- Email: [email protected]
Design and optimization of the levulinic acid recovery process from lignocellulosic biomass
Accepted Manuscript Title: Design and Optimization of the Levulinic Acid Recovery Process from Lignocellulosic Biomass Author: Le Cao Nhien Nguyen Van Duc Long Moonyong Lee PII: DOI: Reference:
S0263-8762(15)00365-2 http://dx.doi.org/doi:10.1016/j.cherd.2015.09.013 CHERD 2024
To appear in: Received date: Revised date: Accepted date:
18-6-2015 31-8-2015 17-9-2015
Please cite this article as: Nhien, L.C., Long, N.V.D., Lee, M.,Design and Optimization of the Levulinic Acid Recovery Process from Lignocellulosic Biomass, Chemical Engineering Research and Design (2015), http://dx.doi.org/10.1016/j.cherd.2015.09.013 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1
Highlights
An innovative TDWC-D was proposed to separate a heterogeneous azeotrope mixture.
3
The TAC could be reduced by 28.1% by the innovative TDWC-D.
4
Many benefits could be achieved by changing process sequence structure.
5
A compact and energy efficient LA production process is proposed.
6
Large potential for commercial LA production from lignocellulosic biomass
cr
ip t
2
7 8
us
Design and Optimization of the Levulinic Acid Recovery Process
9
an
from Lignocellulosic Biomass
10
Le Cao Nhien1, Nguyen Van Duc Long1 and Moonyong Lee*
M
11 12
School of Chemical Engineering, Yeungnam University, Gyeongsan 712-749, South Korea
13
Running title: Design and Optimization of the Levulinic Acid Recovery Process from
15
Lignocellulosic Biomass
Ac ce p
16
te
d
14
17
Submitted to Chemical Engineering Research and Design Journal
18
1
19
*
20
Prof. Moonyong Lee
21
Email: [email protected]
22
Telephone: +82-53-810-2512
These two authors contributed equally to this work.
Correspondence concerning this article should be addressed to:
23
1 Page 1 of 40
23
Abstract Levulinic acid (LA), which is one of the top twelve value-added chemicals from biomass
25
feedstock, has been recognized in a large number of applications. Nevertheless, its capability on an
26
industrial scale has been limited by the high-cost of the raw materials and the lack of detailed process
27
design. This paper reports the simulation, detailed design and optimization of an industrial process
28
recovering LA from biomass, a renewable and inexpensive feedstock. The aqueous feed mixture from
29
the acidic hydrolysis of biomass was fed to the separation process to recover LA as the main product,
30
along with furfural and formic acid as valuable byproducts. Furfural was then used as an extracting
31
solvent, resulting a significant decrease in total cost. Detailed analysis was conducted to assess the
32
feasibility of the separation method. The purification process was intensified by applying an
33
innovative top dividing wall column with a decanter configuration (TDWC-D). The response surface
34
methodology (RSM), which allows the interactions between variables to be identified and quantified,
35
was used to optimize the TDWC-D structure. The predictions by the RSM showed satisfactory
36
agreement with the rigorous simulation results. The results showed that the innovative TDWC-D
37
configuration could save up 16.4% and 20.6% of the overall energy requirements and total annual cost,
38
respectively, compared to the conventional sequence.
40 41
cr
us
an
M
d
te
Ac ce p
39
ip t
24
Keywords: Lignocellulosic Biomass; Levulinic Acid; Furfural; Energy Efficiency Improvement; Dividing Wall Column; Response Surface Methodology
2 Page 2 of 40
42 43
1. Introduction The dependence of the global economy on the utilization of mineral resources has increased in
45
recent years, accounting for 80.9% of the total energy production in 2006 (Koh and Ghazoul, 2008).
46
Nevertheless, these resources have many disadvantages, such as limited reserves, environmentally
47
damaging and responsibility for global warming. Therefore, the interest of biomass, which is the only
48
renewable resource of fixed carbon, has increased considerably. Biomass can not only produce
49
conventional hydrocarbon liquids for transportation, but also petrochemical products (Huber et al.,
50
2006, Petrus et al., 2006, Corma et al., 2007). Biofuels from biomass are viewed by many policy
51
makers as a key to reduce the reliance on foreign oil, lowering the emissions of greenhouse gases and
52
meeting the rural development goals (Koh and Ghazoul, 2008).
an
us
cr
ip t
44
Levulinic acid, which is recognized in a large number of applications, is listed as one of the top-
54
twelve value-added chemicals from biomass in terms of its potential markets, its derivatives and the
55
complexity of the synthetic pathways (Werpy and Petersen, 2004). This compound has frequently
56
been suggested to be a starting material for the production of many industrial and pharmaceutical
57
compounds, such as methyltetrahydrofuran, delta aminolevulinic acid, ethyl levulinate and diphenolic
58
acid. In particular, with MTHF applications only, LA manufacturers could face the potential demand
59
of more than 20,000 kilotons by 2020 (Grand View Research, 2014). Nevertheless, the capability of
60
this process on an industrial scale has been limited by the expensive raw materials and energy-
61
intensive processes (Seibert, 2010).
Ac ce p
te
d
M
53
62
Several processes have been used to produce LA from an inexpensive feedstock, lignocellulosic
63
biomass. Dunlop and Wells (1957) proposed the earliest continuous process for producing LA from
64
corn cob waste. The aqueous feed mixture obtaining from acidic hydrolysis is introduced to an
65
extractor to extract LA and then fed to a distillation column to produce pure LA. Ghorpade and Hanna
66
(1999) presented the LA process based on the reactive extrusion from corn starch. These two
67
processes only considered LA as a product. Recently, Sadhukhan et al. (2014) presented a good
68
example of a LA production process from corn cob, which considered both the reaction and separation
69
stages. Seibert (2010) reported a process that uses furfural to extract LA and formic acid from the 3 Page 3 of 40
acidic hydrolysis of biomass. The advantage of this process is the use of furfural, one of the products,
71
as a solvent. Interestingly, these reports showed that a combination of an extractor and a series of
72
distillation columns is used to purify LA, showing great potential to improve the energy efficiency,
73
economic performance and environmental impact. Furthermore, none of the previous processes
74
considered the detailed design methodology and advanced techniques in the case of commercial scale
75
production. A comprehensive design by intensified configurations with economic evaluation is needed
76
to expand the capacity to an industrial scale as well as to reduce the LA price and environmental
77
impact.
us
cr
ip t
70
Many studies have examined the relative advantages of a dividing wall column (DWC) (as shown
79
in Fig. 1) to improve the energy efficiency and reduce the consumption of utilities (Amminudin and
80
Smith, 2001; Long and Lee, 2013a; Long and Lee, 2013b). These reports have shown that DWC
81
systems can achieve energy savings of up to 30% over conventional direct and indirect distillation
82
sequences (Kiss and Ignat, 2012; Long et al., 2013). This is because DWCs can allow reversible splits
83
with no part of the separation being performed twice, which is the main contributor to their superior
84
energy efficiency over other column configurations (Poth et al., 2004). Additional benefits arise from
85
the fact that only one reboiler and condenser are involved instead of two reboilers and two condensers
86
in a conventional distillation sequence (Premkumar and Rangaiah, 2009).
Ac ce p
te
d
M
an
78
87
To design and optimize DWC, Halvorsen and Skogestad (2003) proposed the use of a minimum
88
vapor flow, or the so called Vmin diagram method, which is a graphical visualization of the minimum
89
energy. On the other hand, it cannot be applied to the separation of azeotropic mixtures. Several
90
rigorous methods have proposed, such as a sequential quadratic programming-based method
91
implemented in Aspen Plus (Kiss and Suszwalak, 2012a), and a MILP approach based on the use of
92
two superstructure levels (Caballero and Grossmann, 2014).
93
Recently, Long and Lee (2012) reported a practical and efficient method based on the response
94
surface methodology (RSM). The RSM is a general technique that is used for an empirical study of
95
the relationships between the measured responses and independent input variables (Box and Wilson,
96
1951). By allowing the interactions between variables to be identified and quantified, RSM can
97
predict reliable results with a little computational effort. Therefore, it can be used to optimize the 4 Page 4 of 40
design and operation conditions of a process (Long and Lee, 2012). On the other hand, when the
99
behavior of the function is highly nonlinear in terms of the design variables, the RSM encounters
100
some difficulties in approximating the sufficient quality on the entire design space (Merval et al.,
101
2006). The optimum may not be defined if the range of factors is too narrow or too wide. Furthermore,
102
critical factors may not be defined or specified correctly. In the case of many factors, the problem
103
becomes complex and the RSM cannot be predicted well.
ip t
98
This paper proposes an economical and environmental process for producing LA from an
105
inexpensive raw material, lignocellulosic biomass, along with the detailed design variables and
106
operating conditions. The simulation work was conducted using the commercial simulator, Aspen Plus.
107
A range of sequences and intensified configurations applying TDWC-D are presented with respect to
108
their energy consumption and economic performance. Furthermore, the RSM was used to optimize
109
the structure of the proposed TDWC-D.
M
2. LA separation process
112
2.1. Separation feasibility
d
110 111
an
us
cr
104
The feed mixture considered was obtained from the acidic hydrolysis process of biomass and was
114
an aqueous solution containing LA, furfural, formic acid, and water. The typical concentrations of the
115
mixture ranged from 3 to 8 wt. % LA, from 1 to 5 wt. % furfural, from 1 to 5 wt. % formic acid, and
116
the balance being water (Reunanen et al. 2013). With this component mixture, although LA did not
117
form an azeotrope with the other components, two azeotrope mixtures still formed. Fig. 2 shows a
118
ternary diagram for the furfural-water-formic acid system. As shown in this figure, the water-furfural
119
mixture has a minimum-boiling heterogeneous azeotrope with a composition of 64.5 wt.% furfural at
120
97.8 oC. Another maximum-boiling homogeneous azeotrope forms between formic acid and water
121
with a higher azeotropic temperature of 106.8 oC. In that ternary map, there are two distillation
122
regions divided by the distillation boundary.
Ac ce p
te
113
123
Fig. 3 presents a simplified feasible flow sheet for producing LA from lignocellulosic biomass
124
(Seibert, 2010). The conventional process sequence consists of one extractor and three distillation
125
columns. The feed mixture and the furfural solvent are introduced into the extractor (EX) to generate 5 Page 5 of 40
a water-rich phase as the top stream and a solvent-rich phase containing mainly furfural, LA and
127
formic acid as the bottom stream. The furfural-rich stream is then fed to the first distillation column
128
(C1) to collect LA as the bottom stream product, whereas the distillate contains furfural, formic acid
129
and water, which are introduced into the second distillation column (C2). The function of C2 is to
130
separate the inlet into furfural, formic acid as the bottom stream and a light boiling furfural-water
131
azeotropic mixture as the top stream. The azeotropic mixture is then delivered to storage before being
132
recycled back to the extractor. The bottom stream of C2 is then subjected to the third distillation
133
column (C3) to recover the formic acid as the distillate product, and the furfural is collected as the
134
bottom stream. The furfural solvent is recycled back to the extractor and reused as the extractant.
us
cr
ip t
126
All processes were performed using the simulator, Aspen Plus V8.6, with the NRTL–HOC
136
(nonrandom two–liquid–Hayden – O’Connell) fluid package. The HOC equation of state as the vapor
137
phase model and NRTL for the liquid phase were used to predict not only the vapor–liquid
138
equilibrium but also the vapor–liquid–liquid equilibrium of these simulations. Dimerization affects
139
the vapor–liquid equilibrium, vapor phase properties, density, and liquid phase properties. The HOC
140
equation reliably predicts the solvation of polar compounds and the dimerization in the vapor phase
141
that occurs with mixtures containing carboxylic acids (Aspen Technology, 2010). The binary pairs of
142
the interaction parameters for LA were estimated using the UNIFAC model.
143
2.2. Conventional distillation sequence (CS)
Ac ce p
te
d
M
an
135
144
The LA production rate considered in this study was 50 kilotons per year with a commercial purity
145
of 98 wt.% (Hayes et al., 2008). The furfural reused as a solvent was kept at 99.9 wt.% purity while
146
the purity of the formic acid product was 85 wt.%, which is considered the industry standard and
147
accounting for the major portion of the global demand (Burke, 2013). Table 1 lists the feed
148
composition, temperature and pressure used in this process.
149
In this study, the CS was designed and optimized first for a fair comparison with the other
150
proposed optimized sequences (Fig. 3). In this process, the component recoveries depend mainly on
151
the efficiency of the extractor or the extraction solvent efficiency. This means that solvent selection is
152
the key in this process. Although formic acid recovery is quite low when selecting furfural as a
153
solvent, there is some trade-off between the total process cost and the formic acid recovery. Note that 6 Page 6 of 40
furfural is one of the products and more effective in the recovery of LA and formic acid than other
155
solvents, such as methyl isobutyl ketone and isopropyl acetate (Seibert, 2010). In the extractor, the
156
recovery of LA, formic acid and furfural also increase with increasing extracting solvent/feed ratio but
157
more water appears in the furfural-rich phase. In this study, a solvent/feed ratio of 1.2 : 1 was selected
158
to obtain 97 wt.% LA recovery. Table 2 lists the conditions and product specifications for each column
159
in the conventional column sequences. Note that high pressure (HP) steam is used in the first column
160
while medium pressure (MP) steam is used in the second and third columns. The rigorous simulation
161
results showed that the energy requirements of C1, C2 and C3 were 22,930, 4,250 and 3,770 kW,
162
respectively.
us
cr
ip t
154
164
an
163
3. Proposed sequences
This section proposes the different configurations to separate LA and co-products. The results were
166
then compared in terms of the energy requirement and total annual cost while obtaining the same
167
requirements of productivity, recovery and purity. The column diameters were determined using the
168
column flooding conditions that fix the upper limit of the vapor velocity. The operating velocity is
169
normally between 70 and 90% of the flooding velocity (Premkumar and Rangaiah, 2009). In this
170
study, all columns were designed with a load near 85% of the load at the flooding point to prevent
171
flooding in the columns. The capital costs were calculated based on Guthrie’s modular method
172
(Biegler et al., 1997). Table 3 lists the utility cost data used in this study. The 10 year-plant lifetime
173
was used to calculate the total annual cost (TAC).
Ac ce p
te
d
M
165
174
All conventional columns were optimized with the number of trays and the feed location as the
175
main variables for maximizing total annual cost saving (Fig. 4). Fig. 5 shows one optimization case of
176
the C3 in the conventional sequence. In TDWC-D optimization, after a preliminary design, the main
177
variables, such as the internal vapor flow to the prefractionator (FV), and the number of trays in the
178
top (N1), bottom (N2) and feed rectifying (N3) sections were then optimized. The aim of the
179
optimization was to maximize the TAC savings. The optimization was constrained by the product
180
purities and recoveries.
7 Page 7 of 40
181
Max(%TAC saving) f ( N1, N 2, N 3, FV )
182
A Box−Behnken design was used under the response surface methodology to determine how the
183
variables interact and to optimize the system in terms of the TAC saving. Box and Behnken (1960)
184
proposed some three-level designs for fitting the response surfaces. These designs are formed by
185
combining 2k factorials with incomplete block designs. The resulting designs are generally very
186
efficient in terms of the number of required runs, and they are either rotatable or nearly rotatable.
187
Rotability is a very important property in the selection of the RSM. Because the purpose of the RSM
188
is optimization and the location of the optimum is unknown prior to running the experiment or
189
simulation, it makes sense to use a design that provides equal estimation precision in all directions
190
(Montgomery, 2001). Note that the Box-Behnken design is a spherical design with all points lying on
191
the sphere of the radius. In addition, the Box-Behnken design does not contain any points at the
192
vertices of the cubic region created by the upper and lower limits for each variable. This could be
193
advantageous when the points on the corners of the cube represent the factor-level combinations that
194
are prohibitively expensive or impossible to test because of the physical process constraints. Moreover,
195
the Box-Behnken designs often have fewer design points; they require less effort to run than central
196
composite designs with the same number of factors (Minitab 16, 2010). After determining the
197
preliminary ranges of the variables through single-factor testing, the simulation run data was fitted to
198
a second-order polynomial model and the regression coefficients were obtained. The generalized
199
second-order polynomial model used in the response surface analysis is as follows:
ip t
cr
us
an
M
d
te
Ac ce p
200
(1)
k
k
i 1
i 1
Y 0 i N i ii N i2 ij N i N j i j
(2)
201
where Y is the predicted response, Ni are the uncoded or coded values of the variables, β0 is a constant,
202
βi, βii, and βij are the coefficients of the linear, quadratic, and interactive terms, respectively, and ε is
203
the error term. Minitab® software was used for response surface fitting to maximize the total annual
204
cost savings.
205
3.1. Improving conventional sequence by a decanter (CS-D)
8 Page 8 of 40
In the C2 of the conventional sequence, there was an amount of furfural loss (2,175 kg/hr) in the
207
C2-top due to a heterogeneous azeotrope between water and furfural (see Fig. 6). As shown in Fig. 2,
208
a significant liquid phase split envelope is clearly shown in the ternary diagram of the furfural-water-
209
formic acid system so that a decanter can be applied to reduce the furfural loss and make the
210
distillation more efficient. In particular, the decanter or gravity settling tank, are the simplest form of
211
equipment used to separate liquid-liquid mixtures. Several studies have reported the benefits of a
212
decanter in solving a heterogeneous azeotrope (Kiss and Suszwalak, 2012b; Wu et al., 2014). In this
213
case, furfural and formic acid are collected as the bottom stream of the C2, while the binary
214
heterogeneous azeotrope water-furfural is sent to the overhead of C2, which is then delivered to a
215
condenser-decanter. Fig. 6 presents a flow sheet of the C2 integrated with a decanter. The azeotrope
216
furfural-water on top of C2 is cooled and then split naturally into two liquid phases by the decanter,
217
following the liquid-liquid equilibrium tie lines. The water-rich phase is designed to be totally
218
delivered to storage while the organic phase, which contains mainly furfural, is refluxed back to C2.
219
Fig. 7 shows the effects of temperature on the liquid-liquid separation in the decanter. In particular,
220
the amount of furfural loss decreases with decreasing temperature of the inlet-decanter stream.
221
Therefore, in the decanter design, the temperature of the C2-overhead is cooled to 60oC to make the
222
separation in the decanter more efficient. As a result, integration of the decanter can not only save the
223
energy requirements in the reboiler duty (2.39% of the C2 and C3), but also increase the furfural
224
recovery. The amount of furfural loss in the C2-top stream was reduced from 2,175 kg/hr to 336 kg/hr,
225
i.e. a decrease of 85% furfural loss compared to that in the column without a decanter.
226
3.2. Intensifying conventional sequence by innovative TDWC-decanter (TDWC-D)
Ac ce p
te
d
M
an
us
cr
ip t
206
227
In addition, to improve the energy efficiency in the columns, minimizing the number of columns is
228
also crucial because it directly affects the overall economics and safety in a LA production process.
229
DWC would be the best option to intensify the process for achieving both tasks simultaneously. In
230
particular, an innovative TWDC-D was considered here to integrate the C2 and C3 to improve the
231
energy efficiency and reduce the investment cost.
232
After a preliminary design for TDWC-D, it was then optimized using the RSM. Table 4 lists the
233
factors and levels used in the RSM based optimization. Fifteen simulations were run to optimize 3 9 Page 9 of 40
parameters of the TDWC-D structure. For each run, the internal vapor flow to the prefractionator was
235
varied to maximize the TAC saving while obtaining the required product purity and recovery. Fig. 8
236
presents a schematic diagram illustrating the optimized TDWC-D that replaces the columns, C2 and
237
C3, of the CS sequence. The rigorous simulation results show that the optimized TDWC-D can save
238
up 30.6 and 28.1% in terms of the operating cost and TAC, respectively, compared to the two columns,
239
C2 and C3, of the base sequence. Compared to the whole CS sequence, the innovative configuration
240
combining TDWC and a decanter can allow operating cost and TAC savings of 7.0% and 7.2%,
241
respectively.
242
3.3. Improving conventional sequence by changing sequence – proposed sequence (PS)
us
cr
ip t
234
Multicomponent mixtures can be separated by different distillation column sequences. Many
244
studies have examined the optimal sequence of simple columns that do not involve heat integration
245
(Nishida et al., 1981; Smith, 2005; Westerberg, 1985). The number of possible distillation sequences
246
increases with increasing number of products (King, 1980). Interestingly, the “rule of thumb” is the
247
most common method used for finding the optimal sequence of simple distillation columns (King,
248
1980). The sequences that remove the lightest components alone one by one in the column overheads
249
should be favored (Smith, 2005), i.e., favor the direct sequence. In addition, in the base case, the first
250
distillation column C1 requires the largest amount of energy and highest level of steam (high pressure
251
steam) compared to the other columns. This column is used to vaporize a large number of furfural and
252
light components as well as purify LA, which has the highest boiling point (257 oC) compared to the
253
other components, furfural (162 oC), formic acid (101 oC), water (100 oC). Based on the well-known
254
heuristics for separation sequence synthesis, a proposed sequence for producing LA was used to
255
enhance the energy efficiency (see Fig. 9).
Ac ce p
te
d
M
an
243
256
An aqueous mixture from the acidic hydrolysis reaction of biomass and the furfural solvent was
257
fed to the extractor (P-EX) to generate a water-rich phase as the top stream and solvent-rich phase
258
containing mainly furfural, LA and formic acid as the bottom stream. In contrast to the CS sequence,
259
the bottom stream of P-EX was fed to the proposed C1 (P-C1), which delivered a light boiling
260
azeotrope of furfural and water as the top stream, whereas the bottom stream contained formic acid,
261
furfural and LA. The bottom of P-C1 was then introduced to the proposed C2 (P-C2) to collect formic 10 Page 10 of 40
acid as the distillate and the bottom stream comprising furfural and LA, which was subjected to the
263
proposed C3 (P-C3). The role of P-C3 is to separate levulinic acid as the bottom product and furfural
264
as the top product. Here, LA is recovered at the third column so the proposed sequence is called the
265
direct sequence. Note that medium pressure steam was used on P-C1 and P-C2, whereas P-C3 used
266
high pressure steam. The simulation results showed that the energy requirement in the reboiler using
267
HP columns is reduced from 22,930 kW in C1 of the CS sequence to 13,850 KW in C3 of the PS
268
sequence. As a result, the PS sequence that changes the order of the distillation columns can reduce
269
the annual operating cost and TAC by up to 16.4 and 16.0%, respectively, compared to the base case.
270
Note that these heuristics are not novel but they can be applied easily and efficiently, which leads to
271
significant TAC savings in the LA process. This process is particularly interesting to many companies
272
that prefer to use conventional columns.
273
3.4. Improving the proposed sequence by decanter (PS-D)
M
an
us
cr
ip t
262
In the PS sequence, the main role of the first column (P-C1) is to remove water as much as
275
possible. On the other hand, the minimum-boiling heterogeneous azeotrope mixture between furfural
276
and water forms, causing a furfural loss of 3,196 kg/hr in the top of this column. Therefore, a decanter,
277
which can split the liquid-liquid mixture naturally, can be integrated in this column as same as the
278
second column in the base case to reduce the loss of furfural. Fig. 10 shows the simplified flow sheet
279
of the P-C1 integrated with a decanter. To improve the separation efficiency in decanter, the overhead
280
of P-C1 was cooled to 60 oC before being introduced to the decanter. Not only can the furfural loss be
281
reduced by up to 87.9%, but also the amount of water removed increases from 2,960 kg/hr to 3,157
282
kg/hr, which is comparable to the column without a decanter. The simulation results showed that not
283
only can the energy requirement in the reboiler duty in P-C1-D column be reduced slightly (1.7%),
284
but there is also less furfural loss compared to the column without a decanter.
285
3.5. Intensifying proposed sequence with innovative TDWC-D (PS-TDWC-D)
Ac ce p
te
d
274
286
In the CS sequence, although the C1 requires most energy, however, TDWC-D can only be used to
287
replace the C2 and C3. On the other hand, in the PS sequence, both the P-C1 and P-C3 are energy-
288
intensive columns so that intensification either the P-C1 or P-C3 can bring more benefit. In particular,
289
to intensify the proposed conventional sequence, the TDWC was used to integrate the first two 11 Page 11 of 40
columns. To optimize the TDWC-D, some preliminary simulation runs were carried out to determine
291
the main optimizing variables and their levels. Corresponding to the changes in the factor value, the
292
response values (TAC saving) were recorded. The presence of curvature suggests that the simulation
293
region is close to the optimum. Table 4 lists the factors and levels used in the RSM-based optimization.
294
The TDWC-D parameters were optimized over 15 simulation runs. For each run, the internal vapor
295
flow to the prefractionator was varied to maximize the TAC savings while obtaining the required
296
product purity and recovery. Fig. 11 presents a contour plot of the interactions between the design
297
variables N1, N2 and N3. In each case, the one remaining parameter was set automatically as its
298
center point value by the software. The resulting second-order polynomial model was as follows:
us
cr
ip t
290
an
Y 10.3743 5.44663N1 3.99257N2 4.11248N3 3.01056N12
299
5.15933N22 1.55725N32 1.20847N1N2 1.96654N1N3 0.582208300 N2 N3
(3)
As shown in Fig. 12, the TAC saving was predicted to be highest (15.3%) with the coded levels of
302
the top section (N1), bottom section (N2) and feed rectifying section (N3) of 0.6970, 0.5152 and -
303
0.9798, respectively. The resulting determination coefficient was R2 = 95.7, which indicates that the
304
model is suitable for prediction within the range of simulated variables. The natural values of the
305
variables can be derived from the coded levels. Recycling of the vapor was then optimized to
306
maximize the TAC savings while achieving the product purity requirements and recovery. Fig. 13
307
compares the three-dimensional response surface plots of the interaction between N1 and N2
308
evaluated by the RSM and by rigorous simulation. The figure shows that the three-dimensional
309
response surface plots by the RSM are similar to the actual ones. Fig. 14 shows a schematic diagram
310
illustrating the optimal TDWC-D from the conventional sequence. The TDWC-D combining a
311
decanter, which can increase the furfural recovery and improve the heterogeneous azeotrope
312
separation, with a TDWC, which can reduce energy requirement and TAC, can bring many benefits in
313
terms of energy saving and cost reduction. The simulation showed that the TDWC-D can save up to
314
13.2% in terms of the TAC compared to the C2 and C3 of the conventional sequence. This rigorous
315
simulation agreed well with the predicted value (15.3%). In addition, using a TDWC-D could save
316
15.4% of the energy requirements in terms of the reboiler duty compared to P-C1 and P-C2 of the PS
317
sequence.
Ac ce p
te
d
M
301
12 Page 12 of 40
318
Table 5 summarizes the key results, while Fig. 15 conveniently illustrates the annual operating cost
319
and TAC for all configurations considered. The results suggest that the TDWC-D can be applied
320
efficiently to improve the distillation processes while requiring less TAC and energy.
322
ip t
321
4. Conclusions
A levulinic acid recovery process from abundant and inexpensive lignocellulosic biomass was
324
designed and optimized to be more energy-efficient. This process was then intensified efficiently
325
using an innovative TDWC-D. After a preliminary design, the TDWC-D structure was then optimized
326
and implemented simply and efficiently by the RSM using Aspen Plus and Minitab. The simulated
327
results were in good agreement with the predicted values. The use of innovative TDWC-D can have
328
many benefits, such as increasing the furfural recovery, increasing the separation efficiency and
329
decreasing the TAC significantly. The incorporation of a decanter into the TDWC is a very attractive
330
option for improving the separation efficiency and energy savings, especially for the separation of
331
heterogeneous azeotrope mixtures. The proposed sequence including TDWC-D showed 21.6% and
332
20.6% decreases in the energy requirement and TAC, respectively, compared to a conventional
333
column sequence. Based on these results, the use of TDWC-D for LA production, which has the
334
limitations in commercial scale of existing processes currently, is technically feasible and particularly
335
interesting in constructing a new Levulinic plant. The process compactness is an additional benefit
336
from TDWC-D applications.
338 339
us
an
M
d
te
Ac ce p
337
cr
323
Acknowledgments
The authors wish to acknowledge the financial support from the R&D Convergence Program of
340
NST (National Research Council of Science & Technology) of Republic of Korea and KITECH
341
(Korea Institute of Industrial Technology) (ES150001).
342 343
Appendix. Column cost correlations
344
a. Capital cost: Guthrie’s modular method was applied (Biegler et al., 1997). The investment cost for
345
conventional distillation is the total cost of the column and auxiliary equipment, such as reboilers and 13 Page 13 of 40
346
condensers. For the DWC, it includes the additional dividing wall cost. In this study, the Chemical
347
Engineering Plant Cost Index of 585.7 (2011) was used for cost updating. Tray stack = (N-1) × tray spacing
(4)
349
Total height = tray stack + extra feed space + Disengagement + skirt height
(5)
350
Updated bare module cost (BMC) = UF × BC × (MPF+MF-1)
(6)
present cos t index ba secos t index
BC is the bare cost, for vessels: BC BC0 (
353
For the heat exchanger: BC BC0 (
354
Area of the heat exchanger, A
(8)
(9)
(10)
M
Q U T
A ) A0
L D ) ( ) L0 D0
cr
352
(7)
us
where UF is the update factor: UF
an
351
ip t
348
where MPF is the material and pressure factor; MF is the module factor (typical value), which is
356
affected by the base cost. D, L and A are the diameter, length and area, respectively.
d
355
Updated bare module cost for the tray stack (BMC) = UF × BC × MPF
358
The material and pressure factor: MPF = Fm + Fs + Fm
360
b. Operating cost (Op):
Ac ce p
359
te
357
Op =Csteam + CCW
361
where Csteam is the cost of the steam and CCW is the cost of cooling water.
362
c. Total annual cost (TAC) (Smith, 2005)
363
TAC capitalcost
364
where i is the fractional interest rate per year and n is the number of years.
i (1 i )n Op (1 i)n 1
(11) (12)
(13)
(14)
365 366
References
14 Page 14 of 40
367
Amminudin, K.A., Smith, R., 2001. Design and Optimization of Fully Thermally Coupled Distillation
368
Columns: Part 2: Application of Dividing Wall Columns in Retrofit. Chem. Eng. Res. Des. 79,
369
716-724. Aspen Technology, 2010. Aspen Plus: User guide, Vol. 1 & 2.
371
Biegler, L.T., Grossmann, I.E. and Westerberg, A.W., 1997. Systematic methods of chemical process
375 376
cr
374
Box, G.E.P., Behnken, D.W., 1960. Some New Three Level Designs for the Study of Quantitative Variables, Technometrics 2, 455-476.
us
373
design. Prentice Hall Inc., New Jersey.
Box, G.E.P., Wilson, K.B., 1951. On the experimental attainment of optimum conditions. J Roy Stat Soc Series B 13, 1-45.
an
372
ip t
370
Burke, R.A., 2013. Hazardous Materials Chemistry for Emergency Responders, third ed. CRC Press.
378
Caballero, J.A., Grossmann, I.E., 2014. Optimal synthesis of thermally coupled distillation sequences
380 381
using a novel MILP approach, Comput. Chem. Eng. 61, 118-135. Corma, A., Iborra, S., Velty, A., 2007. Chemical Routes for the Transformation of Biomass into Chemicals, Chem. Rev. 107, 2411-2502.
d
379
M
377
Dunlop, A.P.; Wells, P.A., 1957. Process for producing levulinic acid. US patent 2,813,900.
383
Ghorpade, V.M.; Hanna, M.A., 1999. Method and apparatus for production of levulinic acid via
385 386 387 388
Ac ce p
384
te
382
reactive extrusion. US patent 5,859,263. Grand View Research, 2014. Levulinic Acid Market Analysis and Segment Forecasts to 2020, ISBN Code: 978-1-68038-068-2.
Halvorsen, I.J., Skogestad, S., 2003. Minimum energy consumption in multicomponent distillation. 2: Three-product Petlyuk arrangements, Ind. Eng. Chem. Res. 42, 605–615.
389
Hayes, D.J., Fitzpatrick, S., Hayes, M.H.B., Ross, J.R.H., 2008. The Biofine Process – Production of
390
Levulinic Acid, Furfural, and Formic Acid from Lignocellulosic Feedstocks, Biorefineries-
391
Industrial Processes and Products. Wiley-VCH Verlag GmbH, 139-164.
392 393 394
Huber, G.W., Iborra, S., Corma, A., 2006. Synthesis of Transportation Fuels from Biomass: Chemistry, Catalysts, and Engineering, Chem. Rev. 106, 4044-4098. King, C.J., 1980. Separation Processes, 2 ed. McGraw-Hill, New York. 15 Page 15 of 40
400 401 402 403 404 405 406
Kiss, A.A. and D. J. P. C. Suszwalak, 2012b. Enhanced bioethanol dehydration by extractive and
ip t
399
wall column, Comput. Chem. Eng. 38, 74-81.
azeotropic distillation in dividing-wall columns. Sep. Purif. Technol. 86(0), 70-78.
Koh, L.P., Ghazoul, J., 2008. Biofuels, biodiversity, and people: Understanding the conflicts and
cr
398
Kiss, A.A., Suszwalak, D.J.-P.C., 2012a. Innovative dimethyl ether synthesis in a reactive dividing-
finding opportunities, Biol. Conserv. 141, 2450-2460.
us
397
production – DWC makes it happen. Applied Energy 99, 146-153.
Long, N.V.D., Lee, M.Y., 2012. Dividing wall column structure design using response surface methodology. Comput. Chem. Eng. 37, 119-124.
an
396
Kiss, A.A., Ignat, R.M., 2012. Enhanced methanol recovery and glycerol separation in biodiesel
Long, N.V.D., Lee, M.Y., 2013a. Optimal retrofit of a side stream column to a dividing wall column for energy efficiency maximization. Chem. Eng. Res. Des. 91, 2291-2298.
M
395
Long, N.V.D., Lee, M.Y., 2013b. Design and optimization of heat integrated dividing wall columns
408
for improved debutanizing and deisobutanizing fractionation of NGL. Korean J. Chem. Eng. 30,
409
286-294.
te
411
Long, N.V.D., Kwon, Y., Lee, M.Y., 2013. Design and optimization of thermally coupled distillation schemes for the trichlorosilane purification process. Appl. Therm. Eng. 59, 200-210.
Ac ce p
410
d
407
412
Merval, A., Samuelides, M., Grihon, S., 2006. Application of response surface methodology to
413
stiffened panel optimization, 47th conference on AIAA/ASME/ASCE/AHS/ASC Structures,
414
Structural Dynamics, and Materials Conference, Newport, Rhode Island.
415 416 417 418 419 420 421 422
Minitab 16 Statistical Software, 2010. [Computer software]. State College, PA: Minitab, Inc. (www.minitab.com)
Montgomery, D.C., Design and Analysis of Experiments, 2001. 5th edition, John Wiley & Sons, New York, pp.427-510. Nishida, N., Stephanopoulos, G., Westerberg, A.W., 1981. A review of process synthesis. AIChE J. 27, 321-351. Petrus, L., Noordermeer, M. A., 2006. Biomass to Biofuels, A Chemical Perspective. Green Chem. 8, 861-867. 16 Page 16 of 40
426 427 428 429 430 431
Premkumar, R., Rangaiah, G.P., 2009. Retrofitting conventional column systems to dividing-Wall Columns. Chem. Eng. Res. Des. 87, 47-60. Reunanen, J., Oinas, P., Nissinen, T., 2013. A process for recovery of formic acid, US Patent
ip t
425
Ing. Tech. 76, 1811-1814.
8,530,695 B2.
Turton, R., Bailie, R.C., Wallace, B.W., Joseph, A.S., Debangsu, B., 2012. Analysis, Synthesis, and
cr
424
Poth, N., Stichlmair, J., Brusis, D., 2004. Minimaler Energiebedarf von Trennwandkolonnen. Chem.
Design of Chemical Processes, fourth ed. Prentice Hall.
us
423
Sadhukhan, J., Siew N.K., Martinez, E.H., 2014. Biorefineries and Chemical Processes, Design, Integration and Sustainability Analysis, first ed. John Wiley & Sons, United Kingdom.
433
Seibert, F., 2010. A method of recovering levulinic acid. WIPO patent, WO/2010/030617 A1.
434
Smith, R., 2005. Chemical Process Design and Integration, John Wiley & Son, West Sussex, England.
435
Westerberg, A.W., 1985. The synthesis of distillation-based separation systems. Comput. Chem. Eng.
M
436
an
432
9, 421-429.
Werpy, T., Petersen, G., 2004. Top Value Added Chemicals from Biomass Volume I—Results of
438
Screening for Potential Candidates from Sugars and Synthesis Gas, National Renewable Energy
439
Laboratory (NREL).
441 442
te
Ac ce p
440
d
437
Wu, Y.C., et al., 2014. Energy-Saving Dividing-Wall Column Design and Control for Heterogeneous Azeotropic Distillation Systems. Ind. Eng. Chem. Res. 53(4), 1537-1552.
17 Page 17 of 40
442 443
List of Figures
445
Fig. 1. Dividing wall columns (DWCs): (a) conventional DWC; (b) top DWC; (c) bottom DWC.
446
Fig. 2. Ternary diagram for the furfural-water-formic acid system (mass basis).
447
Fig. 3. Schematic diagram of the optimal conventional sequence for producing LA from biomass.
448
Fig. 4. Flowsheet of the optimization procedure for a single distillation column.
449
Fig. 5. TAC saving plots for the C3 of the conventional sequence
450
Fig. 6. Schematic diagram of the C2 of the CS integrated with a decanter.
451
Fig. 7. Effect of temperature on furfural loss in the water-rich stream of the C2-decanter.
452
Fig. 8. Schematic diagram of the innovative TDWC-D.
453
Fig. 9. Schematic diagram of the proposed conventional sequence for producing LA from
cr
us
an
M
454
ip t
444
biomass.
Fig. 10. Schematic diagram of the proposed conventional sequence integrated a decanter.
456
Fig. 11. Contour plots of the interactions between the top section (N1), bottom section (N2) and
te
457
d
455
feed rectifying section (N3).
Fig. 12. Optimization plot by the RSM from Minitab.
459
Fig. 13. Three-dimensional response surface plots of the interaction between N1 and N2
460
Ac ce p
458
predicted by the RSM (a) and by the rigorous simulation (b).
461
Fig. 14. Schematic diagram of the proposed sequence intensified by the innovative TDWC-D.
462
Fig. 15. Comparison of the different LA recovery processes in terms of the annual operating cost
463
and TAC.
18 Page 18 of 40
M
an
us
cr
ip t
464
465
d te
467
Fig. 1. Dividing wall columns (DWCs): (a) conventional DWC; (b) top DWC; (c) bottom DWC.
Ac ce p
466
19 Page 19 of 40
467
470 471
Ac ce p
469
te
d
M
an
us
cr
ip t
468
Fig. 2. Ternary diagram for the furfural-water-formic acid system (mass basis).
20 Page 20 of 40
us
cr
ip t
471
472 473
Fig. 3. Schematic diagram of the optimal conventional sequence for producing LA from biomass.
Ac ce p
te
d
M
an
474
21 Page 21 of 40
M
an
us
cr
ip t
474
475
te
d
Fig. 4. Flowsheet of the optimization procedure for a single distillation column.
Ac ce p
476
22 Page 22 of 40
477 b)
478
us
cr
ip t
a)
Fig. 5. TAC saving plots for the C3 of the conventional sequence
Ac ce p
te
d
M
an
479
23 Page 23 of 40
479
an
us
cr
ip t
480
481 482
Fig. 6. Schematic diagram of the C2 of the CS integrated with a decanter.
Ac ce p
te
d
M
483
24 Page 24 of 40
us
cr
ip t
483
te
d
M
Fig. 7. Effect of temperature on furfural loss in the water-rich stream of the C2-decanter.
Ac ce p
485
an
484
25 Page 25 of 40
d
te
489
Fig. 8. Schematic diagram of the innovative TDWC-D.
Ac ce p
487 488
M
an
us
cr
ip t
486
26 Page 26 of 40
us
cr
ip t
489 490
491
Fig. 9. Schematic diagram of the proposed conventional sequence for producing LA from
493
biomass.
an
492
Ac ce p
te
d
M
494
27 Page 27 of 40
Fig. 10. Schematic diagram of the proposed conventional sequence integrated a decanter.
an
495 496
us
cr
ip t
494
Ac ce p
te
d
M
497
28 Page 28 of 40
ip t cr us an M d 499
te
498
Fig. 11. Contour plots of the interactions between the top section (N1), bottom section (N2) and
Ac ce p
497
feed rectifying section (N3).
29 Page 29 of 40
ip t
500
Fig. 12. Optimization plot by the RSM from Minitab.
Ac ce p
te
d
M
an
504
us
503
cr
501 502
30 Page 30 of 40
504 a)
b)
8
0
1 0
N2
cr
TAC saving (%)
ip t
16
-1 0 1
N1
us
d
M
an
predicted by the RSM (a) and by the rigorous simulation (b).
te
506
Fig. 13. Three-dimensional response surface plots of the interaction between N1 and N2
Ac ce p
505
-1
31 Page 31 of 40
Fig. 14. Schematic diagram of the proposed sequence intensified by the innovative TDWC-D.
an
508 509
us
cr
ip t
507
Ac ce p
te
d
M
510
32 Page 32 of 40
an
us
cr
ip t
510
511
Fig. 15. Comparison of the different LA recovery processes in terms of the annual operating cost
513
and TAC.
Ac ce p
te
d
M
512
33 Page 33 of 40
514
List of Tables
516
Table 1. Conditions of the feed mixture
517
Table 2. Design parameters of a conventional sequence for the production of LA from biomass
518
Table 3. Utilities cost data (Turton et al., 2012)
519
Table 4. Coded levels of the factors for TDWC variables of the CS sequence and PS sequence
520
Table 5. Comparison of the different structural alternatives for improving the performance of
cr
process for producing LA from biomass
us
521
ip t
515
522
an
523 524
M
525 526
530 531 532 533 534 535
te
529
Ac ce p
528
d
527
536 537 538 539 540
34 Page 34 of 40
541 542 543
Table 1. Conditions of the feed mixture
Component
Mass fraction (wt.%) 86%
Formic acid
3%
cr
Water
4%
us
Furfural Levulinic acid (LA)
7%
Temperature (oC)
an
25
Pressure (atm)
2 90,000
M
Mass flowrate (kg/hr)
Ac ce p
te
d
544 545
ip t
Feed conditions
35 Page 35 of 40
545 546
Table 2. Design parameters of a conventional sequence for the production of LA from biomass
547
C2
C3
22
27
28
Sieve
Sieve
Sieve
Column diameter (m)
3.6
1.7
Number of flow paths
1
1
Tray type
610
610
Max flooding (%)
87.83
Condenser duty (kW)
17,220
an
us
Tray spacing (mm)
Reboiler duty (kW)
1.8
cr
Number of trays
ip t
C1
22,930
1
610
80.36
85.93
2,699
3,694
4,250
3,770
M
Purity (mass fraction, wt.%)
Levulinic acid (LA)
549
85.0
Ac ce p
548
99.9
te
Formic acid
d
Furfural
98.0
36 Page 36 of 40
Table 3. Utilities cost data (Turton et al., 2012) Utility
Price ($/GJ)
Cooling water
0.35
Medium pressure steam
14.19
High pressure steam
17.70
cr
551
ip t
549 550
Ac ce p
te
d
M
an
us
552
37 Page 37 of 40
552
Table 4. Coded levels of the factors for TDWC variables of the CS sequence and PS sequence Case
Factor
Levels
Proposed sequence
Top section (N1)
22
Bottom section (N2)
5
Feed rectifying section (N3)
14
Top section (N1) Bottom section (N2)
30
10
15
18
22
18
22
26
2
6
10
13
17
20
an
Feed rectifying section (N3) 554
Ac ce p
te
d
M
555
1
26
us
Conventional sequence
0
cr
-1
ip t
553
38 Page 38 of 40
ip t cr
Annual operating cost (US $) Annual operating cost saving (%)
16,332,077
CS-D
PS
PS-D
PS-TDWC-D
15,151,046
13,646,300
13,729,322
12,812,024
0.6%
7.2%
16.4%
15.9%
21.6%
17,308,411
17,214,679
16,103,366
14,541,721
14,640,722
13,736,949
-
0.5%
7.0%
16.0%
15.4%
20.6%
-
Ac
ce pt
Total annual cost saving (%)
CS-TDWC-D
16,234,472
ed
Total annual cost (US $)
CS
M an
Structural Alternative
us
Table 5. Comparison of the different structural alternatives for improving the performance of process for producing LA from biomass
40 Page 39 of 40
Ac
ce
pt
ed
M
an
us
cr
i
Graphical Abstract (for review)
Page 40 of 40
S0263-8762(15)00365-2 http://dx.doi.org/doi:10.1016/j.cherd.2015.09.013 CHERD 2024
To appear in: Received date: Revised date: Accepted date:
18-6-2015 31-8-2015 17-9-2015
Please cite this article as: Nhien, L.C., Long, N.V.D., Lee, M.,Design and Optimization of the Levulinic Acid Recovery Process from Lignocellulosic Biomass, Chemical Engineering Research and Design (2015), http://dx.doi.org/10.1016/j.cherd.2015.09.013 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1
Highlights
An innovative TDWC-D was proposed to separate a heterogeneous azeotrope mixture.
3
The TAC could be reduced by 28.1% by the innovative TDWC-D.
4
Many benefits could be achieved by changing process sequence structure.
5
A compact and energy efficient LA production process is proposed.
6
Large potential for commercial LA production from lignocellulosic biomass
cr
ip t
2
7 8
us
Design and Optimization of the Levulinic Acid Recovery Process
9
an
from Lignocellulosic Biomass
10
Le Cao Nhien1, Nguyen Van Duc Long1 and Moonyong Lee*
M
11 12
School of Chemical Engineering, Yeungnam University, Gyeongsan 712-749, South Korea
13
Running title: Design and Optimization of the Levulinic Acid Recovery Process from
15
Lignocellulosic Biomass
Ac ce p
16
te
d
14
17
Submitted to Chemical Engineering Research and Design Journal
18
1
19
*
20
Prof. Moonyong Lee
21
Email: [email protected]
22
Telephone: +82-53-810-2512
These two authors contributed equally to this work.
Correspondence concerning this article should be addressed to:
23
1 Page 1 of 40
23
Abstract Levulinic acid (LA), which is one of the top twelve value-added chemicals from biomass
25
feedstock, has been recognized in a large number of applications. Nevertheless, its capability on an
26
industrial scale has been limited by the high-cost of the raw materials and the lack of detailed process
27
design. This paper reports the simulation, detailed design and optimization of an industrial process
28
recovering LA from biomass, a renewable and inexpensive feedstock. The aqueous feed mixture from
29
the acidic hydrolysis of biomass was fed to the separation process to recover LA as the main product,
30
along with furfural and formic acid as valuable byproducts. Furfural was then used as an extracting
31
solvent, resulting a significant decrease in total cost. Detailed analysis was conducted to assess the
32
feasibility of the separation method. The purification process was intensified by applying an
33
innovative top dividing wall column with a decanter configuration (TDWC-D). The response surface
34
methodology (RSM), which allows the interactions between variables to be identified and quantified,
35
was used to optimize the TDWC-D structure. The predictions by the RSM showed satisfactory
36
agreement with the rigorous simulation results. The results showed that the innovative TDWC-D
37
configuration could save up 16.4% and 20.6% of the overall energy requirements and total annual cost,
38
respectively, compared to the conventional sequence.
40 41
cr
us
an
M
d
te
Ac ce p
39
ip t
24
Keywords: Lignocellulosic Biomass; Levulinic Acid; Furfural; Energy Efficiency Improvement; Dividing Wall Column; Response Surface Methodology
2 Page 2 of 40
42 43
1. Introduction The dependence of the global economy on the utilization of mineral resources has increased in
45
recent years, accounting for 80.9% of the total energy production in 2006 (Koh and Ghazoul, 2008).
46
Nevertheless, these resources have many disadvantages, such as limited reserves, environmentally
47
damaging and responsibility for global warming. Therefore, the interest of biomass, which is the only
48
renewable resource of fixed carbon, has increased considerably. Biomass can not only produce
49
conventional hydrocarbon liquids for transportation, but also petrochemical products (Huber et al.,
50
2006, Petrus et al., 2006, Corma et al., 2007). Biofuels from biomass are viewed by many policy
51
makers as a key to reduce the reliance on foreign oil, lowering the emissions of greenhouse gases and
52
meeting the rural development goals (Koh and Ghazoul, 2008).
an
us
cr
ip t
44
Levulinic acid, which is recognized in a large number of applications, is listed as one of the top-
54
twelve value-added chemicals from biomass in terms of its potential markets, its derivatives and the
55
complexity of the synthetic pathways (Werpy and Petersen, 2004). This compound has frequently
56
been suggested to be a starting material for the production of many industrial and pharmaceutical
57
compounds, such as methyltetrahydrofuran, delta aminolevulinic acid, ethyl levulinate and diphenolic
58
acid. In particular, with MTHF applications only, LA manufacturers could face the potential demand
59
of more than 20,000 kilotons by 2020 (Grand View Research, 2014). Nevertheless, the capability of
60
this process on an industrial scale has been limited by the expensive raw materials and energy-
61
intensive processes (Seibert, 2010).
Ac ce p
te
d
M
53
62
Several processes have been used to produce LA from an inexpensive feedstock, lignocellulosic
63
biomass. Dunlop and Wells (1957) proposed the earliest continuous process for producing LA from
64
corn cob waste. The aqueous feed mixture obtaining from acidic hydrolysis is introduced to an
65
extractor to extract LA and then fed to a distillation column to produce pure LA. Ghorpade and Hanna
66
(1999) presented the LA process based on the reactive extrusion from corn starch. These two
67
processes only considered LA as a product. Recently, Sadhukhan et al. (2014) presented a good
68
example of a LA production process from corn cob, which considered both the reaction and separation
69
stages. Seibert (2010) reported a process that uses furfural to extract LA and formic acid from the 3 Page 3 of 40
acidic hydrolysis of biomass. The advantage of this process is the use of furfural, one of the products,
71
as a solvent. Interestingly, these reports showed that a combination of an extractor and a series of
72
distillation columns is used to purify LA, showing great potential to improve the energy efficiency,
73
economic performance and environmental impact. Furthermore, none of the previous processes
74
considered the detailed design methodology and advanced techniques in the case of commercial scale
75
production. A comprehensive design by intensified configurations with economic evaluation is needed
76
to expand the capacity to an industrial scale as well as to reduce the LA price and environmental
77
impact.
us
cr
ip t
70
Many studies have examined the relative advantages of a dividing wall column (DWC) (as shown
79
in Fig. 1) to improve the energy efficiency and reduce the consumption of utilities (Amminudin and
80
Smith, 2001; Long and Lee, 2013a; Long and Lee, 2013b). These reports have shown that DWC
81
systems can achieve energy savings of up to 30% over conventional direct and indirect distillation
82
sequences (Kiss and Ignat, 2012; Long et al., 2013). This is because DWCs can allow reversible splits
83
with no part of the separation being performed twice, which is the main contributor to their superior
84
energy efficiency over other column configurations (Poth et al., 2004). Additional benefits arise from
85
the fact that only one reboiler and condenser are involved instead of two reboilers and two condensers
86
in a conventional distillation sequence (Premkumar and Rangaiah, 2009).
Ac ce p
te
d
M
an
78
87
To design and optimize DWC, Halvorsen and Skogestad (2003) proposed the use of a minimum
88
vapor flow, or the so called Vmin diagram method, which is a graphical visualization of the minimum
89
energy. On the other hand, it cannot be applied to the separation of azeotropic mixtures. Several
90
rigorous methods have proposed, such as a sequential quadratic programming-based method
91
implemented in Aspen Plus (Kiss and Suszwalak, 2012a), and a MILP approach based on the use of
92
two superstructure levels (Caballero and Grossmann, 2014).
93
Recently, Long and Lee (2012) reported a practical and efficient method based on the response
94
surface methodology (RSM). The RSM is a general technique that is used for an empirical study of
95
the relationships between the measured responses and independent input variables (Box and Wilson,
96
1951). By allowing the interactions between variables to be identified and quantified, RSM can
97
predict reliable results with a little computational effort. Therefore, it can be used to optimize the 4 Page 4 of 40
design and operation conditions of a process (Long and Lee, 2012). On the other hand, when the
99
behavior of the function is highly nonlinear in terms of the design variables, the RSM encounters
100
some difficulties in approximating the sufficient quality on the entire design space (Merval et al.,
101
2006). The optimum may not be defined if the range of factors is too narrow or too wide. Furthermore,
102
critical factors may not be defined or specified correctly. In the case of many factors, the problem
103
becomes complex and the RSM cannot be predicted well.
ip t
98
This paper proposes an economical and environmental process for producing LA from an
105
inexpensive raw material, lignocellulosic biomass, along with the detailed design variables and
106
operating conditions. The simulation work was conducted using the commercial simulator, Aspen Plus.
107
A range of sequences and intensified configurations applying TDWC-D are presented with respect to
108
their energy consumption and economic performance. Furthermore, the RSM was used to optimize
109
the structure of the proposed TDWC-D.
M
2. LA separation process
112
2.1. Separation feasibility
d
110 111
an
us
cr
104
The feed mixture considered was obtained from the acidic hydrolysis process of biomass and was
114
an aqueous solution containing LA, furfural, formic acid, and water. The typical concentrations of the
115
mixture ranged from 3 to 8 wt. % LA, from 1 to 5 wt. % furfural, from 1 to 5 wt. % formic acid, and
116
the balance being water (Reunanen et al. 2013). With this component mixture, although LA did not
117
form an azeotrope with the other components, two azeotrope mixtures still formed. Fig. 2 shows a
118
ternary diagram for the furfural-water-formic acid system. As shown in this figure, the water-furfural
119
mixture has a minimum-boiling heterogeneous azeotrope with a composition of 64.5 wt.% furfural at
120
97.8 oC. Another maximum-boiling homogeneous azeotrope forms between formic acid and water
121
with a higher azeotropic temperature of 106.8 oC. In that ternary map, there are two distillation
122
regions divided by the distillation boundary.
Ac ce p
te
113
123
Fig. 3 presents a simplified feasible flow sheet for producing LA from lignocellulosic biomass
124
(Seibert, 2010). The conventional process sequence consists of one extractor and three distillation
125
columns. The feed mixture and the furfural solvent are introduced into the extractor (EX) to generate 5 Page 5 of 40
a water-rich phase as the top stream and a solvent-rich phase containing mainly furfural, LA and
127
formic acid as the bottom stream. The furfural-rich stream is then fed to the first distillation column
128
(C1) to collect LA as the bottom stream product, whereas the distillate contains furfural, formic acid
129
and water, which are introduced into the second distillation column (C2). The function of C2 is to
130
separate the inlet into furfural, formic acid as the bottom stream and a light boiling furfural-water
131
azeotropic mixture as the top stream. The azeotropic mixture is then delivered to storage before being
132
recycled back to the extractor. The bottom stream of C2 is then subjected to the third distillation
133
column (C3) to recover the formic acid as the distillate product, and the furfural is collected as the
134
bottom stream. The furfural solvent is recycled back to the extractor and reused as the extractant.
us
cr
ip t
126
All processes were performed using the simulator, Aspen Plus V8.6, with the NRTL–HOC
136
(nonrandom two–liquid–Hayden – O’Connell) fluid package. The HOC equation of state as the vapor
137
phase model and NRTL for the liquid phase were used to predict not only the vapor–liquid
138
equilibrium but also the vapor–liquid–liquid equilibrium of these simulations. Dimerization affects
139
the vapor–liquid equilibrium, vapor phase properties, density, and liquid phase properties. The HOC
140
equation reliably predicts the solvation of polar compounds and the dimerization in the vapor phase
141
that occurs with mixtures containing carboxylic acids (Aspen Technology, 2010). The binary pairs of
142
the interaction parameters for LA were estimated using the UNIFAC model.
143
2.2. Conventional distillation sequence (CS)
Ac ce p
te
d
M
an
135
144
The LA production rate considered in this study was 50 kilotons per year with a commercial purity
145
of 98 wt.% (Hayes et al., 2008). The furfural reused as a solvent was kept at 99.9 wt.% purity while
146
the purity of the formic acid product was 85 wt.%, which is considered the industry standard and
147
accounting for the major portion of the global demand (Burke, 2013). Table 1 lists the feed
148
composition, temperature and pressure used in this process.
149
In this study, the CS was designed and optimized first for a fair comparison with the other
150
proposed optimized sequences (Fig. 3). In this process, the component recoveries depend mainly on
151
the efficiency of the extractor or the extraction solvent efficiency. This means that solvent selection is
152
the key in this process. Although formic acid recovery is quite low when selecting furfural as a
153
solvent, there is some trade-off between the total process cost and the formic acid recovery. Note that 6 Page 6 of 40
furfural is one of the products and more effective in the recovery of LA and formic acid than other
155
solvents, such as methyl isobutyl ketone and isopropyl acetate (Seibert, 2010). In the extractor, the
156
recovery of LA, formic acid and furfural also increase with increasing extracting solvent/feed ratio but
157
more water appears in the furfural-rich phase. In this study, a solvent/feed ratio of 1.2 : 1 was selected
158
to obtain 97 wt.% LA recovery. Table 2 lists the conditions and product specifications for each column
159
in the conventional column sequences. Note that high pressure (HP) steam is used in the first column
160
while medium pressure (MP) steam is used in the second and third columns. The rigorous simulation
161
results showed that the energy requirements of C1, C2 and C3 were 22,930, 4,250 and 3,770 kW,
162
respectively.
us
cr
ip t
154
164
an
163
3. Proposed sequences
This section proposes the different configurations to separate LA and co-products. The results were
166
then compared in terms of the energy requirement and total annual cost while obtaining the same
167
requirements of productivity, recovery and purity. The column diameters were determined using the
168
column flooding conditions that fix the upper limit of the vapor velocity. The operating velocity is
169
normally between 70 and 90% of the flooding velocity (Premkumar and Rangaiah, 2009). In this
170
study, all columns were designed with a load near 85% of the load at the flooding point to prevent
171
flooding in the columns. The capital costs were calculated based on Guthrie’s modular method
172
(Biegler et al., 1997). Table 3 lists the utility cost data used in this study. The 10 year-plant lifetime
173
was used to calculate the total annual cost (TAC).
Ac ce p
te
d
M
165
174
All conventional columns were optimized with the number of trays and the feed location as the
175
main variables for maximizing total annual cost saving (Fig. 4). Fig. 5 shows one optimization case of
176
the C3 in the conventional sequence. In TDWC-D optimization, after a preliminary design, the main
177
variables, such as the internal vapor flow to the prefractionator (FV), and the number of trays in the
178
top (N1), bottom (N2) and feed rectifying (N3) sections were then optimized. The aim of the
179
optimization was to maximize the TAC savings. The optimization was constrained by the product
180
purities and recoveries.
7 Page 7 of 40
181
Max(%TAC saving) f ( N1, N 2, N 3, FV )
182
A Box−Behnken design was used under the response surface methodology to determine how the
183
variables interact and to optimize the system in terms of the TAC saving. Box and Behnken (1960)
184
proposed some three-level designs for fitting the response surfaces. These designs are formed by
185
combining 2k factorials with incomplete block designs. The resulting designs are generally very
186
efficient in terms of the number of required runs, and they are either rotatable or nearly rotatable.
187
Rotability is a very important property in the selection of the RSM. Because the purpose of the RSM
188
is optimization and the location of the optimum is unknown prior to running the experiment or
189
simulation, it makes sense to use a design that provides equal estimation precision in all directions
190
(Montgomery, 2001). Note that the Box-Behnken design is a spherical design with all points lying on
191
the sphere of the radius. In addition, the Box-Behnken design does not contain any points at the
192
vertices of the cubic region created by the upper and lower limits for each variable. This could be
193
advantageous when the points on the corners of the cube represent the factor-level combinations that
194
are prohibitively expensive or impossible to test because of the physical process constraints. Moreover,
195
the Box-Behnken designs often have fewer design points; they require less effort to run than central
196
composite designs with the same number of factors (Minitab 16, 2010). After determining the
197
preliminary ranges of the variables through single-factor testing, the simulation run data was fitted to
198
a second-order polynomial model and the regression coefficients were obtained. The generalized
199
second-order polynomial model used in the response surface analysis is as follows:
ip t
cr
us
an
M
d
te
Ac ce p
200
(1)
k
k
i 1
i 1
Y 0 i N i ii N i2 ij N i N j i j
(2)
201
where Y is the predicted response, Ni are the uncoded or coded values of the variables, β0 is a constant,
202
βi, βii, and βij are the coefficients of the linear, quadratic, and interactive terms, respectively, and ε is
203
the error term. Minitab® software was used for response surface fitting to maximize the total annual
204
cost savings.
205
3.1. Improving conventional sequence by a decanter (CS-D)
8 Page 8 of 40
In the C2 of the conventional sequence, there was an amount of furfural loss (2,175 kg/hr) in the
207
C2-top due to a heterogeneous azeotrope between water and furfural (see Fig. 6). As shown in Fig. 2,
208
a significant liquid phase split envelope is clearly shown in the ternary diagram of the furfural-water-
209
formic acid system so that a decanter can be applied to reduce the furfural loss and make the
210
distillation more efficient. In particular, the decanter or gravity settling tank, are the simplest form of
211
equipment used to separate liquid-liquid mixtures. Several studies have reported the benefits of a
212
decanter in solving a heterogeneous azeotrope (Kiss and Suszwalak, 2012b; Wu et al., 2014). In this
213
case, furfural and formic acid are collected as the bottom stream of the C2, while the binary
214
heterogeneous azeotrope water-furfural is sent to the overhead of C2, which is then delivered to a
215
condenser-decanter. Fig. 6 presents a flow sheet of the C2 integrated with a decanter. The azeotrope
216
furfural-water on top of C2 is cooled and then split naturally into two liquid phases by the decanter,
217
following the liquid-liquid equilibrium tie lines. The water-rich phase is designed to be totally
218
delivered to storage while the organic phase, which contains mainly furfural, is refluxed back to C2.
219
Fig. 7 shows the effects of temperature on the liquid-liquid separation in the decanter. In particular,
220
the amount of furfural loss decreases with decreasing temperature of the inlet-decanter stream.
221
Therefore, in the decanter design, the temperature of the C2-overhead is cooled to 60oC to make the
222
separation in the decanter more efficient. As a result, integration of the decanter can not only save the
223
energy requirements in the reboiler duty (2.39% of the C2 and C3), but also increase the furfural
224
recovery. The amount of furfural loss in the C2-top stream was reduced from 2,175 kg/hr to 336 kg/hr,
225
i.e. a decrease of 85% furfural loss compared to that in the column without a decanter.
226
3.2. Intensifying conventional sequence by innovative TDWC-decanter (TDWC-D)
Ac ce p
te
d
M
an
us
cr
ip t
206
227
In addition, to improve the energy efficiency in the columns, minimizing the number of columns is
228
also crucial because it directly affects the overall economics and safety in a LA production process.
229
DWC would be the best option to intensify the process for achieving both tasks simultaneously. In
230
particular, an innovative TWDC-D was considered here to integrate the C2 and C3 to improve the
231
energy efficiency and reduce the investment cost.
232
After a preliminary design for TDWC-D, it was then optimized using the RSM. Table 4 lists the
233
factors and levels used in the RSM based optimization. Fifteen simulations were run to optimize 3 9 Page 9 of 40
parameters of the TDWC-D structure. For each run, the internal vapor flow to the prefractionator was
235
varied to maximize the TAC saving while obtaining the required product purity and recovery. Fig. 8
236
presents a schematic diagram illustrating the optimized TDWC-D that replaces the columns, C2 and
237
C3, of the CS sequence. The rigorous simulation results show that the optimized TDWC-D can save
238
up 30.6 and 28.1% in terms of the operating cost and TAC, respectively, compared to the two columns,
239
C2 and C3, of the base sequence. Compared to the whole CS sequence, the innovative configuration
240
combining TDWC and a decanter can allow operating cost and TAC savings of 7.0% and 7.2%,
241
respectively.
242
3.3. Improving conventional sequence by changing sequence – proposed sequence (PS)
us
cr
ip t
234
Multicomponent mixtures can be separated by different distillation column sequences. Many
244
studies have examined the optimal sequence of simple columns that do not involve heat integration
245
(Nishida et al., 1981; Smith, 2005; Westerberg, 1985). The number of possible distillation sequences
246
increases with increasing number of products (King, 1980). Interestingly, the “rule of thumb” is the
247
most common method used for finding the optimal sequence of simple distillation columns (King,
248
1980). The sequences that remove the lightest components alone one by one in the column overheads
249
should be favored (Smith, 2005), i.e., favor the direct sequence. In addition, in the base case, the first
250
distillation column C1 requires the largest amount of energy and highest level of steam (high pressure
251
steam) compared to the other columns. This column is used to vaporize a large number of furfural and
252
light components as well as purify LA, which has the highest boiling point (257 oC) compared to the
253
other components, furfural (162 oC), formic acid (101 oC), water (100 oC). Based on the well-known
254
heuristics for separation sequence synthesis, a proposed sequence for producing LA was used to
255
enhance the energy efficiency (see Fig. 9).
Ac ce p
te
d
M
an
243
256
An aqueous mixture from the acidic hydrolysis reaction of biomass and the furfural solvent was
257
fed to the extractor (P-EX) to generate a water-rich phase as the top stream and solvent-rich phase
258
containing mainly furfural, LA and formic acid as the bottom stream. In contrast to the CS sequence,
259
the bottom stream of P-EX was fed to the proposed C1 (P-C1), which delivered a light boiling
260
azeotrope of furfural and water as the top stream, whereas the bottom stream contained formic acid,
261
furfural and LA. The bottom of P-C1 was then introduced to the proposed C2 (P-C2) to collect formic 10 Page 10 of 40
acid as the distillate and the bottom stream comprising furfural and LA, which was subjected to the
263
proposed C3 (P-C3). The role of P-C3 is to separate levulinic acid as the bottom product and furfural
264
as the top product. Here, LA is recovered at the third column so the proposed sequence is called the
265
direct sequence. Note that medium pressure steam was used on P-C1 and P-C2, whereas P-C3 used
266
high pressure steam. The simulation results showed that the energy requirement in the reboiler using
267
HP columns is reduced from 22,930 kW in C1 of the CS sequence to 13,850 KW in C3 of the PS
268
sequence. As a result, the PS sequence that changes the order of the distillation columns can reduce
269
the annual operating cost and TAC by up to 16.4 and 16.0%, respectively, compared to the base case.
270
Note that these heuristics are not novel but they can be applied easily and efficiently, which leads to
271
significant TAC savings in the LA process. This process is particularly interesting to many companies
272
that prefer to use conventional columns.
273
3.4. Improving the proposed sequence by decanter (PS-D)
M
an
us
cr
ip t
262
In the PS sequence, the main role of the first column (P-C1) is to remove water as much as
275
possible. On the other hand, the minimum-boiling heterogeneous azeotrope mixture between furfural
276
and water forms, causing a furfural loss of 3,196 kg/hr in the top of this column. Therefore, a decanter,
277
which can split the liquid-liquid mixture naturally, can be integrated in this column as same as the
278
second column in the base case to reduce the loss of furfural. Fig. 10 shows the simplified flow sheet
279
of the P-C1 integrated with a decanter. To improve the separation efficiency in decanter, the overhead
280
of P-C1 was cooled to 60 oC before being introduced to the decanter. Not only can the furfural loss be
281
reduced by up to 87.9%, but also the amount of water removed increases from 2,960 kg/hr to 3,157
282
kg/hr, which is comparable to the column without a decanter. The simulation results showed that not
283
only can the energy requirement in the reboiler duty in P-C1-D column be reduced slightly (1.7%),
284
but there is also less furfural loss compared to the column without a decanter.
285
3.5. Intensifying proposed sequence with innovative TDWC-D (PS-TDWC-D)
Ac ce p
te
d
274
286
In the CS sequence, although the C1 requires most energy, however, TDWC-D can only be used to
287
replace the C2 and C3. On the other hand, in the PS sequence, both the P-C1 and P-C3 are energy-
288
intensive columns so that intensification either the P-C1 or P-C3 can bring more benefit. In particular,
289
to intensify the proposed conventional sequence, the TDWC was used to integrate the first two 11 Page 11 of 40
columns. To optimize the TDWC-D, some preliminary simulation runs were carried out to determine
291
the main optimizing variables and their levels. Corresponding to the changes in the factor value, the
292
response values (TAC saving) were recorded. The presence of curvature suggests that the simulation
293
region is close to the optimum. Table 4 lists the factors and levels used in the RSM-based optimization.
294
The TDWC-D parameters were optimized over 15 simulation runs. For each run, the internal vapor
295
flow to the prefractionator was varied to maximize the TAC savings while obtaining the required
296
product purity and recovery. Fig. 11 presents a contour plot of the interactions between the design
297
variables N1, N2 and N3. In each case, the one remaining parameter was set automatically as its
298
center point value by the software. The resulting second-order polynomial model was as follows:
us
cr
ip t
290
an
Y 10.3743 5.44663N1 3.99257N2 4.11248N3 3.01056N12
299
5.15933N22 1.55725N32 1.20847N1N2 1.96654N1N3 0.582208300 N2 N3
(3)
As shown in Fig. 12, the TAC saving was predicted to be highest (15.3%) with the coded levels of
302
the top section (N1), bottom section (N2) and feed rectifying section (N3) of 0.6970, 0.5152 and -
303
0.9798, respectively. The resulting determination coefficient was R2 = 95.7, which indicates that the
304
model is suitable for prediction within the range of simulated variables. The natural values of the
305
variables can be derived from the coded levels. Recycling of the vapor was then optimized to
306
maximize the TAC savings while achieving the product purity requirements and recovery. Fig. 13
307
compares the three-dimensional response surface plots of the interaction between N1 and N2
308
evaluated by the RSM and by rigorous simulation. The figure shows that the three-dimensional
309
response surface plots by the RSM are similar to the actual ones. Fig. 14 shows a schematic diagram
310
illustrating the optimal TDWC-D from the conventional sequence. The TDWC-D combining a
311
decanter, which can increase the furfural recovery and improve the heterogeneous azeotrope
312
separation, with a TDWC, which can reduce energy requirement and TAC, can bring many benefits in
313
terms of energy saving and cost reduction. The simulation showed that the TDWC-D can save up to
314
13.2% in terms of the TAC compared to the C2 and C3 of the conventional sequence. This rigorous
315
simulation agreed well with the predicted value (15.3%). In addition, using a TDWC-D could save
316
15.4% of the energy requirements in terms of the reboiler duty compared to P-C1 and P-C2 of the PS
317
sequence.
Ac ce p
te
d
M
301
12 Page 12 of 40
318
Table 5 summarizes the key results, while Fig. 15 conveniently illustrates the annual operating cost
319
and TAC for all configurations considered. The results suggest that the TDWC-D can be applied
320
efficiently to improve the distillation processes while requiring less TAC and energy.
322
ip t
321
4. Conclusions
A levulinic acid recovery process from abundant and inexpensive lignocellulosic biomass was
324
designed and optimized to be more energy-efficient. This process was then intensified efficiently
325
using an innovative TDWC-D. After a preliminary design, the TDWC-D structure was then optimized
326
and implemented simply and efficiently by the RSM using Aspen Plus and Minitab. The simulated
327
results were in good agreement with the predicted values. The use of innovative TDWC-D can have
328
many benefits, such as increasing the furfural recovery, increasing the separation efficiency and
329
decreasing the TAC significantly. The incorporation of a decanter into the TDWC is a very attractive
330
option for improving the separation efficiency and energy savings, especially for the separation of
331
heterogeneous azeotrope mixtures. The proposed sequence including TDWC-D showed 21.6% and
332
20.6% decreases in the energy requirement and TAC, respectively, compared to a conventional
333
column sequence. Based on these results, the use of TDWC-D for LA production, which has the
334
limitations in commercial scale of existing processes currently, is technically feasible and particularly
335
interesting in constructing a new Levulinic plant. The process compactness is an additional benefit
336
from TDWC-D applications.
338 339
us
an
M
d
te
Ac ce p
337
cr
323
Acknowledgments
The authors wish to acknowledge the financial support from the R&D Convergence Program of
340
NST (National Research Council of Science & Technology) of Republic of Korea and KITECH
341
(Korea Institute of Industrial Technology) (ES150001).
342 343
Appendix. Column cost correlations
344
a. Capital cost: Guthrie’s modular method was applied (Biegler et al., 1997). The investment cost for
345
conventional distillation is the total cost of the column and auxiliary equipment, such as reboilers and 13 Page 13 of 40
346
condensers. For the DWC, it includes the additional dividing wall cost. In this study, the Chemical
347
Engineering Plant Cost Index of 585.7 (2011) was used for cost updating. Tray stack = (N-1) × tray spacing
(4)
349
Total height = tray stack + extra feed space + Disengagement + skirt height
(5)
350
Updated bare module cost (BMC) = UF × BC × (MPF+MF-1)
(6)
present cos t index ba secos t index
BC is the bare cost, for vessels: BC BC0 (
353
For the heat exchanger: BC BC0 (
354
Area of the heat exchanger, A
(8)
(9)
(10)
M
Q U T
A ) A0
L D ) ( ) L0 D0
cr
352
(7)
us
where UF is the update factor: UF
an
351
ip t
348
where MPF is the material and pressure factor; MF is the module factor (typical value), which is
356
affected by the base cost. D, L and A are the diameter, length and area, respectively.
d
355
Updated bare module cost for the tray stack (BMC) = UF × BC × MPF
358
The material and pressure factor: MPF = Fm + Fs + Fm
360
b. Operating cost (Op):
Ac ce p
359
te
357
Op =Csteam + CCW
361
where Csteam is the cost of the steam and CCW is the cost of cooling water.
362
c. Total annual cost (TAC) (Smith, 2005)
363
TAC capitalcost
364
where i is the fractional interest rate per year and n is the number of years.
i (1 i )n Op (1 i)n 1
(11) (12)
(13)
(14)
365 366
References
14 Page 14 of 40
367
Amminudin, K.A., Smith, R., 2001. Design and Optimization of Fully Thermally Coupled Distillation
368
Columns: Part 2: Application of Dividing Wall Columns in Retrofit. Chem. Eng. Res. Des. 79,
369
716-724. Aspen Technology, 2010. Aspen Plus: User guide, Vol. 1 & 2.
371
Biegler, L.T., Grossmann, I.E. and Westerberg, A.W., 1997. Systematic methods of chemical process
375 376
cr
374
Box, G.E.P., Behnken, D.W., 1960. Some New Three Level Designs for the Study of Quantitative Variables, Technometrics 2, 455-476.
us
373
design. Prentice Hall Inc., New Jersey.
Box, G.E.P., Wilson, K.B., 1951. On the experimental attainment of optimum conditions. J Roy Stat Soc Series B 13, 1-45.
an
372
ip t
370
Burke, R.A., 2013. Hazardous Materials Chemistry for Emergency Responders, third ed. CRC Press.
378
Caballero, J.A., Grossmann, I.E., 2014. Optimal synthesis of thermally coupled distillation sequences
380 381
using a novel MILP approach, Comput. Chem. Eng. 61, 118-135. Corma, A., Iborra, S., Velty, A., 2007. Chemical Routes for the Transformation of Biomass into Chemicals, Chem. Rev. 107, 2411-2502.
d
379
M
377
Dunlop, A.P.; Wells, P.A., 1957. Process for producing levulinic acid. US patent 2,813,900.
383
Ghorpade, V.M.; Hanna, M.A., 1999. Method and apparatus for production of levulinic acid via
385 386 387 388
Ac ce p
384
te
382
reactive extrusion. US patent 5,859,263. Grand View Research, 2014. Levulinic Acid Market Analysis and Segment Forecasts to 2020, ISBN Code: 978-1-68038-068-2.
Halvorsen, I.J., Skogestad, S., 2003. Minimum energy consumption in multicomponent distillation. 2: Three-product Petlyuk arrangements, Ind. Eng. Chem. Res. 42, 605–615.
389
Hayes, D.J., Fitzpatrick, S., Hayes, M.H.B., Ross, J.R.H., 2008. The Biofine Process – Production of
390
Levulinic Acid, Furfural, and Formic Acid from Lignocellulosic Feedstocks, Biorefineries-
391
Industrial Processes and Products. Wiley-VCH Verlag GmbH, 139-164.
392 393 394
Huber, G.W., Iborra, S., Corma, A., 2006. Synthesis of Transportation Fuels from Biomass: Chemistry, Catalysts, and Engineering, Chem. Rev. 106, 4044-4098. King, C.J., 1980. Separation Processes, 2 ed. McGraw-Hill, New York. 15 Page 15 of 40
400 401 402 403 404 405 406
Kiss, A.A. and D. J. P. C. Suszwalak, 2012b. Enhanced bioethanol dehydration by extractive and
ip t
399
wall column, Comput. Chem. Eng. 38, 74-81.
azeotropic distillation in dividing-wall columns. Sep. Purif. Technol. 86(0), 70-78.
Koh, L.P., Ghazoul, J., 2008. Biofuels, biodiversity, and people: Understanding the conflicts and
cr
398
Kiss, A.A., Suszwalak, D.J.-P.C., 2012a. Innovative dimethyl ether synthesis in a reactive dividing-
finding opportunities, Biol. Conserv. 141, 2450-2460.
us
397
production – DWC makes it happen. Applied Energy 99, 146-153.
Long, N.V.D., Lee, M.Y., 2012. Dividing wall column structure design using response surface methodology. Comput. Chem. Eng. 37, 119-124.
an
396
Kiss, A.A., Ignat, R.M., 2012. Enhanced methanol recovery and glycerol separation in biodiesel
Long, N.V.D., Lee, M.Y., 2013a. Optimal retrofit of a side stream column to a dividing wall column for energy efficiency maximization. Chem. Eng. Res. Des. 91, 2291-2298.
M
395
Long, N.V.D., Lee, M.Y., 2013b. Design and optimization of heat integrated dividing wall columns
408
for improved debutanizing and deisobutanizing fractionation of NGL. Korean J. Chem. Eng. 30,
409
286-294.
te
411
Long, N.V.D., Kwon, Y., Lee, M.Y., 2013. Design and optimization of thermally coupled distillation schemes for the trichlorosilane purification process. Appl. Therm. Eng. 59, 200-210.
Ac ce p
410
d
407
412
Merval, A., Samuelides, M., Grihon, S., 2006. Application of response surface methodology to
413
stiffened panel optimization, 47th conference on AIAA/ASME/ASCE/AHS/ASC Structures,
414
Structural Dynamics, and Materials Conference, Newport, Rhode Island.
415 416 417 418 419 420 421 422
Minitab 16 Statistical Software, 2010. [Computer software]. State College, PA: Minitab, Inc. (www.minitab.com)
Montgomery, D.C., Design and Analysis of Experiments, 2001. 5th edition, John Wiley & Sons, New York, pp.427-510. Nishida, N., Stephanopoulos, G., Westerberg, A.W., 1981. A review of process synthesis. AIChE J. 27, 321-351. Petrus, L., Noordermeer, M. A., 2006. Biomass to Biofuels, A Chemical Perspective. Green Chem. 8, 861-867. 16 Page 16 of 40
426 427 428 429 430 431
Premkumar, R., Rangaiah, G.P., 2009. Retrofitting conventional column systems to dividing-Wall Columns. Chem. Eng. Res. Des. 87, 47-60. Reunanen, J., Oinas, P., Nissinen, T., 2013. A process for recovery of formic acid, US Patent
ip t
425
Ing. Tech. 76, 1811-1814.
8,530,695 B2.
Turton, R., Bailie, R.C., Wallace, B.W., Joseph, A.S., Debangsu, B., 2012. Analysis, Synthesis, and
cr
424
Poth, N., Stichlmair, J., Brusis, D., 2004. Minimaler Energiebedarf von Trennwandkolonnen. Chem.
Design of Chemical Processes, fourth ed. Prentice Hall.
us
423
Sadhukhan, J., Siew N.K., Martinez, E.H., 2014. Biorefineries and Chemical Processes, Design, Integration and Sustainability Analysis, first ed. John Wiley & Sons, United Kingdom.
433
Seibert, F., 2010. A method of recovering levulinic acid. WIPO patent, WO/2010/030617 A1.
434
Smith, R., 2005. Chemical Process Design and Integration, John Wiley & Son, West Sussex, England.
435
Westerberg, A.W., 1985. The synthesis of distillation-based separation systems. Comput. Chem. Eng.
M
436
an
432
9, 421-429.
Werpy, T., Petersen, G., 2004. Top Value Added Chemicals from Biomass Volume I—Results of
438
Screening for Potential Candidates from Sugars and Synthesis Gas, National Renewable Energy
439
Laboratory (NREL).
441 442
te
Ac ce p
440
d
437
Wu, Y.C., et al., 2014. Energy-Saving Dividing-Wall Column Design and Control for Heterogeneous Azeotropic Distillation Systems. Ind. Eng. Chem. Res. 53(4), 1537-1552.
17 Page 17 of 40
442 443
List of Figures
445
Fig. 1. Dividing wall columns (DWCs): (a) conventional DWC; (b) top DWC; (c) bottom DWC.
446
Fig. 2. Ternary diagram for the furfural-water-formic acid system (mass basis).
447
Fig. 3. Schematic diagram of the optimal conventional sequence for producing LA from biomass.
448
Fig. 4. Flowsheet of the optimization procedure for a single distillation column.
449
Fig. 5. TAC saving plots for the C3 of the conventional sequence
450
Fig. 6. Schematic diagram of the C2 of the CS integrated with a decanter.
451
Fig. 7. Effect of temperature on furfural loss in the water-rich stream of the C2-decanter.
452
Fig. 8. Schematic diagram of the innovative TDWC-D.
453
Fig. 9. Schematic diagram of the proposed conventional sequence for producing LA from
cr
us
an
M
454
ip t
444
biomass.
Fig. 10. Schematic diagram of the proposed conventional sequence integrated a decanter.
456
Fig. 11. Contour plots of the interactions between the top section (N1), bottom section (N2) and
te
457
d
455
feed rectifying section (N3).
Fig. 12. Optimization plot by the RSM from Minitab.
459
Fig. 13. Three-dimensional response surface plots of the interaction between N1 and N2
460
Ac ce p
458
predicted by the RSM (a) and by the rigorous simulation (b).
461
Fig. 14. Schematic diagram of the proposed sequence intensified by the innovative TDWC-D.
462
Fig. 15. Comparison of the different LA recovery processes in terms of the annual operating cost
463
and TAC.
18 Page 18 of 40
M
an
us
cr
ip t
464
465
d te
467
Fig. 1. Dividing wall columns (DWCs): (a) conventional DWC; (b) top DWC; (c) bottom DWC.
Ac ce p
466
19 Page 19 of 40
467
470 471
Ac ce p
469
te
d
M
an
us
cr
ip t
468
Fig. 2. Ternary diagram for the furfural-water-formic acid system (mass basis).
20 Page 20 of 40
us
cr
ip t
471
472 473
Fig. 3. Schematic diagram of the optimal conventional sequence for producing LA from biomass.
Ac ce p
te
d
M
an
474
21 Page 21 of 40
M
an
us
cr
ip t
474
475
te
d
Fig. 4. Flowsheet of the optimization procedure for a single distillation column.
Ac ce p
476
22 Page 22 of 40
477 b)
478
us
cr
ip t
a)
Fig. 5. TAC saving plots for the C3 of the conventional sequence
Ac ce p
te
d
M
an
479
23 Page 23 of 40
479
an
us
cr
ip t
480
481 482
Fig. 6. Schematic diagram of the C2 of the CS integrated with a decanter.
Ac ce p
te
d
M
483
24 Page 24 of 40
us
cr
ip t
483
te
d
M
Fig. 7. Effect of temperature on furfural loss in the water-rich stream of the C2-decanter.
Ac ce p
485
an
484
25 Page 25 of 40
d
te
489
Fig. 8. Schematic diagram of the innovative TDWC-D.
Ac ce p
487 488
M
an
us
cr
ip t
486
26 Page 26 of 40
us
cr
ip t
489 490
491
Fig. 9. Schematic diagram of the proposed conventional sequence for producing LA from
493
biomass.
an
492
Ac ce p
te
d
M
494
27 Page 27 of 40
Fig. 10. Schematic diagram of the proposed conventional sequence integrated a decanter.
an
495 496
us
cr
ip t
494
Ac ce p
te
d
M
497
28 Page 28 of 40
ip t cr us an M d 499
te
498
Fig. 11. Contour plots of the interactions between the top section (N1), bottom section (N2) and
Ac ce p
497
feed rectifying section (N3).
29 Page 29 of 40
ip t
500
Fig. 12. Optimization plot by the RSM from Minitab.
Ac ce p
te
d
M
an
504
us
503
cr
501 502
30 Page 30 of 40
504 a)
b)
8
0
1 0
N2
cr
TAC saving (%)
ip t
16
-1 0 1
N1
us
d
M
an
predicted by the RSM (a) and by the rigorous simulation (b).
te
506
Fig. 13. Three-dimensional response surface plots of the interaction between N1 and N2
Ac ce p
505
-1
31 Page 31 of 40
Fig. 14. Schematic diagram of the proposed sequence intensified by the innovative TDWC-D.
an
508 509
us
cr
ip t
507
Ac ce p
te
d
M
510
32 Page 32 of 40
an
us
cr
ip t
510
511
Fig. 15. Comparison of the different LA recovery processes in terms of the annual operating cost
513
and TAC.
Ac ce p
te
d
M
512
33 Page 33 of 40
514
List of Tables
516
Table 1. Conditions of the feed mixture
517
Table 2. Design parameters of a conventional sequence for the production of LA from biomass
518
Table 3. Utilities cost data (Turton et al., 2012)
519
Table 4. Coded levels of the factors for TDWC variables of the CS sequence and PS sequence
520
Table 5. Comparison of the different structural alternatives for improving the performance of
cr
process for producing LA from biomass
us
521
ip t
515
522
an
523 524
M
525 526
530 531 532 533 534 535
te
529
Ac ce p
528
d
527
536 537 538 539 540
34 Page 34 of 40
541 542 543
Table 1. Conditions of the feed mixture
Component
Mass fraction (wt.%) 86%
Formic acid
3%
cr
Water
4%
us
Furfural Levulinic acid (LA)
7%
Temperature (oC)
an
25
Pressure (atm)
2 90,000
M
Mass flowrate (kg/hr)
Ac ce p
te
d
544 545
ip t
Feed conditions
35 Page 35 of 40
545 546
Table 2. Design parameters of a conventional sequence for the production of LA from biomass
547
C2
C3
22
27
28
Sieve
Sieve
Sieve
Column diameter (m)
3.6
1.7
Number of flow paths
1
1
Tray type
610
610
Max flooding (%)
87.83
Condenser duty (kW)
17,220
an
us
Tray spacing (mm)
Reboiler duty (kW)
1.8
cr
Number of trays
ip t
C1
22,930
1
610
80.36
85.93
2,699
3,694
4,250
3,770
M
Purity (mass fraction, wt.%)
Levulinic acid (LA)
549
85.0
Ac ce p
548
99.9
te
Formic acid
d
Furfural
98.0
36 Page 36 of 40
Table 3. Utilities cost data (Turton et al., 2012) Utility
Price ($/GJ)
Cooling water
0.35
Medium pressure steam
14.19
High pressure steam
17.70
cr
551
ip t
549 550
Ac ce p
te
d
M
an
us
552
37 Page 37 of 40
552
Table 4. Coded levels of the factors for TDWC variables of the CS sequence and PS sequence Case
Factor
Levels
Proposed sequence
Top section (N1)
22
Bottom section (N2)
5
Feed rectifying section (N3)
14
Top section (N1) Bottom section (N2)
30
10
15
18
22
18
22
26
2
6
10
13
17
20
an
Feed rectifying section (N3) 554
Ac ce p
te
d
M
555
1
26
us
Conventional sequence
0
cr
-1
ip t
553
38 Page 38 of 40
ip t cr
Annual operating cost (US $) Annual operating cost saving (%)
16,332,077
CS-D
PS
PS-D
PS-TDWC-D
15,151,046
13,646,300
13,729,322
12,812,024
0.6%
7.2%
16.4%
15.9%
21.6%
17,308,411
17,214,679
16,103,366
14,541,721
14,640,722
13,736,949
-
0.5%
7.0%
16.0%
15.4%
20.6%
-
Ac
ce pt
Total annual cost saving (%)
CS-TDWC-D
16,234,472
ed
Total annual cost (US $)
CS
M an
Structural Alternative
us
Table 5. Comparison of the different structural alternatives for improving the performance of process for producing LA from biomass
40 Page 39 of 40
Ac
ce
pt
ed
M
an
us
cr
i
Graphical Abstract (for review)
Page 40 of 40