[!

i5@
----------------
class and home )

Class and Home Problems (CHP) present scenarios that enhance the teaching of chemical
engineering at the undergraduate or graduate level. Submissions must have clear learning
objectives. CHP papers present new applications or adaptations that facilitate learning in
specific ChE courses. Submit CHP papers through [Link]/cee, include CHP in
the title, and specify CHP as the article type.

EXPLORATION OF THE DESIGN OF

DISTILLATION COLUMNS - BEYOND
REFLUX AND TRAY COUNT
AMPARO GOMEZ SIURANA AND ALICIA FONT-ESCAMILLA
University of Alicante • Alicante, Spain, E-03080

INTRODUCTION

T
he starting point of this work can be found in an in- (K-values), and relative volatilities with reference to
teresting problem proposed by Seader et al.P 1 where n-butane (C4 )? (Author's note: It is worth mentioning
some features of multicomponent distillation, from that in the original statement, the requested volatilities
the behavior of the condenser to the thermal condition of the referred to propane (C3), but in this case it has been
feed stream, are posed. The complete answer to the prob- changed because in this type of problem, the relative
lem requires the use of different tools available in chemical volatilities are usually referred to respect to the heavy
processes simulators, as ChemCAD from Chemstations, Inc. key component (C4)).
121 or Aspen HY SYS® and Aspen Plus™ from Aspentech. c3J
[Link] reflux ratio is 1.3 times the minimum reflux, what
Students not only have to be able to recognize and to know is the actual reflux ratio, R? How many theoretical
what unit to choose and when and how to use it, but also to plates, N, are needed in the rectifying and stripping
select the best way to extract and analyze the results. The sections?
process knowledge acquired by solving the problem permits
us to develop a case study in order to perform a comparison d. Compute the separation of species. How will the
among the different ways to select the optimal feed stage, separation differ if a reflux ratio of 1.5, 15 theoretical
as well as to study the effect of the feed temperature on the plates, and feed in the ninth plate are chosen?
column behavior.
Amparo Gomez Siurana is a full professor
The above-mentioned exercise corresponds to problem of Advanced Separation Processes at the
number 10.30 of the textbook by Seader et alY 1 and focuses University of A/icante (Spain). Her research
interests focus on the study of the thermal and
on the design of a depropanizer with the problem statement catalytic pyrolysis of different organic materials.
as follows: She is very interested in the development of
strategies to improve the learning of separa-
A depropanizer distillation column is designed to operate tion processes, considering aspects beyond
the usual contents of conventional courses.
at a feed stage pressure of 315 psia (21.7 bar)for separating ORCID: 0000-0001-8334-0606
a feed into distillate and bottoms with theflowrates shown in
Alicia Font-Escamilla is an associate pro-
Table 1. The feed is 66 mol% vapor at tower pressure. Steam fessor at the University of Alicante (Spain).
at 315 psia (21.7 bar) and cooling waterat65 °F (291 K) are Her research has focused on the experimental
determination of phase equilibria and its applica-
available for the reboiler and condenser, respectively. Assume tion in separation processes. Her research and
a 2-psi (0. 138 bar) column pressure drop. teaching are closely related and, for this reason,
she is trying to deepen the knowledge and skills
For this separation: of separation process students. ORCID: 0000-
0003-2469-6391
a. Should a total condenser be used for this column?
b. What are the feed temperature, the equilibrium ratios © Copyright ChE Division of ASEE 2021

104 Chemical Engineering Education

 e. Compute the temperature and concentration in each The mixture and the separation shown in Table 1 have been
stage. What is the effect offeed plate location? How called Case 1, and four additional cases described in Table
will the results differ if a reflux ratio of 1.5 and 15 2 have also been considered. The flowrate of the consecu-
theoretical plates is used? tive components as well as the separation between the key
components are the same as in Table 1. For each case the
TABLE 1 five thermal conditions of feed in Table 3 have been studied.
Proposed separation corresponding to problem 10.30 The partially vaporized mixture of Case 1 corresponds to the
from Seader et aV 11 with the original units changed original problem statement proposal. q is defined as the liquid
from lbmol/h to kmol/h. fraction of the feedC4 ,SJ and varies from< 0 for a superheated
kmol/h vapor (SHV) to> 1 for a subcooled liquid (SCL).

Component Feed Distillate Bottoms

Methane (Cl) 26 26.0 - INTENDED LEARNING OUTCOMES
Ethane (C2) 9 9.0 -
The problem proposed in this work addresses several objec-
Propane (C3) 25 24.6 0.4 tives. It is an exercise included in an Advanced Separation
n-Butane (C4) 17 0.3 16.7 Processes course that is part of the Chemical Engineering
Master's program of the University of Alicante (Spain). Some
n-Pentane (CS) 11 - 11.0
parts of the exercise also could be useful for undergraduates.
n-Hexane (C6) 12 - 12.0 Multiple problems can be extracted from this material and
Total 100 59.9 40.l used in multiple forms covering different learning strategies.
Instructors can select parts of the analysis in order for the
students to master basic separation concepts or to gain more
In this work the original flowrate units have been changed
in-depth knowledge.
from lbmol/h to kmol/h in order to obtain larger column diam-
eters, in the range of the available correlations for calculating Usually these exercises focus on the search for the opti-
the costs of columns.£2,3l In the following sections the answer mum combination of number of stages and reflux ratio. In
to these five points, as well as the analysis of the best feed this case the influence of the feed stage location and the feed
stage location and the effect of the thermal feed phase, are thermal condition are considered. The proposed case study
provided and analyzed. The detailed calculation procedures allows the students to learn the importance of knowing and
can be found at [Link] In this paper we will comparing the different alternatives for a particular process
focus on the interpretation of the results. and developing conclusions. Moreover, they can discover
the utility of using the results of the simulation software for

TABLE2
Cases to be considered. The flowrates of the consecutive components and the keys recoveries are the same as in Table
1 (98.40 % and 98.24% ofrecovery of light key, LK, in the distillate and heavy key, HK, in bottoms, respectively). Ci
represents a linear hydrocarbon with i carbon atoms. Expected distillate and bottoms temperatures are their bubble
or dew temperatures calculated from the compositions obtained from the preliminary mass balance.
Feed Expected Expected
Column
Condenser bubble/dewpoint distillate bottoms
Case Components LK HK Pressure
type temperatures temperature temperature
(bar)
(K) (K) (K)
Cl, C2, C3, C4,
1 C3 C4 21.7 Partial 217.3 / 389.7 294.1 426.6
C5,C6
C6, C7, C8, C9,
2 C8 C9 2.17 Total 410.7 / 496.7 390.4 479.8
ClO, C15
CS, C6, C7, C8,
3 C7 C8 2.17 Total 376.0 / 426.9 356.2 446.3
C9,Cl0
CS, C6, C7, C8,
4 CS C6 2.17 Total 376.0 / 426.9 333.6 411.6
C9,C10
CS, C6, C7, C8,
5 C9 ClO 2.17 Total 376.0 / 426.9 370.5 478.9
C9,C10
Vol. 55, No. 2, Spring 2021 105
 TABLE3
Feed phase conditions at column pressure (indicated in this Table) taken into account for each case.
aLH,HK is the relative volatility of light key component with respect to the heavy key.

Case 1 Case2 Cases 3, 4 and 5

P=21.7 bar P= 2.17 bar P = 2.17 bar
ULK,HK ULK,HK ULK,HK
Feed phase condition q T(K) ULK,HK T(K) ULK,HK T(K)
(Case 3) (Case 4) (Case 5)
Superheated vapor (SHV) <0 399.8 1.840 520.0 1.488 440.0 1.762 1.808 1.733
Saturated vapor (SV) 0 389.7 1.916 496.7 1.554 426.9 1.826 1.873 1.791
Partially vaporized (PV) 0.34 357.1 2.204 447.2 1.734 410.9 1.913 1.964 1.866
Saturated liquid (SL) 1 217.3 6.362 410.7 1.922 376.0 2.155 2.207 2.088
Subcooled liquid (SCL) >1 205.4 7.277 400.0 1.988 360.0 2.294 2.346 2.200

purposes other than the mere simulation or design of units • The approximate design or simulation of distillation
and the importance of not using these simulation packages columns has been carried out using the Fenske-Un-
as a black box. Similarly, the instructor can highlight the derwood-Gilliland (FUG) method, with the Kirkbride
usefulness of the binary distillation concepts, well known or the Fenske equations for the optimal feed stage
from previous courses, for understanding multicomponent location, and using the Shortcut distillation unit of the
distillation behavior. There are several points useful for above-mentioned ChemCAD version.
reinforcing the idea of the appropriate use of the simulator. • The rigorous simulation of distillation columns uses the
These include the need to know which thermodynamic models rigorous simultaneous correction distillation (SCDS)
should be selected and why, the importance of checking the column of ChemCAD, which is a rigorous multi-stage
fulfillment of assumptions of approximate methods, and how vapor-liquid equilibrium module that simulates any
to locate and export results that are useful for carrying out single column calculation of two-phase or three-phase
additional calculations. distillation systems. It applies a Global Newton or
We recommend that the problem be solved by the whole simultaneous correction method.[61
class with each student or small group of students devoted to It is assumed that students have previously learned how to
one of the different parts of the work. It is most appropriate select the best thermodynamic model or ChemCAD unit
to be assigned as an out-of-classroom activity, guided by the to perform the different calculations involved.
instructor who balances the tasks for each student or team
According to King,[71 for a given number of equilibrium
and takes into account the time needed for their resolution.
stages, the optimal feed stage location is the one requiring
Additional data necessary for some calculations, such as the
the lowest reflux to obtain a given separation between key
cost of columns, should be provided to students, highlighting
components. Moreover, several rules of thumb are widely
the importance of considering aspects such as the equipment
used for determining the optimum feed location stage in
sizes and the required properties of heating or cooling agents.
simple and efficient ways. Among them, Kister[6l suggests
Finally, the results can be shared, analyzed, and discussed in
the following for the optimum feed stage:
the classroom with all the students.
• It is the stage where the ratio of key-component
mole fractions in the liquid on the feed stage (xLK/
METHODOLOGY xHK) is the closest to this ratio in the liquid por-
tion of the feed stream at tower pressure. The key
In order to solve the proposed exercise, the following pro- components, i.e. light key (LK) and heavy key (HK)
cedures and tools have been used:P,4 .Sl components, are defined as those which are being
• The Soave-Redlich-Kwong (SRK) equation of state separated in the column. The values for calculating
(EOS), which is applicable to hydrocarbon mixtures this ratio in the feed stream have been obtained with
at all pressures and non-cryogenic temperatures,[!] has the Flash unit of ChemCAD from the VLE calculation
been selected for calculating the vapor-liquid equilib- of the feed steam at the specified thermal conditions.
rium (VLE) of mixtures. In the liquid phases through the column, xLK/xHK is
calculated from the liquid mole fraction profiles re-
• VLE calculations have been performed with the Flash
sulting from the rigorous simulation using the SCDS
unit of ChemCAD v7.l .l.
unit of ChemCAD.
106 Chemical Engineering Education
 • It is the stage provided by the intersection of the op- • The true optimal feed stage has been considered as
erative lines on a Hengstebeck diagram. The method that yielding the lowest column costs. For a given
of Hengstebeck, which is described by Kister,[61 is an number of theoretical stages, it seems reasonable that
approximated method for designing multicomponent the true optimal feed stage will be that yielding the
distillation columns. It is based on defining equivalent lowest column cost. This work considers that this
binary streams having only the key components in the criterion gives the "true" result, and it is compared
same relative amount as in the original multicomponent with results given by other methods.
mixture. The method of Hengstebeck provides tools All these concepts should be known by the students, who have
to calculate the binary equivalent equilibrium x, - y, previously been instructed in the correct use of the simulator
curve as well as the effective liquid and vapor streams to solve phase equilibrium and multicomponent distillation
in the rectifying (L, and V,) and stripping (L,' and V,') problems, as well as in the different methods for selecting
sections, where molar overflow is considered constant. the feed stage.
After that, the problem is solved in exactly the same
way as the McCabe-Thiele method. In this work the
flowrate and composition profiles resulting from the RESULTS AND DISCUSSION
SCDS-rigorous simulation of the columns have been
used to calculate the equivalent binary equilibrium The following discussion focuses on the results obtained
curve and internal flowrates. and their interpretation. More information about calculations
• It is the stage yielding the most equal slopes on both and details of the results can be found at [Link]
sides of the feed stage in a key ratio plot. The key
ratio plots, i.e. the plots of the log (xLK /xHK) profile
Answers to the Proposed Questions
through the column, are described by Kisterl61 as a tool
to check the appearance of "retrograde" or "reverse" a) Shouul a total condenser be used for this column?
distillation. They indicate when the column operates The condenser requires a sufficient difference of tempera-
near total reflux or minimum reflux, and the feed ture between the available refrigerant, in this case cooling
stage is reflected by the appearance of some extent water at 291 K, and the distillate, at higher temperature, in
of reverse distillation. The reverse distillation is very order to provide the required driving force for the heat transfer.
well explained by Kingl71 and Kisterl61 , and results in Total condensers yield a liquid distillate at its bubble point or
the increase of the HK and the decrease of the LK when as a subcooled liquid. In this case the bubble point tempera-
we go up the column, i.e. the reverse of the expected ture of the distillate in Table 1 at 21. 6 bar is 194. 6 K, lower
behavior. This is due to the presence of multiple heavy than that of cooling water (291.5 K); therefore, the use of a
non-key components in the rectifying section such that total condenser is not possible. The condenser pressure has
the behavior of the HK is that of a light component; been estimated considering that the pressure drop specified
similarly, there may be many light non-key components in the problem statement only occurs between the feed stage
in the stripping section yielding the behavior of the LK and the condenser.
as that of a heavy component. In this work the LK and Partial condensers yield a vapor distillate at its dew point.
HK mole fractions needed for the construction of key They are used when there are very light components, as it oc-
ratios graphics have been obtained from the SCDS- curs in this case with methane and ethane in the feed stream.
rigorous simulation of the columns. Consequently, high column pressures and/or low condenser
• It is the stage resulting from the application of every temperatures are required to condense these very volatile
approximate or rigorous analytic technique. In this components in a total condenser, and the use of a partial con-
work the results of the application of the approximate denser avoids the use of costly refrigeration. In this case the
equations of Fenske and Kirkbride,l 1 ,41 provided by difference between the dew point temperature of the distillate
the Shortcut distillation tool of ChemCAD, have been (294.0 K) and that of the available cooling water is very small
considered. (only 2. 5 K) and it cannot be used as a refrigerant for a partial
condenser either. Thus, we can conclude that this separation
• It is the stage yielding the lowest values of reboiler
will require a partial condenser operated with another cooling
duty or the lower reflux. Students usually consider
agent at a sufficiently lower temperature.
that the optimal feed stage is the one minimizing the
heat supply needed in the reboiler, understanding that b) What are the feed temperature, the equilibrium ratios
it should be the one yielding the lowest operating costs (K-values), and relative volatilities with reference to C4?
and, therefore, the lowest total cost. This approach, as The VLE calculation for the feed stream in Table 1, 66%
well as that suggested by Kingl71based on the location vaporized at 21.7 bar, yields a temperature of 357.1 K and
giving the lowest reflux flowrate, will be also checked. the equilibrium ratios, K;, shown in Table 4. For each
Vol. 55, No. 2, Spring 2021 107
component i, the relative volatilities with respect to the HK pressure and C4 split as Column A, with the exception that
component (C4), calculated as K;I Kc4 , are also shown. The N = 15, NF = 9, and R = I. 5. This column is further referred
main conclusion is that the relative volatility of the LK is high to as Column B. The main differences are that Column B
enough to achieve the separation by distillation. Moreover, yields lower C3 recovery in the distillate and higher heat
the volatilities of the non-key components permit the separa- duties than Column A. The interpretation of these results is
tion suggested in the preliminary mass balance on Table 1. not easy, because Column B has a higher reflux ratio but a
lower number of stages than Column A. Thus, the study
of cases available for the Shortcut module in ChemCAD is
TABLE4 very useful for this comparison. The study of the requested
Equilibrium ratios (K;) and relative volatilities ( ai,c4 ) for separation of C3 in Column A with 15 theoretical stages
the feed stream 66% vaporized at 21.7 bar (T= 357.1 K). needs R/Rmin:::: 2.88, i.e. R:::: 2.6, much higher than the value
specified, R = I. 5. Thus, the lower reflux ratio, the lower
Component Cl C2 C3 C4 CS C6
the separation. The same result is obtained from a sensi-
K; 8.610 2.976 1.351 0.613 0.287 0.137 tivity analysis on Column B, modifying the reflux ratio to
ai,,C4 14.05 4.855 2.204 1.000 0.468 0.223 determine the corresponding C3 flowrate in the distillate.
e) Compute the temperature and concentration in each
stage. What is the effect offeed plate location? How will
c) If the reflux ratio is 13 times the minimum reflux, what the results differ if a reflux ratio of 1. 5 and 15 theoretical
is the actual reflux ratio? How many theoretical plates are plates are used?
needed in the rectifying and stripping sections? The column profiles are obtained from a rigorous simulation
In order to answer this question, the approximate design of the column that provides the values of the variables cor-
of the column was performed using the Shortcut unit of responding to the two phases leaving each theoretical stage.
ChemCAD, using both the Kirkbride and Fenske equations This is accomplished using the SCDS unit within ChemCAD.
for the feed stage location and the needed specifications given Columns A and B have been simulated, specifying the key
in the initial problem statement. The results obtained are: components recoveries shown in Table 1 as separation re-
• Reflux ratio, R = 1.171 quirements. It is worth mentioning that the results obtained
• Number of theoretical stages (condenser and reboiler for the distillate and bottoms products as well as for the
included), N = 23 .9 (it will be rounded up to 24 for the condenser and reboiler duties are very close to those obtained
rigorous simulation) by the approximate Shortcut calculations, thus showing that
the FUG method assumptions (constant molar overflow and
• Feed stage (considering the condenser as first stage)
approximately constant volatilities) are adequately fulfilled.
given by the Kirkbride equation, NF = 12.7 (it will be
rounded up to 13 for the rigorous simulation). Conse- Figure 1 shows the temperature and key component liquid
quently, the column has a partial condenser (stage 1), mole fraction profiles corresponding to Columns A and B. As
11 equilibrium stages in the rectifying section (stages seen in Figure la, the temperature profiles are very similar,
2 to 12), 11 equilibrium stages in the stripping section but Column A shows a higher temperature in the rectifying
(stages 13 to 23, with the feed stage, 13, as the first one section and only slightly lower (or equal) than Column B
in this section), and the reboiler (stage 24). According in the stripping section. This is also reflected by the inner
to Seader et alf 11 ·the Fenske equation should only be streams of Column A having higher C4 and lower C3 contents
used for symmetrical mixtures and separations, which in the rectifying section than Column B , and the opposite in
gives NF = 11.8 (rounded up to 12). the stripping section, as Figure lb reflects. These profiles
can be qualitatively explained based on a McCabe-Thiele
As the students have been instructed, these results provide
diagram; for Column A, with lower reflux ratio, the opera-
the starting point for the rigorous simulation and, moreover,
tive lines are closer to the equilibrium curve. As a result, the
will allow the comparison proposed in the next question. This
driving force for the separation in each equilibrium stage is
column is further referred to as Column A.
lower, and the rectifying section, where streams are enriched
d) Compute the separation of species. How will the separa- in the LK component, allows lower C4 removing and C3
tion differ if a reflux ratio of 1.5, 15 theoretical plates, and enriching. This is the reason why more stages are required
feed in the ninth plate are chosen? to achieve the same separation. Similarly, in the stripping
The separation of components predicted by the Shortcut section, where streams are enriched in the HK component,
module for Column A is practically the same as that shown the lower reflux ratio results in the lower driving force and
in Table 1. The column proposed for comparison has been lower separation, so streams in Column A have higher C3 and
simulated with the Shortcut module in ChemCAD and with lower C4 contents. The students learned the binary separation
the same specifications for the feed stream, condenser, column concepts and calculations in previous courses on Separation
108 Chemical Engineering Education
Processes. Here, the McCabe-Thiele diagram is used to as a light component near the reboiler because the
justify and interpret multicomponent behaviors, which the scarcity of compounds lighter than C4.
instructor can take advantage of to highlight and reinforce • Components lighter than C3 practically do not appear
this previous knowledge. in the stripping section and show a near constant con-
Figure le shows the composition profiles of column A, while centration in the main part of the rectifying section.
Column B exhibits a similar behavior. The obtained profiles Similarly, components heavier than C4 are practically
reflect the expected behavior on a column with good separa- absent in the rectifying section and constant in the
tion between key components, as the following statements stripping section. This is the behavior assumed by
indicate: the Hengstebeck method,[6l which is applied in the
• The maxima of C3 and C4 approaches to the top and next section.
bottom of the column, respectively, which are charac- Figure le also shows that there is some reverse distillation
teristic of key components. [4 ,6l They appear near the around the feed stage due to the existence of so many heavy
condenser due to the scarcity of components heavier components above of the feed stage, producing C4 as a light
than C3, with the consequence that C3 shows the be- component, and so many light components under the feed
havior of a heavy component. Similarly, C4 behaves stage, producing C3 as a heavy component.

15

21 13

-....... 17
11..;
iii'
II
..;
II 9
.!5..
.
.!5. 13
GI
.a
7
GI
.a
E
:I
C
E 9
:I
GI
IIO
5
C
GI "'
t;;
31' 5 3
t;;
1 1
280 330 380 430
T(K)

(a)

15

21 • C3 Column A 13 21
.,... e4 Column A

.
~ 16
•C3 Column B
-•·C4 Column B
11

9
iii'
..;
.
GI
.a
16
..; II
E
II

.
.!5. 11 7
..
.!5.
GI
.a
:I
C
GI
IIO
11
GI
.a
E 5
E
:I "'
t;;
C
ii!
f
6 GI 6
IIO . 'i'
:0 "'
3 t;;
'' *\ ,,,,-·· /
b;
1
0 0.2 0.4 0.6
•
0.8 1
1 1
=--~"
0 0.2 0.4 0.6 0.8 1
Liquid mole fractions

(b) (c)

Figure 1. (a) Comparison between the temperature and (b) the liquid C3 and C4 mole fraction profiles in Columns A and B. (c)
Liquid composition profile in Column A. Stage 1 is the partial condenser and the last stage is the reboiler. The vapor profiles
follow similar patterns to (b) and (c).
Vol. 55, No. 2, Spring 2021 109
Comparison Among Different Criteria to Select the and 11, allowing us to conclude that the optimal feed stage
Optimal Feed Stage appears at NF = 9. It is worth mentioning that this criterion
The analysis of the influence of the operating variables on is highly time consuming because not only do 21 simulations
the behavior of a distillation column is not an easy task be- have to be completed, but also the liquid profiles for each one
cause different results are obtained depending on the variables of them have to be extracted for the keys ratio calculation
selected. For instance, given a column with the same number and comparison.
of theoretical stages where the reflux ratio and the reboiler Feed stage from the intersection of the operative lines
duty are specified and the feed stage changes, different distil- in a Hengstebeck diagram. The method of Hengstebeckf61
late and bottoms products are obtained with similar operating provides tools to determine the equivalent binary flowrates of
costs. Nevertheless, if the specifications are based on mole the streams in each section of the column from the calculated
fractions, flowrates, or recoveries of key components, the limiting compositions of the non-key components, i.e. from
product streams obtained when the feed stage is modified will the composition of non-key components in the zone where
be very similar, but the requirements of reflux and condenser they have constant mole fraction. Figure lb shows that this
and reboiler duties, and consequently the operating column assumption is adequately fulfilled in this case, therefore the
cost, will be different. The study carried out in this section is method of Hengstebeck can be applied.
based on the specification of the number of theoretical stages Once the equivalent binary composition offeed, distillate,
and the required separation, and different ways to locate the and bottoms, the equivalent binary equilibrium curve, and the
optimal feed stage will be studied. The column selected is equivalent operative lines have been calculated, the problem
the one previously designed with the Shortcut method, with is solved in exactly the same way as the McCabe-Thiele
24 theoretical stages to obtain the separation shown in Table 1 method for binary mixtures. The details of the application
with R/Rmin = 1.3. The required simulations for the application of the Hengstebeck method in this example can be found at
of the different criteria for selecting the optimal feed stage http:/ /[Link]/3rQft6X, and the obtained Hengstebeck diagram
have been performed with SCDS, varying the feed stage loca- is shown in Figure 2.
tion in the range of 2-23 (stage 1 is the partial condenser and
As in the McCabe-Thiele method, the feed stage corre-
stage 24 is the reboiler). The complete optimization process
sponds to the intersection of the operative lines. The results
would require the calculation of the cost of a series of columns
from Figure 2 suggest that the optimal column has 16 theoreti-
with different number of stages, each one of them with the
optimal feed stage location. cal stages, with NF = 8. Therefore, considering the same ratio
of the number of stages in each section, it can be concluded
Feed stage from the comparison between xuJxHK in each that for a 24 theoretical stages column, NF = 12. Note that
stage and in the liquid portion of the feed stream. The this method is substantially more time consuming than the
VLE calculation of the feed stream provides the keys ratio previous one and, moreover, requires multiple assumptions.
value in the liquid phase, i.e. 0.2030/0.2284 =0.8888, for the Thus, at least a priori, it should not be selected as the best
comparison with the values obtained from the liquid mole method based on convenience.
concentration profiles in the column. The C3 and C4 concen-
tration profiles corresponding to each simulated column, each
one with a different feed stage location, have been extracted
from the SCDS simulation, and the stage showing the keys 1.0 Operative lines:
ratio closest to 0.8888 has been identified. Consequently, the -rectifying, Ye= 0.2645 + 0.7322 xe
0.8 :- ·- stripping, Ye= -0.0051-1.2168 Xe ,_
column for which this stage was the nearest to the feed stage
has been identified as that having the optimal feed location,
0.6
according to this criterion. As an example, Table 5 shows
the results when the feed stage is moved between stages 7
.::0.4

0.2
TABLES
Location of the stage with the closest keys ratio to O.O Bottoms Feed Distillate
that in the liquid portion of the feed stream when NF 0.0 0.2 0.4 0.6 0.8 1.0
changes from 7 to 11.
Feed stage, NF 7 8 9 10 11
Figure 2. Hengstebeck diagram corresponding to the sepa-
Stage with Xc3 lxc4 closest ration shown in Table I. x, and y, represent the equivalent
12 11 8 8 8
to 0.8888 binary compositions of multicomponent mixtures.

110 Chemical Engineering Education

 Feed stage from a key ratio plot. Using this criterion,
the optimal feed stage is the one yielding the most equal 1.5
slopes on both sides of the feed stage in a key ratio plot.
1.0
Figure 3 shows the key ratio plots obtained when the feed
stage moves from stage 2 to stage 23. When the feed stage 0.5
is between stages 2 and 7, the slope of the key ratio plot in 0.0
-;.
the upper part of the rectifying section is almost horizontal,
thus suggesting the existence of a zone with almost no 1.:! -0.5
11D
changes in the key ratio, i.e. almost no separation between .2 -1.0
keys. Similarly, when the feed stage approaches the bottom
-1.5
section of the column, in the range of 13 to 23 stages, the
last stages of the rectifying section show quite horizontal -2.0
profiles. Therefore, the optimal feed stage should be in the 2 4 6 8 10 12 14 16 18 20 22
range of 8 to 12 stages. The enlarged key ratio plots in this Stage number

range suggest that the optimal location seems to be stage

10. Note that this way of selecting the optimal feed stage is (a)
also very time consuming, and the results are not very clear
because the differences among the cases with NF equal to
9, 10 and 11 are negligible. 0.3
■ Feed stage =8
Feed stage from the Kirkbride and Fenske equations. • Feed stage = 9
These are approximated equations, very well described ... Feed stage = 10
in textbooks focused on multicomponent distillation,[ 1 ,41 0
• Feed stage= 11
which calculate the optimal feed stage exclusively from a
preliminary mass balance, as shown in Table 1. According
to the results described previously, the optimal feed stage
locations provided by equations of Kirkbride and Fenske are, -0.3
respectively, 13 and 12. These calculations are easy and fast -0.4
and can be performed by hand. Nevertheless, the validity is
-0.5
relative because the approach depends on the applicability
6 7 8 9 10 11 12 13 14 15
of the empirical equations to each particular case. Seader Stage number
et alY 1 report a comparison of the results of both equations
with the exact result, showing noticeable deviations. (b)
Feed stage selected as that yiel,ding the lowest reflux
flowrate. This criterion is the first one suggested by Figure 3. Key ratio plots corresponding with (a) columns varying
King.[71 It can be justified taking into account that, as can the feed stage from stages 2 to 23 and (b) enlarged version for
be expected and as shown in the next sections, the lowest columns varying the feed stage from stages 8 to 12.
reflux flowrate also corresponds to the situation with the
lowest energy requirements and the lowest column operat-
ing costs. Consequently, this criterion could be considered so flat indicates that small - or not so small - deviations from
as equivalent to the one giving the true result, i.e. based on the optimal feed stage will not have a noticeable influence on
the most economical column. Moreover, by using chemi- the operation of the column.
cal process simulation software, this criterion can be easily
Feed stage selected as the one yielding the lowest reboiler
applied by performing a sensitivity study where the feed
duty. As in the previous case, this criterion can be easily applied
stage location is selected as the independent variable and
with simulation software by performing a sensitivity study and
the reflux rate as the dependent variable. In this case NF has
choosing reboiler duty as the independent variable. In this work
been varied from 2 to 23 in 21 equal steps, and the calculated
the exchanged heat at the reboiler, Qreboiler , and the condenser,
molar reflux rate has been obtained. Figure 4a shows the
Qcondenser ' have been checked. As expected, the sum ( Qcondenser
results obtained, yielding stage 10 as the optimal feed stage.
+ Q,eboiler) has almost the same value for all the cases because
The pathway observed in Figure 4a will be repeated when
it corresponds to the difference between the enthalpy of the
variables other than the reflux rate are considered, thus ap-
outlet and inlet streams, which is very similar in all the cases
pearing as alternative variables to be taken into account for
because the products are always almost the same. Some devia-
the optimization of the feed stage location. Moreover, the
tions observed when the feed stage approaches the column top
fact that the area surrounding the minimum of Figure 4a was
and bottom could be explained considering that, in such cases,
Vol. 55, No. 2, Spring 2021 Ill
there are not enough theoretical stages to eliminate the heavy order to evaluate the operating costs. In this work the col-
non-key components from the distillate or the light non-key umn sizing has been performed by following the procedure
components from the bottoms product. The results obtained shown of Sinnott and Towler.[ 81 The capital costs have been
are represented in Figure 4b. According to this criterion, the calculated according the equations, figures, and relationships
optimal feed stage is NF = 10, which also yields the lowest provided by Turton et al)41 based on the module factor ap-
condenser duty and reflux flowrate. Moreover, the profiles proach, and the operating costs have been obtained as Ulrich
shown by these variables follow exactly the same pattern of and Vasudevan[51 proposed. The criteria and considerations
variation. Therefore, both criteria could be considered as considered to complete these calculations are shown in http://
equivalent. [Link]/3rQft6X, and the results obtained are represented in
Feed stage selected as the one yielding the lowest total Figure 4c. The lowest cost corresponds to the column with
column cost. Given a theoretical number of stages, the exact the feed stage at stage 10, with very small cost changes when
result for the optimal feed stage location should be the one !he feed stage location moves between stages 8 and 13. In
corresponding to the economic optimum. The column costs this example the utility costs are noticeably higher than the
calculation is a long process that requires the sizing of the fixed costs, which is very common because of noticeably
column, condenser, and reboil er to obtain the capital costs, as high energy costs. Therefore, it can be concluded that the
well as the selection of the adequate utilities, i.e. the adequate feed stage location involving the lowest reflux flowrate and,
cooling and heating agents and their corresponding loads, in thus, requiring the lowest utilities consumption, should also

1200

1000 15000 -
§
g 800
0

-+-Qreboiler

i ~10000
ia
..,...Qcondenser
~
;;::
IC
::, 400
liac: 5000
200

0 -·· 0
1 3 5 7 9 11 13 B D H ll n 1 3 5 7 9 11 13 15 17 19 21 23
Feed Stage, NF Feed Stage, NF

(a) (b)

3.5E+06

3.0E+OG

2.5E+06
... Fixed costs (€/y)

~.,,
2.0E+OG ...-utility costs (€/y)
•Total costs (€/y)
t: l.5E+06
0
u
1.0E+0G

5.0E+05

0.0E+00 ~ I • • • • • ■ ■ ■ •
1 3 5 7 9 11 13 B D H ll n
Feed Stage, NF

(c)

Figure 4. (a) Calculated refiuxfiowrate, (b) Energetic requirements and (c) Costs for columns with N =24 and ranging the
feed stage location from 2 to 23.
112 Chemical Engineering Education
be the one yielding the lowest total cost, i.e. the optimum. As a conclusion, the use of the Kirkbride and Fenske equa-
Moreover, the column having the lowest reflux flowrate also tions should be limited to very preliminary design steps or
corresponds to the one with the lowest reflux and reboil by hand calculations when a chemical processes simulation
ratios (the reboil ratio, VIB, is defined as the ratio between package is not available. In this case, whenever is possible,
the flowrates of vapor from the reboiler and bottoms), liquid we recommend performing a further optimization. This can
and vapor flowrates through the column, column diameter, easily be done by using the sensitivity study tool in the simula-
and heat exchanger sizes. Consequently, it not only has the tion package, with the feed stage as the independent variable
minimum utility cost, but it also shows the minimum fixed and the reboiler duty or the reflux flowrate as the dependent
cost. As seen in Figure 4, the pattern followed by costs is variable. As Figure 4 reflects, the optimal feed stage, at the
exactly the same of that shown by the reflux flowrate and du- minimum of the curves, appears in a wide zone that is almost
ties, as well as by the other above-mentioned variables. Thus, flat. This pattern suggests the possibility that small displace-
despite this pattern not being generalized, it makes sense to ments of the feed stage location near the minimum would not
consider that the three criteria for selecting the optimal feed have noticeable influence on the operation and cost of the
stage could be considered as equivalent. column. Therefore, the deviations involved in the use of the
Final comparison among the different criteria for locat- approximated methods for locating the optimum feed stage
ing the optimal feed stage. Table 6 summarizes the results location should have not a significant impact. This behavior
yielded by the different methods considered for selection of has been also observed when the feed phase condition was
the optimal feed stage. Considering that the true value cor- modified and for other mixtures with different ranges ofVLE
responds to the lowest column costs, criteria based on looking temperatures. Other variables, such as the liquid and vapor
for the lowest reflux flowrate or energetic requirements are flowrates, the diameter of each section of the column, the
completely equivalent. These criteria avoid the need oflong flowrates of cooling and heating agents, the heat exchange
and tedious cost calculations but involve the rigorous simula- areas of condenser and reboiler, and the reflux and reboil ra-
tion of several columns that, in the case of separations showing tios, also exhibit the same pattern of variation. Consequently,
convergence problems or columns with high number of stages, any of the variables described could be selected in order to
could be very time consuming. Despite the fact that these locate the optimal feed stage.
results cannot be generalized, it seems that the Hengstebeck
and keys ratio comparison methods, also based on rigorous Effect of the Thermal Feed Phase Condition
simulation results, tend to give worse results and, moreover,
In order to check the effect of the feed phase condition, we
their application requires significant effort. The method based
performed a study of a distillation column to give the separa-
on the key ratio plots is also time consuming and, although in
tion proposed in the five cases described in Tables 1 and 2
this case yield the correct result, the selection of the adequate
with the five thermal characteristics shown in Table 3. First,
plot is not so clear. The application of the Kirkbride and Fen-
we compare the results for Case 1, at the different thermal
ske equations do not give the best results, but they are within
conditions considered, for a column with N = 24 theoretical
the range of acceptable values, as Figure 4 reflects. Moreover,
stages and the feed stage location at stage NF = 13 to give the
they involve very fast and simple calculations, only needing
required keys recovery of Table 1. Next, we consider the ef-
the preliminary mass balance of the column.
fect of the Case 1 feed temperature on the preliminary column
design, i.e. on the number of stages and the required reflux
ratio to give the specified separation. Finally, the behavior
TABLE6 of Cases 2 to 4, with the five thermal conditions specified in
Optimal feed stage location from different criteria. Table 3, is summarized and analyzed. In all the cases, the
comparison among the patterns of variation of the operation
Method Feed stage, NF variables with the feed stage location has been checked, show-
Comparison of keys ratio 9 ing the same behavior than that of Figure 4.
Hengstebeck method 12 Effect of the feed phase condition for a given number of
theoretical stages and feed stage location. Figure 5 shows
Key ratio plot 10
the temperature profiles and the key ratio plots obtained
Kirkbride equation 13 when a column with 24 theoretical stages (N = 24) with feed
Fenske equation 12 stream entering at stage 13 (NF = 13) is simulated with the
feed stream corresponding to Case 1 (Tables 1 and 2) and the
Lowest reflux flowrate 10 five phase conditions considered in Table 3. The tray tem-
Lowest reboiler duty 10 peratures increase as the feed temperature increases (Figure
Sa). Moreover, the temperature profiles for cases SHV and
Lowest column cost 10
SV (the vapor phase feeds) are nearly coincident, as well as
Vol. 55, No. 2, Spring 2021 113
those for cases SL and SCL (the liquid phase feeds). As expected, does not seem to significantly affect the reverse distillation
the temperatures of stages I (condenser) and 24 (reboiler) are phenomena near the feed stage.
the same for the five cases because the specified separation is Figure 6 shows the liquid and vapor profiles correspond-
also the same. The temperature range in the column is between ing to the separation of mixtures with the five feed phase
294. 2 K and 426. 6 K; the partially vaporized feed temperature conditions of Case 1. As expected, the liquid (subcooled
(Table 2) is within this range and near the mean of the end point or saturated) entering the column mixes with the liquid
values, but the liquid feeds are colder than the distillate, and coming from the rectifying section, giving higher liquid
the vapor feeds are relatively close to the bottoms temperature. flowrate in the stripping section than that corresponding
Therefore, the vapor feeds tend to heat and the liquid feeds tend to merely the sum of both flowrates, because the liquid,
to cool the column. The higher temperature profile corresponds at lower temperature than that of the feed stage, causes
to higher heavy key concentration profiles, i.e. lower Xc3 lxc4 the condensation of a part of the vapor coming from the
values through the column (see Figure Sb). This behavior will stripping section. Consequently, if the bottoms flowrate
have consequences on both the reflux ratio and the rest of related is constant, the vapor flowrate in the stripping section will
variables required for obtaining the specified separation. Ac- be increased.
cording to Figure Sb, in this specific case, the feed temperature
Similarly, when the feed is a saturated or superheated
vapor at higher temperature than that of the feed stage, it
mixes with the vapor coming from the stripping section
23 -
and increases the vapor flowrate in the rectifying section
21 because of the flow of vapor feed and the vaporization
19 of a part of the liquid. Thus, if the distillate flowrate is
17 constant, the rectifying section will show higher liquid
.,._ 15 flowrates, resulting in higher reflux ratios. It is important
t 13 - -«--q<O
to highlight that this behavior is not only related to the VLE
~ 11 _,,_q=O characteristics of the feed stream but also to the difference
Iii 9 *q =0.34 between the temperatures of the feed and of the feed stage.
"' 7
:;; -+-q = 1 The calculated diameters of each section fit with this be-
5 *q>1 havior, and the diameters in the stripping section increase
3 when the feed temperature decreases.
1
290 310 330 350 370 390 410 430 A comparison of the effect of the feed phase condition
T(K) based on the condenser and reboiler duties or on the col-
umn costs should include the energetic or economic costs
(a)
involved in heating or cooling the feed stream to specified
thermal condition. As the feed temperature increases, i.e.
as the enthalpy of the feed stream increases, the energetic
2,0
requirements at the reboiler decrease because part of the
energy required to accomplish the separation is supplied
by the feed stream. At the same time, the condenser duty
increases because the column temperature has increased
but the temperature of the distillate is the same - it has the
same composition at the same pressure. Nevertheless, the
energy savings at the reboiler would be misleading if the
feed stream must be previously heated.
The results obtained reflect that the analysis of the influ-
ence of the feed phase condition for a column where N and
-2.0
NF are specified depend more on the difference between the
1 3 5 7 9 11 13 ~ 17 ~ ll ll
Stage number feed and column temperatures than on the actual phases of
the feed stream. One important effect is related to the flow-
(b)
rates circulating through the column. These flowrates affect
the column diameters and, consequently, the column's fixed
Figure 5. Temperature profiles (a) and key plots (b) correspond- costs. Nevertheless, the total column costs are frequently
ing to a column with N = 24 and NF = 13 ,fed with the mixture dominated by the operating costs and, moreover, the influ-
corresponding to Case 1 with the different phase conditions ence of the diameter on the column cost is usually less than
considered in Table 3. that of the column height. With respect to the total column
114 Chemical Engineering Education
 whether to include the costs required to modify the feed
23 phase condition.
21 Effect of the feed phase condition on the design of the
19 column for a given separation. The preliminary design
17 -
of the columns to achieve the separation given in Table 1
15
with R/Rmin = 1.3 and considering Case 1 (Table 2) with
_g 13
the five feed phase conditions shown in Table 3 has been
5 11 ~- -+-q<O
performed applying the FUG method with the Kirkbride

"'i ~ \\~-- --k-q=O

---q =0.34 equation for the feed stage location and using the Shortcut
-+-q = 1 unit of ChemCAD. The results are shown in Table 7. As
5 ~ -><-q>1
3 .,~ expected, the five cases show the same values of Nmin and
1 ------- - --- - --- --------d---·~~~~ NFIN because the feed phase condition does not play any
0 so 100 150 200 250 role in the Fenske equation for Rmin nor in the Kirkbride
L (kmol/h) equation for NF. However, there is a slight decrease in the
required number of theoretical stages as the feed tempera-
(a)
ture increases, but this decrease is not significant, reflected
in the 2.4 theoretical stage difference between the extreme
cases SHV and SCL. The observed increase of the mini-
23
mum reflux ratio as the feed temperature increases is related
21
19 to the previously described effect of the feed stream on the
17 flowrates circulating through the column.
15
1l 13
5 11 ~q<O TABLE7
C 9 _._q=O Preliminary design of columns giving the separation
ft 7 ---q=0.34 on Table 1 with the five feed phase options on Table 3
:;; 5 ~q=1
-..-q>1 and withR/Rmin = 1.3.
3
1 -Qcondenser Qreboiler
Case Rmfu Nmfu N NF
0 so 100 150 200 250 (MJ/h) (MJ/h)
V(kmol/h) SHV 1.2836 10.7 22.9 12.2 1507.5 655.8
(b) sv 1.2538 10.7 23.0 12.2 1472.5 732.9
PV 0.9008 10.7 23.9 12.7 1058.1 1223.5
Figure 6. Liquid (a) and vapor (b) profiles corresponding to a SL 0.5892 10.7 25.3 13.4 692.5 3120.1
column with N = 24 and NF = 13 ,fed with the mixture corre-
SCL 0.5845 10.7 25.3 13.4 686.9 3234.3
sponding to Case 1 with the different phase conditions considered
in Table 3.
At this point, it is important to remember that the FUG
costs, we must consider whether the feed stream is available at method is an approximate method whose validity is re-
the specified temperature and pressure or if it has to be heated stricted to cases fulfilling the assumptions of the model,
or cooled. In that case, the cost of this energetic change must be i.e. constant molar overflow and small changes of relative
also included in the calculations. Another interesting conclusion volatilities through the column. Figure 6 shows that it
is that the increase of the reflux ratio and decrease of the reboil is reasonable to accept the hypothesis of constant molar
ratio as the feed temperature increases suggest that hotter feeds overflow. With respect to the relative volatilities, Figure
could make the separation between keys in the stripping section 7 shows the profiles of relative volatility of the light key
easier, whereas colder feeds could make the separation in the obtained for the five columns simulated with N = 24 and NF
rectifying section easier.
= 13. Table 3 shows the relative volatility of the light key
component at the five considered feed phase conditions.
When the feed temperature is below the condenser temperature, Despite the assumption of constant volatilities profiles,
the hot oil consumption in the reboiler increases dramatically, another aspect should be highlighted. The Underwood
which is a situation to be avoided. Also, when the feed is at a equations for the minimum reflux calculation[ll assumes
temperature near that of the reboiler, both the reflux ratio and that, for class 2 separations, volatilities are constant in the
the refrigerant consumption increase significantly, which is un- region between the two pinch-zones (at the rectifying and
desirable. With respect to the cost calculation, we must consider
Vol. 55, No. 2, Spring 2021 115
stripping sections) and would be adequately represented by In Cases 2 and 3 the same flowrates appearing in Table 1
the relative volatilities at the feed thermal condition. As are used. The feed component flowrates for Cases 4 and 5
shown in Figure 7, cases SL and SCL do not fulfill this condi- are also the same as in Table 1; however, the distillate and
tion and, therefore, the results obtained with the Shortcut tool bottoms compositions have been established on a preliminary
should be carefully analyzed. The instructor can take advan- mass balance based on the same recovery of key components
tage of these situations in order to emphasize the importance as in the other cases: 98 .40 % of light key in the distillate and
of having adequate knowledge of the characteristics of the 98 .24 % of heavy key in the bottoms. According to the VLE
equations and their application. When they are used as a part characteristics of mixtures corresponding to Cases 2 to 5, the
of a simulation package, students may tend to view them as a column pressure has been fixed at 2.17 bar in order to use cool-
black box without recognizing the underlying assumptions. ing water at the condenser and saturated steam at the reboiler.
The same analysis carried out for Case 1 in order to study
the optimal feed stage location and the effect of the thermal
23 condition of the feed stream has been performed for Cases 2
to 5. The results confirm the previously discussed comments
19
17
and conclusions. Figure 8 shows the evolution with feed tem-
lii 15 perature of some interesting operating variables for the five
e:s
13 considered cases. As expected, the heat exchange area of the
.C

:' 9
11 reboiler (Areboile,) decreases as the feed temperature increases,
as shown in Figure 8.a. The largest drop in heat exchange area
ti 7
appears when the feed changes from liquid to vapor. Case 5
5
shows the more pronounced slope, i.e. the highest difference
3
1 between the extreme values, thus suggesting that this behavior
1.5 1.7 1.9 2.1 2.5 2.7 2.9 could be related to the type of separation that, in this case,
requires obtaining a bottoms product that is nearly pure HK.
Consequently, despite the fact that the bottoms temperature of
Figure 7. Light key relative volatility profiles for columns Cases 2 and 5 is almost the same, the separation is not. Case 5
with N = 24 and NF = 13 ,fed with the five mixtures described requires a significantly higher reboil ratio than Case 2, resulting
in Table 3. in higher reboiler area. The required flowrate of heating agent
(hot oil in Case 1 and steam at different pressures in the other
Can the previously observed behavior be generalized to cases), shown in Figure 8b (mhot oil and ms,eam• respectively)
other mixtures? The answer to this question is not easy follows the same trend. Case 1 is the only exception, and it
because it requires an extensive study of mixtures with dif- shows the previously discussed dramatic increase of heating
ferent VLE properties being separated over a wide range of agent when the feed temperature is significantly lower than the
operational variables. In this work the proposed study has distillate temperature. Likewise, the heat exchange area of the
been extended to other hydrocarbon mixtures covering dif- condenser increases as the temperature increases, with a pat-
ferent VLE temperature ranges and showing different type tern of variation similar to that at the reboiler. As Figures 8c
of separations, shown in Tables 2 and 3. Thus, the following and 8d show, there is a substantial difference between the areas
situations have been considered: obtained for liquid and vapor feeds. The highest differences
• Case 1: The bubble point of the feed is significantly correspond to Case 1, with the highest difference between the
lower than the distillate temperature feed bubble and dew point temperatures, and Case 4, yielding
nearly LK pure distillate. Therefore, it seems that, besides
• Case 2: The dew point of the feed is higher than the
the expected variation of the heat exchange areas, the type of
bottoms temperature
separation is also a factor, and the reboiler and the condenser
• Case 3: The bubble and dew point of the feed are within areas increase when the composition of HK and LK, respec-
the range of the column temperatures tively, approach nearly pure streams. Finally, Figure 8e shows
Moreover, the separation specified in Table 1 could be that the reflux ratio (R) increases when the feed temperature
considered relatively close to a symmetrical separation. increases, which is the biggest difference observed for Case
The study also includes: 4. This variation agrees with the corresponding evolution of
the Rmin values. It not only depends on the effect of the feed
• Case 4: The light key component is the most volatile
stream on the vapor and liquid flows inside the column, but
in the feed mixture
also the ease of separation of the key component, i.e. their
• Case 5: The heavy key component is the least volatile relative volatilities, the feed composition, and the specified
in the feed mixture recoveries or concentrations at the distillate and bottoms.

116 Chemical Engineering Education

 140

120
...case 1
--case 2
3500

3000
--------
-----~~---~-~ l
...-case2 160000
+-Case3
100
-+-Case4 q: 2500 ---case 3
"3' -+-Case 4 120000 :.,,
80 +Cases
~ 2000 +-Case 5
J!
r:.-
..s. 60 i 1500 ...Case 1
80000 I
-~
0 ~

~
~ 1000 'ii

J: le SOO
40000 {

0 ei. 0 0
200 250 300 350 400 450 SOO 550 200 250 300 350 400 450 SOO 550
Feed temperature (K) Feed temperature (K)

(a) (b)

90 90000 40
80
-case 1 q: 80000
"3' ---case 2
35
70 ....case2 tti' 70000 ;::;-
r:;- 60
---case3
--ease4
,..;
ill 60000
---case 3
....-Case4
-,,..Cases
30
25~
."'CU

~so --case 5 "'

_[ 50000
II -Casel
20~
} 40 ~ 40000 :!,.
1,
8
'<t 30 """" 30000
-:
15 ;
.!!P

~
!
20
r
i20000 I 10
J
10 § 10000 I 5
0
e 0
L________ - -·-- ·-- - - - - 0
200 250 300 350 400 450 500 550 200 250 300 350 400 450 SOO 550
Feed temperature (K) Feed temperature (K)

(c) (d)

7 -case 1
-,,...case 2
6 ---case 3
--case4
5 ..... case 5
a: 4
3

1
0
200 250 300 350 400 450 SOO
Feed temperature (K)

(e)

Figure 8. (a) Reboiler heat exchange area, (b) requiredfiowrate of heating agent, (c) condenser heat exchange area, (d) required
cooling agentfiowrate, and (e) reflux ratio versus feed temperature for [Link] cases considered in this section.

Vol. 55, No. 2, Spring 2021 117

CONCLUSIONS variables. The approximate Kirkbride and Fenske equations
do not give sufficiently good results, and their use should be
In this work, a problem statement focusing on the behavior limited to preliminary design steps or hand calculations. The
of a multicomponent distillation column has been used as a results obtained reflect that small changes of the feed stage
starting point in order to perform a deep study of the differ- location around the optimum value do not have a noticeable
ent procedures to select the optimal feed stage location and effect on the column behavior. Therefore, the deviations in-
thermal feed phase. This problem allows us to teach students volved in the use of the approximate equations are acceptable.
the importance of knowing and comparing the different ap- With respect to the study of the influence of the feed phase
proaches to solving a problem and developing their own condition on the operation of a distillation column, situations
conclusions. The discussion of the results emphasizes the in which the feed temperature is not within column tempera-
importance and usefulness of the knowledge of the behavior ture range should be avoided because it results in an unde-
of binary distillation, deeply studied in previous courses, for &irable increase of some operating variables. The observed
interpreting and understanding multicomponent separations. increase of the reflux ratio and decrease of the reboil ratio
The application of approximate and rigorous methods in the as the feed temperature increases allow us to conclude that
design of distillation columns also highlights the importance hotter feeds could make the separation between keys in the
of not using the simulation packages as a black box. It is stripping section easier, whereas colder feeds could make the
critical to check if the assumptions of approximate methods separation in the rectifying section easier. Besides the influ-
are fulfilled and to know the available tools to make use of ence of the feed phase condition on the operating variables,
the column profiles obtained by rigorous methods. the type of separation also affects the behavior of the column,
The analysis performed is based on the assumption that the and the areas of the reboiler and the condenser increase when
true optimal feed stage location is the one corresponding with the required contents of HK and LK, respectively, approach
the lowest column costs. This study indicates that this situ- nearly pure streams.
ation is reached when several variables - such as the reflux
flowrate, reflux ratio, reboiler or condenser duties, reboil ratio,
among others - reach their minimum values. Consequently,
REFERENCES
the true optimal feed stage can also be considered as the one
1. Seader JD, Henley EJ, and Roper DK (2011) Separation Process
giving the lowest value of some of these variables. According Principles. Chemical and Biochemical Operations. John Wiley &
to this criterion, seeking the optimal feed stage requires the Sons, Inc. Hoboken, NJ.
simulation of several columns with different feed stages. This 2. Chemstations Inc, [Link] Accessed De-
cember 6, 2020.
could be time consuming, especially in cases with conver-
3. Aspentech, [Link] .[Link]/ .Accessed December 6, 2020
gence problems or with a high number of stages. Neverthe- 4. Turton R, Bailie RC, Whiting WB, Shaeiwitz JA, and Bhattacharyya
less, it should be considered that other methods, such as the D (2013) Analysis, Synthesis, and Design of Chemical Processes. 4th
key ratio comparison, the key ratio plots, or the Hengstebeck edition. Pearson. Hoboken, New Jersey.
5. Ulrich GD and Vasudevan PT (2006) How to estimate utility costs.
method, are also based on rigorous simulation results but do
Chemical Engineering. April: 66-69. [Link]
not give the best result; moreover, they require additional corn/how-to-estimate utility costs/.Accessed December 6, 2020.
calculations and/or graphical representations. Therefore, if 6. Kister HZ (1992) Distillation Design. McGraw Hill. Boston, MA.
a simulation package is available and the column simulation 7. King CJ (1980) Separation Processes. 2nd edition. McGraw Hill.
New York, NY.
does not show convergence problems, the optimal feed stage 8. Sinnott Rand Towler G (2012) Diseflo en lngenierfa Qufmica. Serie
can be obtained from a sensitivity study where the feed stage de lngenierfa Qufmica Coulson & Richardson. Reverte. Barcelona,
location is selected as the independent variable and the depen- Spain. □
dent variable is selected from among of the above-mentioned

118 Chemical Engineering Education