N3 + N4, to pass through a minimum.
  Therefore, one can write:
V min2
R min ˆ 1 …33† d…N3 ‡ N4 †
D2 ˆ0 (40).
dxAL
By setting the operating reflux ratio of the tower equal to
From Eq. (40), xA,L.and xB,L = 1–xA,L, are easily found. N3
1.2 · Rmin, the other parameters are calculated by applying the
and N4 are calculated by using the values of xA,L and xB,L deter-
material balance equations for each section of the tower. Here,
mined from Eqs. (36) and (37).
a new parameter, r, which is the split of internal liquid over
This procedure can be followed in a similar manner for col-
both sides of the middle wall of the divided wall column, is in-
umn section 3. For section 3, one can write:
troduced as:
0P .P 1
ai xi;s ai xi;s
L2
Log@P i 2 .P i 3 A
a u″ a u″
ˆr …34† ai x ai x
L2 i;L i;L
ai u ″2 ai u ″ 3
N5 ˆ   …41†
u″2
From inspection of Eq. (34), it is seen that r cannot be any Log
u″3
desired value but must be chosen in the following range:
  and
L min;2 L min;3 ‡ S L2 L min;1 ‡ qF
Max ; <r< …35† 0P .P 1
L2 L2 L2 ai x
i;L
ai x
i;L

Log@P a x .P a x A
ai w ″2 ai w ″ 3

i i;W3 i i;W3
By referring to Fig. (2) and using Underwood’s equations ai w ″2 ai w ″ 3

[14], the theoretical number of trays, N3 and N4 are found N6 ˆ   …42†

w″2
from: Log
w″3

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com

Chem. Eng. Technol. 2007, 30, No. 9, 1284–1291 Modeling 1289

HerexiL with i = B and C, are the compositions of the liquid

16
a) stream leaving section 1. Since xCL ˆ 1 xBL , there are two
14
equations and three unknowns (N5, N6 and xBL ). By increasing
Number of Trays

12 N3 + N4 xBL , N5 in Eq. (41) decreases while N6 in Eq. (42) increases,

10 N3 Fig. 5b). Therefore, it is necessary for the graph of the total
N4
8 number of trays in section 3, N5 + N6, to pass through a mini-
6
mum. xBL andxCL can be calculated by writing the necessary
equation for the minimum condition. N5 and N6 are calcu-
4
lated by using these values in Eqs. (41) and (42).
2 The total number of trays for section 2 is N3 + N4 and that
0 of section 3 is N5 + N6. The required number of trays for sec-
0 0.2 0.4 x 0.6 0.8 1 tion 1 (feed stream side) should necessarily be equal to N4 +
AL
N5. Thus:

(N1 + N2) = (N4 + N5) (45)

12

10 b) The Kirkbride equation is used to locate the feed tray stage

Number of Trays

N5
N6 [15]:
8 N5 + N6
 
6 N1 D1 ZA F 0:206
ˆ × …46†
N2 W1 Zc F
4
Solving Eqs. (45) and (46) permits one to find N1 and N2.
2

0
0 0.2 0.4
xB L
0.6 0.8 1 6 Verification of the Model

Figure 5. a) Number of trays vs. xA,L, and b) Number of trays vs. In order to check the validity of the above method, the data of
xB L. Muralikrishna et al. [7] was used, Tab. 3, so that the current
results can be compared with their work. The results of the
present calculations and those of Muralikrishna are shown in
where u″2 ; u″3 ; w″2 and w″3 in the above equations are found
from: Table 3. The data of Muralikrishna et al. [7], which ar e used in
the introductory example to verify the design method.
X ai xi;D D3
V3 ˆ 3
(44† Feed Composition (A, B and C) ZF = [0.333, 0.333, 0.333]
ai u″
f i ˆ B; C g where aB > u″2 > aC > u″3 > 0 Molar Feed F = 30 mol/h
Relative Volatilities aA = 6.773, aB = 2.798, aC = 1
and
Feed Quality q = 0.557
X ai xiW W3 xAD2 ˆ xCW3 ˆ 0:95
V3 ˆ 3
…45† Impurity Specifications
ai w″ xBS ˆ 0:9
f i ˆ B; C g where w″2 > aB > w″3 > aC xBD2 ˆ xBW3 ˆ 0:05
xAS
ˆ1
xCS

Table 4. Comparison of present results with those of Muralikrishna et al. [7].

Number of Feed Number of Trays in Liquid Side Vapor R2 Internal Vmin petlyuk,
Trays in Stage Sections 2 and 3 Recycle Stream Recycle Reflux, r (mol/h)
Section 1 Stage Stage Stage

Muralikrishna 12 10 25 6 12 19 3.46 – –
et al. [7]
Present method 8 4 16 5 10 13 3.46 0.73 31.9532
for Preferred Split
Present method 8 4 17 6 11 14 3.46 0.69 31.9532
for Balanced Split

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com

1290 N. Sotudeh et al. Chem. Eng. Technol. 2007, 30, No. 9, 1284–1291

Table 5. The results of the current calculations using bp = 0.5085 the minimum vapor flow required for the tower are found
and r = 0.73. from the current method but these parameters are absent in
the work of Muralikrishna et al. The current authors believe
D1 = 15.085 D2 = 10 D3 = 4.9150
that the present results are more accurate since the Fenske
V1 = 23.6204 V2 = 41.6126 V3 = 17.9922 equation has not been used in the current calculations. The de-
tails of the results of the calculations form the current investi-
L1 = 8.5354 L2 = 31.6126 L3 = 13.0772
gation are shown in Tab. 5. Figs. 6a) and 6b) are plots showing
V 1 = 10.3304 V 2 = 17.9922 V 3 = 28.3226 the final layouts of the columns based on the current method
L1 = 25.2454 L2 = 23.0772 L3 = 38.3226
and the work of Muralikrishna et al.

W1 = 14.915 W2 = 5.085 W3 = 10
u′1 ˆ 5:2477; u′2 ˆ 2:7434; w′1 ˆ 6:8889; w′2 ˆ 3:4895
xA;L1 ˆ 0:4243; N3 ˆ 4:9418; N4 ˆ 4:2612
u″2 ˆ 2:1184; u″3 ˆ 0:9780; ; w″2 ˆ 2:8583; w″3 ˆ 1:3245
xB;L1 ˆ 0:3715; N5 ˆ 3:7665; N6 ˆ 3:4571
N1 ˆ 4:0185; N2 ˆ 4:0092
Tab. 4. The previous authors have used the Underwood’s,
Fenske and Gilliland equations in their calculations, while the
current work used the aforementioned method. As seen from
Tab. 4, the current method demonstrates a more economical
design when compared with the previous study for both cases
b = bp and b = bb. From Tab. 4, the split of the liquid, r, and
7 Conclusions
A method for design of divided wall columns was developed in
the current work. The method was different from the work of
previous investigators since the Fenske and Gilliland equations
were not used in the design of the tower. In the process of de-
veloping the design equations, it was found that the split of in-
ternal reflux over both sides of the middle wall of the column
was bounded. Setting r equal to its upper bound and using the
proper value of b, resulted in a more efficient and economical
design. The authors believe that this procedure is extendable
to systems with feed streams composed of more than three
components.

D2 = 10
x AD2 = 0.95 D2 = 10
x AD2 = 0.95
x BD2 = 0.05
x BD2 = 0.05
1 1

L1
5 ≈ 0. L1
L2 = not given
6 L2

L2 L2
F = 30mol / hr L2 L2
q = 0.557
1 1 1 9 S = 10
z =[ : : ] 10 12 S = 10
3 3 3 x BS = 0.9
F = 30mol / hr x BS = 0.9
α A = 6.773 x AS
=1 q = 0.557 x AS
α B = 2.798 x BS =1
1 1 1 x BS
αC = 1 z =[ : : ] 16
3 3 3
α A = 6.773
α B = 2.798
13
αC = 1 19

16 25
W3 = 10
W3 = 10 xCW3 = 0.95
xCW3 = 0.95
x BW3 = 0.05
x BW3 = 0.05
(a) (b)
Figure 6. (a) The layout of the column based on the present model, and (b) The layout of the column based on the work of Muralikrishna
et al. [7].

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com

Chem. Eng. Technol. 2007, 30, No. 9, 1284–1291 Modeling 1291

Acknowledgements h′ [–] common (minimum energy)

Underwood root in column section 2
The authors wish to express their gratitude to the Esfahan Oil h″ [–] common (minimum energy)
Refining Company (EORC) for the financial support of this Underwood root in column section 3
project.
Subscripts
A, B, C component A, B, C
Symbols used D top net distillate flow
S side stream flow
A, B, C [mol/h] flow rate of component A, B, C in the W bottom net distillate flow
feed b balanced point
D [mol/h] top net distillate flow i component i
F [mol/h] feed flow rate p preferred split point
L [mol/h] liquid flow rate in rectifying part min minimum reflux condition (infinity trays number)
L [mol/h] liquid flow rate in stripping part 1 column section 1 (prefractinator), also biggest
S [mol/h] side stream flow rate Underwood root
V [mol/h] vapor flow rate in rectifying part 2 column section 2, also root between aB and aC
V [mol/h] vapor flow rate in stripping part 3 column section 3, also smallest Underwood root
W [mol/h] bottom net distillate flow
x [–] mole fraction in liquid phase
y [–] mole fraction in vapor phase References
z [–] mole fraction in feed
q [–] feed quality in column section 1 [1] F. B. Petlyuk, V. M. Platonov, D. M. Slavinskii, Int. Chem.
q′ [–] feed quality in column section 2 Eng. 1965, 5 (3), 555.
q″ [–] feed quality in column section 3 [2] G. Kaibel, Chem. Eng. Technol. 1987, 10, 92.
[3] A. Brugma, US Patent 2 295 256, 1942.
Greek symbols [4] F. Lestak, R. Smith, V. Dhole, Trans. Inst. Chem. 1994, 72, 639.
a [–] relative volatility [5] J. Cerda, W. Westerberg, Ind. Eng. Chem. Process Des. Dev.
b [–] recovery of component B in top 1981, 20, 546.
product [6] C. Triantafyllou, R. Smith, Trans. Inst. Chem. 1992, 70, 118.
u [–] Underwood root in rectifying part of [7] K. Muralikrishna, K. P. Madhavan, S. S. Shah, Trans. IChemE
column section 1 2002, 80, 155.
u′ [–] Underwood root in rectifying part of [8] K. Glinos, M. F. Malone, Chem. Eng. Res. Des. 1988, 66, 229.
column section 2 [9] Z. Fidkowski, L. Krolikowski, AIChE J. 1986, 32, 537.
u″ [–] Underwood root in rectifying part of [10] A. J. V. Underwood, Chem. Eng. Prog. 1948, 44 (8), 603.
column section 3 [11] K.-Y. Li Yaws, C. S. Fang, Chem. Eng. 1981, 88, 153.
w [–] Underwood root in stripping part of [12] J. Stichlmair, in Ullmann’s Encyclopedia of Industrial Chemis-
column section 1 try, Vol. B3, VCH, Weinheim 1988.
w′ [–] Underwood root in stripping part of [13] I. J. Halvorsen, S. Skogestad, Ind. Eng. Chem. Res. 2003, 42,
column section 2 596.
w″ [–] Underwood root in stripping part of [14] A. J. V. Underwood, J. Inst. Pet. 1948, 32, 598.
column section 3 [15] C. G. Kirkbride, Pet. Ref. 1944, 23, 321.
h [–] common (minimum energy)
Underwood root in column section 1

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com