Article

Cite This: Ind. Eng. Chem. Res. 2017, 56, 14557−14564 pubs.acs.org/IECR

Process Synthesis and Optimization of Propylene/Propane

Separation Using Vapor Recompression and Self-Heat Recuperation
Chang Chu En Christopher, Arnab Dutta, Shamsuzzaman Farooq, and Iftekhar A. Karimi*
Department of Chemical and Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4, Singapore 117585
*
S Supporting Information

ABSTRACT: Propylene/propane separation is one of the

most energy-intensive processes in the chemical industry. In
this study, the concepts of mechanical vapor recompression
(MVR) and self-heat recuperation (SHR) are simultaneously
considered for the synthesis of four distillation-based
configurations. Each configuration is optimized for the
minimum total annualized separation cost using a simu-
lation-based optimization framework with Aspen HYSYS as
the process simulator and particle swarm optimization in
MATLAB as the optimizer. The best configuration uses both
MVR and SHR. MVR eliminates the need for reboiler steam,
and SHR utilizes the remaining sensible heat to preheat the
feed and the compressor inlet stream in MVR. This reduces
the energy consumption by 45% and the separation costs by 20% as compared to the values for an MVR design with no SHR.

1. INTRODUCTION splitter (to separate propylene from propane) is almost

Propylene is one of the most important petrochemicals inevitable in any propylene production unit.
According to the U.S. Department of Energy, the separation
produced globally today. Propylene is an essential raw material
of propylene/propane mixtures by distillation is one of the
in the manufacture of a variety of propylene derivatives such as
most energy-intensive commercial distillation processes.4,5
polypropylene, acrylonitrile, cumene, and acrylic acid. These
Separation of propylene/propane mixtures by distillation
derivatives are then processed to produce end products ranging
occurs either at high pressures (>20 bar) or at low pressures,
from furniture, automobile parts, textiles, and packaging films to where the latter requires cryogenic conditions.6,7 At high
common household objects. Thus, it is evident that propylene pressures, the relative volatility of propylene and propane drops
is an important platform chemical. The global annual to about 1.08−1.12, depending on the mixture composition.8
production capacity of propylene was estimated at 114 million This results in high energy requirements at the reboiler, thus
tons in 2015, second behind ethylene among organic making this process highly energy-intensive. On the other hand,
chemicals.1 The global demand for propylene has been although the relative volatility of the components is
projected to rise faster than that of ethylene, growing at a considerably higher at lower pressures, the condenser needs
rate of 5% annually.2 to operate at cryogenic conditions to enable condensation of
About 85% of the global propylene demand is met by the top product. Among these two options, distillation at high
conventional methods such as thermal cracking and fluid pressure is generally favored from a cost perspective, because
catalytic cracking.1 Multiple hydrocarbon products, including cooling water can be used as a cooling medium in the
light alkanes, alkenes, and aromatics, are generated through condenser, instead of refrigerants, which are expensive. Note
thermal cracking at high temperatures, and numerous that, as propylene and propane form a close-boiling mixture, C3
separation steps are required before key products such as splitters can even require up to 240 stages for separation.9,10
propylene and ethylene can be obtained. Other propylene The energy-intensive nature of propylene/propane separa-
production methods include on-purpose technologies (OPTs) tion has prompted researchers to explore alternative separation
such as propane dehydrogenation, olefin metathesis, and technologies such as extractive distillation, 11 membrane
methanol-to-olefins and coal-to-olefins processes.3 Propylene separation,12,13 and adsorption.4,14,15 There are also hybrid
obtained from any of these sources is mixed with propane (and designs that combine different separation technologies, such as
trace amounts of other hydrocarbons) and needs to be further
purified for downstream applications. The manufacture of Received: August 21, 2017
polypropylene, which accounts for about two-thirds of Revised: October 22, 2017
propylene consumption, requires polymer-grade propylene Accepted: November 16, 2017
with a purity of 99.5% propylene by weight.1 Thus, a C3 Published: November 17, 2017

© 2017 American Chemical Society 14557 DOI: 10.1021/acs.iecr.7b03432

Ind. Eng. Chem. Res. 2017, 56, 14557−14564
Industrial & Engineering Chemistry Research Article

hybrid membrane distillation,12 hybrid adsorption distillation drums, and heat exchangers are all assumed to be zero.
schemes,16 and hybrid distillation−pervaporation schemes.17 Therefore, the presence of several pumps (e.g., bottom of flash
Eldridge presented an exhaustive review of different separation drums and condensers) and compressors common to all
technologies that have been applied to olefin/paraffin mixtures configurations is neglected. The minimum approach temper-
such as propylene and propane.18 Despite the proliferation of ature for all heat exchangers is assumed to be 5 °C. Aspen
alternative separation technologies, distillation remains the HYSYS, version 9.0,37 is used to simulate and evaluate alternate
primary means by which propylene and propane are separated configurations for propylene/propane separation with the
in industry. Jana19 and Kiss et al.20 have presented excellent Peng−Robinson equation of state as the fluid package. The
reviews of various energy-efficient distillation technologies for a Peng−Robinson equation of state is appropriate for propylene/
variety of binary and multicomponent separation mixtures. propane simulation as it shows a good fit between the
Mechanical heat pumps (MHPs), which use electrical or experimental and simulation data.38
mechanical energy to move heat from a heat source at a lower 2.2. Configuration 1: Conventional Distillation. A high-
temperature to a heat sink at a higher temperature, have been pressure distillation column with CW as the coolant in the
successfully integrated with distillation columns, providing condenser is conventionally used in the industry for C3
substantial energy savings without the need for major splitting. Figure 1 shows the process flow diagram (PFD) for
modifications to the existing process.21 MHPs can be further
broken down into three basic configurations: mechanical vapor
recompression (MVR), external vapor recompression, and
bottom flashing. Kazemi et al. 22 reduced the energy
consumption of a benzene/toluene distillation process by
modifying the conventional MVR design. Waheed et al.23
enhanced the performance of a de-ethanizer unit by optimizing
the MVR operating parameters. MVR has also been applied to
processes such as desalination24 and food product manufactur-
ing.25 Self-heat recuperation (SHR) is another energy-saving
technology that can improve the energy efficiency of a process
by circulating waste heat within the process streams to satisfy
heating requirements. SHR has been applied to a variety of
distillation processes including distillations of hydrocarbon
systems,26−28 crude oil distillation,29 and cryogenic air
separation,30 each showing approximately 20−50% reduction
in energy requirements. Internally heat integrated distillation Figure 1. Configuration 1: Conventional distillation.
columns (HIDiC), where heat transfer occurs directly between
the stages in the stripping and rectification sections of a
column, have also been extensively applied to energy-intensive
processes such as propylene/propane separation.9,31,32 How- this configuration. With the column at 18 bar and 99.6%
ever, HIDiC has not been implemented on an industrial-plant propylene purity, the number of stages (NDC) and the feed
scale because of design and construction complexities.19,33,34 stage location (FDC) are the only decision variables available for
Recognizing the practical implementation issues with HIDiC optimizing this configuration.
that are yet to be fully resolved, the focus of this article is on 2.3. Configuration 2: Mechanical Vapor Recompres-
MVR and SHR. It is evident from the open literature that MVR sion (MVR). MVR is a specific type of mechanical heat pump
has been studied for the propylene/propane system. However, (MHP) system in which the process fluids (in this case, the top
the synergistic effect of incorporating the concepts of MVR and and bottom products of the C3 splitter) exchange heat with
SHR for a propylene/propane system has not yet been each other, as shown in Figure 2. The vapor stream from the
addressed. Thus, in this study, we propose different process column top at (Pt, Tt) undergoes compression, which increases
configurations for a C3 splitter based on the concepts of MVR its pressure and temperature to (P1, T1). The thermal energy
and SHR. We also optimize each of these proposed process (both latent and sensible) of this stream provides the reboiler
configurations within a simulation-based optimization frame- heat with the condition that T1 > Treb. T1 and the compressor
work to obtain the configuration with the minimum separation
cost for propylene.

2. PROCESS SYNTHESIS
2.1. Design Basis. According to the proprietary data from a
local petrochemical company, a C3 splitter receives an 85:15
mol % liquid propylene/propane feed at 50 °C and 21 bar from
the upstream C3 H4 hydrogenation unit. The required
specifications for the product streams from the C3 splitter are
99.6 wt % propylene purity (polymer-grade propylene9) with
99.0% recovery.35 These result in a 95.0 wt % pure propane
product (HD-5 propane36). The column uses a total condenser
at 18 bar with 32 °C cooling water (CW) and a partial reboiler.
A pressure drop of 0.7 kPa is assumed for each stage in the C3
splitter. Pressure drops across the condenser, reboiler, flash Figure 2. Configuration 2: Mechanical vapor recompression.

14558 DOI: 10.1021/acs.iecr.7b03432

Ind. Eng. Chem. Res. 2017, 56, 14557−14564
Industrial & Engineering Chemistry Research Article

power depend on the compressor pressure ratio ( ) and inlet

P1
Pt
vapor temperature (Tt). After supplying heat to the reboiler, the
exit stream at (P1, T2) is expanded to Pt. This results in a two-
phase mixture that is then separated in a flash drum, with the
vapor being recondensed using CW before being mixed back
with the liquid. A part of this liquid is removed as the propylene
product, and the rest is used as reflux. Thus, NDC, FDC, and the
compressor pressure ratio (PMVR) are the decision variables for
optimizing this configuration.
Now, we develop several modifications of configuration 2
and compare them with configuration 2. To make this Figure 4. Configuration 4: Mechanical vapor recompression with feed
comparison fair and simple, we fix the values NDC and FDC preheating.
for all of the subsequent configurations (configurations 3, 4,
and 5) at the values used for configuration 2. stripping section. Thus, PMVR and the feed vapor fraction (qFP)
2.4. Configuration 3: MVR with Compressor Inlet are the decision variables in the optimization of configuration 4.
Preheating (CIP). In this configuration (Figure 3), the hot 2.6. Configuration 5: MVR with CIP and Feed
Preheating. Configuration 5 (Figure 5) is obtained by

Figure 3. Configuration 3: Mechanical vapor recompression with Figure 5. Configuration 5: Mechanical vapor recompression with
compressor inlet preheating. compressor inlet and feed preheating.

stream exiting the reboiler preheats the column vapor before

combining configurations 3 and 4. The energy from the
the vapor is compressed (this is called compressor inlet
vapor stream exiting the compressor is first used to supply the
preheating or CIP). Thus, the compressor pressure ratio
reboiler duty, and then the sensible heat is used for CIP and
(PMVR) and compressor inlet temperature (TCIP) are the
finally for feed preheating. The rest of the process remains the
decision variables for this configuration. The preheating of the
same. Thus, PMVR, TCIP, and qFP are the decision variables in the
column vapor gives it some of the heat it will eventually need
optimization of configuration 5.
for the reboiler. As the total reboiler duty remains the same, the
compressor now needs to supply less energy to compress the 3. PROCESS OPTIMIZATION
column vapor. Therefore, the compressor needs a lower PMVR
compared to configuration 2 to satisfy the reboiler duty. This We minimize the total annualized cost (TAC) of separating
reduces the operating and capital costs of the compressor. propylene from propane by distillation for a given feed and
Table 2 (below) lists the optimal PMVR values for configurations product specifications as defined in the design basis according
2−5, and one can see that PMVR is indeed lower for to the equation
configurations 3−5. The resultant cooled stream at (P2, T6) min TAC (US$/year) = OPEX + f × (CAPEX) (1)
then follows the same processing steps as in configuration 2 to
deliver liquid propylene. where OPEX is the operating cost per year and CAPEX is the
2.5. Configuration 4: MVR with Feed Preheating. The total grassroots cost for the purchase and installation of all
sensible energy extracted from the reboiler exit stream in major equipment, updated to current-day prices using the
configuration 3 can also be used to preheat the feed to the Chemical Engineering Plant Cost Index (CEPCI) of 556.8 for
distillation column, as shown in Figure 4. Note that both January 2016.39 Equation 2 defines the annualization factor, f,
configurations 3 and 4 utilize the available sensible heat based which is used to capitalize the investment cost over the plant
on the concept of SHR. A preheated feed reduces the reboiler equipment lifetime42
duty and lowers the compression ratio in comparison to that i(1 + i)n
required for configuration 2. However, the pressure ratio is f=
(1 + i)n − 1 (2)
expected to be higher than that of configuration 3 because of
the lower compressor inlet temperature. The upper limit set on where i is fractional interest rate per year and n is the lifespan of
the preheating is that the column feed must not be superheated plant equipment; 0.1 and 20 years are used as the values of i
vapor, which can lead to operating issues such as drying of the and n, respectively. Cost parameters and correlations for both
14559 DOI: 10.1021/acs.iecr.7b03432
Ind. Eng. Chem. Res. 2017, 56, 14557−14564
Industrial & Engineering Chemistry Research Article

OPEX and CAPEX follow the methodology as outlined by through a Component Object Model (COM) in ActiveX,
Turton et al.,40 with the cost of electricity taken as $0.0723/ which allows direct two-way communication between HYSYS
kWh as obtained from the U.S. Energy Information and MATLAB.50 As shown in Figure 6, stream and process data
Administration for the industrial sector in August 2016.41
Note that each equipment unit has a specified range of sizes or
capacities within which the cost correlations are valid, and
outside the given ranges the cost estimates become inaccurate.
To ensure that the cost correlations can be applied without
violating the ranges specified by Turton et al.40 the feed from
the upstream process is split into identical parallel separation
trains. Details about the costing methodology are provided in
the Supporting Information.
3.1. Constraints. The constraints for the different
configurations are listed in Table 1.

Table 1. Constraints for the Optimization Framework

variable constrained value
propylene purity 99.6 wt %
propylene recovery 99.0% Figure 6. HYSYS−MATLAB interface.
minimum approach temperature for heat ≥5 °C
exchangers
compressor 1 < PMVR < 4 obtained from HYSYS are sent to MATLAB, which performs
column feed 0 ≤ vapor fraction ≤ 1 the optimization, and the decision variables obtained at each
iteration of the optimization are sent back to HYSYS to
perform the process simulation.
3.2. Simulation-Based Optimization. Derivative-based An overview of our optimization algorithm is provided in
optimization algorithms require accurate gradient information, Figure 7. NDC and FDC are the discrete decision variables for the
which is not readily available from process simulators. Although distillation column. We optimize these variables for config-
derivatives for the variables can be indirectly obtained through urations 1 and 2 as follows: We fix the continuous variables
perturbation, such an approach involves significantly longer (PMVR, TCIP, qFP) at some nominal values and then compute the
computational times because the flowsheet within the process total annualized cost for each possible combination of NDC and
simulator needs to be solved each time a variable is perturbed. FDC. We choose the combination with the minimum cost. Once
Moreover, the numerical noise inherent in process simulators we have determined the best NDC and FDC values for
affects the accuracy of computed derivatives.43 These problems configuration 2, we then use the same values for configurations
can be avoided through metaheuristic algorithms suitable for 3−5, so no discrete optimization is needed for these
black-box optimization, where optimal solutions can be configurations.
obtained through population-based search techniques without For the optimization of the continuous variables, we use the
requiring any derivative information.44 In metaheuristic PSO algorithm. A key parameter in the PSO framework is the
algorithms, individuals in the population move within a population size or the number of particles in a swarm. A series
predefined search space to locate the optimal solution.43 In of computational trials is performed to analyze the trade-off
this study, particle swarm optimization (PSO) is selected as the between computational time and objective-function value. As
optimization tool because of the relative ease of implementa- population size increases, the optimal solution improves
tion, requiring only a few tuning parameters.45 The PSO marginally, but the computational time increases significantly
algorithm involves an element of learning and communication because more function evaluations must be performed. The
between particles (individuals) within the swarm (population), number of iterations is another key parameter, because PSO
and the movement of each particle is governed by the involves a social learning component that updates the
respective best-known position and the best-known positions movement of the particles toward the best solution obtained
of the entire population. The entire swarm then gradually for each iteration. As the number of iterations increase, the
moves toward the optimal solution with each increasing solutions gradually converge to the optimal solution. A
iteration.46,47 PSO effectively handles exploration and ex- population size of 15 particles with a maximum number of
ploitation through the damping coefficient, w, which provides iterations of 20 is used for all simulations reported in this study.
particles with a high velocity initially so that exploration of the The PSO is initialized by assigning each particle a random
search space can occur and later decreses the velocity to obtain position within the predefined search space bounded by the
solutions geared toward exploitation in the neighborhood of ranges of the decision variables. The objective function is then
the optimal solution.48 However, none of these metaheuristic evaluated for each particle. Equation 1 has been modified to
algorithms can guarantee a global optimal solution. include a penalty term to handle constraint violations, as shown
A simulation-based optimization framework combines the in the equation
benefits of using a process simulator, which seamlessly performs
all thermodynamic calculations, with an external platform in ⎛ US$ ⎞
which rigorous optimization algorithms can be implemented. In min TAC ⎜ ⎟ = OPEX + f × (CAPEX) + z
⎝ year ⎠ (3)
this work Aspen HYSYS as a process simulator is interfaced
with MATLAB, version 2013b, as an external platform.49 A The penalty parameter (z) is of the same order of magnitude as
connection between MATLAB and HYSYS can be established the other terms in eq 3.
14560 DOI: 10.1021/acs.iecr.7b03432
Ind. Eng. Chem. Res. 2017, 56, 14557−14564
Industrial & Engineering Chemistry Research Article

Figure 7. Optimization framework.

Table 2. Comparison of All Five Configurations for Producing 0.45 MTPA of Propylene
configuration
1 2 3 4 5
no. of stages 240 147 147 147 147
feed stage 176 98 98 98 98
compression ratio (PMVR) − 2.15 1.52 1.65 1.49
compressor inlet temperature (°C) (TCIP) − 44.0 54.0 44.0 48.0
feed vapor fraction (qFP) 0.03 0.03 0.03 1.0 1.0
MVR power consumption (MW) − 14.6 8.8 9.7 8.1
thermal energy requirement (GJ/h) 260.3 − − − −
thermal energy equivalent of electricity consumed in MVR−SHR (GJ/h) − 157.7 95.0 104.8 87.5
total annualized cost (106 US$/year) 36.1 20.4 17.5 16.8 16.4
propylene separation cost (US$/ton) 80.2 45.2 38.9 37.3 36.3

The velocity and position of each particle are updated changes. Φ, a dimensionless variable indicating the relative
according to the equations given in Figure 7, which considers position of the feed stage within the column, is defined as
both the best-known positions of the individual particles and
the best-known positions of the entire population. At each Φ = FDC/NDC (4)
iteration, the algorithm updates the best-known position of a
particle if a better position is obtained; otherwise, it retains the Based on the trends in NDC and FDC, the search space for
configuration 1 is constrained to the ranges 200−240 and 0.4−
previous best position, and the particle with the minimum
0.8 for NDC and Φ, respectively. The optimal column design for
function value is considered as the best particle within the configuration 1 has 240 stages with the feed stage at stage 176
entire population. This iterative process continues until a (from the top).
specified termination criterion is met. The terminations of both When MVR is incorporated into the design, a trend of TAC
the exhaustive search and PSO optimization algorithms require increasing with NDC is observed, because consumption of steam
appropriate stopping criteria.51 The exhaustive search is in the reboiler is eliminated. As a result, TAC for configuration
terminated when all combinations of the two discrete variables 2 is dominated by capital cost. Among the components of the
within the search space have been exhausted. For PSO, we use a capital cost, column cost is the most significant. The total cost
distribution-based criterion, in which the algorithm is of the column trays and the total volume of the column are
terminated if the objective-function values for the best and directly proportional to NDC; therefore, total capital cost
worst particles do not differ by more than a threshold in any decreases as NDC decreases. The search space for the
iteration or the limit on the maximum number of iterations has optimization of configuration 2 is constrained to the ranges
been reached. 140−200 and 0.4−0.8 for NDC and Φ, respectively. Beyond the
lower bound of NDC, the specifications for propylene purity and
4. TECHNOECONOMIC ANALYSIS recovery can no longer be achieved. The optimal column
design for configuration 2 has 147 stages with the feed stage at
For configuration 1, TAC decreases monotonically with NDC, stage 98 (from the top). The solution for the discrete variables,
whereas a U-shaped trend is observed for TAC when FDC is NDC and FDC, that are obtained for configuration 2 are also
varied. This is caused by the variation in the reflux ratio as FDC applied to configurations 3−5.
14561 DOI: 10.1021/acs.iecr.7b03432
Ind. Eng. Chem. Res. 2017, 56, 14557−14564
Industrial & Engineering Chemistry Research Article

Because it is stochastic in nature, PSO does not guarantee reduce the costs and energy consumptions for other energy-
that the same optimal solution will be obtained each time PSO intensive distillation processes.
is executed.43,52 The objective function, being nonlinear in
nature, results in slightly different combinations of decision
variables to give solutions with similar objective-function values.
■
*
ASSOCIATED CONTENT
S Supporting Information
Thus, PSO is repeated multiple times to ascertain the validity of The Supporting Information is available free of charge on the
the optimal solution. More information on the performance of ACS Publications website at DOI: 10.1021/acs.iecr.7b03432.
the PSO algorithm is available in the Supporting Information. Stream data for optimal configurations, costing method-
Table 2 presents the optimized results obtained for all
ology, and PSO algorithm results (PDF)

■
configurations. For configuration 1, the operating costs strongly
influence the TAC, whereas for the rest of the configurations
(configurations 2−5), the annualized capital cost is dominant AUTHOR INFORMATION
owing to a reduction of 65−81% in operating costs as the heat Corresponding Author
recovered through MVR completely satisfies the reboiler duty, *E-mail: cheiak@nus.edu.sg. Tel.: +65 6516-6359.
thus eliminating the requirement for steam. The TAC of the ORCID
basic MVR design in configuration 2 is further reduced by Shamsuzzaman Farooq: 0000-0002-6501-5540
incorporating CIP, feed preheating, and a combination of both Iftekhar A. Karimi: 0000-0001-7122-0578
CIP and feed preheating into configurations 3, 4, and 5,
Notes
respectively. In addition to the cost savings obtained from
The authors declare no competing financial interest.

■
configuration 2, these modifications resulted in further
reductions of 14−20% in the TAC. In configuration 3, an
ACKNOWLEDGMENTS
increase in the fluid temperature at the compressor inlet
through the use of CIP results in a reduction in the compressor We acknowledge AspenTech Inc. for allowing the use of
power requirement compared to that of configuration 2. HYSYS under an academic license provided to the National
Implementing the concept of feed preheating in configuration 4 University of Singapore. We also acknowledge MathWorks for
shifts a part of the reboiler duty, which is used to vaporize the providing an academic license for MATLAB to the National
liquid within the column, to the feed preheater. The feed University of Singapore. A.D. thanks the Department of
preheater utilizes the sensible heat (after the required energy Chemical and Biomolecular Engineering, NUS, for providing
has been supplied to the reboiler) to increase the enthalpy of NUS Research Scholarship (NRS). We also acknowledge Mr.
Kuah Wee Chong for his input at different stages of this work.

■
the feed stream. This leads to a reduction in reboiler duty, so
that less energy is to be recovered through MVR. This
ultimately leads to a reduction in the compressor power NOMENCLATURE
requirement compared to that of configuration 2. Configuration Abbreviations
5 makes use of the high-enthalpy stream generated from MVR CAPEX = capital cost (US$/year)
to sequentially transfer heat between the process streams. It CIP = compressor inlet preheating
reaps reductions in TAC coming from the CIP as well as the DV = decision variable
feed preheating and proves to be the best configuration from MHP = mechanical heat pump
the perspectives of both energy and costs. Configuration 5 leads MVR = mechanical vapor recompression
to reductions of 45% in the energy consumption for the OPEX = operating cost (US$/year)
propylene/propane separation process and 20% in the PSO = particle swarm optimization
separation cost as compared to the values for the MVR design SHR = self-heat recuperation
with no SHR (configuration 2). The electricity demands of TAC = total annualized cost (US$/year)
MVR for configurations 2−5 are also converted to a thermal Symbols
energy basis by assuming a 33% efficiency for obtaining f = annualization factor
electricity from thermal energy through the combustion of FDC = feed stage location from the top of the distillation
fuels.53 The thermal energy equivalents of electricity con- column
sumption for all five configurations are presented in Table 2. h = stream enthalpy
i = fractional interest rate per year
5. CONCLUSIONS n = number of years
In this study, four different distillation-based configurations NDC = number of stages in the distillation column
employing various combinations of mechanical vapor recom- Pi = pressure of stream i
pression and self-heat recuperation are proposed, optimized, PMVR = compression ratio of the MVR compressor
and compared in terms of their abilities to reduce the energy qFP = vapor fraction of the feed stream entering the
consumption of a C3 splitter. The two techniques fully distillation column
eliminate the need for low-pressure steam in the reboiler and TCIP = temperature of the suction stream to the MVR
reduce the total energy consumption and separation cost. The compressor
best configuration employs both mechanical vapor recompres- Ti = temperature of stream i
z = penalty parameter
sion and self-heat recuperation and reduces the annualized
separation cost by 20% and the energy consumption by 45%, Greek Symbols
when compared to the configuration with simple mechanical Φ = dimensionless feed stage location
vapor recompression alone. The modifications proposed in this Subscripts
study for the mechanical vapor recompression system can also cond = condenser
14562 DOI: 10.1021/acs.iecr.7b03432
Ind. Eng. Chem. Res. 2017, 56, 14557−14564
Industrial & Engineering Chemistry Research Article

reb = reboiler (23) Waheed, M. A.; Oni, A. O.; Adejuyigbe, S. B.; Adewumi, B. A.;

■
Fadare, D. A. Performance Enhancement of Vapor Recompression
Heat Pump. Appl. Energy 2014, 114, 69.
REFERENCES (24) Aly, N. H.; El-Figi, A. K. Mechanical Vapor Compression
(1) Coombs, D. Propylene: The “Other” Olefin. Presented at the Desalination Systems - A Case Study. Desalination 2003, 158 (1−3),
Goldman Sachs Chemical Intensity Conference, Houston, TX, Mar 15, 143.
2016. (25) Chua, K. J.; Chou, S. K.; Yang, W. M. Advances in Heat Pump
(2) Horncastle, A.; Gotpagar, J.; Ozeir, F.; Singh, S. Global Systems: A Review. Appl. Energy 2010, 87 (12), 3611.
Petrochemical Disruptions: Business Model Innovations for a Dynamic (26) Van Duc Long, N.; Lee, M. A Novel NGL (Natural Gas Liquid)
Market; Strategy&: New York, 2014. Recovery Process Based on Self-Heat Recuperation. Energy 2013, 57,
(3) Moulijn, J. A.; Makkee, M.; van Diepen, A. E. Chemical Process 663.
Technology, 2nd ed.; John Wiley & Sons: Chichester, U.K., 2013. (27) Matsuda, K.; Kawazuishi, K.; Kansha, Y.; Fushimi, C.; Nagao,
(4) Jarvelin, H.; Fair, J. R. Adsorptive Separation of Propylene− M.; Kunikiyo, H.; Masuda, F.; Tsutsumi, A. Advanced Energy Saving
Propane Mixtures. Ind. Eng. Chem. Res. 1993, 32 (10), 2201. in Distillation Process with Self-Heat Recuperation Technology.
(5) Ghosh, T. K.; Lin, H. Da; Hines, A. L. Hybrid Adsorption- Energy 2011, 36 (8), 4640.
Distillation Process for Separating Propane and Propylene. Ind. Eng. (28) Kiuchi, T.; Yoshida, M.; Kato, Y. Energy Saving Bioethanol
Chem. Res. 1993, 32 (10), 2390. Distillation Process with Self-Heat Recuperation Technology. J. Jpn.
(6) Alcántara-Avila, J. R.; Gómez-Castro, F. I.; Segovia-Hernández, J. Pet. Inst. 2015, 58 (3), 135.
G.; Sotowa, K. I.; Horikawa, T. Optimal Design of Cryogenic (29) Kansha, Y.; Kishimoto, A.; Tsutsumi, A. Application of the Self-
Distillation Columns with Side Heat Pumps for the Propylene/ Heat Recuperation Technology to Crude Oil Distillation. Appl. Therm.
propane Separation. Chem. Eng. Process. 2014, 82, 112. Eng. 2012, 43, 153.
(7) Baker, R. W. Future Directions of Membrane Gas Separation (30) Fu, Q.; Kansha, Y.; Song, C.; Liu, Y.; Ishizuka, M.; Tsutsumi, A.
Technology. Ind. Eng. Chem. Res. 2002, 41, 1393. A Cryogenic Air Separation Process Based on Self-Heat Recuperation
(8) Mann, A. N.; Pardee, W. A.; Smyth, R. W. Vapor-Liquid for Oxy-Combustion Plants. Appl. Energy 2016, 162, 1114.
Equilibrium Data for Propylene-Propane System. J. Chem. Eng. Data (31) Shahandeh, H.; Ivakpour, J.; Kasiri, N. Feasibility Study of Heat-
1963, 8 (4), 499. Integrated Distillation Columns Using Rigorous Optimization. Energy
(9) Olujić, Ž .; Sun, L.; de Rijke, A.; Jansens, P. J. Conceptual Design 2014, 74 (C), 662.
of an Internally Heat Integrated Propylene-Propane Splitter. Energy (32) Nakaiwa, M.; Huang, K.; Endo, a.; Ohmori, T.; Akiya, T.;
2006, 31 (15), 3083. Takamatsu, T. Internally Heat-Integrated Distillation Columns: A
(10) Olujic, Z.; Fakhri, F.; De Rijke, A.; De Graauw, J.; Jansens, P. J. Review. Chem. Eng. Res. Des. 2003, 81 (1), 162.
Internal Heat Integration - The Key to an Energy-Conserving (33) Chen, H.; Huang, K.; Wang, S. A Novel Simplified
Distillation Column. J. Chem. Technol. Biotechnol. 2003, 78 (2−3), 241. Configuration for an Ideal Heat-Integrated Distillation Column
(11) Liao, B.; Lei, Z.; Xu, Z.; Zhou, R.; Duan, Z. New Process for (Ideal HIDiC). Sep. Purif. Technol. 2010, 73 (2), 230.
(34) Harwardt, A.; Marquardt, W. Heat-Integrated Distillation
Separating Propylene and Propane by Extractive Distillation with
Columns: Vapor Recompression or Internal Heat Integration?
Aqueous Acetonitrile. Chem. Eng. J. 2001, 84 (3), 581.
AIChE J. 2012, 58 (12), 3740.
(12) Benali, M.; Aydin, B. Ethane/ethylene and Propane/propylene
(35) Wu, W.; Shao, B.; Zhou, X. Dynamic Control of a Selective
Separation in Hybrid Membrane Distillation Systems: Optimization
Hydrogenation Process with Undesired MAPD Impurities in the C3-
and Economic Analysis. Sep. Purif. Technol. 2010, 73 (3), 377.
Cut Streams. J. Taiwan Inst. Chem. Eng. 2015, 54, 28.
(13) Pan, Y.; Li, T.; Lestari, G.; Lai, Z. Effective Separation of
(36) Bryan, P. F. Removal of Propylene from Fuel-Grade Propane.
Propylene/propane Binary Mixtures by ZIF-8 Membranes. J. Membr. Sep. Purif. Rev. 2004, 33 (2), 157.
Sci. 2012, 390−391, 93. (37) Aspen HYSYS for Hydrocarbons; AspenTech Inc.:Cambridge,
(14) Bao, Z.; Alnemrat, S.; Yu, L.; Vasiliev, I.; Ren, Q.; Lu, X.; Deng, MA, 2016. http://www.aspentech.com/products/aspen-hysys/ (ac-
S. Adsorption of Ethane, Ethylene, Propane, and Propylene on a cessed Dec 22, 2016).
Magnesium-Based Metal-Organic Framework. Langmuir 2011, 27 (38) Pandey, S. Simulation and Mulitobjective Optimization of Cold-
(22), 13554. End Separation Process in an Ethylene Plant. Master’s Thesis, National
(15) Khalighi, M.; Karimi, I. A.; Farooq, S. Comparing SiCHA and University of Singapore, Singapore, 2013.
4A Zeolite for Propylene/propane Separation Using a Surrogate-Based (39) Jenkins, S. Current Economic Trends - March 2016, Access
Simulation/optimization Approach. Ind. Eng. Chem. Res. 2014, 53 Intelligence, LLC, Rockville, MD, 2016. http://www.chemengonline.
(44), 16973. com/current-economic-trends-march-2016/?printmode=1 (accessed
(16) Kumar, R.; Golden, T. C.; White, T. R.; Rokicki, A. Novel Nov 28, 2016).
Adsorption Distillation Hybrid Scheme for Propane/Propylene (40) Turton, R.; Bailie, R. C.; Whiting, W. B.; Shaewitz, J. A.;
Separation. Sep. Sci. Technol. 1992, 27 (15), 2157. Bhattacharya, D. Analysis Synthesis and Design of Chemical Processes, 4th
(17) Naidu, Y.; Malik, R. K. A Generalized Methodology for Optimal ed.; Prentice Hall International Series in the Physical and Chemical
Configurations of Hybrid Distillation-Pervaporation Processes. Chem. Engineering Sciences; Prentice Hall: Ann Arbor, MI, 2012.
Eng. Res. Des. 2011, 89 (8), 1348. (41) Average Price of Electricity to Ultimate Customers by End-Use
(18) Eldridge, R. B. Olefin/Paraffin Separation Technology: A Sector, by State, August 2016 and 2015 (Cents per Kilowatthour). U.S.
Review. Ind. Eng. Chem. Res. 1993, 32, 2208. Energy Information Administration, Washington, DC, 2016. http://
(19) Jana, A. K. Advances in Heat Pump Assisted Distillation www.eia.gov/electricity/monthly/epm_table_grapher.cfm?t=epmt_5_
Column: A Review. Energy Convers. Manage. 2014, 77, 287. 6_a (accessed Nov 28, 2016).
(20) Kiss, A. A.; Flores Landaeta, S. J.; Infante Ferreira, C. A. (42) Smith, R. Chemical Process: Design and Integration; John Wiley &
Towards Energy Efficient Distillation Technologies - Making the Right Sons, Ltd: Chichester, U.K., 2005.
Choice. Energy 2012, 47 (1), 531. (43) Javaloyes-Antón, J.; Ruiz-Femenia, R.; Caballero, J.a. Rigorous
(21) Annakou, O.; Mizsey, P. Rigorous Investigation of Heat Pump Design of Complex Distillation Columns Using Process Simulators
Assisted Distillation. Heat Recovery Syst. CHP 1995, 15 (3), 241. and the Particle Swarm Optimization Algorithm. Ind. Eng. Chem. Res.
(22) Kazemi, A.; Hosseini, M.; Mehrabani-Zeinabad, A.; Faizi, V. 2013, 52 (44), 15621.
Evaluation of Different Vapor Recompression Distillation Config- (44) Adams, T. A.; Seider, W. D. Practical Optimization of Complex
urations Based on Energy Requirements and Associated Costs. Appl. Chemical Processes with Tight Constraints. Comput. Chem. Eng. 2008,
Therm. Eng. 2016, 94, 305. 32 (9), 2099.

14563 DOI: 10.1021/acs.iecr.7b03432

Ind. Eng. Chem. Res. 2017, 56, 14557−14564
Industrial & Engineering Chemistry Research Article

(45) Khan, M. S.; Lee, M. Design Optimization of Single Mixed

Refrigerant Natural Gas Liquefaction Process Using the Particle
Swarm Paradigm with Nonlinear Constraints. Energy 2013, 49 (1),
146.
(46) Eberhart, R.; Kennedy, J. A New Optimizer Using Particle
Swarm Theory. Proc. Sixth Int. Symp. Micro Mach. Hum. Sci. 1995, 39.
(47) Eberhart, R. C.; Shi, Y. Particle Swarm Optimization:
Developments, Applications and Resources. Proc. of the 2001 Congress
on Evolutionary Computation IEEE Press 2001, 81.
(48) Kim, M. J.; Song, H.-Y.; Park, J.-B.; Roh, J.-H.; Lee, S. U.; Son,
S.-Y. An Improved Mean-Variance Optimization for Nonconvex
Economic Dispatch Problems. J. Electr. Eng. Technol. 2013, 8 (1), 80.
(49) MATLAB, version 2013b; The MathWorks, Inc.: Natick, MA,
2013. https://www.mathworks.com/products/matlab.html (accessed
Jan 6, 2017).
(50) Navarro-Amorós, M. a.; Ruiz-Femenia, R.; Caballero, J. a.
Integration of Modular Process Simulators under the Generalized
Disjunctive Programming Framework for the Structural Flowsheet
Optimization. Comput. Chem. Eng. 2014, 67, 13.
(51) Zielinski, K.; Laur, R. Stopping Criteria for a Constrained
Single-Objective Particle Swarm Optimization Algorithm. Informatica
2007, 31 (1), 51.
(52) Brits, R.; Engelbrecht, A. P.; van den Bergh, F. Locating
Multiple Optima Using Particle Swarm Optimization. Appl. Math.
Comput. 2007, 189 (2), 1859.
(53) Kiran, B.; Jana, A. K.; Samanta, A. N. A Novel Intensified Heat
Integration in Multicomponent Distillation. Energy 2012, 41 (1), 443.

14564 DOI: 10.1021/acs.iecr.7b03432

Ind. Eng. Chem. Res. 2017, 56, 14557−14564