Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Journal of the Taiwan Institute of Chemical Engineers 117 (2020) 1925

Contents lists available at ScienceDirect

Journal of the Taiwan Institute of Chemical Engineers


journal homepage: www.elsevier.com/locate/jtice

Development of two plant-wide glycerol carbonate production


processes: Design, optimization and environmental analysis
Bor-Yih Yua,b,*
a
Department of Chemical and Materials Engineering, Chang Gung University, Taoyuan City 333, Taiwan
b
Division of Pediatric Allergy, Asthma and Rheumatology, Department of Pediatrics, Chang Gung Memorial Hospital Linkou Branch, Taoyuan City 333, Taiwan

A R T I C L E I N F O A B S T R A C T

Article History: This work firstly presents the rigorous modeling of the two plant-wide glycerol carbonate (GC) production
Received 12 November 2020 processes through the transesterification reaction of glycerol (GLY). One is the Separated Reaction and Distil-
Revised 12 December 2020 lation (SRD) process, and the other is the Reactive Distillation-based (RD) process. This work covers the ther-
Accepted 14 December 2020
modynamic and kinetic verification, the rigorous simulation of the two processes, and optimization of the
Available online 24 December 2020
proposed processes based on simulated annealing method. From optimization, it is found that the RD process
reduced about 33.1% cost from the SRD process. Also, carbon emission analyses are performed on the two
Keywords:
optimized processes, in which the CO2 emission amount per unit amount of GC formation (CO2-e, in kg CO2/
Glycerol
Glycerol carbonate
kg GC) are calculated. It is reported that the RD process reduces 36.1% of CO2 emission (CO2-e=0.207) com-
Process design pared with the SRD process (CO2-e=0.322).
Simulated annealing © 2020 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Reactive distillation
Carbon emission

1. Introduction To date, the development of this process has been limited to con-
ceptual design [3,8], albeit already having many contributions discus-
At present, large amount of glycerol (GLY) has been generated as sing the catalyst/kinetics [57] and thermodynamics properties [4,9].
the side product from the bio-diesel production. This has caught The only simulation study found in literature was contributed by Xu
wide attentions of converting GLY into value-added chemicals. As and Xu [10], in which the GC production process was investigated as
glycerol carbonate (GC) has many favorable properties (i.e. low vola- part of an integrated bio-diesel process, yet with their focuses on the
tility, low toxicity, high bio-degradability, good reactivity), producing exergy analysis and the overall process economics.
it from GLY is considered promising [1]. Here, two plant-wide GC production processes, the Separated
Many routes for converting GLY into GC have been reported [1], such Reaction and Distillation (SRD) process and the Reactive Distillation-
as reacting GLY with urea [2], phosgene, carbon monoxide (CO), and car- based (RD) process, are rigorously studied and compared. The prelim-
bon dioxide (CO2). However, drawbacks exist in each of them. The urea inary design and the parametric studies of the RD process were
route requires the highly-vacuumed operating condition in order to recently mentioned by our group in a conference proceeding [11].
remove the side product, ammonia. The reactants in the phosgene route This process is further clarified and refined with the current work.
and the CO route are toxic. The stable structure of CO2 makes the direct Firstly, special attentions are paid on regressing both the thermody-
conversion process thermodynamically unfavorable, albeit the overall namic and the kinetic parameters. Then, two processes are proposed,
process seems greener and more atom efficient [13]. and optimized using simulated annealing approach. Finally, the envi-
An emerging route to produce GC is through the transesterifica- ronmental analyses on the optimized processes are performed, with
tion reaction between GLY and dimethyl carbonate (DMC) [37]. the CO2 emission amount per unit amount of GC formation (CO2-e, in
Being endothermic and reversible, this reaction was reported to be kg CO2/kg GC) compared. Through this study, a baseline for further
suitable to operate under modest temperature and pressure with studies (i.e. scale-up, control, operation and economical evaluation)
DMC in excess. Besides, generating methanol (MeOH) as the side could be provided.
product, the further separation of the azeotropic mixture DMC/
MeOH is required. 2. Overview

* Correspondence to: Department of Chemical and Materials Engineering, Chang


Seven components are included in the processes. They are glyc-
Gung University, Taoyuan City 333, Taiwan. erol (GLY), dimethyl carbonate (DMC), methanol (MeOH), glycidol
E-mail address: boryihyu@mail.cgu.edu.tw (GLC), glycerol carbonate (GC), carbon dioxide (CO2) and aniline

https://doi.org/10.1016/j.jtice.2020.12.016
1876-1070/© 2020 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
20 B.-Y. Yu / Journal of the Taiwan Institute of Chemical Engineers 117 (2020) 1925

(ANI). All of them can be found in the Aspen Plus database. The simu- Note that reaction rates were obtained in between 120 and 140 °
lation work was performed in Aspen Plus V10, selecting NRTL as the C, which is beyond the normal boiling point of DMC (90.0 °C) and
thermodynamic model. MeOH (64.7 °C). This indicates that the reaction was operated under
elevated pressure. As the reactions occur in the liquid phase, the
2.1. Thermodynamics operating pressure should be in between the vapor pressure of DMC
and MeOH under the corresponding temperatures. This feature is
Previously, it was reported that the reaction starts from a bi-pha- considered in the following up simulation studies.
sic liquid mixture of GLY and DMC, and becoming mono-phasic with
MeOH and GC being produced [4,9]. Hence, special attentions are 3. Process development
paid on the behavior in between these components. The liquid-liquid
equilibrium (LLE) data, and the required binary interaction parame- This work targets on developing a 50 (kTon/y) GC production pro-
ters in this quaternary system were investigated by Wang and Lu [9], cess, operating at 9095% of its full capacity. For the products, the
and Esteban et al. [4]. However, Wang and Lu [9] regressed a single commercial grade of 99 mol% (i.e. 99.2 wt%) GC, and 99.9 mol% (i.e.
set of binary pairs which only fitted to all LLE data, but failed to 99.8 wt%) MeOH are specified.
describe the vapor-liquid equilibrium (VLE). On the other hand, Este-
ban et al. regressed different sets of parameters under various tem- 3.1. The separated reaction/distillation (SRD) process
peratures [4], which are difficult to use in the current study. The
details are collected in Section A1 of the supplemental material. The optimized SRD process is depicted in Fig. 1, with the optimi-
Here, the interaction parameters in between GLY/DMC/GC/MeOH zation procedure mentioned in Section 4. The process begins with
are revisited, targeting at obtaining a parameter set capable of pre- mixing 50 kmol/h of fresh GLY with 50 kmol/h fresh DMC, and the
dicting both VLE and LLE. To this end, the LLE data [4], and VLE data recycled DMC, to reach ER of 2.0. This combined feed is pumped to
for the binary system of DMC/MeOH, GLY/MeOH, and GLY/DMC 4.25 atm, heated to 140 °C, and sent to the reactor. Since the reactor
(accessed from the Aspen built-in NIST database) are used together handles both vapor and liquid, a continuous stirred-tank reactor
for re-regression. The detailed descriptions and the verification of the (CSTR) is used (REACT). This reactor is 0.632 m3 in volume, and
re-regression are also collected in Section A1. All the binary interac- packed with 63.2 kg catalyst. From which, the reaction proceeds with
tion parameters used in this work are listed in Table S1. 81.45% per-pass conversion.
The liquid stream from the reactor is depressed to 1 atm and
2.2. Reaction kinetics flashed. The flashed liquid is sent to the downstream column (C1) to
recover the unreacted DMC from the distillate. On the other hand,
The kinetics proposed by Singh et al. is referenced [5], which is the flashed vapor (rich in DMC and MeOH) is mixed with the reactor
comprised of three reactions: vapor, totally condensed, further mixed with the C1 distillate, and
sent to the pre-concentrator (PRE). In PRE, a DMC/MeOH mixture of
r1
GLY þ DMC ! GC þ 2MeOH ð1Þ ca. 20 mol% DMC is obtained from the distillate, which is sent to the
extractive distillation section. On the other hand, a high purity DMC
r2 stream is obtained from the PRE bottom and is recycled.
GC þ 2MeOH ! GLY þ DMC ð2Þ
The typical two-column extractive distillation process configura-
r3 tion with a feed-effluent-heat-exchanger (FEHE) is used here. Aniline
GC ! GLC þ CO2 ð3Þ
is used as the entrainer. This configuration has been proved the most
Eqs. (1) and (2) represent the reversible main reactions, and Eq. (3) is economically attractive one for this system [12]. The MeOH product
the side reaction of GC decomposition. A very active heterogeneous cata- comes out from the distillate of the extractive distillation column
lyst (Zn4La1) made up of ZnO (density=5.61 g/cm3) and La2O3 (EDC), while a high purity DMC stream is obtained from the distillate
(density=6.51 g/cm3) was used. The reaction system was experimentally of the solvent recovery column (SRC) and recycled.
studied in a batch reactor between 120 to 140 °C, and DMC excess ratio The C1 bottom enters C2 for further purification. The side product
(denoted as “ER” hereafter) from 2 to 6, using 0.5 wt% of catalyst. GLC is separated from the C2 distillate, while GC and the unreacted
The reaction rates are described in the power-law form, based on GLY is obtained from the C2 bottom, and then sent to C3. Finally, the
the liquid phase mole fraction of components [5]. However, the veri- GC product of 99 mol% is obtained from the C3 bottom, while the
fication results are unsatisfactory, especially the GC selectivity. There- unreacted GLY is recovered from the C3 distillate, and then recycled.
fore, the kinetic parameters were re-regressed based on the reported Since some CO2 is formed by GC dissociation, partial vapor-liquid
results [5]. The kinetic expressions for Eqs. (1)(3) with the re-calcu- condensers are equipped in PRE and EDC, with each setting 0.2%
lated parameters are listed in Eqs. (4)(6). vapor vented from the column to avoid CO2 accumulation. Also, C1,
    C2 and C3 are operated under vacuumed condition, since the normal
kmol 144; 400
r1 ¼ 1:31  1016 exp XDMC XGLY ð4Þ boiling point of GLY and GC are too high to use the high pressure
kgcat s RT
steam as the reboiler heat source. Other detailed settings of this pro-
    cess can be found in Table S2.
kmol 60; 013
r2 ¼ 6:06  104 exp XGC XMeOH 2 ð5Þ
kgcat s RT
3.2. The reactive distillation-based (RD) process
   
kmol 113; 000
r3 ¼ 1:63  109 exp XGC ð6Þ Reactive distillation is a promising technique of process intensifi-
kgcat s RT
cation to improve the performance of the equilibrium-limited reac-
In the above three equations, the activation energies are in units tions. Its applications have continued to attract the attentions
of kJ/kmol, and X represents mole fraction of the components in the [1315]. The optimized flowsheet of the RD process is depicted in
liquid phase. In Singh et al.’s contribution, the phase transition during Fig. 2, with the optimization details in Section 4. The detailed simula-
the reaction was not specifically mentioned [5]. Therefore, the mole tion settings are collected in Table S3. First, the fresh GLY of
fraction (X) of the combined liquid phase is used for the kinetic 50 kmol/h, and the combined DMC stream of 100 kmol/h, enter at
expression. The detailed regression steps as well as the verification the 5th and 18th tray of the reactive distillation column (RDC),
results are collected in Section A2 of the supplemental material. respectively. The RDC has a total of 19 trays, operated at 3.09 atm,
B.-Y. Yu / Journal of the Taiwan Institute of Chemical Engineers 117 (2020) 1925 21

Fig. 1. The optimized flowsheet of the SRD process.

with a 0.0068 atm pressure drop per tray assumed throughout the vapor-liquid-liquid check in RDC, no phase split is observed. This
column. The reactive section is in between the feed locations, with ensures the successful continuous operation of reactive distillation.
25.46 kg catalyst packed on each tray. Note that the catalyst loading The liquid composition profile is shown in Fig. 3(c). It could be clear
(Wcat, in kg) is determined by the following equation: that MeOH in the middle/upper section of RDC prevents the liquid
phase from phase splitting. On the other hand, the GLY at the lower
pD2
Wcat ¼  WH  0:9  Ccat ð8Þ section is too less to have phase separation.
4
The first three term of Eq. (8) (p4D  WH 0:9) indicates the total
2

liquid volume on the tray (in m3), where D is the column diameter 4. Optimization
(in m); WH is the weir height (set at 0.1016 m), and the factor 0.9
indicates a downcomer area of 10%. The Ccat is the bulk catalyst den- 4.1. Optimization settings
sity, defined as the mass amount of catalyst per unit volume of the
reacting medium. It is determined to be 419.20 kg/m3 from optimiza- Here, the modified version of the total annual cost (TAC) [1618],
tion. defined by Eq. (9), is minimized through optimization:
The RDC combines the function of the preheater, reactor and PRE TCC X  X 
in the SRD process. A DMC/MeOH mixture with 13.7 mol% DMC is Obj: Func ¼ þ TOC þ Ri Ri;base UPR  Pj Pj;base UPP ð9Þ
PBP i j
obtained from the RDC distillate, and is sent to the extractive distilla-
tion section, in which the identical configuration as in the SRD pro- This equation considers the annualized capital cost (the 1st term),
cess is used to separate DMC and MeOH. On the other hand, the RDC the total operating cost (the 2nd term), and the stream cost differen-
bottom is depressed to 1 atm and flashed. Besides, 99% conversion of ces (the 3rd and 4th term). The annualized capital cost is determined
GLY is reached in RDC. Also, a partial vapor-liquid condenser is by dividing the total capital cost (TCC) by the payback period (PBP, 3
equipped, setting 0.2% vapor vented from the column, so as to avoid years). The total operating cost (TOC) refers to the utilities and cata-
CO2 accumulation. lyst used in the production, calculated based on 8000 h of annual
Similar to the SRD process, the flashed liquid is sent to down- operation. The cost data is referenced from Luyben’s book, and is col-
stream columns (C1 and C2) for GC recovery. Note that reaching 99% lected in Section A3 of the supplemental meterial [19]. Besides, in
conversion in RDC eliminates the use of the C3 here. On the other order to reach the vacuumed conditions in the downstream columns,
hand, the flashed vapor is total condensed, mixed with the C1 distil- the liquid ring pump system is suggested [20]. However, the related
late, and recycled to RDC. costs only account for a very small portion of the objective function,
The temperature, composition and reaction profiles inside the and is not considered in optimization.
RDC are illustrated in Fig. 3. From Fig. 3(a), it is noted that the tem- The stream cost difference terms consider the costs originated
peratures for some reaction stages lies outside 120 to 140 °C, on from varying reactant and product flowrates, majorly resulted from
which the kinetic expressions are verified. However, the deviation is the varying reaction conversion and the GC dissociation in the cur-
less than 2% (lowest: 118.2 °C, highest: 142.7 °C). Hence, it is consid- rent systems. With this concept, it keeps the objective function from
ered acceptable. The reaction profile could be seen in Fig. 3(b). The being diluted by the massive contribution of a full stream cost. While
reaction is more active in the lower part of the column, which reason- perfoming optimization runs, the base flowrates for measuring the
ably corresponds to the higher temperature over there. difference are fixed (see Table 1). This enables the optimization
As addressed, the reaction starts from being bi-phasic, and turns results to be comparable. In Eq. (9), the third term is for the raw
mono-phasic in a batch reactor. However, after performing the materials (i=GLY or DMC), while the last term is for the products
22 B.-Y. Yu / Journal of the Taiwan Institute of Chemical Engineers 117 (2020) 1925

Fig. 2. The optimized flowsheet of the RD process.

(j=GC or MeOH). For the unit prices (UPi and UPj), 1.036 [21], 0.685 reaction conversion and results in higher operation cost. Secondly,
[12], 2.40 [22] and 0.32 [12] (USD/kg) are assigned to GLY, DMC, GC the rectifying section of EDC in both proesses is fixed at 5, which is
and MeOH, respectively. large enough to prevent the entrainer from going upward. Also, in
The proposed processes are optimized using the simulated the RD process, 1 stripping stage is assumed below DMC feeding
annealing (SA) method. Developed by Kirkpatrick et al. [23]. based on point for maintaining the temperature at the RDC bottom. Further
the work by Metropolis et al. [24], SA compares the optimization to increasing the stripping lower section and length leads to higher bot-
the annealing process of metallurgical material, which is popularly tom temperature and also the reboiler duty of RDC.
used in optimization with process simulator [2527]. Compared The main parameters used in SA are listed in Table 1. The pro-
with the frequently used sequential iterative methond [2830], SA is posed algorithm and the detailed information of the variables are
much more efficient in computation. In this work, the SA algorithm is illustrated in Fig. S1 and Table S4 for the SRD process, and in Fig. S2
coded in MATLAB, which is linked with the Aspen simulation file via and Table S5 for the RD process. Other detailed information can be
ActiveX server. found in Section A4 of the supplemental material. The optimization
In the SRD process, a total of 21 variables are optimized. There are for each process was performed three times, starting from different
the reactor temperature (TR) and volume (VR), the ER, the DMC con- base cases, in order to confirm the consistency of the optimization
tent in the PRE distillate (DDMC), the molar ratio of entrainer to feed results.
(FE/FF), total number trays of the extractive section (NEXT) and the
stripping section (NSTR) of EDC, the total number of trays as well as 4.2. Optimization results
the feed location of SRC, PRE, C1, C2 and C3 (NTSRC, NFSRC, NTPRE,
NFPRE, NTC1, NFC1, NTC2, NFC2, NTC3, NFC3), the operating pressure of Fig. 4(a) tracks the change of the objective function when optimiz-
C1, C2 and C3 (PC1, PC2, PC3), and the reflux ratio of C3 (RR3). ing the SRD process. Since both the discrete and continuous variables
On the other hand, 17 variables are optimized for the RD process. are optimized in the same algorithm, the optimized configurations
Those variables different from the SRD process are the RDC operating from different runs may not be completely identical. This is com-
pressure (PRDC), feed location of GLY (NFGLY) and DMC (NFDMC) into monly observed when performing SA, especially as the number of
RDC, the DMC content in the RDC distillate (DDMC), and the bulk mass variables becomes large. However, the objective function settles to
concentration of the catalyst (Ccat). The remaining variables are ER, very similar final states, which adequately indicates the consistency
FE/FF, NEXT, NSTR, NTSRC, NFSRC, NTPRE, NFPRE, NTC1, NFC1, NTC2, and NFC2. of the algorithm. The configuration with the lowest objective func-
Some less influential variables are not optimized for enhancing tion among the three runs (i.e. the case 1) is selected for demonstra-
the convergability of simulation runs. Firstly, the reactor pressure in tion in Fig. 1. Similarly, Fig. 4(b) tracks the objective function while
the SRD process is set to 1.1 times of the DMC vapor pressure under optimizing the RD process, also with the consistency of the algorithm
corresponding temperatures (TR) during optimization. Setting a proved. The configuration corresponding to the lowest objective
higher one prevents MeOH from evaporating, which retards the function (i.e the case 3) is depicted in Fig. 2.
B.-Y. Yu / Journal of the Taiwan Institute of Chemical Engineers 117 (2020) 1925 23

Fig. 4. Tracking of the objective function. (a) The SRD process. (b) The RD process.

Fig. 5 depicts the itemized cost based on the major process units
of the selected optimized processes. Other detailed results for opti-
mizing two processes are collected in Tables S6 and S7, respectively.
Some important results of the selected optimized configuration are
listed in Table 2.
It is clear that the RD process outperforms the SRD process, with
about 33.1% reduction in the objective function. This significant dif-
ference is mainly from the integration of the preheater, reactor, and
PRE into a single RDC, which handles the energy more efficiently.

Fig. 3. Profiles in the RDC. (a) Temperature; (b) Reaction and (c) liquid-phase composi-
tion.

Table 1
The parameters used in the SA algorithm.

Values

Initial temperature (T0) 80


Final temperature (T) 0.001
Energy state (E) The objective function in Eq. (8)
Annealing rate (k) 0.85
Equilibrium number of iteration (EqNT) 63 (for the SRD process)
51 (for the RD process)
GCbase (kmol/h)a 5789.01
MeOHbase (kmol/h)a 3072.91
GLYbase (kmol/h)a 4604.74
DMCbase (kmol/h)a 4488.13
a
Used in Eq. (8) for optimizing both processes. Fig. 5. Itemized costs for the optimized SRD and RD processes.
24 B.-Y. Yu / Journal of the Taiwan Institute of Chemical Engineers 117 (2020) 1925

Table 2
Main results from optimization.

The SRD process The RD process

Objective Function (kUSD) 2334.38 1561.75

From 713.38 (QPre heater, LPS) 1965.77 (QrRDC, MPS)


Reaction 1186.01 (QREACT, LPS)

1712.70 (QrPRE, 2-bar steam) 1314.61 (QrEDC, MPS)


Energy 1120.47 (QrEDC, MPS) 455.31 (QrSRC, HPS)
Demand From 586.32 (QrSRC, HPS) 373.55 (QrC1, HPS)
Separation 645.55 (QrC1, HPS) 85.95 (QrC2,HPS)
79.25 (QrC2, HPS)
651.27 (QrC3, HPS)

GC Production Rate (kg/h) 5785.84 5798.96


*“Qr” denotes the reboiler duty, “Q” denotes the utility in heat exchangers.

This integration leads to a 37.6% cost saving as can be calculated from


Fig. 5. (i.e. 2334.38 £ 42.3% and 1561.75 £ 39.4%). Fig. 6. The calculated CO2-e for both processes.
Another distinct difference in between the two processes is that
the SRD process favors an incomplete per-pass conversion of 81%
(see Table S6), comparing with the 99% conversion in the RD process. reaching its lower limit. Hence, the reaction performance at an even
This indicates the equilibrium-limited nature of the reaction. Hence, lower ER could be interesting. Currently, this idea is not carried out,
a further separation column (C3) is required for separating the since the kinetics is only verified at ER greater than 2.
unreacted GLY, contributing 11.9% of the total cost. Furthermore, the Besides, the charm of using reactive distillation is the possibility to
optimized conversions in the SRD process are reported to be lower reach a high reaction conversion, by just feeding the reactants at their
than those in the base cases in all the optimization runs. It reveals the stoichiometric ratio. Given that this is feasible in the current system, a
unfavorable economic performances resulted from GC decomposition high-purity GC stream is obtained at the RDC bottom. This indicates
at longer residence time. the requirement of a higher grade heat source (i.e. electricity) to the
Besides, in the SRD process, the cost in the extractive distillation reboiler, owing to the unfavorably high boiling point of GC (ca 354 °C
(i.e. EDC+SRC) rises after optimization. This penalty is offset by the at 1 atm). Also, the GC decomposition may occur under this situation.
saving in PRE, as the optimized DMC content in the PRE distillate is To resolve these problems, operating the RDC under vacuumed con-
farther away from the azeotropic point (ca. 20 mol%). This illustrates dition could be considered, albeit it inevitably lowers the reaction
the interactions in between the reaction and the separation sections. temperature. Hence, the related experimental efforts in the future
could be valuable.
4.3. Carbon emission analysis
5. Conclusion
Carbon emission analysis considers the indirect CO2 emission
resulted from consuming steam and electricity, which are derived In this work, two plant-wide glycerol carbonate (GC) production
from carbonaceous fuels [16,17,28,30,31]. The corresponding CO2 processes, the separated reactor/distillation (SRD) process and the
emission generated from using unit amount of different heating utili- reactive distillation based (RD) process, were firstly investigated. The
ties are collected in Table 3 [25]. Here, the reported index is the kg of research scope covers regressing of the thermodynamic and kinetic
net CO2 emission per kg of GC formed (denoted as “CO2-e” hereafter). parameters, designing of two processes, optimizing by simulated
The CO2-e from the two optimized processess, shown in Figs. 1and annealing method, and performing carbon emission analysis. From
2 are compared, with the reuslts shown in Fig. 6. In general, the RD the generated results, the RD process reduces ca. 33.1% cost and
process reduces 36.1% of CO2 emission (CO2-e=0.207) from the SRD 36.1% CO2 emission from the SRD process. Finally, it is recommended
process (CO2-e=0.322). The major reason for this significant reduction to put more experimental efforts in developing the reaction kinetics
is because the RDC reboiler duty requires only about half of the total under a lower temperature and less excess amount of DMC. This
heating demand from the pre-heater, reactor and PRE in the SRD pro- would reduce the energy requirement in separation, and lead to less
cess. This corresponds to a 41.1% reduction in carbon emission. (i.e. CO2 emission.
0.322*49.0% and 0.207*45.0%). Also, the emission from the additional
651.3 kw HPS in the C3 reboiler contributes 12.6% of the total emis- Declaration of Competing Interest
sion from the SRD process.
The author declare that he has no known competing financial
4.4. Further discussions interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
The possible further improvement on the RD process is discussed
here. Note that the current optimized RD process is operated at ER=2, Acknowledgments

Table 3 The research funding from the Ministry of Science and Technology
Information of the utilities used throughout the work [25]. of R. O. C. under grant no. MOST 108-2218-E-182 -009 -MY3 is
greatly appreciated.
Utilities Unit CO2 emission amount (kg-CO2/GJ)

Steam (2 bar, 120 °C) 66.68 Supplementary materials


Steam (6 bar, 160 °C) LPS 72.86
Steam (11 bar, 184 °C) MPS 76.60
Steam (42 bar, 254 °C) HPS 91.14 Supplementary material associated with this article can be found
in the online version at doi:10.1016/j.jtice.2020.12.016.
B.-Y. Yu / Journal of the Taiwan Institute of Chemical Engineers 117 (2020) 1925 25

References [16] Wang C, Wang C, Guang C, Gao J, Zhang ZS. Hybrid reactive distillation using poly-
octylmethylsiloxane membrane for isopentyl acetate production from mixed PVA
[1] Sonnati MO, Amigoni S, de Givenchy EPT, Darmanin T, Choulet O, Guittard F. Glyc- by products. J Chem Technol Biot 2019;94(2):527–37.
erol carbonate as a versatile building block for tomorrow: synthesis, reactivity, [17] Wang C, Zhang ZS, Zhang XK, Gao J, Stewart B. Energy-saving hybrid processes
properties and applications. Green Chem 2013;15(2):283–306. combining pressure-swing reactive distillation and pervaporation membrane for
[2] Lertlukkanasuk N, Phiyanalinmat S, Kiatkittipong W, Arpornwichanop A, Aiou- n-propyl acetate production. Sep Purif Technol 2019;221:1–11.
ache F, Assabumrungrat S. Reactive distillation for synthesis of glycerol carbonate [18] Zhang ZS, Wang C, Guang C, Stewart B. Cost-saving and control investigation for
via glycerolysis of urea. Chem Eng Process 2013;70:103–9. isopentyl acetate ionic liquid catalyzed synthesis through conventional and divid-
[3] Li JB, Wang T. Coupling reaction and azeotropic distillation for the synthesis of ing-wall reactive distillation. Process Saf Environ 2019;129:89–102.
glycerol carbonate from glycerol and dimethyl carbonate. Chem Eng Process [19] Luyben WL. Principles and case studies of simultaneous design (Chapter 5). New
2010;49(5):530–5. York: Wiley; 2011.
[4] Esteban J, Ladero M, Molinero L, Garcia-Ochoa F. Liquid-liquid equilibria for the [20] Seider WD, Seader JD, Lewin DR. Product and process design principles: synthesis,
ternary systems DMC-methanol-glycerol, DMC-glycerol carbonate-glycerol and analysis, and evaluation 3rd ed. (Chap 22: cost accounting and capital cost esti-
the quaternary system DMC-methanol-glycerol carbonate-glycerol at catalytic mation) S., W.,. USA: John Wiley & Sons, Inc.; 2009.
reacting temperatures. Chem Eng Res Des 2014;92(12):2797–805. [21] ICIS pricing: Glycerine (US Gulf), 8th January. 2014.
[5] Singh D, Reddy B, Ganesh A, Mahajani S. Zinc/lanthanum mixed-oxide catalyst for [22] Nguyen N, Demirel Y. Economic analysis of biodiesel and glycerol carbonate pro-
the synthesis of glycerol carbonate by transesterification of glycerol. Ind Eng duction plant by glycerolysis. J Sustain Bioenergy Syst 2013;3:209–16.
Chem Res 2014;53(49):18786–95. [23] Kirkpatrick S, Gelatt CD, Vecchi MP. Optimization by simulated annealing. Science
[6] Devi P, Das U, Dalai AK. Production of glycerol carbonate using a novel Ti-SBA-15 1983;220:671–80 No. 4598.
catalyst. Chem Eng J 2018;346:477–88. [24] Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH, Teller E. Equation of
[7] Li YJ, Liu JX, He DH. Catalytic synthesis of glycerol carbonate from biomass-based state calculations by fast computing machines. J Chem Phys 1953;21:1087–92
glycerol and dimethyl carbonate over Li-La2O3 catalysts. Appl Catal A  Gen No. 6.
2018;564:234–42. [25] Yu BY, Wu PJ, Tsai CC, Lin ST. Evaluating the direct CO2 to diethyl carbonate (DEC)
[8] Wang HJ, Pang L, Yang C, Liu YH. Production of glycerol carbonate via reactive dis- process: Rigorous simulation, techno-economical and environmental evaluation.
tillation and extractive distillation: an experimental study. Chin J Chem Eng J Co2 Util 2020;41:101254.
2015;23(9):1469–74. [26] Yang XL, Ward JD. Extractive distillation optimization using simulated annealing
[9] Wang HJ, Lu PF. Liquid-liquid equilibria for the system dimethyl carbonate plus and a process simulation automation server. Ind Eng Chem Res 2018;57
methanol plus glycerol in the temperature range of (303.15 to 333.15) K. J Chem (32):11050–60.
Eng Data 2012;57(2):582–9. [27] Cui Y, Zhang Z, Shi X, Guang C, Gau J. Triple-column side-stream extractive distil-
[10] Xu CX, Xu Q. Novel design for simultaneous production of biodiesel and glycerol lation optimization via simulated annealing for the benzene/isopropanol/water
carbonate from soybean oil. Ind Eng Chem Res 2018;57(49):16809–16. separation. Sep Purif Technol 2020;236:1–8.
[11] Yu BY, Chen WJ. Rigorous simulation and optimization of a plant-wide glycerol [28] Li GJ, Wang C, Guang C, Zhang ZS. Energy-saving investigation of hybrid reactive
carbonate (GC) production process. Comput Aided Chem Eng 2020;48:283–8. distillation for n-butyl acetate production from two blending feedstocks. Sep Purif
[12] Yu BY, Chen MK, Chien IL. Assessment on CO2 utilization through rigorous simu- Technol 2020;235.
lation: converting CO2 to dimethyl carbonate. Ind Eng Chem Res 2018;57 [29] Zhu ZY, Li GX, Yang JW, Dai Y, Cui PZ, Wang YL, Xu DM. Improving the energy effi-
(2):639–52. ciency and production performance of the cyclohexanone ammoximation process
[13] Oksal IN, Kaymak DB. Dynamic controllability comparison of reactive distillation via thermodynamics, kinetics, dynamics, and economic analyses. Energ Convers
columns with single and double reactive sections for two-stage consecutive reac- Manag 2019;192:100–13.
tions. Chem Eng Res Des 2018;129:391–402. [30] Yang X, Wang S, Li GX, Zhao F, Feng Z, Chen XN, Zhu ZY, Wang YL, Gao J. Process
[14] He RN, Dong YB, Zou Y, Zhao JH, Yaseen M, Mu CX, Tong ZF. Simulation and opti- design and comprehensive analysis of the ethanol amination process to improve
mization of reactive distillation for the production of ethyl acetate using [BMIM] acetonitrile production. Ind Eng Chem Res 2020;59(11):5047–55.
HSO4 as catalyst. Chem Eng Res Des 2020;161:218–31. [31] Gadalla MA, Olujic Z, Jansens PJ, Jobson M, Smith R. Reducing CO2 emissions and
[15] Kaymak DB, Unlu H, Ofkeli T. Control of a reactive distillation column with double energy consumption of heat-integrated distillation systems. Environ Sci Technol
reactive sections for two-stage consecutive reactions. Chem Eng Process 2005;39(17):6860–70.
2017;113:86–93.

You might also like