Professional Documents
Culture Documents
Chemical Engineering & Processing: Process Intensi Fication: Articleinfo
Chemical Engineering & Processing: Process Intensi Fication: Articleinfo
A R T I C L E I N F O A B S T R A C T
Keywords: Isopropanol, also known as isopropyl alcohol (IPA) is an important solvent used in many industries. It is mostly
Isopropyl alcohol produced by direct hydration of propylene involving reversible reaction and azeotropic distillation in separate
Catalytic distillation equipment. There have been many studies on separation of IPA and water azeotropic mixture but very few
Reactive distillation studies on the use of reactive distillation (RD), for improving overall efficiency of the entire process. RD can be
Propylene–propane separation
used for IPA production in two different ways: with excess propylene to avoid azeotropic separation, and with
Extractive distillation
excess water to achieve near complete conversion of propylene. The present work investigates both these options
by design and optimization of complete IPA process for minimizing capital and operating costs. The results show
that the RD process with excess water has lower costs compared to the RD process with excess propylene; but the
former incurs small loss of propylene.
⁎
Corresponding author.
E-mail address: chegpr@nus.edu.sg (G.P. Rangaiah).
http://dx.doi.org/10.1016/j.cep.2017.04.021
Received 21 March 2017; Accepted 30 April 2017
Available online 08 May 2017
0255-2701/ © 2017 Elsevier B.V. All rights reserved.
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
Fig. 1. RD-splitter process flow diagram with stream data for the optimal solution with the lowest TAC.
Fig. 2. RD-ED process flow diagram with stream data for the optimal solution with the lowest TAC.
109
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
Table 1
Decision variables with their bounds for the RD-splitter and RD-ED processes for IPA.
1. Number of equilibrium stages in the RD column (range: 18–40) 1. Number of equilibrium stages in the RD column (range: 16–20)
2. Number of equilibrium stages in the splitter column (range: 160–200) 2. Number of equilibrium stages in the ED column (range: 25–40)
3. Feed stage of the splitter column (range: 70–110) 3. Number of equilibrium stages in the ER column (range: 15–30)
4. Distillate flow rate of the RD column (range: 170–190 kmol/h) 4. IPA-water feed stage of the ED column (range: 20–25)
5. Distillate flow rate of the splitter column (range: 150–180 kmol/h) 5. DMSO-water feed stage of the ER column (range: 6–14)
6. Reflux ratio of the RD column (range: 15–25)
7. Distillate flow rate of the RD column (range: 20–30 kmol/h)
Notes: IPA mole fraction of 0.999 in the RD bottoms is achieved by varying reflux ratio, Notes: IPA mole fraction of 0.999 and recovery of 0.9999 in the ED distillate are
and propylene mole fraction of 0.95 in the splitter distillate is achieved by varying achieved by varying distillate rate and reflux ratio. DMSO mole fraction of 0.999 and
reflux ratio, both within Aspen Plus simulation. recovery of 0.9999 in the ER bottoms are achieved by varying distillate rate and
reflux ratio. All these are within Aspen Plus simulation.
RD in the process scheme of Xu et al. [11]. Recently, Niu and Rangaiah the splitter column for separation. Propylene of purity similar to the
[13] proposed a heuristic-based retrofitting methodology for intensify- fresh feed is obtained from the splitter top and sent back to the RD as
ing existing processes, and applied it to the conventional IPA process. recycle.
They developed two retrofit solutions – one using RD with excess The alternate RD-ED process uses excess water to produce IPA in a
propylene and another using RD with excess water. In another study, RD column. Similar to the RD-splitter, the RD column with 3 reactive
Niu and Rangaiah [14] explored retrofitting the IPA process based on stages is used. As shown in Fig. 2, water and propylene are fed into the
RD with excess propylene, and developed retrofit solutions involving RD column at the second and fourth stages (counting from the column
RD with excess water. Both these studies are in the context of top) respectively, which are the two ends of the reactive stages. An IPA/
retrofitting an existing process. water mixture is collected as the bottom product. The conventional
Thus, RD can be used in two different ways for IPA production. One two-column ED set-up is added after the RD, to produce high purity IPA
alternate is RD with excess propylene to produce pure IPA without using dimethyl sulfoxide (DMSO) as the low-boiling entrainer. This is
requiring the downstream separation of azeotropic mixture of IPA and based on Arifin and Chien [7], who showed that ED is better than HAD
water but this requires the difficult and expensive separation of for separation of 77 wt.% (50.1 mol%) IPA in water, with 32% lower
propylene–propane for recycling propylene. Another alternate is RD TAC. As shown in Fig. 2, DMSO is recovered in the entrainer recovery
with excess water, which requires further separation of IPA and water (ER) column for reuse in the ED column; make-up entrainer is quite
involving an azeotrope but avoids propylene–propane separation. It is small (∼0.4 kmol/h or ∼3 kg/h).
of particular interest to compare these two alternatives for new plants, For the present study, both the RD-splitter and RD-ED processes are
and this is the main objective of this study. The two RD-based processes simulated rigorously in Aspen Plus version 9. Feed to both processes is
for IPA production are simulated and optimized for capital and the chemical-grade propylene of 95 mol% purity (with propane as the
operating costs by multi-objective optimization (MOO). Results show impurity). Its flow rate is 400 kmol/h, which results in the production
that the IPA process using RD with excess water is more economical of around 375 kmol/h. The propylene feed mixture is assumed to be
than that using the RD with excess propylene. obtained as a fully refrigerated [15] saturated liquid at 1.05 bar and
The rest of this paper is organized as follows. The next section −46.7 °C. By preheating the propylene feed in a heat exchanger using
describes the two IPA processes, one using RD with excess propylene cooling water at 30 °C, 1600 kmol/h of chilled water at 5 °C can be
and another using RD with excess water. MOO problem and its solution produced. Thus, the final temperature of fresh propylene feed is 24.9 °C.
methodology are outlined in Section 3. Optimization results for the two Taking molecular mass of isopropanol of 60.1 g/mol and plant opera-
IPA processes are presented and discussed in Section 4. Finally, Section tion of 8000 h/year, the production capacity of either process is around
5 summarizes the main findings of this study. 180,000 ton/year.
For the RD-splitter process, the water feed rate is fixed at 380 kmol/
2. Process alternatives for IPA production h. In Fig. 1, the molar ratio of propylene in the combined feed to the
water feed is 1.39:1, which is close to the 1.5:1 molar ratio used by
In this section, the two alternatives for IPA production are Wang and Wong [12] in their simulations. In the RD-ED process, the
described. First, IPA process using RD with excess propylene is propylene to water ratio is 1:2.05 to ensure excess water in the RD
presented. This process requires propylene–propane separation by column. Further, the recycled DMSO entrainer flow rate is fixed at
distillation (splitter) for recovering excess propylene, and hence it is 673 kmol/h, as shown in Fig. 2. This gives a DMSO to IPA/water molar
referred to as the RD-splitter process. Then, IPA process using RD with feed ratio of around 0.864:1, compared to 1.025:1 ratio used by Arifin
excess water is described; it is referred to as the RD-ED process since it and Chien [7] in their optimized IPA dehydration process design.
involves separation of IPA–water mixture by ED. For simulating the RD-splitter process in Aspen Plus, UNIFAC and
The RD-splitter process (Fig. 1) consists of a RD column followed by Peng–Robinson models are chosen for the RD and splitter column,
a splitter column to separate unreacted propylene from propane inert, respectively, as in Xu et al. [11], who verified their suitability by
for recycling. It is a modification of the design by Wang and Wong [12] comparison with the experimental data. All the distillation columns
to perform reaction(s) in a RD column using excess propylene. Based on including RD are simulated rigorously using RADFRAC block in Aspen
preliminary analysis, a RD column with 3 reactive stages is used. As Plus and assuming vapor–liquid equilibrium on each stage. Further,
shown in Fig. 1, water and propylene are fed into the RD column at the reactive stages in the RD column are simulated assuming equilibrium-
third and fifth stages (counting from the top of the column) respec- reaction model. This simulation approach follows that of Xu et al. [11]
tively, which are the two ends of the reactive stages. As water is the and Wang and Wong [12].
limiting reactant, practically all the water fed to the column is Pressure of the RD column is selected as 20 bar, which is concluded
consumed in the reaction. This helps to avoid the problem of separating to be optimal by Xu et al. [11]. Splitter column pressure is taken to be
azeotropic mixture of IPA and water. IPA of high purity (99.9 mol%) is 18 bar. These pressures allow the use of cheaper cooling water in the
obtained as the bottom product of the RD column. Distillate of the RD condensers of RD and splitter columns. RD has a partial condenser so
column consists of excess propylene and propane. This stream is sent to that the vapor distillate stream can be fed to the splitter column in order
110
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
Fig. 3. Pareto-optimal solutions and the corresponding decision variables for the RD-splitter process. Square symbol refers to the Pareto-optimal solution having the minimum TAC.
to reduce reboiler duty of the latter column in the RD-splitter process variables and constraints. These are discussed, one by one, for IPA
and to remove lighter gases in the RD-ED process. NRTL is chosen as the process optimization. Cost is a major consideration for setting up any
property package for ED and ER columns, as in Arifin and Chien [7], new plant. Two objectives frequently analyzed are Annualized Capital
and [24]. Both ED and ER columns in the RD-ED process are at Cost (ACC) and Annual Operating Cost (AOC). Although these two can
atmospheric pressure for easier separation and to avoid high pressure, be combined into a single objective (namely, TAC), they are considered
and yet cooling water can be used in their condensers. Condensers of here as two separate objectives for simultaneous minimization. This
splitter, ED and ER columns are assumed to be total for obtaining liquid approach provides many optimal solutions for further analysis and
distillate streams. selection. Thus, IPA process with either the RD-splitter or RD-ED is
optimized for the following two objectives:
3. Multi-objective optimization problem and its solution Minimize f1 : Annualized Capital Cost, ACC ($/year). (3a)
methodology
Minimize f2 : Annual Operating Cost, AOC ($/year). (3b)
A typical MOO problem has two or more objectives, decision
111
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
Fig. 3. (continued)
Here, ACC is estimated based on all major equipment in the process. for condensers and reboilers of the distillation columns, electricity for
Capital cost of the distillation columns, heat exchangers and pumps in pumps and cost of catalyst replacement in the RD column. Unit cost
the process are estimated using the Capital Cost Estimation Program ($/GJ) of utilities are: 8.58 for low pressure steam (3.5 bar, 148 °C),
(CCEP), developed by Feng and Rangaiah [16] based on cost correla- 10.05 for medium pressure steam (10 bar, 185 °C), 12.92 for high
tions in Seider et al. [25]. Data required for cost estimation like column pressure steam (30 bar, 237 °C), 0.29 for cooling water and 16.8 for
diameter, operating temperature and pressure are obtained directly electricity; these values are average of those in Seider et al. [25] and
from the Aspen Plus simulation results. Tray efficiency of the columns is Turton et al. [26]. For costing the catalyst in the RD column,
estimated using the correlation of Kessler and Wankat [17], which is commercially available Amberlyst 36 catalyst is assumed. Mass of this
based on O’Connell's curve; for this, geometric mean of relative catalyst needed per stage (M kg/stage) is estimated using:
volatility of key components and liquid viscosity values for all stages
in the column is used. Annualization factor for calculating ACC is 0.2 πD 2h
M= × ρ × (1 − ε ) × n.
based on interest rate of 15% and plant life of 10 years. 4 (4)
The second objective, AOC is estimated based on utilities required
Here, D is the diameter of tray (m), h is the height of catalyst packing on
112
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
Fig. 4. Temperature and liquid composition profiles in the RD column corresponding to the Pareto-optimal solution with the lowest TAC for the RD-splitter process.
the tray (m) and ρ is catalyst density (= 1200 kg/m3), ε is void fraction 4. Results and discussion
of catalyst and n is the fraction of tray area occupied by the catalyst. ε
(= 0.36) is based on the minimum bulk porosity of sphere packings for In this section, results for the RD-splitter and RD-ED processes are
uniform-sized spheres Zhang et al. [27]. For conservative estimation presented in Sections 4.1 and 4.2, respectively. Then, performance of
and achieving reaction equilibrium, h is assumed to be 0.2 m (about one the two processes is compared in Section 4.3.
third of typical tray-to-tray spacing) and n is assumed to be 0.85.
Catalyst life is assumed to be 2 years, based on methyl tertiary-butyl
ether (MTBE) production process, which uses similar acidic cation 4.1. RD-splitter process
exchange resin [18]. Taking catalyst cost of $531/kg from Sigma-
Aldrich Co. LLC [19], total annual cost of catalyst for all reactive stages Pareto-optimal solutions including optimal values of decision vari-
can be calculated. ables obtained for the RD-splitter process, using EMOO program are
In designing a distillation column, three important variables are the presented in Fig. 3. Fig. 3a shows that AOC decreases substantially from
number of trays, feed stage and reflux ratio. Capital cost of a column is around $10 million/year to $8 million/year while ACC increases
related to the number of trays, and operating cost depends on the feed slightly from $3.2 million/year to $3.4 million/year. As shown in
stage and reflux ratio. Note that a large reflux ratio also results in a Fig. 3b and c, further increase in ACC above $3.4 million/year is mainly
larger column diameter, reboiler and condenser. Thus, there is trade-off due to increasing number of stages in the splitter column, which has
between number of trays and reflux ratio. Operating cost of a distilla- negligible impact on AOC. TAC can be calculated by simple addition of
tion column is often more significant than capital cost (e.g., see Table 4 AOC and ACC for the optimal solutions. The Pareto-optimal solution
in [20]). In summary, decision variables and their bounds for optimiz- having the lowest TAC is shown as a rectangle in Fig. 3.
ing the RD-splitter and RD-ED processes are given in Table 1. These and As can be seen in Fig. 3d and e, ACC increases while AOC decreases
the bounds for each variable are chosen based on typical values in the as the number of stages in the RD column increases from 21 to 33. This
literature, to ensure simulation convergence and to provide scope for is expected since a larger number of stages, while incurring higher
optimization. capital cost, allows the column to be operated at a lower reflux ratio,
For optimizing the RD-ED process, two additional constraints are which lowers operating cost. For the RD-splitter process, savings in
imposed. Temperature of the RD column is lower than that in the RD- AOC from increasing the number of stages in the RD column outweigh
splitter set-up. This will reduce reaction rate of propylene hydration. A the slight increase in ACC. Optimal values of the feed stage for the
temperature of 130 °C in the reactive stages [11] is recommended to splitter is found to be scattered between 78 and 86, with no clear
ensure reasonable reaction kinetics. Hence, a minimum stage tempera- relationship between the splitter feed stage and ACC or AOC. Hence,
ture of 125 °C is imposed for the top most reactive stage, which is the these results are not shown for brevity.
reactive stage with the lowest temperature in the column. Secondly, as Fig. 4 shows the temperature and liquid composition profiles in the
the vent gas is not recycled back to the reaction stage, some propylene RD column corresponding to the Pareto-optimal solution with the
is wasted through the vent and IPA production is reduced since lowest TAC for the RD-splitter process. As expected, temperature in
propylene is the limiting raw material. Based on preliminary tests, a the column gradually increases from the top (stage 1) to bottom (stage
minimum IPA flow of 372 kmol/h in the RD bottoms is imposed. 33). Composition profiles confirm the production of pure IPA from the
Sharma et al. [21] have developed an Excel-based multi-objective bottom of the RD column. Within the column, DIPE is formed and its
optimization (EMOO) program based on the elitist non-dominated mole fraction is around 0.35 on stages 3–23. However, owing to
sorting genetic algorithm (NSGA-II). They tested the program for a reversible reactions and separation in the RD column, DIPE is very
number of benchmark functions and applied to MOO of a falling-film small in both the distillate and bottoms streams.
evaporator for milk concentration. Subsequently, EMOO program has The RD column distillate rate has a more complex effect on the
been applied to different applications and also enhanced with real objectives. Fig. 3f and g show that ACC decreases and AOC increases
coding and performance-based termination criteria [22,23]. In the when the RD column distillate rate increases from 178 kmol/h to
present study too, this program has been successfully employed for 185 kmol/h. A similar pattern can be seen for the splitter column
design optimization of two IPA processes based on RD. NSGA-II distillate rate (Fig. 3h and i), which ranges from 156 kmol/h to
algorithm parameters used are: real coding, non-uniform mutation, 163 kmol/h. In fact, a linear relationship is found between the RD
tournament selection, population size of 100, crossover probability of and splitter distillate rates; this is attributable to nearly constant flow
0.9, mutation probability of 0.1 and maximum number of generations rate of splitter bottoms at around 22 kmol/h for all Pareto-optimal
of 50. solutions. The splitter (and indirectly the RD column) distillate rate is
related to the combined feed flowrate into the RD column due to the
113
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
Fig. 5. Pareto-optimal solutions and the corresponding decision variables for the RD-ED process. Square symbol refers to the Pareto-optimal solution having the minimum TAC.
recycle stream from the splitter column to the RD column. As a result, a 4.2. RD-ED process
higher splitter (and RD column) distillate rate increases propylene feed
into the RD column, which will increase RD column diameter, reboiler The Pareto-optimal solutions obtained by EMOO program, for the
and condenser duties. Thus, both AOC and ACC should increase as the MOO of the RD-ED process are shown in Fig. 5a . As AOC decreases
splitter/RD column distillate rate increases. As shown in Fig. 3f–i, AOC from around $6.5 million/year to $6.16 million/year, ACC increases
increases with splitter/RD column distillate; however, ACC decreases from $1.95 million/year to $2.25 million/year. With reference to the
with splitter/RD column distillate, which is due to confounding by optimal solution having the lowest TAC (square point in Fig. 5), further
another decision variable, namely, number of ideal stages in RD reduction in AOC increases ACC disproportionately.
column. As shown in Fig. 3j and k, number of ideal stages in RD Fig. 5b and c show that ACC increases and AOC decreases when the
column decreases as RD distillate rate increases. number of ideal stages in the ED column increases from 27 to 35. This
trend is not found for the number of ideal stages in RD and ER columns,
which have optimal values 17 or 18 stages, and between 15 and 18
114
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
Fig. 6. Temperature and liquid composition profiles in the RD column corresponding to the Pareto-optimal solution with the lowest TAC for the RD-ED process.
4.3. Comparison [1] P. Dutia, Isopropyl alcohol: a techno-commercial profile, Chem. Week. (2012)
211–216. April 24, Retrieved from www.chemicalweekly.com/Profiles/Isopropyl_
Alcohol.pdf.
The optimal results corresponding to the lowest TAC for both the [2] J.I. Kroschwitz, Isopropyl Alcohol in Kirk-Othmer Encyclopedia of Chemical
RD-splitter and RD-ED processes are compared in Table 2. It can be seen Technology, 5th ed., Wiley-Interscience, New York, 2004.
that the RD-ED process has better values for all cost metrics. In terms of [3] I.L. Chien, K.L. Zeng, H.Y. Chao, Design and control of a complete heterogeneous
azeotropic distillation column system, Ind. Eng. Chem. Res. 43 (2004) 2160–2174.
ACC, the high cost of the splitter column in the RD-splitter process
115
W.J. Chua et al. Chemical Engineering & Processing: Process Intensification 118 (2017) 108–116
[4] S. Arifin, I.L. Chien, Combined pre-concentrator/recovery column design for pdf.
isopropyl alcohol dehydration process, Ind. Eng. Chem. Res. 46 (2007) 2535–2543. [16] Y. Feng, G.P. Rangaiah, An evaluation of capital cost estimation methods and
[5] W.T. Chang, C.T. Huang, S.H. Cheng, Design and control of a complete azeotropic programs, Chem. Eng. 118 (8) (2011) 22–29.
distillation system incorporating stripping columns for isopropyl alcohol dehydra- [17] D.P. Kessler, P.C. Wankat, Distillation design: correlations for column parameters,
tion, Ind. Eng. Chem. Res. 51 (2012) 2997–3006. Chem. Eng. 26 (September) (1988) 71–74.
[6] W.L. Luyben, Plantwide control of an isopropyl alcohol dehydration process, AIChE [18] US International Trade Commission, Methyl tertiary-butyl ether (MTBE): conditions
J. 52 (2006) 2290–2296. affecting the domestic industry, US International Trade Commission, 2-2, 1999
[7] S. Arifin, I.L. Chien, Design and control of an isopropyl alcohol dehydration process Retrieved from https://www.usitc.gov/publications/docs/pubs/332/pub3231.pdf.
via extractive distillation using dimethyl sulfoxide as an entrainer, Ind. Eng. Chem. [19] Sigma-Aldrich Co. LLC, Amberlyst 36, (2016) Retrieved from http://www.
Res. 47 (2008) 790–803. sigmaaldrich.com/catalog/product/aldrich/436712?lang=en®ion=SG.
[8] S. Sommer, T. Melin, Design and optimization of hybrid separation processes for the [20] C.M. Lek, G.P. Rangaiah, K. Hidajat, Distillation: revisiting some rules of thumb,
dehydration of 2-propanol and other organics, Ind. Eng. Chem. Res. 43 (2004) Chem. Eng. September (2004) 50–55.
5248–5259. [21] S. Sharma, G.P. Rangaiah, K. Cheah, Multi-objective optimization using MS excel
[9] P.K. Sharma, P.C. Wankat, Distillation-adsorption hybrid processes to separate with an application to design of a falling-film evaporator system, Food Bioprod.
binary liquid mixtures with homogeneous azeotrope, Sep. Sci. Technol. 48 (2013) Process. 90 (2) (2012) 123–134.
1–14. [22] J.Y.Q. Wong, S. Sharma, G.P. Rangaiah, Design of shell-and-tube heat exchangers
[10] Y. Liu, S.D. Feist, C.M. Jones, D.R. Armstrong, Isopropanol dehydration by hot-gas for multiple objectives using elitist non-dominated sorting genetic algorithm with
pressure swing adsorption: experiments, simulations and implementation, Ind. Eng. termination criteria, Appl. Therm. Eng. 93 (2016) 888–899.
Chem. Res. 53 (2014) 8599–8607. [23] S. Sharma, G.P. Rangaiah, F. Maréchal, Multi-objective optimization programs and
[11] Y. Xu, K.T. Chuang, A.R. Sanger, Design of a process for production of isopropyl their application to amine absorption process design for natural gas sweetening, in:
alcohol by hydration of propylene in a catalytic distillation column, Chem. Eng. G.P. Rangaiah (Ed.), Multi-Objective Optimization: Techniques and Applications in
Res. Des. 80 (2002) 686–694. Chemical Engineering, second ed., World Scientific, 2017.
[12] S.J. Wang, D.S. Wong, Control of reactive distillation production of high-purity [24] K. Liang, W. Li, H. Luo, M. Xia, C. Xu, Energy-efficient extractive distillation process
isopropanol, J. Process Control 16 (2006) 385–394. by combining preconcentration column and entrainer recovery column, Ind. Eng.
[13] W.M. Niu, G.P. Rangaiah, Process retrofitting via intensification: a heuristic Chem. Res. 53 (2014) 7121–7713.
methodology and its application to isopropyl alcohol process, Ind. Eng. Chem. Res. [25] W.D. Seider, J.D. Seader, D.R. Lewin, Product and Process Design Principles, 3rd
55 (2016) 3614–3629. ed., Wiley, New York, 2010.
[14] W.M. Niu, G.P. Rangaiah, Retrofitting an isopropyl alcohol process based on [26] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, Analysis, Synthesis, and Design
reactive distillation and propylene-propane separation, Chem. Eng. Process.: of Chemical Processes, 3rd ed., Prentice Hall, New Jersey, 2009.
Process Intensif. 108 (2016) 164–173. [27] W. Zhang, K.E. Thompson, A.H. Reed, L. Beenken, Relationship between packing
[15] American Chemistry Council, Propylene product stewardship guidance manual, structure and porosity in fixed beds of equilateral cylindrical particles, Chem. Eng.
(2007) Retrieved from https://www.shell.com/content/dam/shell/static/ Sci. 61 (24) (2006) 8060–8074.
chemicals/downloads/products-services/propylene-productstewardshipmanual.
116