Accepted Manuscript

Title: A New Sulfolane Aromatic Extractive Distillation

Process and Optimization for Better Energy Utilization

Authors: Qin Wang, Jing Y. Chen, Ming Pan, Chang He,

Chang C. He, Bing J. Zhang, Qing L. Chen

PII: S0255-2701(17)31059-0
DOI: https://doi.org/10.1016/j.cep.2018.04.011
Reference: CEP 7253

To appear in: Chemical Engineering and Processing

Received date: 21-10-2017

Revised date: 20-3-2018
Accepted date: 10-4-2018

Please cite this article as: Wang Q, Chen JY, Pan M, He C, He CC, Zhang BJ, Chen QL,
A New Sulfolane Aromatic Extractive Distillation Process and Optimization for Better
Energy Utilization, Chemical Engineering and Processing - Process Intensification
(2010), https://doi.org/10.1016/j.cep.2018.04.011

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 A New Sulfolane Aromatic Extractive Distillation Process
and Optimization for Better Energy Utilization

Qin Wang, Jing Y. Chen, Ming Pan, Chang He, Chang C. He, Bing J. Zhang*, Qing L. Chen*

T
School of Chemical Engineering and Technology, Guangdong Engineering Centre for Petrochemical Energy Conservation, Sun Yat-Sen

IP
University, No. 135, Xingang West Road, Guangzhou, 510275, China

R
SC
Graphical Abstract

Graphical Abstract

FS3=13.106 ×103 kg/h U

N
xA in NA 1w.%
E3 FS6=12.726 ×103 kg/h
D1
xA,S6 1w.% Nonaromatic
Nonaromatic RT2=0.5 S6
A
23.95 kg/h product
rectifying column NS3=35 FS7=356.05 kg/h
T2 QR, T2=3.07 GJ/h xH2O,S7 99.9w.%
Extractive S3 S7 Dirty water
crude feed TS2=117.8 °C distillation column E4
M

S1 S2 NS16=8 E5
T1 NS2=42 S4 FS8=43.035 ×103 kg/h
22.72 w.% A E1 FS13=36.604 ×103 kg/h D2
77.28 w.% NA QR, T1=49.47 GJ/h x =6.3 w.% xNA, S8 1500ppm Aromatic
3 S13 S,S13 S8 product
55.716 ×10 kg/h
E2 FS9=2139.13 kg/h
Entrainer RT3=0.3
recovery column xH2O ,S9 99.9w.% Dirty water
N =5 S9
F1 PF1=PT3,S5 T3 NS13=25
ED

S5 S14
3
FS16=60×10 kg/h FS14=285.919 ×103 kg/h QR, T3=11.21 GJ/h
xCH,S16=3.3 w.%, TS16=98 °C E10 xS,S14 =96.6 w.% FS11=2531.13 kg/h
S16 S15 S14 Stripping
S11 stream
Semi-lean solution cooler
FS12=279.88×103 kg/h E6
xCH,S12=1000 ppm, TS12=98 °C E7
S12 S10
PT

Lean solution cooler

E
CC
A
Highlights:

 A new sulfolane aromatic extractive distillation process is presented based on thermodynamic analysis

and process analysis.

 A strategy using a combination of MATLAB and Aspen Plus is proposed to optimize the significant

operating parameters.

T
 Process simulation with thermodynamic parameter modification for the aromatic ED process is carried

IP
out.

R
 The new extractive distillation process reduces the total annual cost by 13.04%.

SC
U
N
Abstract
A
Extractive distillation (ED) processes are widely used for the separation of aromatic and non-aromatic
M

hydrocarbons. Approximate boiling points of components and azeotropes in the mixture need solvents to aid the
ED

separation processes. A considerable mass flowrate and recovery of solvents in separation processes leads to

significant energy requirement. This study provides a new extractive distillation process for the efficient
PT

separation of aromatic and non-aromatic hydrocarbons aided by sulfolane solution. A flash tank and a semi-lean
E

solution stream are introduced to modify existing extractive distillation processes to reduce the reboiler heat duty
CC

of the entrainer recovery column and improve the separation performance of the extractive distillation column.

The NRTL-RK property method and rigorous unit models in Aspen Plus are used to simulate new and existing
A

processes. No-databank model parameters are regressed to improve the accuracy of simulations. A coordinative

strategy is proposed to optimize the significant operating parameters for new and existing processes by

combining Aspen Plus with MATLAB. Compared with the optimal existing extractive distillation process, the
new extractive distillation process reduces the operating costs by 8.9% when heat integration is not considered,

and 13.04% of the total annual cost can be reduced when a heat exchanger network is considered.

Keywords: Extractive distillation; Sulfolane; Process simulation; Optimization; Energy saving

Nomenclature

T
IP
𝐺𝑖𝑗 coefficient characterizing interactions of components i and j

R
N number of components

SC
T temperature

𝑥𝑗 overall mole fraction of substance 𝑗

𝑅 U
N
ideal gas constant
A
Greek Letters
M

𝛾 activity coefficient
ED

𝛼𝑖𝑗 non-randomness interaction parameter for components i and j

𝜏𝑗𝑖 dimensionless interaction parameter

PT

Subscripts
E

𝑖, 𝑗 components
CC
A

1. Introduction

BTXs (benzene, toluene and xylene isomers) are extremely important basic chemicals for the oil, gas and

petrochemical industries with global productions above 40 million metric tonnes per year for benzene and

xylenes and roughly half for toluen [1] BTXs used as starting points for derived downstream products are
popularly produced from petroleum naphtha, coking naphtha, and pyrolysis gasoline through hydro-treating and

continuous catalytic reforming (CCR) processes. CCR processes produce high-quality oil streams that include

aliphatic, alicyclic and aromatic hydrocarbons between C5 to C8.[2] Certain components in the oil streams

exhibit approximate boiling points and form azeotropes. Hence, it is difficult and energy-intensive to separate

aromatics from the oil streams via regular distillation.

T
Some improved distillation processes are adopted to separate the mixture such as azeotropic distillation[3][4]

IP
reactive distillation,[5][6] pressure swing distillation[7][8] solvent extraction(SE),[9][10] and extractive

R
distillation (ED) [11][12][13] SE, ED and azeotropic distillation are the most common processes used to separate

SC
the mixtures of aromatic and non-aromatic hydrocarbons [14]. The selection of a separation method depends on

U
the aromatic content of the hydrocarbon feed. The conventional processes for the separation of these aromatic
N
and aliphatic hydrocarbon mixtures are liquid extraction, suitable for the range of 20-65 wt.% aromatic content,
A
extractive distillation for the range 65-90 wt.% aromatics and azeotropic distillation for high aromatic content, >
M

90 wt.% [15]. Polar solvents are often used to separate the slightly polar aromatics from the non-aromatic
ED

hydrocarbons. The solvents include N-formylmorpholine (NFM),[16][17] N-methyl-2-pyrrolidinone (NMP) [18]

glycol,[19] and sulfolane.[20][21] Among these solvents, sulfolane is very popular because of the highest
PT

selectivity, higher dissolving capacity, higher boiling point and higher thermal stability.
E

According to Weissermel and Arpe [Error! Bookmark not defined.], ED processes are available for the separation of
CC

aromatic and aliphatic hydrocarbons in the range 65-90 wt.% aromatics in the feed mixture. We are focusing on

the separation of aromatic hydrocarbons from the feed stream of naphtha, which may contain 65-90 wt.%
A

aromatics. ED uses an entrainer to increase the relative volatility of the light and heavy components. Thus, the

light components can be drawn out from the top of the column as a product; the heavy components can be drawn

out from the bottom of the column as another product. This implicates additional distillation steps to separate the
entrainer from both the extract and raffinate phases and to purify the solvent with, consequently, additional

investment and energy consumption. Nevertheless, global issues of environmental and energy have motivated

the efficient increase in energy utilization to produce chemicals with lower energy consumption. This study

focuses on the sulfolane ED process for the separation of the mixtures of aromatic and non-aromatic

hydrocarbons and for lower energy requirement.

T
In recent years, the use of process simulation software for process design, control and optimization has gained

IP
increased interest because of the development of the computer.[22] Aspen Plus is a process simulation software

R
commonly used in chemical engineering applications that uses basic engineering relationships such as mass and

SC
energy balance, phase and chemical equilibrium, and reaction kinetics to predict the behaviour of a

U
thermodynamic process or chemical reaction.[23] The procedures for process simulations, based on Aspen Plus,
N
consist of defining components, appropriate thermodynamic models, required equipment and operating
A
conditions. A reliable thermodynamic model with accurate parameters is important for the simulation results.
M

So far, many thermodynamic models have been developed to estimate and predict the phase and chemical
ED

equilibrium of multicomponent chemical mixtures. They are also commonly used in process simulation

programs to calculate the mass balance in and around separation units. For the aromatic recovery ED process,
PT

due to the presence of polar compounds such as NFM, NMP, glycol and sulfolane, the non-random-two-liquid
E

(NRTL) model proposed by Renon and Prausnitz [24] and the universal quasi-chemical (UNIQUAC) model
CC

established by Abrams and Prausnitz [25] can be used to calculate the activity coefficient of the liquid phase in the

simulation. These models require some basic underlying parameters which must be known for all chemicals and
A

can be derived from experimental activity coefficients, or from phase diagrams, from which the activity

coefficients themselves can be calculated. Hence, for better design and optimization of the ED process to

separate aromatics from hydrocarbon mixtures, liquid-liquid equilibrium (LLE) and liquid-vapour equilibrium
(LVE) data containing sulfolane in a wide range of temperatures are essential [26]. Chen et al.[27] measured

vapour-liquid equilibria and density of the binary mixtures containing toluene, benzene, o-xylene, m-xylene,

sulfolane and nonane at 333.15 and 353.15K. Cassell et al. [28] measured the ternary LLE data for the systems

of sulfolane-toluene-pentane and sulfolane-benzene-pentane at 290.15, 298.15 and 323.15K by gas

chromatographic analyses and correlated the NRTL and UNIQUAC equations. Ko et al. [29] obtained LLE data

T
for the binary system containing sulfolane and cycloalkanes (cyclohexane and cyclooctane) over a temperature

IP
range around 300 K to the near upper critical solution temperature (UCST) using circulation-type equipment

R
with an equilibrium view cell and on-line gas chromatography. The binary LLE data sets were correlated well

SC
with the NRTL and UNIQUAC models using temperature-dependent parameters. Santiago et al. [30] measured

U
the LLE data of the three quaternary systems containing sulfolane, nonane + undecane + benzene + sulfolane,
N
nonane + undecane + toluene + sulfolane and nonane + undecane +m-xylene + sulfolane, at 298.15 and 313.15K.
A
The calculated compositions based on the NRTL model were found to be in better agreement with the
M

experimental data than those based on the UNIFAC-LLE model. Rappel et al.][31] presented the LLE data for the
ED

ternary systems: sulfolane + p-xylene + cyclohexane, sulfolane + p-xylene + n-hexane and sulfolane + toluene +

n-hexane at 308.15 and 323.15 K using a jacketed glass cell. The binary interaction parameters for UNIQUAC
PT

and NRTL activity coefficient models were fitted from the experimental data. All of these researchers found that
E

UNIQUAC and NRTL models show very good agreement with the experimental data. These experimental data
CC

also provided data for regression of basic underlying parameters which can improve the accuracy of process

simulation and phase calculation of the processes containing sulfolane.

A

With an increasing awareness of the environmental impacts and practical limitations on energy supply,

researchers aim to increase the efficiency of energy utilization. Process simulation and optimization[[32][33][34]

process improvement,[35][36] process coupling][37][38] and heat integration [39][40][41][42][43]are all useful
approaches to save energy as well as increasing the efficiency of the current energy utilization. Process

technology improvement, energy savings and energy utilization efficiency improvement were also studied for

the sulfolane aromatics process, including solvent mechanism research[Error! Bookmark not defined.],
[44],

phase-equilibrium data measurement and thermodynamic model study[Error! Bookmark not defined.-Error! Bookmark not

defined.],
[45], process simulation and optimization, process improvement, dynamic simulation and process control

T
etc. Choi et al. [46] established a model and optimization system for the sulfolane extraction plant based on

IP
process simulation. The parametric models obtained from the steady-state simulations were employed to develop

R
an optimization system based on the sequential quadratic programming (SQP) scheme. Optimal values of recycle

SC
streams were determined from the present optimization system. Optimum recycle rates were found and an almost

U
10% increase in the purity of benzene could be achieved by the application of the optimization system.
N
Lek-utaiwan et al.][47] proposed a practical methodology for the design and optimization of extractive distillation
A
using Aspen Plus by illustrating via an industrial separation of C8-aromatics mixture. The selection result may be
M

inconclusive due to the lack of accurate or missing parameters in the property model. The authors carried out
ED

experimental verification and the property parameter determinations were proved to be necessary as an

additional step to achieve a successful and reliable design. The practical methodology proposed in the study was
PT

proven to be accurate and effective in the design of extractive distillation for close-boiling mixtures. The
E

research showed that process simulation and optimization for the sulfolane separation process, based on
CC

computer calculation, can provide guidance for the product quality improvement and an efficient increase in

energy utilization. Li et al. [48] proposed several retrofitted processes to make full use of the sensible heat of the
A

heavy entrainer adopted in the extractive distillation that focuses on the separation of a benzene/cyclohexane

mixture using sulfolane as entrainer. In the basic case study, a side reboiler that utilizes self-heat or external heat

as the heat source is introduced in the separation system to reduce the energy requirement. The simulation results
show that the extractive distillation with heat integration, which utilizes the entrainer stream to heat the side

reboiler and the bottom reboiler, is the most energy-conserving sequence. In addition, this design sequence also

shows apparent benefits from both economic and environmental aspects. In the research of Qin et al. [49], the

steady state of an extractive distillation system was simulated using the RadFrac model in Aspen Plus, and the

controllability of this system with feed flow rate and feed composition disturbances was studied in Aspen

T
Dynamics based on the case study of benzene and cyclohexane separation via extractive distillation using

IP
sulfolane as an entrainer. Two control structures were demonstrated in the article. Each control structure has

R
three main temperature controllers, two in the extractive distillation column and one in the entrainer recovery

SC
column.

U
With the development of research, several ionic liquids (ILs), as well as polar organic solvents and mixtures
N
of them, have been evaluated as possible extraction agents in this aromatic ED process. Meindersma et a][50]
A
developed a conceptual process for the extraction of aromatic hydrocarbons with the ionic liquid
M

4-methyl-N-butylpyridinium tetrafluoroborat. The result shows that the energy requirement of an extraction
ED

process with an ionic liquid as the solvent is much lower than that with sulfolane, due to the lower process

[51]
temperature and the simpler recovery of the ionic liquid. Ferro et al. applied COSMO-based process
PT

simulations using Aspen Plus and Aspen HYSYS to the conceptual design of the two main unit operations for
E

the separation of aromatic and aliphatic hydrocarbons, with ILs as extracting solvents. For the first time,
CC

multi-component (up to 28 components) mixture feeds were taken into account to model the naphtha in the

process design. Binary model (n-hexane + benzene, n-heptane + toluene, n-octane + ethylbenzene, n-octane +
A

m-xylene) mixture feeds were also considered to validate the computational procedure. COSMO-based models

exhibited a reasonably good predictability of both the thermo-physical properties of the pure (hydrocarbons and

ILs) components and the LL and VL equilibria of their mixtures.

 Current computer programs for separation processes with highly non-ideal materials often lack the basic

underlying parameters of the thermodynamic model. Some researchers have addressed this shortcoming by

program and software integration. The program uses Component Object Model (COM) technologies available in

the Microsoft Windows operating system to provide interoperability of COM-aware applications as well as a

highly effective approach for process simulation and optimization. Jiang et al. [52] carried out steady-state and

T
dynamic simulations of a hydrocracking fractionator using the process simulator Aspen Plus with industrial

IP
process data. To obtain more economic benefits, the hydrocracking process needed to be optimized. The on-line

R
soft sensor and optimal quality control strategy are integrated by MATLAB CAPE-OPEN and Aspen Plus with

SC
an object linking and embedding (OLE) for the process control server. Gutiérrez-Guerra et al.[53] proposed a

U
novel stochastic optimization algorithm known as the Boltzmann Univariate Marginal Distribution Algorithm
N
(BUMDA) with constraints handling to optimize the heat-integrated distillation column (HIDiC) sequences. The
A
evaluation was performed by applying a new robust methodology using the interface Matlab-Excel-Aspen Plus.
M

Results show that the interface can be used to solve this kind of complex optimization problem successfully. In
ED

this paper, the Radfrac model is used in Aspen Plus to carry out the full process simulation. The sensitivity

analysis and minimization of the operating costs are established as the objective function in the MATLAB
PT

program.
E

The overall framework of this work is illustrated in Fig. 1. We first conducted the energy analysis and
CC

material balance analysis of an existing process and the relative volatility analysis of the ternary

aromatic-non-aromatic-sulfolane system. A new aromatic ED process was then presented for better energy
A

utilization based on the analysis of the ratio of unities, the material balance lines and relative volatility curves

discussed in section 2.2. The new process introduced a new flash tank and a semi-lean solution stream to

intensify the separation of aromatic and non-aromatic hydrocarbons. Aspen Plus was utilized to predict the mass
and energy balance as well as the phase equilibrium of the aromatic ED process. The significant parameters of

the new ED process including the temperature, feed location and mass flow rate of the semi-lean solution are

optimized by combining Aspen Plus with MATLAB. Heat exchanger networks (HENs) are designed for the

optimal new and existing processes. The comparison between the optimal new and existing processes

demonstrates that the operating cost is decreased by 8.9% when heat integration is not considered, and 13.04%

T
of the total annual cost can be reduced when a heat exchanger network is considered.

IP
2. Aromatic extractive distillation

R
2.1. Conventional ED process

SC
A conventional aromatic ED process from a refinery in China is shown in Fig. 2. The aromatic ED process

U
includes an extraction distillation column (T1), a non-aromatic purification column (T2), an entrainer recovery
N
column (T3) and heat exchangers. The common extraction distillation column is usually divided into three parts:
A
the solvent recovery section (rectifying section), extractive section and stripping section. [Error! Bookmark not defined.]
M

The solvent recovery section is essentially trays which are functionally equivalent to the non-aromatic
ED

purification column in Fig. 2. The solvent regeneration column is often installed to remove the tars and residues

produced in the process and can be ignored during the steady simulation and analysis[54][55]. Column T1 uses the
PT

entrainer to separate the crude feed into the raffinate stream (S3) and extraction stream (S5). The raffinate stream
E

is further refined in the non-aromatic purification column. The extraction stream is introduced into column T3 to
CC

separate aromatic components from the entrainer. Stripping steam S11 is used to form some azeotrope with

aromatics. The quality specification of aromatic and non-aromatic products is that the non-aromatic content in
A

the aromatics product stream S8 should not be greater than 1500 ppm, the sulfolane content in the aromatic

product stream S8 (xS,S8) should not be greater than 2 ppm and the aromatics content in the non-aromatic product

stream should not be greater than 1% (w/w).

 The total mass flowrate of the feed mixture is 55,716 kg/h. The supply temperature of the mixture is 40 °C.

Table 1 shows the composition of crude feed. The aromatics content in the feed is 77.13%. In Table 1, letters P

and N denote paraffin and naphthenic hydrocarbons, respectively; the atomic number of carbon in a component

is abbreviated to Cn (n=4, 5, 6, 7, 8); BEN, TOL, PX, MX and OX represent benzene, toluene, p-xylene,

m-xylene and o-xylene, respectively. In order to characterize the crude feed effectively, there is a simple

T
hypothesis that the paraffin is n-paraffin, and the naphthenic hydrocarbons are cyclohexane or

IP
n-alkyl-cyclohexanes[56][57]. The aromatic hydrocarbons are characterized in detail using real composition. Table

R
2 shows the operating conditions of the existing process.

SC
2.2. Energy and process analysis

U
Energy analysis: Fig. 3 represents the ratio of hot unities for each column in the existing process. We can find
N
that 97% of the total hot utility is consumed by columns T1 and T3. 71% and 16% of the hot utility in column T1
A
are used to heat the entrainer and crude feed, respectively; 13% of hot duty is used for the vaporization of the
M

bottom stream. The feed temperature of column T1 is very low, leading to the highest amount of hot utility used
ED

for the entrainer and feed heating. The temperature curve of column T1 shown in Fig. 4 also demonstrates that

there is a great drop at the feed location. Hence, the optimization of column feed conditions can obtain better
PT

energy performance. 8% of the hot utility in column T3 is used to heat the lean solution and 92% of hot duty for
E

the vaporization of the bottom stream. Most of the hot utility in column T3 is used for the separation process of
CC

solvent and aromatics. Hence, lowering the feed flowrate, improving feeding conditions and reducing the

purification of the lean solvent from the bottom of column T3 can decrease the hot utility of column T3.
A

Separation effect analysis: The calculation of the relative volatility expressed in Eq. (1) plays an important

role in the chemical process of separation and vapour liquid equilibrium characterization. The paraffin and

naphthenic hydrocarbons are treated as non-aromatics. BEN, TOL, PX, MX and OX are treated as aromatics.
The composition data was assessed on a H2O/sulfolane-free basis. The relative volatility of non-aromatic and

aromatic hydrocarbons varying with sulfolane solution flowrate at each operating condition is plotted for the

normalized mole fractions of the binary system (non-aromatic + aromatic) as depicted in Fig. 5. In any case,

maximizing the flowrate of sulfolane solution can maximize the relative volatility.

𝑦𝐴 𝑥𝐵
𝛼𝐴𝐵 = (1)
𝑦𝐵 𝑥𝐴

T
Material balance analysis: Material balance line diagrams are widely used in the conceptual design of

IP
columns for the ED process or the azeotropic distillation process [58]. The material balance lines of the existing

R
ED process are plotted for the normalized mass fractions of the ternary system (non-aromatic + aromatic +

SC
solvent), as demonstrated in Fig. 6. The composition data of streams was assessed on a H2O-free basis, as

depicted in Table 3.
U
N
As shown in Fig. 2, the point of F, B1, B2, D1, D2 in Fig. 6 represents the composition of the streams of
A
crude feed, rich solvent, lean solution, aromatic hydrocarbons and non-aromatic hydrocarbons, respectively.
M

With the inner material balance envelops as shown in Fig. 2, the 2nd material balance line of rich solvent (B1) to
ED

produce aromatic hydrocarbons (D2) at column T3 top and lean solution (B 2) at column T3 bottom are illustrated

in Fig. 6 by D2B1B2. In the same figure, the 1st inner material balance lines of adding F and the lean solvent is
PT

illustrated by FCB2. The production of rich solvent and pure non-aromatic hydrocarbons can also be illustrated
E

by D1CB1. According to the lever principle, the length of segment FC indicates the flowrate of the lean solution.
CC

When we move point B2 to B2', the purification of the lean solution is reduced. As a result, the hot utility of

column T3 can be decreased when the flowrate of crude feed is a fixed value. However, the length of FC' is also
A

increased, resulting in the increase of the hot utility of column T1.

According to the above analysis, in order to improve the feed conditions of column T1, by lowering the feed

flowrate and heat duty of column T3, a semi-lean solution stream is presented in this study to address these
problems. We can find the utilization of semi-lean solution in other fields. Cho et al.[59] proposed a

superstructure-based optimization approach to simultaneously identify the most appropriate arrangement and

operating conditions of acid gas removal units. A 15.9% reduction in utility cost was obtained by the semi-lean

process modification. The semi-lean solution is also used in post combustion CO2 capture processes and the

decarbonization system in ammonia plants [60]. For the aromatic ED process, the flowrate of the original lean

T
solution can be reduced through the semi-lean solution; this results in the decrease of reboiler duty in entrainer

IP
recovery columns. The reboiler duty of the ED column is increased if other operating parameters remain

R
unchanged. In order to obtain the maximum energy saving, how to balance the increase of ED column heat duty

SC
and the decrease of entrainer recovery column heat duty is also significant after introducing a semi-lean stream

U
in the ED process. Additionally, heat integration is worthy of serious consideration if there are appropriate heat
N
sinks in the process.
A

3. New ED process
M

A semi-lean solution stream and a flash tank were introduced to improve the existing ED process. The
ED

specifications of non-aromatic and aromatic products remain the same as the existing process. The new process

flow diagram is shown in Fig. 7, including the distillation units of the existing process and a new flash tank. The
PT

columns remained and functioned as the original attributes of the existing process. The columns are operated at
E

the same pressures as in the existing process. Nevertheless, the stream directions are varied. The bottom product
CC

of column T1 is fed to the new flash tank (F1) that operates at a pressure within [-0.05 MPag, -0.025 MPag]. The

semi-lean solution is obtained from the liquid phase of F1 and partly sent to column T1. Another part of the
A

semi-lean solution is fed to column T3 for the separation of solution and aromatics. The vapour stream from the

new flash tank is introduced into the upper section of column T3.
 As shown in Fig. 8, in the new process the rich solvent (B1, S5) is decomposed into a semi-lean solution

stream containing a certain amount of aromatics (B2', S15) and the feedstock of column T3 (B 1', S13 & S14)

through the new flash tank. The semi-lean solution stream (S16) at point B2' can be obtained without energy

requirement. Column T3 then separates part of the semi-lean solution stream at point B1' into the aromatic stream

at point A and the lean solution at B2. Hence, the aromatic concentration of column T3 feedstock is increased

T
and the flowrate of feedstock is decreased. At the same time, a double-feed of vapour and liquid streams which

IP
conform to the gas-liquid distribution in column T3 is adopted.

R
4. Process simulation and optimization

SC
4.1. Thermodynamic model and process simulation

U
An appropriate property method is critical to the accuracy of the simulation results. In this study, the NRTL
N
model expressed in Eq. (2) is applied to calculate the activity coefficient of component i in the liquid phase, and
A
the interaction parameters expressed in Eqs. (3-7) bear the same meanings as in the article by Renon and
M

Prausnitz[Error! Bookmark not defined.].

ED

The general equation for ln 𝛾𝑖 of species 𝑖 in a mixture of n components is:

∑j 𝑥𝑗 𝜏𝑗𝑖 𝐺𝑗𝑖 𝑥𝑗 𝐺𝑖𝑗 ∑𝑚 𝑥𝑚 𝜏𝑚𝑗 𝐺𝑚𝑗
ln 𝛾𝑖 = ∑𝑘 𝑥𝑘 𝐺𝑘𝑖
+ ∑𝑗 ∑ (𝜏𝑖𝑗 − ∑𝑘 𝑥𝑘 𝐺𝑘𝑗
) (2)
𝑘 𝑥𝑘 𝐺𝑘𝑗
PT

with:
E

𝐺𝑖𝑗 = 𝑒𝑥𝑝(−𝛼𝑖𝑗 𝜏𝑖𝑗 ) (3)

CC

𝜏𝑖𝑗 = 𝑎𝑖𝑗 + 𝑏𝑖𝑗 ⁄𝑇 + 𝑒𝑖𝑗 ln 𝑇 + 𝑓𝑖𝑗 𝑇 (4)

𝛼𝑖𝑗 = 𝑐𝑖𝑗 + 𝑑𝑖𝑗 (𝑇 − 273.15𝐾) (5)

A

𝜏𝑖𝑖 = 0 (6)

𝐺𝑖𝑖 = 1 (7)

where x is the mole fraction; 𝑐𝑖𝑗 and 𝑑𝑖𝑗 are the symmetrical fitting parameters; 𝑎𝑖𝑗 , 𝑏𝑖𝑗 , 𝑒𝑖𝑗 and 𝑓𝑖𝑗 are the

unsymmetrical fitting parameters. The binary interaction parameters for the NRTL activity coefficient model
built in the Aspen Plus database are non-complete for the system investigated in this study. Some no-databank

parameters are regressed from VLE and LLE experimental data sets that are extracted from the literature [61],[62]

[63] The maximum likelihood function is used as the objective for the regression, expressed in Eq. (8). Table 4

lists the missing binary interaction parameters extracted or regressed by the non-linear regression method based

on the maximum likelihood.

T
2 2 2 2
𝑁𝑃 𝑇𝑒,𝑖 −𝑇𝑚,𝑖 𝑃𝑒,𝑖 −𝑃𝑚,𝑖 𝑥𝑒,𝑖,𝑗 −𝑥𝑚,𝑖,𝑗 𝑦𝑒,𝑖,𝑗 −𝑦𝑚,𝑖,𝑗
Q = ∑𝑁𝐷𝐺
𝑛=1 𝒲𝑛 ∑𝑖=1 [( ) +( ) + ∑𝑁𝐶−1
𝑗=1 ( ) + ∑𝑁𝐶−1
𝑗=1 ( ) ] (8)
𝜎𝑇,𝑖 𝜎𝑃,𝑖 𝜎𝑥,𝑖,𝑗 𝜎𝑦,𝑖,𝑗

IP
The Redlich-Kwong equation-of-state, presented by Redlich and Kwong [64], is applicable for systems at low

R
SC
to moderate pressures (maximum pressure 10 atm.) and with slight vapour-phase non-ideality. The RK model

expressed in Eq. (9) is used to calculate the vapour phase thermodynamic properties in this work:

𝑝=
𝑅𝑇
𝑉𝑚 −𝑏
−
𝑎⁄ 0.5
𝑇
𝑉𝑚 (𝑉𝑚+𝑏) U (9)
N
A
where a and b are constants of the RK equation for the mixture and are related to the properties of the fluid.
M

The simulation software Aspen Plus V8.8 is used for rigorous process simulation using the equilibrium

model. The columns are simulated using the rigorous distillation model RadFrac. A rigorous vapour-liquid valid
ED

phase calculation is used in column T1 and vapour-liquid-dirty water condenser calculation in columns T2 and
PT

T3. The “strongly non-ideal liquid” option is used in the distillation blocks. The convergence method for tear

streams uses the Wegstein Method and the tolerance is set to 1.0 × 10−4 [65]. In column T1, the lean solution
E

rate and distillate rate were used as the operating variables. In column T3, the distillate rate and flow rate of the
CC

stripping steam were used as the operating variables.

A

To assess the accuracy of the simulation, some results obtained from the simulation software Aspen Plus are

compared to the operating data extracted from the real ED process, as shown in Table 5. The comparison of

distillate rates and temperatures shows that the maximal relative deviation between the operating data and

simulation results is 3.13%, indicating that the model is reliable for engineering simulation.
4.2. Optimization objective and framework

Process optimization plays a crucial role in process analysis and evaluation. As in all the aforementioned

works, the steady-state optimization of the new process is discussed using the annual operating cost (CV) as the

objective. The CV expressed in Eq. 10 includes the various operating energy costs required in the process. [Error!

Bookmark not defined.], [ [66]

T
min𝐶𝑉 = 8400 ∑ 𝑐𝑖 𝑞𝑖 (10)

IP
subject to:

𝑥𝐴,𝑆6 ≤ 0.01 (11)

R
𝑥𝑁𝐴,𝑆8 ≤ 0.0015 (12)

SC
𝑥𝑆,𝑆8 ≤ 2ppm (13)

𝑃𝐹1 ≤ 𝑃𝑆13,𝐹𝐿 (14)

U
N
where CV is the annual operating cost; ci and qi are the price and utility consumption per hour, respectively. The
A
product of ci and qi indicate the cost of utility i per hour. The utilities considered for columns T1, T2 and T3 are
M

steam and cooling water. The annual operating time was set to 8400 hours. The utility prices taken from the

software Aspen Energy Analyzer V8.8 are used to calculate the operating cost and listed in Table 6. The quality
ED

specification of aromatic and non-aromatic products is that the non-aromatic content in the aromatics product
PT

stream S8 (xNA,S8) should not be greater than 1500 ppm, the sulfolane content in the aromatic product stream S8

(xS,S8) should not be greater than 2 ppm and the aromatics content in the non-aromatic product S6 (xA,S6) stream
E
CC

should not be greater than 1% (w/w). Moreover, the pressure of the new flash tank (PF1) must not be lower than

that at the feed location of S13 in column T3 (PS13, FL).

A

Note that columns T1, T2 and T3 are kept the same in both the new and existing processes. A new flash tank

was only introduced in the new process. Compared with the columns, the cost of the flash tank is very small.

Hence, we do not compare the equipment costs of the two processes.

 In order to obtain the minimum CV of the new and existing processes, the significant operating parameters are

analyzed and optimized. The operating pressure of column T3 is set to - 0.05 MPag because of the limitation of

cooling water temperature and the thermostability of sulfolane to enable the use of cheap cooling water in the

condenser. The operating pressure of column T1 is assumed to be 0.07 MPag with the limitation of the cooling

water temperature in column T2. Pressure drop of 0.25 MPa and 0.08 MPa are set for column T1 and T3 to

T
coincide with the real operating conditions. The feed location of S12 should be set as 1th when column T1 is

IP
without a condenser.

R
In the new process, Columns T1 and T3 and the new flash tank needed to be reconsidered to optimize the

SC
operating cost. Column T1 in the new process with three feed streams and two product streams is complicated.

U
For column T1, there is a minimum value of lean solution circulation (S12) to guarantee the desired purity of the
N
non-aromatic product which is set as the variable of design specification of column T1. The lower the operating
A
pressure of the new flash tank is, the purer the semi-lean solution is. However, the operating pressure of the flash
M

tank is limited by column T3. The vapour stream from the new flash tank is expected to automatically enter
ED

column T3. As a result, the pressure of the new flash tank must not be lower than that at the feed location of S13

in column T3.
PT

The operating variables of the existing process include the feeding location of the crude feed (N S2) and rich
E

solution (NS5), the temperature of lean solution (TS12), the hydrocarbon concentration in lean solution (xHC,S12)
CC

and the reflux ratio of column T3 (RT3). The new process is more complicated when compared with the

conventional extractive distillation process. Therefore, except for the operating parameters listed above, several
A

other operating parameters need to be further optimized. The adjustable parameters in the new process can be

summarized as follows: (1) the operating parameters of column T1, including temperatures and stages of feed

streams (NS16, TS16, NS2, TS2); (2) the pressure of F1 (PF1); (3) the operating parameters of column T3, including
stages of feed streams (NS13, NS14) and reflux ratio (RT3); (4) the operating parameters of cycle streams, including

the flow rate of the semi-lean solution stream (FS16) and the temperature and hydrocarbon concentration of lean

solution (TS12, xHC,S12). The hydrocarbon concentration of the semi-lean solution is determined by PF1. The lean

solution rate (FS12), stripping steam flow rate (FS11) and distillate rates (DT1, DT3) are simulated as operating

variables to satisfy the separation requirements.

T
From Figs. 2 and 7 the path from the existing process to the new process, it is clear to see that when the

IP
semi-lean solution stream and flash tank are introduced, it should be exactly T1, T2 and T3 in the existing

R
process. Since there are quite a few design and operating variables needed to be determined, several variables are

SC
fixed to examine the impacts of other unfixed variables generated by the flash tank. Also from the analysis and

U
comparison of the operating parameters above, xCH,S12, TS12 and RT3 in the existing process was directly
N
retrofitted to the new process. The process optimization for ED processes dealt with both continuous and discrete
A
variables. TS2, PF1, and FS16 are the continuous variables whereas N S2, NS16 and NS13 are the discrete variables.
M

The simultaneous optimization with continuous and discrete variables can be achieved by mixed linear
ED

programming. At present, with Aspen plus as the process simulator, both direct and continuous variables could

not be optimized simultaneously. Therefore, a sequential iterative optimization search was used with N S2, NS16
PT

and NS13 as inner loop, FS16 as middle loop and TS2 as outer loop. The optimal calculation procedure for the
E

adjustable parameters is presented based on the iterative simulation of the ED process in MATLAB and Aspen
CC

Plus integration based on COM technology, as shown in Fig. 9.

4.3 Optimization results

A

4.3.1 Existing process

 The operating variables are optimized to minimize Cv ($/h) of the existing process. The operating variables

include the feeding location of the crude feed (NS2) and rich solution (NS5), the temperature of lean solution

(TS12), the hydrocarbon concentration in lean solution (xHC,S12) and the reflux ratio of column T3 (RT3).

First inner loop: Optimization of NS2 and TS12 fixing: The impact of the feed location of S2 and temperature of

S12 are explored by increasing the numbers from 30 to 50 and 94°C to 112°C respectively. Fig. 10 shows the

T
reboiler duty of column T1 (QR, T1) with NS2 and TS12 as the variates under the conditions that T S2 = 95°C,

IP
NS5 = 21, xHC,S12 = 1500 ppm and RT3 = 0.7. We can see from Fig. 10 that the QR, T1 shows a decreasing and then

R
increasing trend with the NS2 increasing as well as TS12. From this plot, the best NS2 and TS12 are NS2 = 36th and

SC
TS12 = 98°C. Fig. 11 shows the heat duty of the reboiler in column T1 varying with TS12 at each xCH,S12 under the

U
conditions that TS2 = 95°C, NS5 = 21 and RT3 = 0.7. Particularly, NS2 has been optimized for a specified pair of
N
TS12-xCH,S12. The heat duty QR, T1 presents a decreasing and then increasing trend with the increase of TS12. As
A
observed from Fig. 11, the lowest heat duty QR, T1 can be obtained when TS12 = 98 °C under the above conditions.
M

Assume that the low temperature exhaust heat in the process is used for the heating of S2 or S12, the impact on
ED

the operating cost (Cv, $/h) of the two cases is depicted in Fig.12. For the same amount of low temperature

exhaust heat, different designs will have different effects on Cv, an efficiency utilized design is the heating for
PT

S2 which can achieve a lower Cv.

E

Notice that the tray number is counted from the top to the bottom of the column, and there is no condenser in
CC

T1. From Figs. 10-12, we can deduce the following: (1) The feed location of the crude feed moves down the

column from 30 to 40 for the lowest QR, T1 exemplifies the well-known fact that an optimal value of NS2 exists
A

while the other variables are kept constant. Setting NS2 optimization as an inner loop unit in the optimization

procedure can guarantee an optimal N S2 can be found for each iteration. (2) Increasing the temperature of S12

from 98°C to 112°C caused an increase in QR, T1. The same amount of low temperature exhaust heat used for the
heating of S2 or S12 will have different effects on Cv, an efficiency utilized design is the heating for S2. A fixed

value of TS2 can be set for the process.

Second inner loop: Optimization of NS5 and RT3: Fig. 13 shows the results of the operating costs (Cv) varying

with RT3 and NS5 under the conditions that T S2 = 95°C, NS2 = 36, xHC,S12 = 1500 ppm and TS12 = 98°C. QR, T3 and

FS11 showed similar progressive tendencies with N S5 and RT3 increasing. RT3 should have been determined to

T
minimize Cv and simultaneously satisfy the separation requirement that the sulfolane content in the product

IP
stream S8 (xS,S8) is below 2 ppm. While for higher purity of the aromatic hydrocarbons, R T3 is optimized by the

R
inflection point of each curve. For instance, R1 increases slightly with NS5 moving down the column from 5 to

SC
25 when RT3 = 0.3 and a lower sulfolane content in S8 can be obtained for further down. From this plot, RT3 and

U
NS5 can be set as 0.3 and 25 when xCH,S12 = 1500 ppm, TS12 = 98 °C and NS2 = 36.
N
Outer loop: Optimization of xCH,S12: All the optimization procedure carried out above for columns was run as
A
inner loop units to find the best value for the above operating parameters. However, the real process should
M

implement an entrainer recycle stream FS12 bearing some impurities which influence the operating parameters.
ED

This highlights the importance of optimizing the hydrocarbon concentration of lean solution together with the

operating parameters of columns.

PT

To solve this problem, we adjust xCH,S12 for the simulation and obtain the results for the operating cost (Cv)
E

and the minimum flow rate of the entrainer variation with xCH,S12. Fig.14 illustrates the results for the operating
CC

cost (Cv) and the minimum flow rate of the entrainer which varied with xCH,S12 when TS12 = 98 °C, NS2 = 36,

RT3 = 0.3, and NS5 = 25 for the existing process. FS12 demonstrates a line variety with xCH,S12, while Cv shows an
A

exponential decay with xCH,S12 increasing. Obviously, the Cv value gets even smaller with xCH,S12 increase.

However, Cv decreases slightly, FS12 increases drastically with xCH,S12 increase when xCH,S12 > 1000 ppm.
 We also conclude from Fig.14 that if a lean solution stream with lower hydrocarbon concentration is fed to

the extractive column, a greater F S12 is needed to achieve the same product purity. Meanwhile the concentration

of entrainer fed to the regeneration column decreases due to mass balance, and less energy is used to recycle the

entrainer, leading to a decrease of Cv. From this plot, the xCH,S12 can be set to 1000 ppm.

According to the above analysis, the optimal operating parameters for the existing process are listed in Table

T
7. Based on the optimized results N S2 = 36, NS5 = 25, RT3 = 0.3 and xCH,S12 = 1000 ppm are obtained for

IP
CV = 2.381 × 106$/year.

R
SC
4.3.2 New process

The new process introduces a new flash tank, a new feed stream to column T1, a new feed stream to column

U
T3 and the semi-lean solution stream. Therefore, except for the operating parameters listed in Table 7, several
N
other operating parameters need to be further optimized. The parameters are (1) the feed stage and temperature
A
of streams (NS2, TS2, NS16) at column T1; (2) the pressure of flash tank F1 (PF1); (3) the feed stage of stream S13
M

(NS13) at column T3; (4) the flow rate of the semi-solution stream (FS16). The iterative optimization procedure
ED

with three-loop units described in Fig. 9 is used for the optimization of the new process. Since there are quite a

few design and operating variables needing to be determined, TS12, TS16, NS14, RT3 are fixed to examine the
PT

impact of other unfixed variables.

E

Inner loop: Optimization of the feed locations and PF1: Fig. 15 demonstrates the results for the new process
CC

with PF1 and NS13 as the design variables under the condition that RT3 = 0.3, NS15 = 25, xCH,S12 = 1000 ppm,

TS12 = TS16 = 98 °C, FS16 = 20 × 103 kg/h and NS2 = 36. Fig. 15 shows that further reducing the pressure to the
A

pressure of the feed stage of S13 enables an additional reduction of the energy consumption materialized by the

objective function Cv decrease. First the lowest Cv is achieved when only NS13 is optimized. But a smaller Cv
value can be obtained when PF1 is optimized simultaneously. We also observed the close interrelation between

Cv, PF1 and NS13.

Because the vapour stream from the new flask tank F1 (S13) is expected to automatically enter column T3,

the pressure at the new flash tank must not be lower than that at the feed location of column T3. An operable

area divide line is shown in Fig. 15 by the black dashed line. Operation in the bottom right corner of the divide

T
line is not allowed because of the limitation of the feed pressure. Hence, the pressure of F1 is set to equal the

IP
feed stage pressure of S13. When the NF,S13 is between 2~4, the sulfolane content in the overhead aromatics

R
product always exceeds 2ppm, as shown in Fig. 15 by the red dots. When the NF,S13 is not lower than 5, the

SC
sulfolane content in the overhead aromatics product is always in trace amounts. Hence, the best feed stage of

U
stream S13 NS13 is the 5th. Consequently, the best PF1 is the pressure of stage 5.
N
Middle loop: Optimization of FS16: All the optimization procedure carried out above for feed locations and PF1
A
was run as an inner loop unit to find the best value for the above operating parameters based on a given set of
M

FS16-TS12. However, the new process implements a semi-lean solution recycle stream with no regeneration cost
ED

and bearing some impurities. The impact of the flowrate of the semi-lean solution and the hydrocarbon

concentration of lean solution are explored by increasing the numbers from 20 to 80 and 0.1 to 1000 ppm
PT

respectively. Fig. 16 shows the results of Cv varying with xCH,S12 and FS16 without HENs.
E

It is observed that with the flowrate of S16 increasing, Cv shows a decreasing trend. This goes along with our
CC

intuition based on the analysis of the ternary map of material balance lines discussed in sections 2.2 and 3. When

we move point B2 to B2' by adding a flash tank, the purification of the lean solution is reduced. As a result, the
A

hot utility of column T3 can be decreased. However, the length of FC' is also increased, resulting in an increase

in the hot utility of column T1. A lower energy consumption can be achieved when the impact on T3 is greater

than T1. Therefore the variables xCH,S12 and FS16 are worth optimization to find a minimum Cv. The optimization
of FS16 was set as the middle loop unit to avoid local minimum. However, the maximum flowrate of S16 is

limited by the upper bound of the liquid flowrate of T1. Cv shows an exponential decay with xCH,S12 increasing

till xCH,S12 = 1000 ppm. FS12 demonstrates a line variety with xCH,S12, similar to the one in Fig. 14, xCH,S12 can be

fixed as 1000 ppm.

Outer loop: Optimization of TS12: The temperature curve of column T1 shown in Fig. 4 demonstrates that

T
there is a great drop at the feed location. Hence, the optimization of column feed conditions can obtain better

IP
energy performance. We have proved that for the same amount of low temperature exhaust heat, different

R
designs will have different effects on Cv, an efficiency utilized design is the heating for S2 which can achieve a

SC
lower Cv. Fig. 17 presents the results of Cv varying with FS16 and TS2 without HENs. It is observed that with FS16

U
and TS2 increasing, Cv shows a decreasing trend which goes along with the analysis of the ternary map of
N
material balance lines discussed in sections 2.2 and 3. However, a feasibility dividing line of operation exists and
A
is depicted by the broken black line in Fig. 17. Operation above the dividing line can simultaneously satisfy the
M

aromatic product separation requirement (xS, S8 ≤ 2 ppm) and limitation of FS12. Hence, the optimum FS16 and TS2
ED

for the new process can be found at 60 × 103 kg/h and 117.8 °C.

The final optimal process flow diagram for the new process is shown in Fig. 18. The optimal operating
PT

parameters of the new process are listed in Table 8. From Tables 7 and 8, we can observe the following: (1) The
E

reboiler duty of T3 decreased from 16.81 GJ/h in the optimized existing process to 11.21GJ/h in new process,
CC

showing that improvement was possible due principally to the adding of a flash tank and semi-lean solution

stream. (2) The reboiler duty of T1 slightly increased from 48.75 GJ/h in the optimized existing process to
A

49.47GJ/h in the new process due to the increase of the total flow of semi-lean solution and lean solution. (3)

Meanwhile, the flowrate of the stripping steam fed to T3 savings reached 55.9% and 65.2% due to the decrease

in the lean solution flow rate and optimization of feed locations.

4.3.3 Results comparison

Table 9 lists the heat duty and the minimum Cv results for the optimal new and existing ED processes. In

summary: (1) The reboiler duty decreased from 82.03 GJ/h in the existing process to 68.63GJ/h in the optimal

existing process by the optimization of the significant operating parameters, showing that improvement was

possible due principally to a combination of a lower reflux ratio and a shift of feed locations. The extra reduction

T
to 63.75 GJ/h for the new process proved the usefulness of adding the flash tank and the semi-lean solution

IP
stream, as deduced from the analysis of material balance lines and effects of xCH,S12 on Cv. The total reboiler duty

R
of the new process is 4.88 GJ/h less than that of the optimal existing process, which is about a 7.11% saving in

SC
total reboiler energy consumption. (3) The flowrate of the stripping steam fed to T3 savings reach 55.9% and

U
65.2% due to the decrease in the lean solution flow rate and the optimization of feed locations. The flowrate of
N
the stripping steam in the new process is 680.5 kg/h less than that of the optimal existing process, which is about
A
a 21.19% saving of steam. (2) Meanwhile, energy consumption underlying the Cv value is reduced by 27.3% and
M

33.7% compared to the existing process, respectively. The extra 5.4% reduction is mostly attributed to the
ED

application of the semi-lean solution and flash tank. The new process leads to an 8.9% reduction in the total

annual operating costs compared with the optimal existing process.

PT

5. Heat integration and the HEN design

The software Aspen Energy Analyzer V8.8 was used to carry out pinch analysis and economic evaluation for
E
CC

HENs in the optimal new and existing processes. The stream data are extracted from Aspen Plus V8.8, which

includes heat capacity, flowrate and temperature details, as shown in Table 10. The difference in stream data
A

between the two processes includes process streams SC, E5, SR, E4, SR, E2, S11, S10-S12 and S15-S16, which are

raised by the new flash tank F1 and the new semi-lean solution stream. The default cost and heat transfer

coefficient (HTC) embedded in the database of the Aspen Energy Analyzer were used to design the HENs. Total

annual cost (TAC) of the HEN was used to evaluate the optimal ΔTmin and compare the two processes in this
study. The TAC is the combination of capital cost and operating cost associated with the heat exchangers as

expressed in Eq. (10):

TAC     C i  C V
(10)

where Λ is the annualization factor and was calculated using Eq. (11):

ROR
(1 
PL
)
100
 

T
PL (11)

IP
The rate of return (ROR) is set at 15% and the plant life (PL) is set to 5 years in this study. The investigation

R
of TAC in the two processes on varying ΔTmin shows that the smallest TAC for the optimal existing process is

SC
obtained when ΔTmin is at 7.5°C and that for the new process is 8°C, as shown in Fig. 19.

U
The Aspen Energy Analyzer can automatically generate HEN designs with the objective function of
N
minimizing the TAC. Figs. 20 and 21 give the optimal HEN designs for the two ED processes. Table 11 lists the
A
details of operating and capital costs for the two designs. Although the operating heating costs were higher for
M

the new process, the capital cost of the new process is less than that of the optimal existing process. The new

process saves 13.04% of TAC.

ED

6. Conclusions
PT

We have looked at improving the design and operation of an extractive distillation process for the separation

of the aromatics-non-aromatics with sulfolane. The process flow sheet includes both the extractive distillation
E
CC

column and the entrainer regeneration column. By using insight from the analysis of the energy utilization,

relative volatility curves and material balance lines we have noticed the beneficial effect of lowering the purities
A

of the lean solution. A lower purity cycle stream increases the minimum amount of lean solution and the relative

volatility of aromatics-non-aromatics. As a result, the hot utility of column T3 can be decreased while the hot

utility of column T1 is increased. Based on the common aromatic ED process, a new process was presented

through introducing a semi-lean solution stream and a flash tank.

 We then ran an optimization aimed at minimizing the annual operating cost as objective function Cv. Cv

includes the various operating energy costs required in the process. Rigorous simulations were done in all cases.

For the sake of comparison, we have kept the total number of trays identical to those used in the existing process.

Other variables have been optimized for new and existing processes. The total annual operating cost was

calculated for the two processes. Two important issues have emerged. First, the full simulation model for the

T
new and existing processes was established in Aspen Plus. Some no-databank model parameters were regressed

IP
from the literature. The relative deviation between the real operating data and simulation results is less than 5%.

R
Second, an optimal calculation procedure for the adjustable parameters is presented based on the iterative

SC
simulation of the ED process in MATLAB and Aspen Plus integration based on COM technology. The iterative

U
optimization search was used with NS2, NS16 and NS13 as inner loop, FS16 as middle loop and TS2 as outer loop.
N
The new process has a lower operating cost and TAC. The important quantitative observations of this paper
A
are summarized as below:
M

(1) Compared with the existing ED process, the operating cost of the optimal existing process is reduced
ED

from $2.381×106 to 1.731×106 per year, and the energy requirement is reduced from 82.03 to 68.63GJ/h.

33.7% savings in operating costs have been achieved compared to the existing process thanks to the
PT

optimization of operating parameters.

E

(2) Compared with the optimal existing ED process, the operating cost is reduced from $1.731×106 to
CC

1.578×106 per year, and the energy requirement is reduced from 68.63 to 63.75 GJ/h. The further 5.4%

reduction is mostly attributed to the application of the semi-lean solution and flash tank.
A

(3) Compared with the optimal existing ED process, the operating cost is reduced from $1.38 × 106 to

1.2×106 per year. The new process saves 13.04% of TAC compare with the optimal existing process.

(4) A parameter optimization procedure for the ED processes is put forward to optimize the significant operating
 parameters. The operating variables of the existing process include the feeding locations (NS2, NS5), the

temperature of lean solution (TS12), the hydrocarbon concentration in lean solution (xHC,S12) and the

reflux ratio of column T3 (RT3). Their optimal values are 36, 25, 98°C, 1000 ppm and 0.3, respectively. For

the new process, the parameters are (1) the feed location and temperature of streams (NS2, TS2, NS13, NS16);

(2) the pressure of flash tank F1 (PF1); (3) the flow rate of the semi-solution stream (FS16). Their optimal

T
values are 42, 117.8°C, 5, 8, - 0.048 Mpag and 60 × 103 kg/h, respectively.

R IP
SC
Acknowledgements
U
N
This research is supported by the National Natural Science Foundation of China (Nos. 21776323 and
A
U1462113).
M
ED

Reference

[1] UNEP. Global Chemical Outlook: Towards sound management of chemicals. 2012..
PT

[2] A. Chauvel, G. Lefebvre, Petrochemical Processes, Editions Technip, Paris, 1989.

[3] I. Rodriguez-Donis, J. Acosta-Esquijarosa, V. Gerbaud, E. Pardillo-Fondevila, X. Joulia, Separation of n-hexane-ethyl
E

acetate mixtures by azeotropic batch distillation with heterogeneous entrainers, Chem. Eng. Process. 44.1 (2005) 131-137.
[4] W.L. Luyben, I.L. Chien, Design and Control of Distillation Systems for Separating Azeotropes. John Wiley & Sons, New
CC

York, 2010.
[5]
C.L. Chen, Y.H. Chung, H.Y. Lee, Design and Control of Reactive Distillation Process for the Production of Methyl
Valerate, Ind. Eng. Chem. Res. 55.5 (2016) 1347-1360
A

[6] C. Noeres, E.Y. Kenig, A. Górak, Modelling of reactive separation processes: reactive absorption and reactive distillation,
Chem. Eng. Process. 42.3 (2003) 157-178.
[7] Y. Li, C.J. Xu, Pressure-Swing Distillation for Separating Pressure-Insensitive Minimum Boiling Azeotrope
Methanol/Toluene via Introducing a Light Entrainer: Design and Control, Ind. Eng. Chem. Res. 56.14 (2017) 4017-4037.
[8] E. Lladosa, J.B. Montón, M.A.C. Burguet, Separation of di-n-propyl ether and n-propyl alcohol by extractive distillation
and pressure-swing distillation: Computer simulation and economic optimization, Chem. Eng. Process. 50.11-12 (2011)
1266-1274.
[9] A.I. Papadopoulos, P. Linke, Integrated solvent and process selection for separation and reaction-separation processes,
Chem. Eng. Process. 48.5 (2008) 1047-1060.
[10] T. Zhou, Z. Wang, Y. Ye, L. Chen, J. Xu, Z. Qi, Deep Separation of Benzene from Cyclohexane by Liquid Extraction
Using Ionic Liquids as the Solvent, Ind. Eng. Chem. Res. 51.15 (2012) 5559-5564.
[11] X.G. Lei, C.Y. Li, B.H. Chen, Extractive distillation: A review, Sep. Purif. Technol. 32.2 (2003) 121−213.
[12] G. Li, P. Bai, New Operation Strategy for Separation of Ethanol–Water by Extractive Distillation, Ind. Eng. Chem. Res.
51.6 (2012) 2723-2729.
[13] C.E. Torres-Ortega, J.G. Segovia-Hernandez, F.I. Gomez-Castro, S. Hernandez, A. Bonilla-Petriciolet, B.G. Rong, M.
Errico, Design, Optimization and Control of an Alternative Process Based on Extractive Distillation for the Sweetening of
Natural Gas, Chem. Eng. Process. 74.74 (2013) 55-68.
[14] [H.G. Franck, J.W. Staldelhofer, Industrial Aromatic Chemistry, Springer-Verlag, Berlin, 1988.

T
[15] K. Weissermel, H. J. Arpe, Industrial Organic Chemistry (4th Completely Revised edition), Wiley-VCH, Weinheim, 2003.
[16] Z.R. Wang, S.Q. Xia, P.S. Ma, (Liquid + liquid) equilibria for the ternary system of (N-formylmorpholine + ethylbenzene

IP
+ 2,2 4-trimethylpentane) at temperatures (303.15, 313.15, and 323.15) K, Fluid Phase Equilib. 328 (2012) 25-30.
[17] H. Mahmoudjanloo, A.A. Izadpanah, S. Osfouri, A.H. Mohammadi, Modeling liquid–liquid and vapor–liquid equilibria

R
for the hydrocarbon+N-formylmorpholine system using the CPA equation of state, Chem. Eng. Sci. 98 (2013) 152-159.

SC
[18] G.R. Vakili-Nezhaad, H. Modarress, G.A. Mansoori, Solvent extraction of aromatic components from lube-oil cut by
N-methylpyrrolidone (NMP), Chem. Eng. Technol. 22.10 (1999) 847-853.
[19] [M. Nicolae, F. Oprea, E.M. Fendu, Dipropylene glycol as a solvent for the extraction of aromatic hydrocarbons. Analysis

U
and evaluation of the solvency properties and simulation of the extraction processes, Chem. Eng. Res. Des. 104 (2015)
287-295.
N
[20] M.L. Fu. Use of organic sulfones as the extractive distillation solvent for aromatics recovery, Ind. Eng. Chem. Pro. Des.
Dev. 25.4 (1986) 949-957.
A
[21] [M.L. Fu, D.M. Coombs, Two-liquid-phase extractive distillation for aromatics recovery, Ind. Eng. Chem. Res. 26.3
(1987) 564-573.
M

[22] W.X. Meng, S. Banerjee, X. Zhang, R.K. Agarwal, Process simulation of multi-stage chemical-looping combustion using
Aspen Plus, Energy 90 (2015) 1869-1877.
[23]
ED

Aspen Plus, Aspen Plus user guide, Aspen Technology Inc, Cambridge, 2000.
[24] H. Renon, J.M. Prausnitz, Local compositions in thermodynamic excess functions for liquid mixtures, AIChE J. 14.1
(1968) 135-144.
[25] D.S. Abrams, J.M. Prausnitz, Statistical thermodynamics of liquid mixtures: A new expression for the excess Gibbs
PT

energy of partly or completely miscible systems, AIChE J. 21.1 (1975) 116-128.

[26] M.A. Noriega, P.C. Narváez, A.D. Imbachi, J.G. Cadavid, A.C. Habert, Liquid-liquid equilibrium for
biodiesel-glycerol-methanol or ethanol systems using UNIFAC correlated parameters, Energy 111 (2016) 841-849.
E

[27] W.K. Chen, K.J. Lee, J.W. Ko, C.M.J. Chang, H. Daina, L.S. Lee, Vapor–liquid equilibria and density measurement for
CC

binary mixtures of toluene, benzene, o-xylene, m-xylene, sulfolane and nonane at 333.15K and 353.15K, Fluid Phase Equilib.
287.2 (2010) 126-133.
[28] G.W. Cassell, N. Dural, A.L. Hines, Liquid-liquid equilibrium of sulfolane-benzene-pentane and
A

sulfolane-toluene-pentane, Ind. Eng. Chem. Res. 28 (1989) 1369-1374.

[29] M. Ko, J. Im, J.Y. Sung, H. Kim, Liquid-Liquid Equilibria for the Binary Systems of Sulfolane with Cycloalkanes, Chem.
Eng. Data 51 (2006) 636-638.
[30] R.S. Santiago, M. Aznar, Liquid–liquid equilibria for quaternary mixtures of nonane + undecane + (benzene or toluene or
m-xylene) + sulfolane at 298.15 and 313.15K, Fluid Phase Equilib. 253.2 (2007) 137-141.
[31]R. Rappel, L.M.N.D. Góis, S. Mattedi, Liquid–liquid equilibria data for systems containing aromatic + nonaromatic +
sulfolane at 308.15 and 323.15 K, Fluid Phase Equilib. 202.2 (2002) 263-276.
[32] H.M.S. Lababidi, S.H.A. And, M.A. Fahim, Optimization of Aromatics Extraction of Naphtha Reformate by Propylene
Carbonate/Diethylene Glycol Mixed Solvent, Ind. Eng. Chem. Res. 45.14 (2006) 5086-5097.
[33] Y. Lei, X. Qi, B.J. Zhang, Q.L. Chen, Simultaneous Optimization of the Complex Fractionator and Heat Exchanger
Network Considering the Constraints of Variable Heat Removals in Delayed Coking Units, Ind. Eng. Chem. Res. 53.33
(2014) 13073-13086.
[34] M.H. Khademi, M.R. Rahimpour, A. Jahanmiri, Simulation and optimization of a six-effect evaporator in a desalination
process, Chem. Eng. Process. 48.1 (2009) 339-347.
[35]V.H.G. Díaz, G.O. Tost, Energy efficiency of a new distillation process for isopropanol, butanol, and ethanol (IBE)
dehydration, Chem. Eng. Process. 112 (2017).
[36] X.G. Liu, C. He, C.C He, J.J. Chen, B.J. Zhang, Q.L. Chen, A new retrofit approach to the absorption-stabilization
process for improving energy efficiency in refineries, Energy 118 (2017) 1131-1145.

T
[37] [1] B.J. Zhang, X.L. Luo, X.Z. Chen, Q.L. Chen, Coupling Process Plants and Utility Systems for Site Scale Steam
Integration, Ind. Eng. Chem. Res. 52.41 (2013) 14627-14636.

IP
[38] S.J. Wang, Optimal Design of Reactive Distillation and Azeotropic Distillation Systems with Incorporation of
Dividing-Wall Columns, Chem. Eng. Process. 85.11 (2014) 108-124.

R
[39]B.J. Zhang, X.L. Luo, Q.L. Chen, Hot Discharges/Feeds between Plants to Combine Utility Streams for Heat Integration,

SC
Ind. Eng. Chem. Res. 51.44 (2012) 14461-14472.
[40] A.A. Kiss, Žarko Olujić, A review on process intensification in internally heat-integrated distillation columns, Chem. Eng.
Process. 86 (2014):125-144.
[41]

U
Z. Yi, A.O. Oyedun, M. Wang, C.W. Hui, Simultaneous optimization of a heat integrated coal gasification process, Chem.
Eng. Res. Des. 98 (2015) 136-146.
N
[42]L.X. Kang, Y.Z. Liu, Retrofit of Heat Exchanger Networks for Multiperiod Operations by Matching Heat Transfer Areas
in Reverse Order, Ind. Eng. Chem. Res. 53.12 (2014) 4792–4804.
A
[43]T. Chen, B.J. Zhang, Q.L. Chen, Heat integration of fractionating systems in para-xylene plants based on column
optimization, Energy 72.7 (2014) 311-321.
M

[44] T.L Ho, M. Fieser, L. Fieser, Fieser and Fieser's Reagents for Organic Synthesis: Tetramethylene sulfone (sulfolane), John
Wiley & Sons Inc, New York, 2006.
[45]
ED

R. Wittig, J. Lohmann, J. Gmehling, Prediction of phase equilibria and excess properties for systems with sulfones,
AIChE J. 49.2 (2010) 530-537.
[46] Y.J. Choi, K.W. Cho, B.W. Cho, Y.K. Yeo, Optimization of the Sulfolane Extraction Plant Based on Modeling and
Simulation, Korean. J. Chem. Eng.17.6 (2000) 712-718.
PT

[47] P. Lek-utaiwana, B. Suphanit, P.L. Douglas, N. Mongkolsiri, Design of extractive distillation for the separation of
close-boiling mixtures: Solvent selection and column optimization, Comput. Chem. Eng. 35.6 (2011) 1088-1100.
[48] L.M. Li, Y.Q. Tu, L.Y. Sun, Y.F. Hou, M.Y. Zhu, L.J. Guo, Q.S. Li, Y.Y. Tian, Enhanced Efficient Extractive Distillation
E

by Combining Heat-Integrated Technology and Intermediate Heating, Ind. Eng. Chem. Res. 55.32 (2016) 8837–8847.
CC

[49] J. Qin, Q. Ye, X. Xiong, N. Li, Control of Benzene-Cyclohexane Separation System via Extractive Distillation Using
Sulfolane as Entrainer, Ind. Eng. Chem. Res. 52.31 (2013) 10754-10766.
[50]G.W. Meindersma, A.B. Haan, Conceptual process design for aromatic/aliphatic separation with ionic liquids, Chem. Eng.
A

Res. Des. 86.7 (2008) 745-752.

[51] V.R. Ferro, J.D. Riva, D. Sanchez, E. Ruiz, J. Palomar, Conceptual design of unit operations to separate aromatic
hydrocarbons from naphtha using ionic liquids. COSMO-based process simulations with multi-component “real” mixture
feed, Chem. Eng. Res. Des. 94 (2016) 632-647.
[52] Q. Jiang, Y. Cai, S. Jia, Z. Cao, B. Chen, H. Zhou, Optimal Product Quality Control in a Hydrocracking Fractionator with
Process Simulation Approaches, Ind. Eng. Chem. Res. 54.17 (2015) 4805-4814.
[53] R. Gutiérrez-Guerra, J. Cortez-González, R. Murrieta-Dueñas, J.G. Segovia-Hernández, S. Hernández, A.
Hernández-Aguirre, Design and Optimization of Heat-Integrated Distillation Column Schemes through a New Robust
Methodology Coupled with a Boltzmann-Based Estimation of Distribution Algorithm, Ind. Eng. Chem. Res. 53.27 (2014)
11061-11073.
[54] P.G.T And, V.V. Brunt, Extractive Distillation Solvent Characterization and Shortcut Design Procedure for
Methylcyclohexane - Toluene Mixtures, Ind. Eng. Chem. Res. 39.6 (2000) 1614-1623.
[55] L. Berg, Separation of benzene and toluene from close boiling non-aromatics by extractive distillation, AIChE J. 29.6
(1983) 961-966.
[56] B. Coto, R. Vangrieken, J. Pena, J.J. Espada, A model to predict physical properties for light lubricating oils and its
application to the extraction process by furfural, Chem. Eng. Sci. 61.13 (2006) 4381-4392.
[57]R.V. Grieken, B. Coto, E. Romero, J.J. Espada, Prediction of liquid−liquid equilibrium in the system furfural + heavy
neutral distillate lubricating oil, Ind. Eng. Chem. Res. 44.21 (2005), 8106-8112.

T
[58] W. L. Luyben, I L. Chien, Design and Control of Distillation Systems for Separating Azeotropes, John Wiley & Sons, Inc.,
2010.

IP
[59] H.B. Cho, M. Binns, K.J. Min, J.K. Kim, Automated process design of acid gas removal units in natural gas processing,
Comput. Chem. Eng. 83 (2015) 97-109.

R
[60] M. Karimi, M. Hillestad, H.F. Svendsen, Capital costs and energy considerations of different alternative stripper

SC
configurations for post combustion CO2 capture, Chem. Eng. Res. Des. 89.8 (2011) 1229-1236.
[61] ] M. Ko, J. Im, J.Y. Sung, H. Kim, Liquid-Liquid Equilibria for the Binary Systems of Sulfolane with Alkanes, J Chem.
Eng. Data 52.4 (2007) 1464-1467.

U
[62] J. Im, H. Lee, S. Lee, H. Kim, Liquid–liquid equilibria for the binary systems of sulfolane with branched cycloalkanes,
Fluid Phase Equilib. 52.4 (2007) 34-38.
N
[63] Y.X. Yu, M.Y. He, G.H. Gao, Z.C. Li, Boiling points for five binary systems of sulfolane with aromatic hydrocarbons at
101.33 kPa, Fluid Phase Equilib. 190.1 (2001) 61–71.
A
[64] O. Redlich, J.N. Kwong, On the thermodynamics of solutions; an equation of state; fugacities of gaseous solutions,
Chem. Rev. 44.1 (1949) 233-244.
M

[65] S. Hu, B.J. Zhang, X.Q. Hou, D.L. Li, Q.L. Chen, Design and simulation of an entrainer-enhanced ethyl acetate reactive
distillation process, Chem. Eng. Process. 50.11-12 (2011) 1252-1265.
ED

[66] V. Aniya, D. De, B. Satyavathi, Comprehensive Approach toward Dehydration of tert-Butyl Alcohol by Extractive
Distillation: Entrainer Selection, Thermodynamic Modeling and Process Optimization, Ind. Eng. Chem. Res. 55.25 (2016)
6982-6995.
E PT
CC
A
Figure captions:

Fig. 1 Overall framework of this work

Fig. 2 Flowsheet of the existing ED process

Fig. 3 Components and ratios of hot utilities for the existing process

T
Fig. 4 Temperature and liquid composition profile results of column T1

IP
Fig. 5 Effect of sulfolane solution flowrate on the relative volatility of non-aromatic and aromatic hydrocarbons

R
Fig. 6 Material balance lines for the existing ED process

SC
Fig. 7 Flowsheet of the new aromatic ED process

Fig. 8 Material balance lines for the new ED process

U
N
Fig. 9 Optimization procedure for the adjustable parameters of the new process
A
Fig. 10 Effect of NS2 and TS12 on QR, T1 for the existing process
M

Fig. 11 Effect of TS12 on QR, T1 for the existing process

Fig. 12 Effect of the exhaust heat used for the heating of S2 or S12
ED

Fig. 13 Effect of RT3 and NS5 on Cv for the existing process

PT

Fig. 14 Effect of xCH,S12 on Cv and FS12 for the existing process

Fig. 15 Effect of PF1 and NS13 on Cv for the new process.

E
CC

Fig. 16 Effect of FS16 and xCH,S12 on Cv for the new process

Fig. 17 Effect of FS16 and TS2 on Cv for the new process

A

Fig. 18 The final optimal process flow diagrams for the new process

Fig. 19 The theoretical TAC in the two processes on varying ΔTmin

Fig. 20 The possible optimal HEN design for the optimal existing processes

Fig. 21 The possible optimal HEN design for the new processes
 For an existing solfolane ED process : Energy analysis,
separation effect analysis and material balance analysis

Energy analysis: Energy is Separation effect analysis: Material balance analysis: a

mainly used for the vaporization Maximizing the flowrate of lower purity lean solution
of the crude feed and the heating lean solution can maximize stream increases the
of the lean solution the relative volatility energy efficiency.

Optimization of columns’ A new process with a flash tank

operating parameters to decompose the rich solution

Thermodynamic

T
model modification

IP
process simulation process simulation Optimization program
and optimization and optimization depicted in Fig.8

R
Result for the Result for the new
existing process process

SC
Comparison

Fig. 1 Overall framework of this work.

U
N
A
M

Outer material
1st Inner material balance envelope
balance envelope E3 D1 D1
Non-aromatic S6
purification column Nonaromatic product
T2 S7
ED

S3 Dirty water

E4 nd
2 Inner material
S4 balance envelope

F E5
Extractive distillation
PT

S1 S2 T1 column D2 D2
Crude feed E1 S8
Aromatic product
S9
T3 Dirty water
E2 Entrainer recovery
B1
E

column
S5
S11

E7 E6
CC

S12 B2
S10
Lean solution cooler

Fig. 2 Flowsheet of the existing ED process.

A
 Fig. 3 Components and ratios of hot unities for the existing process.

T
R IP
SC
U
N
A
Fig. 4 Temperature and liquid composition profile results of column T1.
M
ED
E PT
CC
A

Fig. 5 Effect of sulfolane solution flowrate on the relative volatility of non-aromatic and aromatic hydrocarbons.
 T
IP
Fig. 6 Material balance lines for the existing ED process.

R
SC
Outer material
balance envelope
E3 D1
1st Inner material D1
balance envelope Nonaromatic S6
rectifying column
S3
T2
S7
U Nonaromatic product
Dirty water
N
E4
S12 S4
2nd Inner material
A
F Extractive distillation balance envelope
S1 S2 column E5
Crude feed E1 T1 D2
M

D2
S8
S13 Aromatic product
S9
Dirty water
E2
B1 Entrainer recovery
E10 S5 F1 T3 column
ED

B2'
S14 S11
Semi-lean solution cooler

E7 B2 E6
S10
Lean solution cooler
PT

Fig. 7 Flowsheet of the new aromatic ED process.

E
CC
A
 T
R IP
Fig. 8 Material balance lines for the new ED process.

SC
U
N
A
M
ED
EPT
CC
A
 Fix PT1, PT2, PT3 , TS12, TS16, RT3, NS14
Fixed base on conventional
Initial guess of NS2, NS16, NS13, PF1, FS16 , TS2 , xCH,S12 optimal process

Fix xCH,S12
Initial set and guess
Set TS2

Set FS16

T
Set PF1 =PT3,FS13

IP
Adjust of column operating parameters to
achieve the desired product specification

R
SC
Change NS16 and NS2 till Cv minimized

T1 optimization Optimized by
Is the product No Aspen Plus
specification satisfied?

Yes U
N
Adjust of column operating parameters to
achieve the desired product specification
A
M

Change NS13 till Cv minimized

T3 optimization
Is the product No
ED

specification satisfied?

Yes
No
PT

Is PF1 = PT3,FS13?

Yes
E

Is CV minimal with No Iterated by MATLAB

CC

FS16 fixed? program

Yes

Is CV minimal with No
A

TS2 fixed?
Yes
Opimized results

Fig. 9 Optimization procedure for the adjustable parameters of the new process.
 (GJ/h)
50

R,T1
Reboiler duty of T1 / Q 49.5

49

48.5

T
50
45
110 112

IP
Fee 40 108
d 104 106
stag
e of S 35 102 ( C)
100 /T S12
2/N 98 lutio n
30 94
96
re of lean so
S2 ratu
Temp e

R
SC
Fig.10 Effect of NS2 and TS12 on QR, T1 for the existing process.

U
N
49.8
(GJ/h)

A
49.6
R,T1

49.4
pp m
M
Reboiler duty of T1 / Q

00
=1 5
49.2 pp m
x CH ,S1 2 1 00 0
=
49 pm
x CH ,S1 2 5 00 p
=
ED

48.8 x CH ,S1 2

48.6

48.4
PT

94 96 98 100 102 104 106 108 110 112

T emperature of lean solution / T S12 (C)

Fig. 11 Effect of TS12 on QR, T1 for the existing process.

E
CC
A
 T
IP
Fig. 12 Effect of the exhaust heat used for the heating of S2 or S12.

R
SC
U
N
A
M
ED

Fig. 13 Effect of RT3 and NS5 on Cv for the existing process.

EPT
CC
A
 280 286

Flowrate of entrainer / F S12 (10 kg/h)

Operating cost per unit time / C ($/h)
Operating cost per unit time / Cv ($/h)

3
v
Flow rate of entrainer / FS12 (103kg/h)
260 284

240 282

T
220 280

IP
200 278

R
0 500 1000 1500 2000 2500
Hydrocarbon concentration of lean solution / x (ppm)
CH,S12

SC
Fig.14 Effect of xCH,S12 on Cv and FS12 for the existing process

U
N
207
P F1=P T3,S20
A
P F1=P T3,S18
Operating cost per unit time / C ($/h)

P F1=P T3,S16
206.5
P F1=P T3,S14
v

P F1=P T3,S12
206

P F1=P T3,S10
ED

205.5
P F1=P T3,S8
PT

205 Operable area

P F1=P T3,S6
Operable area div ide line
P F1=P T3,S4 Substandard operating points
204.5
4 6 8 10 12 14 16 18 20
E

Feed stage of S13 / N S13

CC

Fig. 15 Effect of PF1 and NS13 on Cv for the new process

A
 Operating cost per unit time / C ($/h)
250
v
240

230

220

210

T
200

IP
20
Sem 30
i-le 0 )
an s olu
40
200 (p pm
50 x CH S12
tion tion /

R
400
flo w 60 600 n s olu
rate 70 o f lea
/F 800
en tr ation
(1 0 3
80 1000 con c
arbo n

SC
S1 6
k g/
h) H yd roc

Fig. 16 Effect of FS16 and xCH,S12 on Cv for the new process.

U
N
A
M
Operating cost per unit time / C ($/h)

205
ED
v

200

195
PT

190

185
E

20
Sem 30
i-le 40 98
CC

an s olu 50 102
tion 60 106
( C)
flo w 70 110
rate d /T S2
/F 80 114 of fee
90 118 p e ratu re
S1 6 (1 0 3 Te m
k g/
h)
A

Fig. 17 Effect of FS16 and TS2 on Cv for the new process.

 FS3=13.106 ×103 kg/h
xA in NA 1w.%
E3 FS6=12.726 ×103 kg/h
D1
xA,S6 1w.% Nonaromatic
23.95 kg/h Nonaromatic RT2=0.5 S6 product
rectifying column NS3=35 FS7=356.05 kg/h
T2 QR, T2=3.07 GJ/h xH2O,S7 99.9w.%
Extractive S3 S7 Dirty water
crude feed TS2=117.8 °C distillation column
S1 E4
S2 NS16=8 E5
T1 NS2=42 S4 FS8=43.035 ×103 kg/h
22.72 w.% A E1 FS13=36.604 ×103 kg/h D2
77.28 w.% NA QR, T1=49.47 GJ/h x =6.3 w.% xNA, S8 1500ppm Aromatic
3 S13 S,S13 S8 product
55.716 ×10 kg/h
E2 FS9=2139.13 kg/h
Entrainer RT3=0.3
recovery column xH2O ,S9 99.9w.% Dirty water
N =5 S9
S5 F1 PF1=PT3,S5 T3 NS13=25
S14
3
FS16=60×10 kg/h FS14=285.919 ×103 kg/h QR, T3=11.21 GJ/h
xCH,S16=3.3 w.%, TS16=98 °C E10 xS,S14 =96.6 w.% FS11=2531.13 kg/h
S16 S15 S14 Stripping
S11 stream
Semi-lean solution cooler

T
FS12=279.88×103 kg/h E6
xCH,S12=1000 ppm, TS12=98 °C E7
S12 S10
Lean solution cooler

IP
Fig. 18 The final optimal process flow diagrams for the new process.

R
SC
230

TAC for conventional optimal process

TAC per unit time/ Cv ($/h)

220
TAC for improved process

U
N
A
210
M

200
4 6 8 10 12
ED

ΔTmin / °C

Fig. 19 The theoretical TAC in the two processes on varying ΔTmin.

E PT
CC
A
 T
IP
Fig. 20 The possible optimal HEN design for the optimal existing processes.

R
SC
U
N
A
M

Fig. 21 The possible optimal HEN design for the new processes.
ED
E PT
CC
A
Tables:

Table 1 Composition and flowrate of crude feed

Table 2 Typical operating conditions of the existing ED process

Table 3 Composition data of streams on a H2O-free basis

T
Table 4 Binary interaction parameters containing sulfolane

IP
Table 5 Plant data and simulation results

R
Table 6 Unit prices of utilities

SC
Table 7 Optimal operating parameters and quality of products for existing process

Table 8 Optimal operating parameters and quality of products for the new process

U
N
Table 9 Heat duty, FS11 and minimized CV results for the three processes
A
Table 10 Stream information of the optimal existing and new process
M

Table 11 The details of operating costs and capital costs for the two designs
ED
E PT
CC
A
 Table 1 Composition and flowrate of crude feed

Components Mass Fraction

PC4 49ppm

PC5 0.016

PC6 0.108

T
PC7 0.055

IP
PC8 0.014

R
SC
NC6 0.019

NC7 0.012

NC8
U0.004
N
BEN 0.378
A

TOL 0.387
M

PX 0.002
ED

MX 0.004

OX 304ppm
EPT
CC
A
 Table 2 Typical operating conditions of the existing ED process

Column Number of stage Feed stage Reflux ratio Ptop/MPag Pbot/MPag Ttop/°C Tbot/°C TF/°C

T1 80 1/41 NA 0.07 0.15 114 170 108/95

T2 12 12 0.5 0.06 0.07 96 140 114

T3 34 21 0.7 -0.05 -0.025 64 176 170

T
R IP
SC
U
N
A
M
ED
E PT
CC
A
 Table 3 Composition data of streams on a H2O-free basis

Stream NO.

F D1 D2 B1 B2
Composition

xNA 0.227 0.99 0.9985 0.0002 0

xA 0.773 0.01 0.0015 0.115 0.0015

T
xS 0 0 2ppm 0.8848 0.9958

R IP
SC
U
N
A
M
ED
EPT
CC
A
 Table 4 Binary interaction parameters containing sulfolane

Pair ij aij aji bij bji cij eij eji AAD% Temp Range/K Ref.

[
PC5-SUL 182.16 -175.98 -4130.99 9713.05 0.20 -28.46 25.42 0.19 304.31-380.35 60]

PC6-SUL -135.41 -140.77 11035.23 8725.38 0.20 18.48 20.09 1.95 300.3-429.93 [60]

PC7-SUL -30.65 -186.39 4437.96 11035.05 0.20 3.74 26.79 1.46 302.09-429.24 [60]

T
PC8-SUL -136.82 29.57 9921.62 -215.64 0.20 19.08 -4.55 1.42 298.15-373.15 [60]

IP
NC6-SUL -20.24 17.42 1512.98 746.09 0.20 3.23 -3.06 0.13 303.1-452.67 [28]

R
[6 ]

SC
NC7-SUL -143.49 -22.46 10000.00 2340.59 0.20 20.20 2.95 1.94 303.96-464.91 1

NC8-SUL 1.63 22.68 1483.83 317.53 0.20 -0.50 -3.64 0.11 314.29-451.14 [28]

EB-SUL -80.00 80.00 -5240.82 -7906.16 0.30 15.71 -10.20

U
0.88 411.69-492.15 [62]
N
PX-SUL -16.91 -11.83 4824.47 -1246.68 0.30 1.16 2.55 0.32 412.97-452.45 [62]
A

MX-SUL -63.73 -71.06 9527.07 1010.88 0.30 7.05 11.45 0.35 412.81-445.89 [62]
M

OX-SUL -14.04 -10.18 4577.22 -1717.93 0.30 0.73 2.45 0.22 419.16-468.49 [62]
ED
E PT
CC
A
 Table 5 Plant data and simulation results

T1 T3
Item

Ttop/°C Tbot/°C D/kg/h Ttop/C Tbot/C D/kg/h

Operating data 114 170 13120 64 176 42925

Simulation results 115.9 172 13302 62 176 43028

T
AAD% 1.67 1.18 1.39 3.13 0 0.24

R IP
SC
U
N
A
M
ED
EPT
CC
A
 Table 6 Unit prices of utilities

Utility stream Inlet T/°C Outlet T/°C Parity/$∙GJ-1

Cooling water 20 25 0.2125

Low pressure steam 125 124 1.9

Medium pressure steam 175 174 2.2

T
High pressure steam 250 249 2.5

R IP
SC
U
N
A
M
ED
EPT
CC
A
 Table 7 Optimal operating parameters and quality of products for existing process

Unit/Stream Item Value

Feed stage of S2 36

Temperature of S12 / S2, ℃ 98 / 95

T1
Top/ bottom pressure of column, MPag 0.07 / 0.15

T
Reboiler duty of T1, GJ/h 48.75

IP
T2 Top/ bottom pressure of column, MPag 0.06 / 0.07

R
SC
Feed stage of S5 / S11 25 / 35

Top / bottom pressure of column, MPag 0.035/0.075

T3 Mass reflux ratio

U
0.3
N
Flowrate of striping water, kg/h 3211.6
A

Reboiler duty of T3, GJ/h

M

16.81

Aromatic product stream xS, S8≤2ppm, xN, S8≤1500ppm (w/w)

ED

Stream Non-aromatic product stream xA, S6≤1% (w/w)

PT

entrainer xCH,S12=1000ppm, FS12=281.3×103kg/h

E
CC
A
 Table 8 Optimal operating parameters and quality of products for the new process

Unit/Stream Item Value

Feed stage of S16 / S2 8 / 42

Temperature of S12 / S16 / S2 98 / 98 / 117.8

T1 Top/bottom pressure of column, MPag 0.07 / 0.15

T
Lean/semi-lean solution circulation, 103kg/h 60 / 279.88

IP
Reboiler duty of T1, GJ/h 49.47

R
SC
T2 Top/ bottom pressure of column, MPag 0.06 / 0.07

Feed stage of S13 / S14 / S11 5 / 25 / 35

Top/bottom pressure of column, MPag

U
-0.05 / -0.025
N
T3 Mass reflux ratio 0.3
A

Flowrate of striping water, kg/h

M

2531.13

Reboiler duty of T3, GJ/h 11.21

ED

Aromatic product stream xS, S8≤2ppm, xN, S8≤1500ppm

xA, S6≤1% (w/w)

PT

Stream Non-aromatic product stream

entrainer xCH,S12=1000ppm, xCH,S16=0.033

E
CC
A
 Table 9 Heat duty, FS11 and minimized CV results for the three processes

Existing Optimal existing process

Item New process
process

QR, GJ/h 82.03 68.63 63.75

QC, GJ/h 108.26 77.33 68.06

T
Stripping steam, kg/h 7278 3211.6 2531.1

IP
CV, 106$/year 2.381 1.731 1.578

R
SC
U
N
A
M
ED
EPT
CC
A
 Table 10 Stream information of the optimal existing and new process

Optimal existing process New process

Streams Tsupply / °C Ttarget /°C H / GJ·h-1 Streams Tsupply / °C Ttarget /°C H / GJ·h-1

S1-S2 40 95 6.56 S1-S2 40 117.8 17.35

S11 40 176 8.55 S11 40 176 6.78

T
Cold
SR, E2 133.3 168.2 48.75 SR, E2 140.1 173.7 49.49

IP
streams
SR, E4 148.8 176 16.8 SR, E4 157.3 176 11.21

R
SC
SR, E6 114.6 140 3.07 SR, E6 114.6 140 3.07

SC, E3 97.7 40 10.14 SC, E3 97.8 40 10.11

Hot SC, E5 69.2 40 34.18

U
SC, E5 70.8 40 32.15
N
streams S10-S12 176 98 39.57 S10-S12 176 98 39.33
A
M

S15-S16 157.7 98 6.38

ED
E PT
CC
A
 Table 11 The details of operating costs and capital costs for the two designs

Item CV H, 105$/year CV C, 104$/year CV, 105$/year CI, 105$ TAC, 106$/year

Optimal existing process 9.65 8.85 10.54 8.44 1.38

New process 10.08 -5.84 9.50 6.48 1.20

T
R IP
SC
U
N
A
M
ED
E PT
CC
A