DMO Synthesis1
DMO Synthesis1
DMO Synthesis1
a r t i c l e i n f o a b s t r a c t
Article history: The rigorous design and optimization of a complex, plant-wide dimethyl oxalate (DMO) synthesis pro-
Received 22 January 2018 cess from syngas is firstly developed in this work. The whole process can be roughly divided into two
Revised 2 June 2018
sections. The first one is to produce DMO through the coupling reaction, in which CO is reacted with
Accepted 17 August 2018
the intermediate, methyl nitrite (MN). This section includes the production, and also the purification of
Available online 23 August 2018
DMO. The second one is the regeneration reaction to get back MN through a packed column reactor,
Keywords: which is simulated as a modified version of reactive distillation column. After the basic design, system-
Process design atic optimization is investigated through minimizing total annual cost, and the potential heat integration
Optimization strategy is also proposed. During optimization, we found that the methanol circulation rate inside the
Dimethyl Oxalate process is the most influential variable, and a higher methanol flow rate (defined as MeOH/NOs) within
Coupling reaction an acceptable region leads to better economic performance.
Regeneration reaction
© 2018 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.compchemeng.2018.08.025
0098-1354/© 2018 Elsevier Ltd. All rights reserved.
86 B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100
Table 1
Thermodynamic parameters in this work (NRTL-RK model).
Table 2
Information of reaction and kinetics.
Kinetic parameters
Equilibrium Constant
ri = [kmol/kgcat-s].
In Ji’s work, a series of sensitivity test on the reaction condi- The unit of reaction rate is originally in mol/m3 -cat/s, and it is
tions (temperature, CO to MN feed ratio and residence time) were transferred to be kmol/kg-cat/s in this work in order to be used
performed, and two critical points in designing the coupling reac- in Aspen Plus. In order to fit the original experimental data better,
tion to form DMO were pointed out. Firstly, MN undergoes self- some of the parameters in Eq. (6) are re-regressed in MATLAB. All
dissociation at the temperature higher than 140 ( °C). The prod- the parameters for this expression are recorded in Table 2. Besides,
ucts from self-dissociation are methanol, acetaldehyde, and NO. the formation rate of DMC was not provided in this expression,
Because higher temperature benefits the reaction rate, the upper and it is reasonably assumed to be one-forty fifth of the DMO pro-
limit of temperature implies that temperature control should be duction rate throughout the reactor. Also, the self-dissociation and
very important. Secondly, MN undergoes catalytic dissociation to catalytic dissociation of MN is ignored in this reaction system due
form methyl formate, methanol, water and NO. MN competes with to the low quantity.
CO for the active sites on the catalyst, thus MN should not be in
excess in order to inhibit the catalytic dissociation reaction. How- 3.2. Background and settings in regeneration reaction
ever, if CO is in too much excess, it overly occupies the active site,
and causes reaction rate to become lower. Thus, Ji suggested that The regeneration reaction is to react NO with the added
the optimal range of CO to MN molar ratio in the feed (CO/MN) methanol and oxygen to form MN. However, according to litera-
should be in between 1 and 2. With proper selections on reaction ture, the reactions involved are quite complicated, including many
conditions, the undesired MN self-dissociation and catalytic disso- sequential and parallel reactions (Li et al., 2013; Liu et al., 2013),
ciation can be effectively inhibited. as the following:
Comparing with the side products from MN dissociation, DMC
is much greater in amount. However, the yield of DMC revealed 2NO + O2 ↔ 2N O2 (7a)
slight change with varying temperature, CO/MN, and space velocity
in the reasonable ranges. Although the existence of DMC cannot N O2 + NO ↔ N2 O3 (7b)
be ignored, the selectivity of DMO and DMC was generally very
high (larger than 40). Thus, the coupling reaction still favors the
production of DMO. 2C H3 OH + N2 O3 ↔ 2C H3 ON O(MN ) + H2 O (7c)
Besides, because of the explosive nature of MN, its composition
inside the reactor should also be carefully considered. According to
literature, nitrogen or carbon dioxide can be fed into reactor with 2N O2 ↔ N2 O4 (7d)
the reactants to serve as diluent. Considering both the safety of
the reactor and the fact that too much inert components lower
C H3 OH + N2 O4 ↔ C H3 ON O(MN ) + HN O3 (7e)
the reaction rate and increase the system loading, the combined
composition of CO and MN (FRACCO+MN ) should be better in the
range between 30 and 40 mol% (Chen, 1995; Tahara et al., 1984; C H3 OH + O2 ↔ 2HCH O + H2 O (7f)
Meng et al., 2004; Yao, 2013). The kinetic expression of coupling
reaction to form DMO provided by Ji et al is illustrated as Eq. (6) in
the following. C H3 OH + N O2 ↔ HCH O + NO + H2 O (7g)
(formula: HCHO). Essentially, among the desired and undesired re- Table 3
Parameters required in regeneration reaction.
actions, the key components are N2 O3 and N2 O4 , respectively. Thus
in operation, it is designed to enhance the reaction (7b), and in the Henry’s constant (bar) In form of: ln(Hi ) = A + B/T
meantime to inhibit reaction (7d). The undesired reactions can be A B
effectively inhibited by selecting the suitable operating conditions.
HNO 8.55 −918.76
Firstly, moderate temperature (30–45 °C) and pressure (around nor- HNO2 10.90 −1275.80
mal pressure) are required for this reaction network. This is be- HMN 9.92 -2133.90
cause low temperature and high pressure favor the equilibrium of
Diffusivity (cm2 /s) In form of: ln(Di ) = A + B/T
reaction (7b). Secondly, the methanol to NO feed ratio should be in
excess to suppress the formation of N2 O4 . Generally, the methanol A B
concentration should be kept at above 70 wt% for this purpose. NO −5.1955 1428.1
Note that the formation of MN always accompanies with forma- NO2 −5.0138 1428.1
tion of water, which dilutes the methanol concentration in the re-
actor. However, the unreacted methanol is separated with water in
a following distillation column. Hence, methanol should not be in
too much excess in order to avoid wasting energy. Also, an overly-
into Aspen Plus in unit of pressure. Besides, the henry’s constant
high methanol concentration accelerates the side reactions. A suit-
between MN and methanol is also needed for describing the sep-
able range of methanol to NO feed ratio is in between 1.5 and 2.2,
aration performance, and it was referenced from Liu et al. in this
as suggested in literature (Liu et al., 2013). The feed molar ratio
work (Liu et al., 2010a). Because the solubility of MN into methanol
between NO and O2 (NO/O2 ) is also important. From the view-
is very small, it is possible to separate these two components by
point of inhibiting the formation of undesired N2 O4 and reaction
condensing the methanol.
to form formaldehyde, and to maximize the reaction conversion,
Based on the above-mentioned characteristics of the regener-
NO/O2 should be large. However, if the remaining NO is sent back
ation reaction, it is not suitable to simulate the regeneration re-
with MN into coupling reactor (R-CP), it will cause the coupling
actor into a plug-flow reactor (module “RPLUG”) in Aspen Plus.
reaction to become slower. Thus, the optimal NO/O2 is at around
Instead, a packed column reactor with packing inside to provide
stoichiometric ratio (=4) from the process understanding. In addi-
large amount of gas-liquid interface and a counter-current flow
tion, due to the explosive nature of MN, nitrogen is often used as
type is more suitable. Thus, this can also be viewed as a modified
inert in the reactor for safety concerns.
version of reactive distillation process, which is a conventionally,
By combining Eqs. (7b) and (7c), we can get the following
widely-used process intensification technique to reach high con-
Eq. (8) to express the reactions which produces MN and water
version for those liquid-phase, equilibrium-limited reactions. The
from methanol, NO, and NO2 :
design, control, optimization, and different reaction configuration
2C H3 OH + N O2 + NO ↔ 2C H3 ONO + H2 O (8) of various reactive distillation process can be extensively found
Note that the overall regeneration reaction mentioned in in literature (Cao et al., 2017; Chua et al., 2017; Harvianto et al.,
Eq. (2) can then be obtained by further combining Eqs. (7a) and 2017; Kaewwisetkul et al., 2017; Shi et al., 2017; Zang et al., 2017;
(8). Norkobilov et al., 2017). In most of those studies, the built-in mod-
In this work, the rate expression of NO oxidation reaction is ule RADFRAC in Aspen Plus can successfully describe the perfor-
cited from Pradhan et al. (1997), which is listed as in Eq. (9). On mance of a reactive distillation column. So here in this work, RAD-
the other hand, the rate expression of regeneration reaction is cited FRAC unit is also selected to express the packed bed reactor for
from Li et al. (2013), and is listed as in Eq. (10). Because the re- simulation purpose. Note that there is a distinct difference of this
generation reaction occurs on the interface of gas–liquid, this set packed column reactor from the conventional reactive distillation
of reaction kinetic was derived from their experimental data in a column. In this process, the regeneration reaction occurs at the
packed bubble column reactor using two-film theory. These com- gas-liquid interface, in which the reaction holdup should be spec-
plicated rate expressions were established with the aid of Fortran ified based on the area of the interface. The simulation study of
subroutine in this study. this kind of reactive distillation process has also been scarce in the
1 652.1
literature.
rNO = − 4.747 PO2 PNO 2 (9) The simulation of packed column is described in detail here.
RT T Firstly, it is modelled as a column with equilibrium stages, and
then we assume a reasonable height equivalent to a theoretical
PNO2
rREG = plate (HETP) to a single stage. After that, pressure drop and col-
HNO2
23725 P
umn diameter can be estimated by the built-in “packing sizing”
NO DNO2 PNO2 and “packing rating” functions. In this work, the authors assume
DNO2 3.7×107 exp − − CMeOH (10) that the column is packed with 1-mm Raschig ring randomly. The
RT HNO 3DNO HNO2
surface area for packing is 220 (m2 /m3 ), the void fraction of this
The unit of rNO is kmol/m3 /s, and the unit of rREG is in
packed column is 0.92, and the HETP is 2 ft. Because the unit of
mmol/cm2 /s. Pi is the partial pressure of components (i = O2 , NO,
reaction is in an area-based unit, thus the reaction holdup should
and NO2 ), while Hi and Di are the henry’s constant and diffusivity
be the actual gas-liquid contacting surface area that the packing
of the species (i = NO and NO2 ). For the remaining notations, CMeOH
provided. The reaction holdup can be calculated by the following
is the methanol concentration, R is the ideal gas constant, and T is
Eq. (11):
temperature in unit of K.
In Li et al.’s work, henry’s constant and diffusivity were listed
in temperature range of 298–318 (K), thus the expressions can be Holdup(area2 ) = π /4 (DIAM )2 (HEP T )(SA )(1 − V OID ) (11)
obtained by simple regression (Li et al., 2013). All the parame-
ters required in Eqs. (9) and (10) are listed in Table 3. Note that
in the original work by Li et al, the Henry’s constants were re- Where DIAM is column diameter in meter, SA is specific surface
ported in unit of kPa-cm3 /mmol. Therefore some necessary conver- area of the packing in m2 /m3 , and VOID is the dimensionless void
sion should be made, as the Henry’s constants can only be input fraction inside the packed column.
B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100 89
4. Steady state process design nature of coupling reaction and the narrow range of feasible op-
erating temperature as discussed previously, fluidized-bed reac-
The process flowsheet with the optimized result is illustrated tor is better to be adopted. A fluidized-bed reactor has proper-
in Fig. 1. The optimization of this process is discussed in later ties such as being able to reach high conversion, having good
Section 5. The fresh feed is assumed based on our previous poly- heat removal ability, and providing a near-uniform temperature
generation work, with 10% of total amount of syngas sent to the inside (Soundararajan et al., 2001; Alwahabi and Froment, 2004;
EG production process (Yu and Chien, 2015). The Syngas-to-EG pro- Turton et al., 2008; Yu and Chien, 2016; Chang et al., 2013). Hence,
cess is a dual-staged process, with formation of DMO as in the first it is especially suitable for the coupling reaction.
stage, and hydrogenation of DMO to EG in the section stage. Note In order to simplify the simulation of reactor without losing the
that hydrogen causes the deactivation of coupling catalyst in the capability of qualitatively illustrating the reaction performance, an
first stage, and CO causes the deactivation of hydrogenation cata- isothermal plug flow reactor (RPLUG module) is assumed here in
lyst in the second stage. Thus, the raw syngas is separated by a this work. Under this mode in Aspen Plus, the heat duty which
pressure swing adsorption (PSA) process to become three streams. should be removed to operate the reactor at the targeted temper-
One is the 99.99 mol% CO stream (remaining: H2), another one is ature will be calculated. For the pressure drop of this tubular re-
the 99.99 mol% H2 stream (remaining: CO), and the other one is actor, Eq. (12) was used based on the fact that the fluidized-bed
the tail gas stream. The recovery of each component in this PSA reactor is operated in the fluidized or bubbling region in which
unit is set to be 90%. (Peking University Pioneer Co., 2016) After the gravity force of solid equals to the drag force.
that, the CO stream is assumed to undergo selective oxidation to p = g(ρs − ρg )(1 − ε )h (12)
become hydrogen-free, and is sent to DMO synthesis process. From
calculation, the fresh CO feed is 256.3 (kmol/h), with 99.88 mol% Where ࢞p is the pressure drop in pascal, g is gravitational ac-
purity and the remainder to be water. celeration in m/s2 , ρ s is the density of solid particle in kg/m3 , ρ g
The fresh CO feed is first mixed with fresh nitrogen and the is the density of gas in kg/m3 , ε is the void fraction inside the bed
MN-rich gas recycled from regeneration section to become a com- (-), and h is the bed height in meter.
bined feed, then is preheated and sent to the coupling reactor (R- In designing the fluidized-bed reactor, the catalyst density is as-
CP). As discussed in Section 3.1, the CO to MN feed ratio (CO/MN) sumed as 880 (kg/m3 ), and the bed voidage inside the reactor is
is set at 1.5, and the total composition of CO and MN (FRACCO+MN ) 0.6. The pressure of reactor is set at 2.5 bar. Considering that high
is set at 0.4 in the combined feed. Considering the exothermic temperature enhances the reaction rate, thus the reactor tempera-
ture is set at 135 °C, at which self-dissociation and catalytic disso-
90 B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100
70 0.85
45.2
60 0.80
45.0
50 44.8
0.75
40 44.6
30 44.4 0.70
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
Height (m) Height (m) Height (m)
Fig. 3. Conversion, temperature and methanol concentration profile inside PCR.
portant for the performance in MSC, as mentioned previously. Thus other degree-of-freedom to be specified, and this value is deter-
this temperature should be determined by optimization, and is fur- mined based on optimization. Note that as the methanol purity is
ther discussed in Section 5. not specified to be ultra-high, as the energy consumption increases
From the previous section, the design of PCR has to meet some with higher targeted purity. It is suggested by open literature that
requirements. The design results are collected from Fig. 3(a) to 95–98 mol% is suitable in this situation, and 95 mol% is selected in
(c). Fig. 3(a) illustrates the conversion profile in PCR, with the this work (Tahara et al., 1984).
height of PCR calculated from bottom to top. Because PCR is mod-
eled by specifying the number of equilibrium stages and HETP, the
5. Process optimization
height is calculated by the multiplication of them. Those points in
Fig. 3(a) represent the average values within a given stage. As the
5.1. Overview of optimization
result reveals, the conversion increases quickly as height goes up
initially, and the increment become less as the height increases.
The first step of optimization is to identify and to categorize
This is because the amount of NO and NO2 become less in the
the variables. Table 4 lists all the variables which need to be deter-
higher portion of the PCR. However, in order to react NO more
mined in the process. In this table, the variables are categorized
efficiently and to avoid NO and O2 recycle back to R-CP, 15 equi-
into three types, Fixed Variables (FVs), Adjusted Variables (AVs)
librium stages (equivalent to 9.14 m of packing) are used to make
and Optimization Variables (OVs). FVs are those variables fixed
the conversion of NO reach 97%. Fig. 3(b) illustrates the temper-
at given values based on overall process understandings. AVs are
ature profile inside PCR. From the result, the temperature of the
those variables adjusted to reach some specifications, and the ba-
whole column is in the range between 44.4 and 45.8 °C. This is
sis for setting AVs are mostly related to separation performances
consistent with the suggestion that the maximum operating tem-
like purity and recovery. OVs are remaining variables that cannot
perature should be around 45 °C in order not to enhance the un-
be categorized to be FVs or AVs, and are treated as design and op-
desired side reactions. Besides, the variation of temperature is not
erating variables in this system. OVs have distinct features com-
large, which may be concluded from the fact that the regeneration
pared with FVs and AVs. Firstly, they cannot be determined only by
is just slight exothermic in nature. The result in Fig. 3(b) also sup-
overall process understandings, nor can they be limited by given
ports the simplification in the work that modeled the regeneration
constraints. Secondly, the changing of those variables lead to ob-
reactor as a CSTR module. Fig. 3(c) illustrates the composition pro-
vious influences on the process performances, typically in energy
file of methanol inside PCR. Due to the large amount of reaction at
consumption or cost. Thus, they have to be determined through
the bottom part of PCR, there is large amount of water generated,
optimization.
which dilutes the methanol concentration. From this sub-figure, it
Before optimization starts, the calculation of capital and oper-
reveals that methanol concentration is above 70 wt% throughout
ating costs should be clarified. As mentioned previously, the cal-
the column. This is consistent with the assumption that side re-
culation of costs for the major equipment and the required util-
actions (Eq. (7f) to (7 h)) are effectively inhibited and thus are ig-
ity is based on the correlation provided by Luyben (2011). Besides,
nored in this work.
the heat source of C2 reboiler is 15-psig steam generated from
The bottom stream from PCR contains mostly methanol and
the waste heat in R-CP. There is no cost data of steam with this
water, and slight amount of DMC is further separated in another
quality in Luyben’s book. According to Douglas, the cost of steam
downstream distillation column (C2). This column is equipped with
with this quality is at 2.284 (USD/GJ) (Douglas, 1988). Because Luy-
a partial vapor-liquid condenser with three outlet streams, thus it
ben’s correlation is adopted throughout this work, thus this price
has three degrees-of-freedoms to be specified. The vapor distillate
is prorated with high pressure steam cost (at 42 bar and 254 °C)
contains the non-condensed light gases, the liquid distillate is the
provided by Luyben, and with 10% penalty from generating steam.
pure methanol that is recycled to PCR, and the bottom is the aque-
Thus the cost for the 15-psig steam is set at 3.6 (USD/GJ) in this
ous waste. In order to save energy, the specification of liquid distil-
work. If the amount of steam generated exceeds the requirement
late is set at 95 mol% methanol, and the bottom is set at 99 mol%
in C2 reboiler, the remainder is viewed as a bonus from the process
water. Reflux ratio and reboiler duty are adjusted to reach these
with the same price. On the contrary, the industrial low-pressure
two specifications. Condenser temperature of C2 is selected as the
steam (6 bar, 154 °C, unit price = 7.78 USD/GJ) is used to make up
92 B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100
Table 4
Status of variables in this process.
the difference in reboiler duty. Last but not least, the cost of cata- convenient and reasonable analysis can be provided. In the results
lyst is assumed to be 25 (USD/kg). of this work, the final optimal point obtained in later Section 5.3 is
It is also important to consider the costs of the raw material selected as the reference point, with all the data already corrected.
into TAC calculation. In this work, the raw materials used for re-
actions are CO, methanol, NO and O2 . The amount of CO input to
the process is assumed to be identical to all the optimized cases
studied, so that its cost is then ignored. For methanol, there is 5.2. Selection of optimization variables
a major input (stream MeOH-MKP on Fig. 1), and some possible
loses (streams PUG-1, PUG-2, and PUG-3 on Fig. 1) in the pro- The whole process contains five major units (R-CP, MSC, C1, PCR
cess. The major loss is through PUG-1. However, as mentioned and C2) with a major recycle stream to provide MN to the coupling
in Section 3, this stream is used for regeneration of methanol, reactor (R-CP). Thus in optimization, all these units should be con-
which is not studied in this work. Therefore, methanol which goes sidered at the same time to account for the interaction behavior
out through this stream is actually not lost. On the other hand, among units. Through the changing of optimization variables, the
losses through PUG-2 and PUG-3 are slight, which can be inferred recycle stream should be always remained connected. For the pur-
from Fig. 1. Hence, the raw material cost of methanol can also pose of convenience during optimization, the recycle stream from
be neglected from calculation of TAC. For oxygen, the required top of PCR to coupling reaction (stream REC in Fig. 1) was broken,
amount is not much. In this work, we assume that the syngas- but with aspen built-in “Transfer” function applied to make the
to-EG process is part of a poly-generation process, so the source flowrates, temperature and compositions of two broken streams
of oxygen is from an air separation unit, which is operated at a identical. Through this way, the essence of connecting the recycle
large scale. So the raw material cost of O2 can also be reasonably stream can be retained.
ignored. Due to the large amount of variables to be determined, some
Hence in this study, only the cost for the makeup NO is consid- preliminary tests were performed to appropriately simplify the op-
ered in the stream cost when analyzing TAC. Because there is no timization problem. In R-CP, CO/MN and FRACCO+MN are two im-
cost data from open literature, thus the NO cost was set at 3 times portant variables that affect reaction performance. CO/MN affects
of the cost of the raw material ammonia, to be 0.93 (kUSD/MT). the amount of catalyst required for a given targeted reaction con-
The way of calculating this makeup cost is first to set a reference version. In the range between 1.5 and 2.0 suggested by litera-
case, and then the makeup flow rate in each other case is sub- ture (Ji, 2010), the required catalyst amount dropped as CO/MN
tracted from that in the reference case and then calculate the cor- increased, which implied that the reaction rate was higher un-
responding cost difference. Finally, this cost difference is included der lower CO/MN value. Thus in this work, CO/MN is fixed at 1.5
in the operating cost of TAC. Note that although the value of calcu- throughout the optimization work. Also, FRACCO+MN was suggested
lated TAC may be different as the reference case changes, the rel- to be in range of 30–40 mol%. From the preliminary analysis, it
ative relationship among cases do not change. Through this way, a reveals that TAC decreases as FRACCO+MN increases. This is be-
cause of the fact that less inert species (nitrogen in this work) is
B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100 93
(6 bar, 7.78 USD/GJ). On the other hand, all of the 2-psig steam re-
covered from R-CP is viewed as the bonus from this process, and
the earning is subtracted from the TAC. The TAC of this case is under different RMeOH from 0.82 to 0.84, Tc-C2 from 47 to 62 (°C),
3797.5 USD/y. Then, the heat integration strategy is applied with- and NT-MSC from 3 to 5. For investigating Tc-C2 , a 5 (°C) interval
out changing any of the design variables, and it is found that 13.5% was firstly adopted to take an overall and quick glance at the im-
reduction in TAC can be obtained. The calculation results are listed pact of this variable on TAC. Generally, as Tc-C2 increases, TAC de-
in the first two columns of Table 5. From this calculation, it is con- creases first and then increases. For the cases under a low Tc-C2 ,
cluded that the process can be largely benefited by this simple the capital cost of condenser in C2 is large, because the tempera-
heat integration. ture difference of distillate and cooling water is quite small. Also,
Then, the optimization is performed, with the heat-integrated a larger reboiler duty is required to reach the specification men-
base case serving as the starting point. The major optimization re- tioned in Table 4. As Tc-C2 increases, the capital cost of condenser
sults are illustrated from Figs. 5 to 9, while the other detailed re- and the utility cost in reboiler both decreases. However, there are
sults are included in the appendix. Fig. 5 describes the variation two negative effects from a high Tc-C2 . The first one is that a high
of TAC (in kUSD) and methanol content at PCR bottom (in wt%) Tc-C2 increases the cooling duty required in PCR condenser, as the
94 B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100
(a) (b)
3240 0.7025
NT-MSC = 3 NT-MSC = 3
3220 NT-MSC = 4 NT-MSC = 4
0.7020
NT-MSC = 5 NT-MSC = 5
3200
0.7015
MeOH (wt%)
TAC (kUSD)
3180
0.7010
3160
0.7005
3140
3120 0.7000
3100 0.6995
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc- C2 ( C) Tc- C2 ( C)
(c) (d)
3260 0.708
NT-MSC = 3 NT-MSC = 3
3240 0.707
NT-MSC = 4 NT-MSC = 4
NT-MSC = 5 0.706 NT-MSC = 5
3220
0.705
3200
MeOH (wt%)
TAC (kUSD)
0.704
3180 0.703
3160 0.702
0.701
3140
0.700
3120 0.699
3100 0.698
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc- C2 ( C) Tc- C2 ( C)
(e) (f)
3280 0.712
NT-MSC = 3 NT-MSC = 3
3260 NT-MSC = 4 0.711 NT-MSC = 4
NT-MSC = 5 0.710 NT-MSC = 5
3240
0.709
MeOH (wt%)
TAC (kUSD)
3220
0.708
3200 0.707
3180 0.706
3160 0.705
0.704
3140
0.703
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc- C2 ( C) Tc- C2 ( C)
Fig. 5. Optimization results at MeOH/NOs = 2.0
(a)(b) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.82
(c)(d) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.83
(e)(f) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.84.
B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100 95
Table 5
TAC comparison between the base cases (with and without heat integration) and the optimized case.
Base case (w/o heat integration) Base case (w/ heat integration) Optimized
liquid distillate is majorly methanol and is recycled back to the top MeOH/NOs = 2.0 and 2.1 occur at the boundary. Once the RMeOH
of PCR. The second one is that the purge amount in vapor distillate becomes even lower, the methanol concentration of 70 (wt%) can-
increases with Tc-C2 . In the purge stream, there is some loss of the not be satisfied.
valuable NO which is not reacted in PCR. The penalty of NO loss Fig. 7 describes the variation of TAC (in kUSD) and methanol
from C2 purge cannot be ignored. Those reasons mentioned above content at PCR bottom (in wt%) at different RMeOH from 0.73 to
combined to form the trade-off in TAC. And from Fig. 5, it is clear 0.75, Tc-C2 from 47 to 62 (°C), and NT-MSC from 3 to 5 under
that the suitable Tc-C2 is likely to be in the range between 52 or 57 MeOH/NOs = 2.2. The major conclusions are again similar to those
(°C). The detailed investigation on TAC in this region is illustrated obtained from Fig. 5. However, a distinct difference comparing
later. with cases under MeOH/NOs = 2.0 and 2.1 is that the optimal point
Two other trends can also be concluded from Fig. 5. The first is no longer at the boundary. This means that under this situ-
one is that cases of greater RMeOH have higher TAC, and this is ation, the amount of methanol circulated in the process is rela-
because larger amount of cooling duty is required at PCR con- tively abundant, which does not tend to let the MeOH concentra-
denser. However, the requirement of MeOH content higher than tion lower than the constrained value during optimization.
70 wt% cannot be achieved once the RMeOH is too low. Also, a For the above-mentioned three cases, the detailed information,
higher RMeOH implies that less amount of methanol is recycled to including the TAC and the methanol concentration, is listed from
coupling reaction and MSC. Thus, the required condenser temper- Table A1 to A3, while the optimization results for NT1, NT2, NF1,
ature in MSC becomes lower under a given DMO recovery, which NF2 under each MeOH/NOs are illustrated from Fig. A1 to A6.
increases the cooling duty requirement. The second trend is that As mentioned earlier, it is also found that for the optimal cases
comparing with RMeOH , NT-MSC has a relatively small impact on under MeOH/NOS at 2.0, 2.1, and 2.2, the TACs at Tc-C2 = 52 and
TAC. 57 (°C) are at the relative low points, and are very close. Hence, a
Fig. 6 describes the variation of TAC (in kUSD) and methanol deeper look into this region is necessary. The TAC calculation re-
content at PCR bottom (in wt%) at different RMeOH from 0.77 to sults between Tc-C2 = 52 and 57 (°C) under each MeOH/NOS with
0.79, Tc-C2 from 47 to 62 (°C), and NT-MSC from 3 to 5 under the optimal RMeOH are listed in Fig. 8(a)–(c). From these results, it
MeOH/NOs = 2.1. The major conclusions obtained from Fig. 5 re- is found that the optimal Tc-C2 is at 54 °C, although the variation is
tains in Fig. 6, but with lower optimal RMeOH than that un- still not very large.
der MeOH/NOs = 2.0. The reason is due to the larger amount of For even lower MeOH/NOs value, such as 1.9, it becomes more
methanol existed in PCR, which directly reduces the cooling utility difficult to satisfy the constrained 70 wt% of MeOH composition. As
required in MSC under the same DMO recovery, and also relieves there is now less MeOH inside PCR, the PCR condenser should be
the limitation to maintain the methanol concentration at above operated under a much lower temperature in order to condense
70 wt% in the process. Moreover, both of the optimal RMeOH under more MeOH back into PCR. This leads to the need of the higher
96 B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100
(a) (b)
3240 0.7045
NT-MSC = 3 NT-MSC = 3
3220 NT-MSC = 4 0.7040 NT-MSC = 4
NT-MSC = 5 NT-MSC = 5
3200 0.7035
3180 0.7030
TAC (kUSD)
MeOH (wt%)
3160 0.7025
3140 0.7020
0.7015
3120
0.7010
3100
0.7005
3080
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc-C2 ( C) Tc-C2 ( C)
(c) (d)
3260 0.712
NT-MSC = 3 NT-MSC = 3
3240 0.711 NT-MSC = 4
NT-MSC = 4
3220 NT-MSC = 5 0.710 NT-MSC = 5
3200 0.709
MeOH (wt%)
TAC (kUSD)
3180 0.708
3160 0.707
3140 0.706
3120 0.705
3100 0.704
3080 0.703
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc-C2 ( C) Tc-C2 ( C)
(e) (f)
3280
NT-MSC = 3 0.715 NT-MSC = 3
3260 NT-MSC = 4 NT-MSC = 4
NT-MSC = 5 0.714 NT-MSC = 5
3240
3220 0.713
MeOH (wt%)
TAC (kUSD)
3200 0.712
3180 0.711
3160 0.710
3140
0.709
3120
0.708
3100
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc-C2 ( C) Tc-C2 ( C)
Fig. 6. Optimization results at MeOH/NOs = 2.1
(a)(b) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.77
(c)(d) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.78
(e)(f) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.79.
B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100 97
(a) (b)
0.7095
3280 NT-MSC = 3 NT-MSC = 3
NT-MSC = 4 NT-MSC = 4
3260 NT-MSC = 5 0.7090 NT-MSC = 5
3240
MeOH (wt%)
TAC (kUSD)
0.7085
3220
3200 0.7080
3180
0.7075
3160
3140 0.7070
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc-C2 ( C) Tc-C2 ( C)
(c) (d)
3280 0.7130
NT-MSC = 3 NT-MSC = 3
3260 NT-MSC = 4 NT-MSC = 4
NT-MSC = 5 NT-MSC = 5
3240 0.7125
MeOH (wt%)
TAC (kUSD)
3220
0.7120
3200
3180
0.7115
3160
3140
0.7110
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc-C2 ( C) Tc-C2 ( C)
(e) (f)
3280 0.7170
NT-MSC = 3 NT-MSC = 3
3260 NT-MSC = 4 NT-MSC = 4
NT-MSC = 5 NT-MSC = 5
3240 0.7165
MeOH (wt%)
TAC (kUSD)
3220
3200 0.7160
3180
3160 0.7155
3140
46 48 50 52 54 56 58 60 62 64 46 48 50 52 54 56 58 60 62 64
o o
Tc-C2 ( C) Tc-C2 ( C)
Fig. 7. Optimization results at MeOH/NOs = 2.2
(a)(b) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.73
(c)(d) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.74
(e)(f) TAC and MeOH content under varying Tc-C2 and NT-MSC at RMeOH = 0.75.
98 B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100
tration at PCR bottom was 70.08 (wt%), and the TAC is 3233.4
3116 (kUSD).
To this point, the optimization is completed. The overall op-
3114 timization results are collected in Fig. 9. The optimal operating
condition is at MeOH/NOS = 2.1, RMeOH = 0.77, Tc- C2 = 54 (°C), NT-
3112 MSC = 4, NT2 = 18, NF2 = 12, NT1 = 13, and NF1 = 5. The calculated
TAC is 3080.9 (kUSD).
3110 The itemized TAC of the optimized case is also listed in
Table 5 in order for direct comparison. We found that 6.2% and
52 53 54 55 56 57
18.87% reduction on TAC can be reached compared with the heat-
Tc-C2 (oC) integrated base case, and the original base case, respectively. Two
(b) comments can be made from this discovering. The first one is that
3110 the reduction of TAC from the heat-integrated base case to the fi-
Min. TAC nal optimized point is not very obvious. This may be because the
3105 suitable operating region for this process as concluded from the
literature has been narrow, which led to another fact that the se-
3100 lected starting point of the optimization is pretty close to the opti-
Min. TAC (kUSD)
mal point. The second one is that larger MeOH/NOs lead to lower
3095 TAC. This conclusion is different from those summarized in open
literatures, which suggested that lower MeOH/NOs is better. This
3090 difference is likely to be arisen from the utility cost in C2 reboiler
and cooling cost of PCR. In this work, the utility used in PCR con-
3085
denser is chilled water, with the unit cost to be 4.43 USD/GJ as
suggested by Luyben. This is close to the steam price used in C2
3080
reboiler (3.60 USD/GJ), which is generated from the waste heat of
3075 R-CP. If the heating source changes, the optimization results are
52 53 54 55 56 57 likely to shift based on the new steam unit cost. In this work, the
Tc-C2 (oC) reason for the lower steam cost is because of the inclusion of a
primary heat integration scheme, which is also a practical strategy
(c) in industry when trying to become more energy-saving. Thus, the
authors consider the optimization work provided in this paper to
3152 Min. TAC be reasonable.
3150 5. Conclusion
Min. TAC (kUSD)
(a) (b)
3150 MeOH/NOs = 2.0 MeOH/NOs = 2.1
3120
Min. TAC (kUSD)
3130
3100
3120
3090
3110
3080
3155 3200
3150 3160
3120
3145
3080
3140
0.73 0.74 0.75 1.9 2.0 2.1 2.2
RMeOH (-)
MeOH/NOs (-)
Fig. 9. Overall optimization results
(a) Minimized TAC under MeOH/NOs = 2.0
(b) Minimized TAC under MeOH/NOs = 2.1
(c) Minimized TAC under MeOH/NOs = 2.2
(d) Total minimized TAC.
Acknowledgment Douglas, J.M., 1988. Conceptual Design of Chemical Processes. McGraw-Hill, New
York.
Gao, X., Zhu, Y.P., Luo, Z.H., 2011. CFD modeling of gas flow in porous medium
The research funding from the Ministry of Science and Technol- and catalytic coupling reaction from carbon monoxide to diethyl oxalate in
ogy of R. O. C. under grant no. MOST 105-2221-E-002-210 is greatly fixed-bed reactors. Chem. Eng. Sci. 66 (23), 6028–6038.
appreciated. Harvianto, G.R., Ahmad, F., Lee, M., 2017. A hybrid reactive distillation process with
high selectivity pervaporation for butyl acetate production via transesterifica-
tion. J. Membr. Sci. 543, 49–57.
Supplementary materials Hickman, D.A., Schmidt, L.D., 1991. Modeling catalytic gauze reactors - ammonia
oxidation. Ind. Eng. Chem. Res. 30 (1), 50–55.
Hsu, K.Y., Hsiao, Y.C., Chien, I.L., 2010. Design and control of dimethyl carbon-
Supplementary material associated with this article can be ate-methanol separation via extractive distillation in the dimethyl carbonate re-
found, in the online version, at doi:10.1016/j.compchemeng.2018. active-distillation process. Ind. Eng. Chem. Res. 49 (2), 735–749.
Ji, Y., 2010. Study of catalyst, mechanism, and intrinsic kinetics of CO catalytic cou-
08.025. pling to dimethyl oxalate (DMO) Ph. D. Thesis. East China University of Science
and Technology.
Ji, Y., Liu, G., Li, W., Xiao, W.D., 2009. The mechanism of CO coupling reaction
References
to form dimethyl oxalate over Pd/alpha-Al2O3. J. Mol. Catal. Chem. 314 (1-2),
63–70.
Alwahabi, S.M., Froment, G.F., 2004. Conceptual reactor design for the methanol– Ji, Y., Zhang, B., Li, W., Xiao, W.D., 2010. Study on intrinsic kinetics of coupling re-
to-olefins process on SAPO-34. Ind. Eng. Chem. Res. 43 (17), 5112–5122. action of carbon monoxide to dimethyl oxalate. Nat. Gas Chem. Ind. (35) 20–24
Cao, Y., Huang, K.J., Yuan, Y., Chen, H.S., Zhang, L., Wang, S.F., 2017. Dynamics and (In Chinese).
control of reactive distillation columns with double reactive sections: feed-split- Jiang, C.W., Zheng, Z.W., Zhu, Y.P., Luo, Z.H., 2012. Design of a two-stage fluidized
ting influences. Ind. Eng. Chem. Res. 56 (28), 8029–8040. bed reactor for preparation of diethyl oxalate from carbon monoxide. Chem.
Chang, J., Zhang, K., Chen, H.G., Yang, Y.P., Zhang, L.M., 2013. CFD modelling of the Eng. Res. Des. 90 (7), 915–925.
hydrodynamics and kinetic reactions in a fluidised-bed MTO reactor. Chem. Eng. Kaewwisetkul, P., Simasatitkul, L., Arpornwichanop, A., 2017. Design and optimiza-
Res. Des. 91 (12), 2355–2368. tion of dimethyl ether production from crude glycerol in a reactive distillation.
Chen, Y. 1995. Gas Phase Catalytic Synthesis of Oxalate Continuous Process, Chem. Eng. Process. 117, 80–88.
CN1027257C (In Chinese). Li, Z., Qian, Z., Zhao, X., Xiao, W., 2004. Studies on hydrogenation of dimethyl ox-
Chua, W.J., Rangaiah, G.P., Hidajat, K., 2017. Design and optimization of isopropanol alate on Cu/SiO2 catalyst. Chem. React. Eng. Technol. 20 (2), 121–127 (In Chi-
process based on two alternatives for reactive distillation. Chem. Eng. Process. nese).
118, 108–116. Li, Z.H., Wang, W.H., Lv, J., Ma, X.B., 2013. Modeling of a packed bubble column for
Crandall, J.W., Deitzler, J.E., Kapicak, L.A., Poppelsdorf, F. 1987. Process for the hy-
drolysis of dialkyl carbonates, US 4663477.
100 B.-Y. Yu et al. / Computers and Chemical Engineering 119 (2018) 85–100
methyl nitrite regeneration based on reaction kinetics and mass transfer. Ind. Tan, J., Ji, Y., Xiao, W. 2008. IR Study and Kinetic Research on CO Catalytic Coupling
Eng. Chem. Res. 52 (8), 2814–2823. to Dimethyl Oxalate Guangdong Chemical Industry 6:7-9 (In Chinese).
Liu, G., Fu, J., Ji, Y., Jin, L., Li, W., 2010a. Measurement of methyl nitrite Henry’s law Turton, R., Bailie, R.C., Whiting, W.B., Shaeiwitz, J.A., 2008. Analysis, Synthesis, and
constants in methanol. Chem. Eng. 38 (6), 61–63 (In Chinese). Design of Chemical Processes. Pearson Education, Boston.
Liu, G., Ji, Y., Li, W., 2010b. Kinetic study on methyl nitrite synthesis from methanol Wang, B., Wang, D., Li, Y., Xu, C. 2015. Process and Device System for Producing
and dinitrogen trixiode. Chem. Eng. J. 157 (2-3), 483–488. Dimethyl Oxalate by High-pressure Carbonylation of Industrial Synthesis Gas
Liu, Y., Jiao, W., Shen, H., Qi, G. l., Yuan, Z. 2013. Process Method and Method for and Producing Ethylene Glycol by Hydrogenation, WO 2015184677 A1.
Regenerating Methyl Nitrite in Production of Dimethyl Oxalate, CN 102008922B. Wang, C.Z., Chen, P.J., Li, Y.K., Zhao, G.F., Liu, Y., Lu, Y., 2016. In situ DRIFT’S study
(In Chinese). of CO coupling to dimethyl oxalate over structured Al-fiber@ns-AlOOH@Pd cat-
Luo, N., Ji, Y., Mao, Y., Zhang, B., 2012. Syn-gas-based mono ethylene glycol synthesis alyst. J. Catal. 344, 173–183.
in Pujing Chemical. Appl. Petrochem. Res. 2, 23–26. Wang, S.J., Yu, C.C., Huang, H.P., 2010. Plant-wide design and control of DMC synthe-
Luyben, W.L., 2011. Principles and Case Studies of Simultaneous Design, Chap. 5. sis process via reactive distillation and thermally coupled extractive distillation.
Wiley, New York. Comput. Chem. Eng. 34 (3), 361–373.
Meng, F.D., Xu, G.X., Guo, R.Q., Yan, H.F., Chen, M.Q., 2004. Kinetic study of carbon Xu, Z.N., Sun, J., Lin, C.S., Jiang, X.M., Chen, Q.S., Peng, S.Y., Wang, M.S., Guo, G.C.,
monoxide coupling reaction over supported palladium catalyst. Chem. Eng. Pro- 2013. High-Performance and long-lived Pd nanocatalyst directed by shape effect
cess. 43 (6), 785–790. for CO oxidative coupling to dimethyl oxalate. ACS Catal. 3 (2), 118–122.
Norkobilov, A., Gorri, D., Ortiz, I., 2017. Comparative study of conventional, reac- Yao, L., 2013. Study on process and simulation of preparation of ethylene glycol by
tive-distillation and pervaporation integrated hybrid process for ethyl tert-butyl syngas Master Thesis. Zhengzhou University.
ether production. Chem. Eng. Process. 122, 434–446. Yu, B.Y., Chien, I.L., 2015. Design and economic evaluation of a coal-based polygener-
Peking University Pioneer Co., 2016. http://www.pioneer-pku.com/en/product.aspx? ation process to coproduce synthetic natural gas and ammonia. Ind. Eng. Chem.
ProductsCateId=78&CateId=78. (Accessed August 2016, In Chinese) Res. 54 (41), 10073–10087.
Peng, S.Y., Xu, Z.N., Chen, Q.S., Wang, Z.Q., Chen, Y.M., Lv, D.M., Lu, G., Guo, G.C., Yu, B.Y., Chien, I.L., 2016. Design and optimization of the methanol-to-olefin pro-
2014. MgO: an excellent catalyst support for CO oxidative coupling to dimethyl cess part I: steady-state design and optimization. Chem. Eng. Technol. 39 (12),
oxalate. Catal. Sci. Technol. 4 (7), 1925–1930. 2293–2303.
Peng, S.Y., Xu, Z.N., Chen, Q.S., Wang, Z.Q., Lv, D.M., Sun, J., Chen, Y.M., Guo, G.C., Yue, H.R., Zhao, Y.J., Ma, X.B., Gong, J.L., 2012. Ethylene glycol: properties, synthesis,
2015. Enhanced stability of Pd/ZnO catalyst for CO oxidative coupling to and applications. Chem. Soc. Rev. 41 (11), 4218–4244.
dimethyl oxalate: effect of Mg2+ doping. ACS Catal. 5 (7), 4410–4417. Zang, X.X., Huang, K.J., Yuan, Y., Chen, H.S., Zhang, L., Wang, S.F., Wang, K., 2017.
Pradhan, M.P., Suchak, N.J., Walse, P.R., Joshi, J.B., 1997. Multicomponent gas absorp- Reactive distillation columns with multiple reactive sections: a case study on
tion with multiple reactions: modelling and simulation of NOx absorption in the disproportionation of trichlorosilane to silane. Ind. Eng. Chem. Res. 56 (3),
nitric acid manufacture. Chem. Eng. Sci. 52 (24), 4569–4591. 717–727.
Schmidt, L.D., Huff, M., Bharadwaj, S.S., 1994. Catalytic partial oxidation reactions Zhang, Q.R., Peng, J.Y., Zhang, K., 2017. Separation of an azeotropic mixture of
and reactors. Chem. Eng. Sci. 49 (24a), 3981–3994. dimethyl carbonate and methanol via partial heat integration pressure swing
Shi, L., Wang, S.J., Huang, K.J., Wong, D.S.H., Yuan, Y., Chen, H.S., Zhang, L., Wang, S.F., distillation. Asia-Pac. J. Chem. Eng. 12 (1), 50–64.
2017. Intensifying reactive dividing-wall distillation processes via vapor recom- Zhao, T.J., Chen, D., Dai, Y.C., Yuan, W.K., Holmen, A., 2004. Synthesis of dimethyl ox-
pression heat pump. J. Taiwan Inst. Chem. E 78, 8–19. alate from CO and CH3 ONO on carbon nanofiber supported palladium catalysts.
Song, H.Y., Jin, R.H., Kang, M.R., Chen, J., 2013. Progress in synthesis of ethylene gly- Ind. Eng. Chem. Res. 43 (16), 4595–4601.
col through C-1 chemical industry routes. Chin. J. Catal. 34 (6), 1035–1050. Zhao, X.G., Lin, Q., Xiao, W.D., 2005. Characterization of Pd-CeO2 /alpha-alumina cat-
Song, K., Ji, Y., Xiao, W., 2007. intrinsic kinetic research on vapor-phase catalytic alyst for synthesis of dimethyl oxalate. Appl. Catal. Gen. 284 (1-2), 253–257.
coupling of CO and methyl nitrite to DMO. Guangdong Chem. Ind. 34 (6), 12 (In Zhu, Y., Yang, Z., Xu, C., Wang, G., W., G. 2015. Catalyst for CO gas phase synthesis
Chinese). of dimethyl oxalate and preparation method for catalyst, CN patent 104741116
Soundararajan, S., Dalai, A.K., Berruti, F., 2001. Modeling of methanol to olefins A (In Chinese).
(MTO) process in a circulating fluidized bed reactor. Fuel 80 (8), 1187–1197. Zhu, Y.P., Tu, S., Luo, Z.H., 2012. Modeling for the catalytic coupling reaction of car-
Tahara, S., Fujii, K., Nishihira, K., Matsuda, M., Mizutare, K. 1984. Process for contin- bon monoxide to diethyl oxalate in fixed-bed reactors: reactor model and its
uously preparing ethylene glycol, US 4453056. applications. Chem. Eng. Res. Des. 90 (9), 1361–1371.