Brazilian Journal

of Chemical ISSN 0104-6632

Printed in Brazil
Engineering www.abeq.org.br/bjche

Vol. 32, No. 01, pp. 283 - 291, January - March, 2015
dx.doi.org/10.1590/0104-6632.20150321s00002825

MODELING AND SIMULATION OF A BENZENE

RECOVERY PROCESS BY EXTRACTIVE
DISTILLATION
L. B. Brondani*, G. B. Flores and R. P. Soares*
Universidade Federal do Rio Grande do Sul, (UFRGS), Departamento de Engenharia Química,
R. Eng. Luis Englert, s/n. Campus Central, CEP: 90040-040, Porto Alegre - RS, Brazil.
Phone: + (55) (51) 3308 3528, Fax: + (55) (51) 3308 3277
E-mail: {lauren, rafael}@enq.ufrgs.br

(Submitted: July 10, 2013 ; Revised: March 20, 2013 ; Accepted: March 27, 2014)

Abstract - Extractive distillation processes with N-formylmorpholine (NFM) are used industrially to separate
benzene from six carbon non-aromatics. In the process studied in this work, the stream of interest consists of
nearly 20 different hydrocarbons. A new set of NRTL parameters was correlated based on literature
experimental data. Both vapor-liquid equilibrium as well as infinite dilution activity coefficient data were
taken into account; missing parameters were estimated with the UNIFAC group contribution model. The
extractive distillation process was simulated using ASPEN Plus®. Very good agreement with plant data was
obtained. The influences of the main operational parameters, solvent to feed ratio and solvent temperature,
were studied. Theoretical optimum operating values were obtained and can be implemented to improve the
industrial process. Extreme static sensitivity with respect to reboiler heat was observed, indicating that this
can be the source of instabilities.
Keywords: Extractive distillation; N-formylmorpholine; NFM; Process simulation; Benzene.

INTRODUCTION and non-aromatics with the same carbon number,

conventional distillation processes are impractical
Benzene has being widely used in the chemical and extractive distillation processes become eco-
industry as a raw material for innumerous com- nomically interesting.
pounds. Due to its high toxicity, it is being replaced According to Emmrich et al. (2001), Krupp Uhde
by other hydrocarbons with similar characteristics. originally developed the Morphylane process (ex-
This movement is also driven by stricter regulations. tractive distillation process) for the recovery of high-
For instance, in Brazil, the commercialization of purity benzene from hydro-refined coke oven ben-
finished products containing more than 0.1% ben- zole, the reason being that the existing liquid-liquid
zene by volume in their composition is not permit- methods of extraction were unable to process coke
ted. However, benzene is still widely employed in, oven benzole due to its high aromatic content. The
for example, styrene polymers; phenol for resins and liquid-liquid extraction techniques exploit the differ-
adhesives (via cumene); and cyclohexane, used in ent solubilities of aromatics and non-aromatics in a
nylon manufacture. polar solvent.
This motivates the petrochemical industry to better Different solvents can be used for the separation
understand and control high purity separation proc- of aromatics from non-aromatics (such as sulfolane,
esses. Due to the similar boiling point of aromatics N-methylpyrollidone, diethylene glycol, etc.). In the

*To whom correspondence should be addressed

284 L. B. Brondani, G. B. Flores and R. P. Soares

process studied in this paper, NFM (N-formylmor- per column (T03) where the separation of benzene,
pholine, shown in Figure 1) is used to accomplish the at the top, from the solvent, at the bottom, occurs.
separation of benzene from a stream containing other The column studied, T02, frequently does not
hydrocarbons, mainly with six carbons. specify the desired products with the desired stabil-
ity. It should simultaneously specify both products:
the overhead product (non-aromatic C6) and the
bottom product (benzene). Therefore, the main goal
of the work was to develop a model which agrees
with plant data. A new set of NRTL parameters was
estimated based on experimental data available in the
Figure 1: NFM molecule (N-formylmorpholine). literature. The insight obtained from this model will
hopefully provide the directions for a more stable
According to ZHIGANG et al. (2005), the addi- operation.
tion of polar solvents to hydrocarbon mixtures re-
sults in increased volatilities of paraffins relative to
naphthenes, olefins, diolefins and alkynes, and in THERMODYNAMIC MODEL
increased volatilities of naphthenes relative to aro-
matics. Therefore, the addition of a polar solvent Since extractive distillation processes are strongly
reduces the effort in the separation by distillation of affected by the non-ideal interactions between the
certain mixtures which otherwise would be impracti- entrainer and the stream of interest, one of the main
cal. Thus, among other desired characteristics, its difficulties when simulating these processes is the
polarity and its high boiling point make NFM a good correct thermodynamic representation of the mixture
solvent for the required separation. behavior in contact with the solvent. Thus, most of the
Figure 2 shows a schematic drawing of the unit work was devoted to the thermodynamic modeling.
studied. First, a literature survey for phase equilibrium and
The C6C8 stream from the pyrolysis gasoline hy- infinite dilution activity coefficient (IDAC) data with
drogenation process is first fractionated into C6 and NFM and the substances involved was accomplished.
C7+ cuts in the distillation column (T01). The C6 cut Information for mixtures with NFM and the following
is sent to the extractive distillation column (T02). In compounds was found in the literature: cyclopentane
this column, NFM is added as the entrainer to extract (CP), n-hexane (n-HX), methylcyclopentane (MCP),
benzene, exiting at the bottom of the column. The benzene (BZ), cyclohexane (CH), n-heptane (n-HP)
overhead stream contains C6 non-aromatics (over- and methylcyclohexane (MCH). Table 1 shows a sum-
head product). The bottom stream is fed into a strip- mary of the collected information.

Benzene C6 non-arom
Fraction V02
Feed V03
V04
V01

Benzene
T03

P02
T02

NFM

P01 NFM + Benzene

C7+ arom

Figure 2: Benzene extractive distillation with a NFM recovery unit, schematic drawing.

Brazilian Journal of Chemical Engineering

 Modeling and Simulation of a Benzene Recovery Process by Extractive Distillation 285

Table 1: Available binary mixture data of the components of interest with NFM,
Component / NFM Liquid-liquid equilibrium Vapor-liquid equilibrium Inf. Dil. Activity Coef.
CP Ko et al., 2003 DDBST
n-HX Cincotti et al., 1999; DDBST; Krummen &
Ko et al., 2002 (b) Gmehling, 2004
MCP Ko et al., 2003 Knoop et al., 1989 and
Weidlich et al., 1987
BZ DDBST and Yin, 2010 DDBST
CH Ko et al., 2003 DDBST
n-HP Cincotti et al., 1999; DDBST; Krummen &
Qattan et al., 1995; Gmehling, 2004
MCH Ko et al., 2003 DDBST DDBST

From this information the parameters for the that used existing vapor-liquid equilibrium data to
NRTL (Non-random two-liquid) model were esti- build the thermodynamic model with the NRTL
mated and the resulting correlation compared with equation and simulate the system with Aspen Plus®.
experimental data for different temperatures. The The author used the NRTL equation and set binary
choice of this model was based on its ability to rep- parameters for NFM for the following pairs: pentane,
resent deviations from ideality at low pressure and hexane, heptane, octane, cyclopentane, cyclohexane,
because it was already used successfully in similar methylcyclopentane, methylcyclohexane, dimethyl-
works, like Ko et al., 2002 (a), Zhu et al., 2007 e cyclohexane, benzene, toluene and ethylbenzene. In
Dongchu et al., 2007. the study it was concluded that the NRTL equation
The NRTL model variant used in this paper is was efficient to simulate the extractive distillation
given by: with NFM.
In their work, Ghaee et al. (2008) used the model
∑ j x j τ jiG ji x j Gij ⎛ ∑m xm τmjGmj ⎞ proposed by Ko et al. (2002 a) to build a dynamic
ln( yi ) =
∑k xk Gki
+ ∑∑ ⎜⎜
k xk Gkj ⎝ ∑k xk Gkj
⎟⎟ (1)
⎠
simulation of an extractive distillation column for
j benzene. On the other hand, Zhu et al. (2008) pre-
dicted vapor-liquid equilibrium data of NFM with
with the other quantities given by: hydrocarbons with an improved UNIFAC group
contribution method modified by Dortmund. They
Gij = exp( −αij τij ) (2)
also found their model to be in agreement with real
industrial operation data.
bij Table 2 shows the availability of binary mixture
τij = aij + + eij ln T + f ijT (3)
T parameters for each component with NFM for the
different models.
αij = cij + dij (T − 273.15K ) (4)
Table 2: Binary mixture parameter availability for
For components with vapor-liquid equilibirum the different models.
(VLE) and/or liquid-liquid equilibrium (LLE) data
available in the literature (compounds showed in Component/ VLE and/ IDAC Ko et al., Zhu et al.,
NFM or LLE regression 2002 (a) 2008
Table 1), the NRTL parameter regression was ac- regression
complished using this equilibrium information, per- CP X X X X
formed by the built-in data regression functionality n-HX X X X X
of Aspen Plus®. For those same components Aspen MCP X X X
Plus® was used in order to obtain estimated NRTL BZ X X X X
CH X X X X
parameters using IDAC data from the literature.
n-HP X X X X
Thus, two different groups of NRTL binary interac- MCH X X X
tion parameters for each component with NFM were
estimated, one based on phase equilibrium data (re- Since the non-aromatic compounds are present in
ferred as VLE and/or LLE regression) and the other low concentration in the NFM phase, the agreement
based on IDAC (referred as IDAC regression). Fur- with experimental data for the activity coefficients at
ther, the NRTL parametrizations from KO et al., infinite dilution is preferred. Thus, the IDAC regres-
2002 (a) and ZHU et al., 2008 were also compared sion parameters were used to assemble the model
with the collected experimental data. proposed in this paper for NFM/non-aromatics. For
Ko et al. (2002 a) was the only reference found benzene, which is found in higher concentration in
Brazilian Journal of Chemical Engineering Vol. 32, No. 01, pp. 283 - 291, January - March, 2015
286 L. B. Brondani, G. B. Flores and R. P. Soares

the NFM-rich liquid phase, the selected model was The overall performance of the model developed
the one estimated from experimental vapor-liquid in this work, compared with experimental IDAC data,
equilibrium data. is shown in Figure 7, while the resulting binary
As expected, in all cases, the model that best fit parameter values are given in Table 3.
the experimental IDAC values was the one estimated Table 3 presents the results of chosen the model
with the same IDAC data. Figures 3, 4 and 5 show for the pairs of components with experimental data.
the model verification with IDAC data for the inter- For benzene the following parameters were also
action of n-heptane, methylcyclohexane and benzene considered: dij = -0.11, eij = -2.42 e eji = 1.666.
with NFM, respectively. They were built by com- The UNIFAC model was used to estimate NRTL
paring the experimental IDAC values for different binary parameters for the components without equi-
temperatures with the ones calculated by the NRTL librium data available with the solvent NFM. These
equation using the available binary parameters of compounds are: 2,3-dimethylbutane (2,3DBM), 2-
each model. Thus, the closer the points are to the methylpentane (2MP), 3-methylpentane (3MP), 2,4-
line, the better that model represents the experimental dimethylpentane (2,4DMP), 2-methylhexane (2MH),
data. 2,3-dimethylpentane (2,3DMP), 1,1-dimethylcyclopen-
Although, naturally, the model estimated from tane (1,1DMCP), 3-methylhexane (3MH), cis-1,3-
IDAC experimental values is superior when com- dimethylcyclopentane (c1,3DCMP), trans-1,3-dimethyl-
pared with the same IDAC data, for benzene the cyclopentane (t1,3DMCP) and trans-1,2-dimethylcy-
parameters obtained from vapor-liquid equilibrium clopentane (t1,2DMCP). According to the laboratory
data were selected for the simulation. As mentioned analyses, they represent about 6wt% of a typical feed
before, this is mainly because benzene is far from for the process under study. In order to increase the
infinite dilution in the process studied. In this situa- reliability of the estimated parameters, the new
tion, the model built from the vapor-liquid equilib- UNIFAC model interaction parameters for -CH2 and
rium experimental data was assumed to be more ACH groups with NFM, proposed by Zhao et al.
appropriate for the final application. The VLE results (2001) were used. The results obtained from UNIFAC
for this mixture can be seen in Figure 6. are shown in Table 4.

5 5

4
ln (γ∞ experimental)

4
ln (γ∞ experimental)

3 3
2
2
1
1
0
0 1 2 3 4 5 0
ln (γ∞ calculated) 0 1 2 3 4 5
ln (γ∞ calculated)
Ko et al., 2002 model Reference
IDAC regression model LLE regression model LLE / VLE regression model Reference
Zhu et al., 2008 model IDAC regression model Zhu et al., 2008 model

Figure 3: Model verification with IDAC experi- Figure 4: Model verification with IDAC experi-
mental data for the interaction of n-heptane with mental data for the interaction of methylcy-
NFM. clohexane with NFM.
1.4 540
520
1.2
ln (γ∞ experimental)

500
Temperature (K)

1 480
0.8 460
440
0.6 420
0.4 400
380
0.2
360
0 340
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0.0 0.2 0.4 0.6 0.8 1.0
ln (γ∞ calculated) Benzene molar concentration in NFM
Ko et al., 2002 model Reference X1 - Experimental Y1 - Experimental
IDAC regression model (this work) VLE regression model (this work) x1 - VLE regression y1 - VLE regression
Zhu et al., 2008 model x1 - IDAC regression y1 - IDAC regression

Figure 5: Model verification with IDAC experi- Figure 6: Comparison of the model generated
mental data for the interaction of benzene with from VLE data and from IDAC data for the
NFM. benzene/NFM mixture.
Brazilian Journal of Chemical Engineering
 Modeling and Simulation of a Benzene Recovery Process by Extractive Distillation 287

ln (γ∞ experimental)
3

0
0 1 2 3 4 5
ln (γ∞ calculated)

MCP n-HX CH n-HP BZ CP MCH Reference

Figure 7: Comparison between gamma infinity data

calculated by the model with experimental data.

Table 3: NRTL binary interaction parameters adjusted in this work.

Comp. i CP n-HX MCP BZ CH n-HP MCH

Comp. j NFM NFM NFM NFM NFM NFM NFM
aij 0.791 0.16 0.543 17.09 0.827 0.069 0.501
aji -1.51 -2.13 -1.35 -12.3 -2.29 -2.39 -2.32
bij 891.5 1006 937.0 -1000 847.9 1036 922.8
bji 863.3 1380 932.1 952.7 1235 1563 1350
cij 0.300 0.300 0.300 0.300 0.300 0.300 0.300

Table 4: NRTL binary parameters estimated by UNIFAC.

Comp. i 2,3DMB 2MP 3MP 2,4DMP 2MH 2,3DMP 1,1DMCP 3MH c1,3DMCP t1,3DMCP t1,2DMCP
Comp. j NFM NFM NFM NFM NFM NFM NFM NFM NFM NFM NFM
bij 1101 1101 1101 1070 1070 1070 1101 1070 1109 1109 1109
bji 520 521 521 639 640 639 522 640 507 507 507
cij 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3

PROCESS MODELING AND SIMULATION Theoretical Plate) given by the packing vendor. Table
5 shows the Radfrac simulation specifications.
With the thermodynamic model parameters set, a
simulation in the commercial simulator Aspen Plus®
was assembled for the extractive distillation column,
as schematically shown in Figure 8. V03
The extractive distillation column simulation was Solvent
carried out using the Radfrac model. This is a rigor-
ous model for simulating all types of multistage va- C6 non-aromatics
por-liquid fractionation operations. These operations
include: ordinary distillation, absorption, reboiled Feed
absorption, stripping, reboiled stripping and extrac- 1
2
tive and azeotropic distillation. The model assumes
32
thermodynamic equilibrium and rigorous mass and 33
energy balances in all stages. It was considered that
the extractive distillation column consisted of 71
stages. Of these, 33 refer to the trays section, 12 are
the equivalent stages of the random packing section
above the solvent feed, 24 are the equivalent stage of
the structured packing section between the solvent
Solvent + Benzene
and hydrocarbon feeds, one is the condenser and the
other is the reboiler. The equivalent stages were cal- Figure 8: Schematic drawing to the extractive distil-
culated from the HETP (Height Equivalent to a lation column.

Brazilian Journal of Chemical Engineering Vol. 32, No. 01, pp. 283 - 291, January - March, 2015
288 L. B. Brondani, G. B. Flores and R. P. Soares

Table 5: Main specifications for the extractive dis- tion. The simulation results were almost insensitive
tillation column simulated with the Radfrac model. with respect to the efficiency in the other packing
sections and trays. Further, this small sensitivity
Configuration Number of stages 71 could not be verified with the currently measured
Condenser Total
Reboiler Thermosiphon
data, due to the small number of temperature sensors
Valid phases VLL available (see Figure 9). Thus, the following effi-
Convergence Strongly non- ciency values were assumed: randow packing from
ideal liquid stages 2 to 13, 100%; structured packing from stages
Streams Solvent Above-stage 14 26 to 37, 35%; and tray section, 80%.
Feed Above-stage 38 Table 6 shows one of the unit feedstock compo-
Pressure (barg) Top stage pressure 1.05
2-13 section ∆P 0.02
sitions studied (feed composition used in the simula-
14-37 section ∆P 0.04 tion was the same as that analyzed in the laboratory),
38-70 section ∆P 0.13 the top composition sampled during operation, the
Bottom stage 1.24 composition obtained with the simulation and the
pressure absolute error (value measured – calculated value).
Condenser (°C) Both reflux and 51.5 As can be seen in Table 6, the simulation results
liquid distillate are
sub cooled agree very well with plant data. The bigger compo-
Reboiler (°C) Temperature 32 sition errors were in the n-hexane and methylcy-
change clopentane values. These components represent 15
and 45% of the stream, respectively, and an error of
Another typical assumption in this work is adia- approximately 0.2 percent cannot be considered sig-
batic column operation, i.e., the metal thermal ca- nificant. Nevertheless, the actual plant mass balance
pacitance is negligible and heat exchange occurs also contains errors, probably due to measurement
only in the condenser and reboiler. inacurracies and inadequate sampling or laboratory
analysis. No additional treatment, like data recon-
ciliation, was applied to the measured plant data.
RESULTS In Figure 9 it is possible to see the temperature
and pressure profiles obtained and compare with the
The proposed plant model was adjusted to match
measured operating temperatures, the first stage be-
actual process conditions by varying the Murphree
ing the condenser and the 71st stage the reboiler. The
efficiencies. The efficiency was considered to be
solvent NFM is added in the 14th stage and the feed
constant at each packing bed and tray section. Three
is above the 38th stage.
different operating conditions were considered to
As NFM is not fed into the top of the column,
determine the average efficiencies in the different
zones. This step was accomplished manually by stages 2 to 13 only have the purpose of not letting
comparing the plant temperature profile and the the overhead product be contaminated with NFM,
simulated values. The result was satisfactory; how- and therefore this region has practically constant
ever, some discrepancy was observed in the bottom temperature. The main influence of NFM on benzene
temperature. occurs just below stage 14, where it is possible to
This difference is probably due to a small lack of notice a large increase in the temperature, even with
fit of the benzene and NFM mixture vapor pressure. the assumption of adiabatic operation. Note that this
Since the pressure is specified in the model, discrep- temperature profile is quite different from a conven-
ancies in the temperature prediction will directly tional distillation process, with it being possible to
impact the model response. Another possible reason observe higher temperatures in intermediate regions
for the temperature difference found in the bottom of of the column. This is due to the phase changes pro-
the column is the feed composition used. Laboratory moted by the entrainer, making almost all of the
analyses are not reliable for the heaviest components benzene coming from the bottom vapor be trans-
in low concentration. Thus, the model included only ferred preferably to the liquid between stages 14 and
components with up to seven carbons. However, this 37. In the 38th stage a slight temperature decrease can
discrepancy affects only the bottom temperature be seen, due to a colder feed inlet. Between stages 39
profile and should not be considered relevant for the and 70, we have a flatter profile because, in this re-
benzene/non-aromatics separation, which is the focus gion, we have a low benzene concentration in the
of this work. vapor phase and not a lot of components are chang-
In the absorption section (between stages 14 and ing phases. Because the bottom mixture is mainly
25), a small efficiency was found (aprox. 35%), indi- composed of solvent with benzene, the bottom sec-
cating a poor liquid and/or vapor internal distribu- tion of the column has its boiling temperature.

Brazilian Journal of Chemical Engineering

 Modeling and Simulation of a Benzene Recovery Process by Extractive Distillation 289

Table 6: Feed composition, top composition and error found between simulation and sampled plant data
(laboratory analyses).

Components Feed mass Top mass Top mass Top composition Bottom mass
composition in % composition (%) composition (%) - absolute error fraction - model
(plant = model) – plant model (%)
Cyclopentene 0.11 0.41 0.37 0.03 0.00
2,3-dimethylbutane 0.14 0.44 0.49 -0.05 0.00
2-methylpentane 1.84 6.31 6.33 0.02 0.00
3-methylpentane 1.36 4.66 4.66 0.00 0.00
n-hexane 4.41 15.28 15.13 0.16 0.00
Methylcyclopentane 13.26 45.34 45.51 -0.17 0.00
2,4-dimethylpentane 0.10 0.33 0.33 0.01 0.00
Benzene 71.87 3.50 3.49 0.01 10.54
Cyclohexane 3.12 10.62 10.71 -0.08 0.00
2-methylhexane 0.58 2.05 1.99 0.06 0.00
2,3-dimethylpentane 0.05 0.18 0.18 0.00 0.00
1,1-dimethylcyclopentane 0.24 0.83 0.82 0.01 0.00
3-methylhexane 0.47 1.66 1.63 0.04 0.00
cis-1,3-dimethylcyclopentane 0.66 2.29 2.27 0.02 0.00
trans-1,3-dimethylcyclopentane 0.47 1.63 1.61 0.01 0.00
trans-1,2-dimethylcyclopentane 0.65 2.24 2.23 0.01 0.00
n-heptane 0.54 1.91 1.85 0.06 0.00
methylcyclohexane 0.12 0.40 0.41 -0.01 0.00
NFM 0.00 0.00 0.00 0.00 89.46

Figure 9: Extractive distillation column temperature

and pressure profiles.

With the plant model adjusted, it was possible to 0.21 GJ/h of reboiling heat. Considering only the
study the column behavior by performing sensitivity exchange of latent heat from a steam of 18barg
analysis. The temperature profile of this column is very (1859 kJ/kg), this corresponds to a variation of only
sensitive to the heat supplied. Along with other ma- 112 kg/h of steam. Thus, it is clear that small varia-
nipulated variables it determines the region where tions in the reboiler heat are enough to bring the
benzene absorption will occur. Figure 10 shows the products out of specification. It can also be noted
static temperature profile change when the reboiler that the major temperature variations for each re-
heat is varied. In these simulations the feed rates and boiling heat are between stages 14 and 70. This hap-
reflux ratio were kept constant. It should be noted pens because above stage 14, no solvent is present
that the simulation results of Figure 10 differ by only and the temperature is controlled by the non-aromatic
Brazilian Journal of Chemical Engineering Vol. 32, No. 01, pp. 283 - 291, January - March, 2015
290 L. B. Brondani, G. B. Flores and R. P. Soares

mixture boiling point. On the other hand, the bottom From this analysis it can be seen that there is an
temperature is related to the boiling temperature of optimal solvent-to-feed ratio for proper column op-
the solvent/benzene mixture. Hence, with reboiling eration. Above this value the benzene recovery does
heating variation, we are actually dislocating the not improve substantially, only operational costs are
stage where the benzene will change phase and, con- increased. The same applies to solvent temperature.
sequently, modifying the temperature profile mainly Temperatures below 115 °C, under the same reflux
in the rectifying section. ratio and reboiler heat, are recommended to obtain a
In Figures 11 and 12 simulation results for benzene better benzene recovery.
recovery (defined as benzene at the bottom / benzene
at the feed) with respect to the solvent-to-feed ratio
and solvent temperature are shown, respectively.
CONCLUSIONS
1 90
1 80
1 70
In the present work an industrial system to sepa-
Temperature (°C)

1 60
1 50 rate benzene from a non-aromatic stream through
1 40
1 30
extractive distillation with NFM as the solvent was
1 20 studied. First, literature data were collected and dif-
110
1 00
ferent sets of parameters for the NRTL model were
90 compared. For many pairs of substances a need to re-
1 4 7 1 0 1 3 1 6 1 9 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70
Stages
estimate the NRTL model parameters was identified.
For the substance pairs with no experimental data
1 5,1  GJ/h 1 5,5 GJ/h 1 5,9 GJ/h 1 6,3 GJ/h 1 6,7 GJ/h
available in the literature, the UNIFAC group contri-
Figure 10: Effect of reboiler heat on the temperature bution model was used to estimate the missing pa-
profile. rameters. As a result of this step a new set of NRTL
parameters was obtained for processes involving
1 00%
NFM, benzene and other non-aromatics.
95%
90%
Using the proposed NRTL parameter set, a plant
85% model was assembled within ASPEN Plus®. Murphree
Benzene Recovery

80% efficiencies were then adjusted for better agreement

75%
70%
with plant data using three different operating points.
65% The model proposed represented satisfactorily the
60% operation of the extractive distillation column and
55%
can be used to study the behavior of most compo-
50%
nents present in the column feed. From the model
5.0
4.4

5.4
4.8
4.9

5.6

5.8
5.9

6.6

6.8
4.3

5.2
5.3

6.2
6.3
4.5

6.5
4.7

5.7

6.7
6.1

Solvent / feed ratio analysis it is possible to demonstrate that a great

difficulty of the extractive distillation column studied
Figure 11: Effect of the solvent-to-feed ratio on the
is its extreme static sensitivity to variations of the
column benzene recovery.
supplied heat. The operational challenge is to control
1 00% the contact point of benzene with NFM within a
95% region that allows the simultaneous specification of
90%
the overhead product (non-aromatic C6) and bottom
85%
Benzene recovery

80% product (benzene). This behavior should be further

75% investigated so that a new control strategy can be
70%
proposed, it perhaps being necessary to add addi-
65%
60% tional thermocouples along the column.
55% Based on the sensitivity analysis of the column
50%
90 95 1 00 1 05 1 1 0 1 1 5 1 20 1 25 1 30 1 35 1 40 1 45
and the optimal operation point, it is possible to im-
Solvent temperature (°C) plement improvements in the operation site. How-
ever, in order to have a complete mapping of the
Figure 12: Effect of the solvent temperature on the possible operating problems, additional hydraulic
benzene recovery. analysis and internals verification are desirable.

Brazilian Journal of Chemical Engineering

 Modeling and Simulation of a Benzene Recovery Process by Extractive Distillation 291

NOMENCLATURE 926 (2002)b.

Ko, M., Na, S., Lee, S. and Kim, H., Liquid-liquid
aij , b ij , c ij non-symmetrical constants [-] equilibria for the binary systems of N-formylmor-
pholine with cycloalkanes. J. Chem. Eng. Data,
and f ij
48, p. 249-252 (2003).
xi component i mole fraction in [-] Ko, M., Na, S., Kwon, S., Lee, S. and Kim, H.,
the mixture Liquid-liquid equilibria for the binary systems of
T absolute temperature [K] N-formylmorpholine with branched cycloalkanes.
ΔP pressure drop [kgf/cm2g] J. Chem. Eng. Data, 48, p. 699-702 (2003).
yi activity coefficient of a [-] Krummen, M. and Gmehling, J., Measurement of
component activity coefficients at infinite dilution in N-
methyl-2-pyrrolidone and N-formylmorpholine
and their mixtures with water using the dilutor
REFERENCES technique. Fluid Phase Equilibria, 215, p. 283-294
(2004).
Cincotti, A., Murru, M., Cao, G., Marongiu, B., Qattan, M., Al-Sahhaf, T. and Fahim, M., Liquid-
Masia, F. and Sannia, M., Liquid-liquid equilibria liquid equilibria in some binary and ternary mix-
of hydrocarbons with N-formylmorpholine, J. tures with N-formylmorpholine. J. Chem. Eng.
Chem. Eng. Data, 44, p. 480-483 (1999). Data, 40, p. 88-90 (1995).
Dortmund Data Bank, DDBST GmbH, Available at Weidlich, U., Roehm, H. and Gmehling, J., Activity
http://www.ddbst.com/ (Acessed on September, coefficients at infinite dilution (NFM - Methylcy-
2011). clopentane). J. Chem. Eng. Data, 32(4), p. 450-
Dongchu, C., Hongqi, Y. and Hao, W., Liquid-liquid 453 (1987).
equilibria of methylcyclohexane-benzene-N-for- Yin, W., Ding, S., Xia, S., Ma, P., Huang, X. and
mylmorpholyne at several temperatures. Fluid Zhu, Z., Cosolvent selection for benzene–cyclo-
Phase Equilibria, 255, p. 115-120 (2007). hexane separation in extractive distillation. J.
Emmrich, G., Gehrke, H. and Ranke, U., Working Chem Eng. Data, 55, p. 3274-3277 (2010).
with an extractive distillation process. Digital- Zhao, Q., Zheng, Y. and Zhao, W., New UNIFAC
refining, p. 1-9 (2001). group-interaction parameters of N-formylmor-
Gahee, A., Sotude-Gharebagh, R. and Mostoufi, R., pholine. Supplied by The British Library (2001).
Dynamic optimization of the benzene extractive Zhingag, L., Biaohuca, C. and Zhongwei, D., Special
distillation unit. Brazilian Journal of Chemical Distillation Processes. Elsevier, Amsterdam (2005).
Engineering, 25(4), p. 765-776 (2008). Zhu, H., Shi, X. and Zhou, W., Process simulation
Knoop, C., Tiegs, D. and Gmehling, J., Activity and parameter optimization of separating aromat-
coefficients at infinite dilution (NFM - Methyl- ics and non-aromatics by extractive distillation
cyclopentane). J. Chem. Eng. Data, 34(2), p. 240- with N-formylmorpholine. Supplied by The British
247 (1989). Library (2007).
Ko, M., Na, S., Cho, J. and Kim, H., Simulation of Zhu, H., Shi, X. and Zhou, W., Process Simulation
the recovery process by extractive distillation. Ko- and parameter optimization of separating aromat-
rean J. Chem. Eng., 19(6), p. 996-1000 (2002)a. ics and non-aromatics by extractive distillation
Ko, M., Lee, S., Cho, J. and Kim, H., Liquid-liquid with N-formylmorpholine. Journal of East China
equilibria for binary systems containing N-for- University of Science and Technology (Natural
mylmorpholine. J. Chem. Eng. Data, 47, p. 923- Science Edition), 34, p. 309-314 (2008).

Brazilian Journal of Chemical Engineering Vol. 32, No. 01, pp. 283 - 291, January - March, 2015