Reactive Distillation for Esterification of an Alcohol Mixture Containing

Department of Chemical Engineering, National Taiwan University Taipei, 10617, Taiwan, and ... National Taiwan University of Science and Technology...
0 downloads 0 Views 7MB Size
7186

Ind. Eng. Chem. Res. 2009, 48, 7186–7204

Reactive Distillation for Esterification of an Alcohol Mixture Containing n-Butanol and n-Amyl Alcohol Hao-Yeh Lee,† Ling-Ting Yen,† I-Lung Chien,‡ and Hsiao-Ping Huang*,† Department of Chemical Engineering, National Taiwan UniVersity Taipei, 10617, Taiwan, and Department of Chemical Engineering, National Taiwan UniVersity of Science and Technology, Taipei 10607, Taiwan

Manufacturing processes in the semiconductor and pharmaceutical industries often produce alcohol mixture byproducts. Therefore, the esterification of alcohol mixtures may be an important step in reusing wastes from these industries. There are two alternative methods for using the alcohol mixtures as feed for reactive distillation (RD). The first method separates this mixture into pure alcohols first and then follows with esterification using the RD column. The second method uses direct esterification of the alcohol mixture in a RD column, and then separates the mixed-ester products. This paper discusses the esterification of a n-butanol (BuOH) and n-amyl alcohol (AmOH) mixture with acetic acid (HAc). This study presents two important results based on optimizing the total annual cost (TAC). First, the mixed BuOH/AmOH system, with direct esterification with RD followed by product separation, is more economical than the system that first separates the mixture. Second, this study proposes a novel economical indirect-sequence design flowsheet with aqueous reflux. Another important issue in this study is the choice of the relative feed location, because the boiling point of acid lies between that of the two alcohols (i.e., BuOH < HAc < AmOH). Reaction kinetics is an important factor to be considered in determining the feed location of the alcohol mixture. 1. Introduction Manufacturing processes in the semiconductor and pharmaceutical industries often produce alcohol mixture byproducts. Therefore, the esterification of alcohol mixtures may be an important step in reusing wastes from these industries. This paper proposes that an alcohol mixture of BuOH and AmOH can be used as a feed stream for further processing. Manufacturing plants often use BuOH and AmOH alcohols as organic solvents for paint, in adhesives, in creating artificial leather, and processing textiles and spices. However, existing literature on this topic focuses on the single esterification of either BuOH or AmOH with acid (e.g., acetic acid) to obtain buytl acetate (BuAc) or amyl acetate (AmAc) in a RD system. Discussion on the esterification of an alcohol mixture as a feed stream is relatively scarce. Several papers in the literature discuss RD in the esterificaton of a single alcohol. Steinigeweg and Gmehling1 used RD experiments and simulations to explore the effects of steady state design variables for a BuOH esterification system. Gangadwala et al.2,3 and Singh et al.4 developed a steady-state column model with reaction kinetics and compared it with experimental data from the literature. Specifically, their model compares three different column configurations for the production of BuAc, preventing the formation of dibutyl ether byproduct and achieving a high purity product. Furthermore, Gangadwala and Kienle5 applied a mixed integer nonlinear programming optimization method to the BuAc RD system. Their optimized designs offer significant economic improvements in production. Chiang et al.6 explored RD system design for a steady state AmOH esterification system and compared it with a coupled reactor/column structure. Their RD system is more efficient with a reduced total annual cost. Tang et al.7 studied the esterification of HAc with five different alcohols, * To whom correspondence should be addressed. Tel.: 886-2-23638999. Fax: 886-2-2362-3935. E-mail: [email protected]. † National Taiwan University Taipei. ‡ National Taiwan University of Science and Technology.

ranging from C1 to C5 with process flowsheets classified into types I, II, and III for these five systems. Their results show that BuOH and AmOH esterification systems have similar thermodynamic properties and are classified as type III systems. On the basis of reaction types, RD systems can be classified as single reversible reactions, series reactions, or parallel reactions. In the case of parallel reactions, such mixtures as mixed isopropyl alcohol and BuOH and mixed ethanol and BuOH react with HAc in the esterification process. These esterification processes with mixed alcohols are only mentioned in several patents.8-11 The mixed BuOH and AmOH feeds of this paper also fall into the case of parallel reactions. In this study, the esterification reactions of BuOH and AmOH demonstrate the use of an alcohol mixture as a feed stream in the RD process. Based on thermodynamic properties and reaction kinetics, this study presents two alternative process designs with three kinds of configurations for the overall RD system. 2. Thermodynamic and Kinetic Models Chiang et al.6 and Venimadhavan et al.12 used the NRTL and UNIQUAC thermodynamic models for the BuOH and AmOH esterification processes, respectively. To simulate a mixed alcohols system, it is first necessary to merge these thermodynamic models into a single model, and then regress the additional three binary pairing parameters. This paper chooses the NRTL model to describe the liquid phase and the Hayden-O’Connell model13 to describe vapor phase. This approach corrects fugacity due to association as dimer in acetic acid at the atmospheric pressure. This system contains six components and 15 binary parameters. The HAc/AmAc, HAc/ H2O, AmOH/AmAc, and AmAc/H2O binary parameters are from Chiang et al.6 The BuOH/AmOH, BuOH/AmAc, BuAc/ AmAc are obtained from the Aspen data bank, and the remaining binary parameters are regressed from experimental data.14-23 These binary parameters are regressed by maximumlikelihood method which can be found in the Aspen Properties

10.1021/ie801891q CCC: $40.75  2009 American Chemical Society Published on Web 07/06/2009

0 0 254.47 2221.5 0.2 -5.519 -6.472 2428.1 3774.4 0.3116 0 0 128.17 -113.6 0.3 0 0 57.328 1424.8 0.2869 0 0 -144.8 320.65 0.3009

[ ] mj

k

k

j



xjGij

∑x G

kj

τij-

m

k

k

m mj

kj

∑x τ G ∑x G

regression Aspen regression Chiang et al.6 regression

-0.039 0.0734 793.01 -391.9 0.2662 0 0 161.48 1186.4 0.3962 0 0 203.57 138.54 0.3

Aspen Plus NRTL:

a

where Gij ) exp(-Rijτij), τij ) aij + bij/T, Rij ) cij, τii ) 0, and Gii ) 1.

0 0 -110.6 424.02 0.2987 0 0 -37.94 214.55 0.2 -2.909 3.0379 1194.7 -1000 0.301 13.513 -8.415 -4129 2580.4 0.18 aij aji bij (K) bji (K) Rij

0 0 41.866 150.87 0.3

Chiang et al.6 regression

regression

regression

Chiang et al.6

i

j

k

k

j ji

∑xτ G ln γ ) ∑x G

ki

ji

+

1.2268 -2.861 -442.3 1264.2 0.3112 0 0 125.45 -110.6 0.3

BuOH BuAc BuOH AmOH HAc H2O HAc AmAc HAc BuAc HAc AmOH HAc BuOH

To compare the minimum TAC, this section discusses two types of process design for a mixed alcohols system. The first is “separation first,” which first separates the mixed alcohols from the upstream, and then feeds the BuOH and AmOH into different RD columns for HAc esterification. The second is “reaction first” which feeds the mixed alcohols stream directly into a RD column, and then separates the two ester products and water. All the conceptual design of the alternative process flowsheets are shown in Figure 3. 3.1. “Separation First” Design. The “separation first” process flowsheet includes one traditional distillation column and two RD columns. The first column separates the BuOH and AmOH from the feed stream as shown in Figure 3a. Then the BuOH and AmOH are fed into their own RD columns and react with HAc. The final products of AmAc and BuAc are

component i component j

3. Conceptual Design of the Process Flowsheet

Table 1. NRTL Model Parameters for BuOH/AmOH Mixture Systema

Table 3 shows the reaction kinetics equations from Gangadwala et al.2 and Lee et al.24 The BuOH esterification rate equation is expressed as a pseudo-homogeneous model by activity, and the AmAc system is expressed as a quasihomogeneous system by concentration. BuOH esterification is an exothermic reaction, and AmOH esterification is an endothermic reaction. In all the design studies followed, the software package of Aspen Plus was used to conduct the rigorous simulation. Equilibrium stages were assumed in the column simulation. Total material, component, and energy balances were calculated in each column tray via RadFrac module in Aspen Plus. The modified Newton’s method of Broyden is chosen to obtain the steady-state solution for the system of algebraic equations. As for the reactive section of the column, catalyst volume in each tray is assumed to occupy half of the total holdup. Catalyst cost is assumed to be $3.5/lb with its density assumed to be 800 kg/m3.

BuOH AmAc

k4

regression

(1)

source

k3

HAc + AmOH y\z AmAc + H2O

Aspen

BuOH H2O

k2

regression

k1

HAc + BuOH y\z BuAc + H2O

Aspen

AmOH BuAc

AmOH AmAc

AmOH H2O

BuAc AmAc

BuAc H2O

AmAc H2O

software. Table 1 shows all NRTL binary parameters used in the simulation. The calculation of the vapor pressure of all pure components can be found in Appendix A of the supplemental file. On the basis of the thermodynamic model, Figure 1 displays all results of 15 binary pairings. In Figure 1, T-xy plot is placed on left-hand side and the x-y plot is placed on the right-hand side. Furthermore, the liquid-liquid equilibrium results of ternary systems are shown in Figure 2. In Figure 2, the gray with dot area is two liquid regions and the straight line connected between two hollow circles is the tie line. Available experimental data are also displayed in Figure 1a-g and Figure 2 with square and diamond symbols. Figures 1 and 2 show that the model results agree quite well with binary and ternary experimental data. Table 2 displays the boiling points rankings for pure components and the azeotropic temperatures of all the azeotropes from experimental data and predictive results. Esterification is a reversible reaction catalyzed by an acidic catalyst. The system described in this paper uses Amberlyst 15 as a solid catalyst. Amberlyst 15 is an acid ion exchanged resin that has the benefit of no environmental problems than homogeneous catalysts. The parallel and competing esterification reactions in this system occur as the following equations.

Chiang et al.6

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

7187

7188

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

7189

7190

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

7191

7192

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Figure 1. T-xy and x-y experiment and predicted diagrams under 1 atm. (a) Gmehling et al.,15 Part 2b, p 138; (b) Gmehling et al.,17 Part 5, pp 147-151; (c) Gmehling et al.,15 Part 2b, pp 194-197; (d) Koichi and Hitoshi;19 (e) Cho et al.;20 (f) Gmehling et al.,16 Part 2d, p 432; (g) Gmehling et al.,14 Part 1a, p 383.

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

7193

Figure 2. LLE experiment and predicted diagram for the ternary system under 1 atm: (a,b) Ruiz Bevia et al.,21 (c) Tan and Aravinth,22 (d) Esquı´el and Bernardo-Gil.23 Table 2. Boiling Points and Azeotropes of BuOH/AmOH Mixture System under 1 atm experimental data component

mole fraction a

BuOH/BuAc/H2O BuAc/H2Oa BuOH/AmAc/H2Oa BuOH/H2Oa AmOH/AmAc/H2Oa AmAc/H2Oa AmOH/H2Oa H2O BuOH/BuAc BuOH HAc HAc/BuOH/BuAc HAc/BuOH BuAc AmOH AmAc

b

(0.0726, 0.224, 0.7034) (0.299, 0.701)c

computed data temp °C

mole fraction

temp °C

90.38 91

(0.0912, 0.2142, 0.6946) (0.2871, 0.7129) (0.1882, 0.0823, 0.7295) (0.2458,0.7542) (0.0418, 0.1352, 0.823) (0.1696, 0.8304) (0.1471, 0.8529) 1 (0.7614, 0.2386) 1 1 (0.4283, 0.4754, 0.0963) (0.4174, 0.5826) 1 1 1

90.44 90.98 91.82 92.85 94.76 94.9 96 100.02 117.01 117.68 118.01 122.43 122.48 126.01 137.68 147.71

(0.247, 0.753)d (0.046, 0.107, 0.847)e (0.166,0.834)f (0.146,0.854)g 1 (0.773, 0.227)h 1 1

92.8 94.9 95.2 95.8 100.02 117 117.68 118.01

(0.404, 0.596)I 1 1 1

122.6 126.01 137.68 147.71

a Heterogeneous azeotrope. b Gmehling et al.18 part II, p 863. c Gmehling et al.18 part I, p 96. d Gmehling et al.18 part I, p 80-81. e Gmehling et al.18 part II, p 867. f Gmehling et al.18 part I, p 102. g Gmehling et al.18 part I, p 91. h Gmehling et al.18 part I, p 531-532. I Gmehling et al.18 part I, p 291-292.

obtained at the bottoms of the two RD columns and H2O is emitted from the outlet at the top of the decanter. According to the boiling point and azeotrope information presented in Table 2, the BuOH and AmOH mixed alcohols feed stream does not contain any azeotropes. This means that this “separation first” process flowsheet can be used because the BuOH and AmOH can be easily separated by a distillation column. Notice that if the mixed reactants contain an azeotrope, this design would

not be economical due to the requirement of more columns. The results of Tang et al.7 provide an optimal design of BuAc and AmAc RD processes for pure BuOH and AmOH. Therefore, determining the optimal design of the first distillation column is the only additional finding required. This study makes several assumptions in the following simulation: The mixed alcohols feed stream (BuOH/AmOH) is set to equal molar, and the total flow rate is 100 kmol/h. The feed stream

7194

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

of pure reactant HAc is also set at a 100 kmol/h flow rate. The decanter temperature is set at 40 °C. Product specifications are set as follows: xBuAc ) 99 mol %; xAmAc ) 99 mol %; and the xHAc in the BuAc product stream must be less than 50 ppm. The tray weir height of the distillation column is set to be 0.0508 m. TAC calculations for optimal system design are based on information provided by Douglas:25

TAC ) operating cost + (capital cost/payback period) (2) The operating cost includes the costs of heating and cooling water, and the capital cost includes the costs of the columns, trays, reboilers, and condensers. The payback period is assumed to be three years, and the wastewater treatment cost is assumed to be negligible in the overall TAC.

Table 3. Reaction Kinetics of BuOH/AmOH Mixture Systema forward reaction rate constant (T ) 402 K)

Keq (T ) 402 K)

(1) BuAc (Pseudo-homogeneous Model)

r ) mcat(k1aHAcaBuOH - k2aBuAcaH2O)

k1 ) 2.23 × 10-3 [kmol/(kgcat · s)]

9.75

k1 ) 3.3856 × 106 exp(-70660/RT), k2 ) 1.0135 × 106 exp(-74241.7/RT) (2) AmAc (Quasi-homogeneous Model)

r ) mcat(k3CHAcCBuOH - k4CBuAcCH2O)

k3 ) 5.9 × 10-6 [m3/(kmol · kgcat · s)]

k3 ) 31.1667 exp(-51740/RT) k4 ) 2.2533 exp(-45280/RT) a

R ) 8.314 (kJ/kmol/K), T (K), r (kmol/s), mcat (kgcat), Ci (kmol/m3cat). (1) Gangadwala et al.2; (2) Lee et al.24

Figure 3. (a) “Separation first” flowsheet; (b1) “direct-sequence”; and (b2) “indirect-sequence” of “reaction first” flowsheet.

2

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

There are two design variables for the distillation column: total number of stages and feed tray location. The reflux ratio and reboiler duty serve as manipulation variables to meet the BuOH and AmOH specifications at the top and bottom of the RD columns, respectively. High-purity BuOH and AmOH are required to use the optimal design proposed by Tang et al.7 for the remaining two RD columns. Therefore, the BuOH and AmOH purity levels must be 99.98 mol % at the top and bottom of the RD columns, respectively. The optimization procedure for the distillation column can be described as follows: 1. Guess a total number of stages (NT). 2. Guess the feed tray location (NF). 3. Change the reboiler duty and the reflux ratio until the two product specifications are met. 4. Go back to step 2 and change the NF until TAC is minimized. 5. Go back to step 1 and change the NT until TAC is minimized. Based on this procedure, the result of NT is 56 stages, and the minimum TAC point occurs at the NF ) 28th stage (counts from top to bottom). The values of the reflux ratio and reboiler duty are 2.63 and 1938.5 kW, respectively. The minimum TAC of this distillation is $588.1 ($1000/year). Upon combining the results of the other two RD columns from Tang et al.,7 the energy consumption of the overall “separation first” process is 6677.7 kW and the minimum TAC of the overall process is $1971.1 ($1000/year). The “separation first” design contains three columns, including one distillation column and two RD columns with decanters. Therefore, the TAC of the “separation first” design may be greater than that in the “reaction first” design, which uses only one RD column with a decanter for process esterification and one distillation column for product separation. The following section investigates the “reaction first” design. 3.2. “Reaction First” Design. Before the simulation of the “reaction first” design can be conducted, two important conditions must be considered. The first consideration is the location of the feed tray for the mixed alcohols stream in relation to the HAc feed stream. To increase contact between the reactants in the reaction zone, the feed tray should generally be located in such a way that reactants feeding to create counter-current flow inside the column determined by the boiling points. According to Table 2, the reactant boiling point is BuOH < HAc < AmOH. As described by Tang et al.,7 the BuOH and AmOH also have different feed tray locations in relation to HAc. To reduce the number of simulations, the “reaction first” design assumes that the feed tray location of the mixed alcohols is higher than the HAc. This is because the difference between the AmOH and HAc boiling points is greater than that between BuOH and HAc. Under equal molar feed conditions, the boiling point of the alcohol mixture should be higher than that of HAc. The second consideration is that there are two types of configurations in the “reaction first” process flowsheet: “directsequence” and “indirect-sequence”. The RD columns of these flowsheets are shown in Figure 3, panels b1 and b2. Table 2 clearly shows that the heaviest product, AmAc, can easily be obtained at the bottom of the RD column, and according to Figure 4 the lightest product, H2O, can be obtained from the top of decanter’s aqueous outlet. However the middle-boiling product, BuAc, has two possible directions to go. In the “directsequence” design, the BuAc is obtained together with the AmAc at the bottom of the RD column and then an additional distillation column is used to separate the two esters. In the “indirect-sequence” design, the organic stream at the top of

7195

Figure 4. RCM and LLE of BuOH-BuAc-H2O ternary system under 1 atm.

the RD decanter is fed to a subsequent distillation column, and the BuAc is obtained at the bottom of this distillation column. 3.2.1. “Direct-Sequence” Configuration. Figure 3b1 shows the configuration of the “direct-sequence” design. Since the RD column of system characteristics is similar to the type III in Tang et al.,7 the major concept of this type III system is extended with an additional distillation column. In this system, both esters are easily obtained at the bottom of the RD columns. Furthermore, Figure 4 displays a minimum boiling BuOH/BuAc/H2O azeotrope in the large two-liquid region. In this liquid-liquid region, one end of the tie-line is directed to the pure water. The overhead stream of the RD column enters the decanter, and the organic rich stream is totally refluxed back to the RD column. The bottom stream enters another distillation column to separate the two esters. The heaviest component, AmAc, is thus easily obtained at the bottom of the distillation column and the BuAc product is obtained at the top of this column. This study makes several assumptions in the following simulation: The mixed alcohols feed stream is set to equal molar. As with the HAc feed stream, the total flow rate of mixed alcohols feed stream is set at 100 kmol/h. The decanter temperature is again set at 40 °C. The BuAc and AmAc product specifications are the same as in the “separation first” design. However, the constraint of the xHAc content in the BuAc product being less than 50 ppm is lifted to simplify process design. The tray weir heights of the RD and distillation columns are set to be 0.1016 and 0.0508 m, respectively. In this system, the feed location of the alcohol mixture is assumed to be higher than that of the HAc feed stream in the reaction zone because the boiling point of the BuOH-AmOH mixture is greater than that for HAc. The seven design variables of this case are the stage numbers of the reactive zone (Nrxn), the rectifying stage numbers of RD (NR), the stripping stage numbers of RD (NS), the HAc feed location (NFHAc), the mixed alcohols feed location (NFOH), the total stage numbers of the distillation column (NT), and the feed location (NF) in the distillation column. Three manipulating variables, which include reboiler duty of both columns and the reflux ratio of the distillation column, can be used to meet product specifications. Note that there are two product specifications, but three manipulation variables. The additional manipulation variable

7196

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Figure 5. Optimal results of “direct-sequence” configuration.

is usually used for optimization. However, if the HAc composition of the RD bottom stream is greater than 0.3 mol %, the reactant residue affects product purity levels, and product specifications cannot be met in the following distillation column. Therefore, the RD reboiler duty is used to meet the HAc composition requirement of 0.3 mol % at the bottom of the RD columns. Then the reflux ratio and reboiler duty of the distillation column are used to meet the BuAc and AmAc specifications, respectively. The catalyst cost is added to the TAC objective function: TAC ) catalyst cost + operating cost + (capital cost/payback period)

(3)

The detailed calculation of individual terms in the TAC can be found in Appendix B of the Supporting Information. The optimization procedure is described in the following steps: 1. Guess Nrxn. 2. Guess NR and NS. 3. Guess NFHAc and NFOH. 4. Guess NT and NF. 5. Change the RD reboiler duty to meet the RD bottom HAc requirement of less than 0.3 mol % and change the reflux ratio and reboiler duty of the distillation column until the product specifications are met. 6. Go back to step 4 and change NT and NF until the TAC is minimized. 7. Go back to step 3 and change NFHAc and NFOH until the TAC is minimized. 8. Go back to step 2 and change NR and NS until the TAC is minimized. 9. Go back to step 1 and vary Nrxn until the TAC is minimized. Figure 5 and Table 4 show the optimal results of the directsequence configuration after completing the steps above. Figure 5 shows that the Nrxn is 27 stages, the NS is 90 stages in the RD column, and the reactant feed tray locations NFOH and NFHAc are on the first and fifth stages in the upper section of the RD column. These results are very similar to the results for the single

Table 4. Optimal TAC and Overall Reboiler Duties of Each Case reaction first case Qr (kW) Qr saving capital cost ($1000/year) operating cost ($1000/year) TAC ($1000/year) TAC saving

separation direct-sequence indirect-sequence: first (ignore HAc spec.) aqueous reflux 6677.7 1244.0 727.1 1971.1

6010.4 10% 1505.5 781.2 2286.7 -16%

4616.5 31% 945.8 597.9 1543.7 22%

BuAc or AmAc RD system. The total number of stages and the feed tray location in the distillation column are 43 stages and the 24th stage. There is no need for NR in the RD column because of the large liquid-liquid envelop (LLE) region at the top of the decanter, and further, because the organic phase stream of the decanter is totally refluxed. The overhead composition can easily fall into the LLE region, and almost pure H2O is obtained from the aqueous phase stream of the decanter because one end of the tie line is near the pure water corner. One problem with this design flowsheet is that it requires a very large NS, as Figure 5 indicates. According to Figure 6, the temperature and composition profiles display only slight changes in the stripping section. Figure 6b clearly shows that more NS is required because the HAc composition decreases very slowly. This can be explained by the closeness of the boiling point temperature between HAc and BuAc in Table 2. Figure 7 displays the reaction rate profiles in the RD column, showing that the reaction rate profile for BuAc is sharper and narrower than for AmAc. This means that the BuAc reaction is faster in these two competing reactions. Table 4 shows that the overall reboiler duty is 6010.4 kW, which represents an energy savings of almost 10% compared with the “separation first” design. However, the total TAC is $2286.7 ($1000/yr), which is 16% more than the “separation first” design. This is because there are too many stages in the RD column, which effectively negates the advantage of reduced energy consumption. This design has another weak point in that the HAc specifications for the BuAc product stream are not included. Figure 6(b) shows that the BuAc composition in the

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

7197

Figure 6. RD (a) temperature and (b) composition profiles of “direct-sequence” configuration.

stripping section of the RD column initially increases, but then decreases when approaching the bottom of the RD columns. This unwanted remixing effect at the bottom of the RD columns also makes the “direct-sequence” configuration not competitive with other design alternatives. The following section considers another type of configuration, the “indirect-sequence” design. 3.2.2. “Indirect-Sequence” Configuration with Aqueous Reflux. Figure 3b2 shows another possible configuration, the “indirect-sequence” design. Unlike the “direct-sequence” configuration, where two esters accumulate at the bottom of the RD columns, the “indirect-sequence” configuration only collects the heaviest product, AmAc, at this location. The other product, BuAc, is obtained at the bottom of another distillation column, which is fed by the organic stream from the top of the first RD decanter.

Figure 7. BuAc and AmAc reaction rate of “direct-sequence” configuration.

The configuration of partial organic reflux can be found in the design of the esterification RD process7,26 where both organic and aqueous products are drawn out from the RD system. The purpose of the organic reflux is to put ester back into the RD column, forcing water to the top by forming the ternary minimum boiling azeotrope of alcohol-ester-water.

7198

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Figure 8. Distillation column mass balance line of “indirect-sequence” configuration.

Typical applications include the ethyl acetate and isopropyl acetate RD processes, which comprise type II RD systems in Tang et al.7 In this paper, we demonstrate a similar column configuration but with different reflux policy of the RD system with a decanter. In this proposed design, we need to force BuAc to go out from the top of the RD column. Thus, putting this component back into the RD column (as organic reflux) does not make any sense. Instead, this study proposes a novel design configuration shown in Figure 3b2. In this design configuration, the role of water (considered to be a light component) is to extract the heavy component, BuAc, to the top of the RD column by forming a ternary BuOH-BuAc-H2O minimum-boiling azeotrope. In other words, this study proposes a design configuration of aqueous reflux with no organic reflux. Figure 8 shows that pure BuAc can be obtained from the bottom of the second distillation column. The top composition of this distillation column does not fall into the LLE region. Thus, an obvious choice for the overhead recycle stream is to put it back to the RD column, and not the decanter. Figure 3b2 depicts the overall design concept. For a fair comparison with the “separation first” design, the optimal design of this proposed “aqueous reflux” design must also require the impurity specification of acetic acid to be less than 50 ppm. However, this specification cannot be met at the bottom of the distillation column with an equal molar mixed alcohols feed. This is because from Figure 1b the HAc and BuAc, although form no azeotrope, are hard to separate due to the xy line very close to the 45° line at the pure BuAc end. Therefore, to make the bottom of the distillation column meet the HAc purity requirement of 50 ppm, the HAc in the RD column must be almost entirely reacted. This means that the decanter’s organic phase should contain only very small amounts of HAc when entering the distillation column. To achieve this purpose, a tiny feeding excess of alcohol must be added. Additional tiny amounts of BuOH can increase the HAc conversion rate and meet the HAc specifications. Figure 9 displays different excess BuOH feed flow rates under the

preliminary results of the simulation. Results show that a small excess of BuOH in the feed stream (50.3 kmol/h) produces the minimum TAC which also meets all product specifications. Even with this feed condition, however, there are still 10 design variables which need to be optimized. Compared with the “direct-sequence” design, additional design variables include the recycle stream feed location from the overhead of the distillation column (NFRecycle), the aqueous reflux split ratio (SR), and the distillation column reflux ratio (RR). These 10 design variables include two continuous variables of SR and RR. To simplify the optimization step, SR and RR intervals of 0.01 and 0.5, respectively, are used to find the optimization results. The AmAc purity requirement is set to 99 mol % at the bottom of the RD columns by varying the RD reboiler duty. The specifications for the bottoms of the distillation columns are set at BuAc g 99 mol % and HAc ) 50 ppm. This is because simply varying the distillation bottom flow rate can produce a BuAc composition greater than 99 mol %, with a HAc composition of 50 ppm. Notice that the reboiler duty of this column was not used as a manipulation variable to avoid the convergence problem in the simulation. The optimization procedure of indirect-sequence is shown in the following steps: 1. Guess NR. 2. Guess SR. 3. Guess NFRecycle. 4. Guess NFHAc and NFOH. 5. Guess Nrxn and NS. 6. Guess NT and NF. 7. Guess RR. 8. Change the RD reboiler duty to meet the RD bottom AmAc requirement of 99 mol %. Change the bottom flow rate of the distillation column until the HAc specification and BuAc product purity requirements are met. 9. Go back to (7) and change RR until the TAC is minimized. 10. Go back to (6) and change NT and NF until the TAC is minimized.

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

7199

Figure 9. Optimal feed stream amount of BuOH excess.

Figure 10. Optimal results of “indirect-sequence” with aqueous reflux configuration.

11. Go back to (5) and change Nrxn and NS until the TAC is minimized. 12. Go back to (4) and change NFHAc and NFOH until the TAC is minimized. 13. Go back to (3) and change NFRecycle until the TAC is minimized. 14. Go back to (2) and change SR until the TAC is minimized. 15. Go back to (1) and change NR until the TAC is minimized. Figure 10 shows the optimal results of the indirect-sequence design after running the optimization procedure. This figure shows that the minimum TAC occurs when NR is equal to 26 stages. Clearly, the rectifying section stages in the direct- and indirect-sequence configurations are very different. For the

direct-sequence configuration, only water needs to be separated out at the top of the RD column. This can easily be achieved by a decanter due to a larger LLE region, thus fewer rectifying section stages are required. In the indirect-sequence design, both water and BuAc must be drawn out from the top of the RD column. The HAc impurity in the organic phase of the decanter would accumulate at the bottom of the distillation columns, thus messing up the HAc specification in the BuAc product stream. As a result, more rectifying section stages would be required to force HAc back into the RD column. Figure 10 also shows that the total number of stages in the distillation column is much less than in the direct-sequence configuration. This is because the distillation column in the indirect-sequence configuration does not separate esters with similar boiling points. As a result, only 21 stages are required

7200

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Figure 11. RD (a) temperature and (b) composition profiles of “indirect-sequence” with aqueous reflux configuration.

for the feed location at the sixth stage. The conceptual design of the second column in Figure 8 shows that it is better to bring the top composition of the distillation column close to the distillation boundary so that less BuAc component will be recycled back to the RD column. However, as this composition approaches the distillation boundary by increasing the reflux ratio, the reboiler duty increases also. As a result, the best reflux ratio is 2.5 to achieve the minimum TAC. Figure 11 shows the temperature and composition profiles of the RD column. Compared with Figure 6a, these shape and temperature range profiles are quite different. For the indirectsequence configuration, there is an obvious temperature break at the 26th stage, and the temperature range is wider than the direct-sequence configuration. Composition profiles reveal two phenomena in the RD column. Compared with direct-sequence configuration, the remixing effect disappears toward the bottom of the RD column in this design. The other phenomenon is that double peaks of HAc composition appear at the 25th and 47th stages. The high HAc peak at the 25th stage is caused by larger stages of the rectifying section suppressing the HAc composition

in the top of the RD column, which forces the heavier HAc down to the end of this section. Below this 25th stage, the HAc reacts with the recycle stream, which contains rich BuOH component. As a result, the HAc composition decreases after this stage. The reaction rate profiles in Figure 12 also show that BuAc esterification exhibits two higher peaks near the recycle stream and the mixed alcohols feed locations. Table 4 shows that the indirect-sequence of the “reaction first” design is better than the “separation first” design. The overall savings of the reboiler duty and TAC are 31% and 22%, respectively, in comparison with the “separation first” design. For the mixed BuOH and AmOH esterification system, the “indirect-sequence” with aqueous reflux is a better process configuration than the “direct-sequence” configuration. 4. Effects of Reaction Kinetics on Relative Feed Tray Location Another interesting issue is the relative feed location of the reactants. The mixed alcohols esterification system can be

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Figure 12. BuAc and AmAc reaction rate of “indirect-sequence” with aqueous reflux configuration.

classified into two categories based on a ranking of the reactant boiling points. One is called “in series,” which means that the boiling points of both alcohols are higher or lower than HAc. The feed locations for this design are easy to determine by feeding heavier reactants at higher stages and lighter reactants at lower stages. This forms a counter-current flow that increases contact between the reactants. However, this heuristic method

7201

experiences difficulty when the HAc boiling point lies between that of the two alcohol reactants. This condition is always advantageous to one alcohol but disadvantageous to the other alcohol, no matter which relative feed location is chosen. Furthermore, there are still two competitive reactions in the system. Reaction kinetic effects should be considered in determining the relative feed location of the system. One conjecture is that the relative feed location should be determined by the esterification reaction with the lower reaction rate to promote this weaker reaction. Another conjecture is to calculate the boiling point of the mixed alcohols to determine the relative feed location of reactants. However, the boiling point of the mixed alcohols is always influenced by the ratio of the mixed alcohols. If we can understand which factors have the greatest effect on the relative feed locations of the reactants in alternating system, the half of the optimization steps can be saved. The previous section uses the boiling point of mixed alcohols to determine the relative feed location of the two reactants. Because the BuOH boiling point is very close to the HAc boiling point in the BuOH/AmOH mixing system, the feed location of the mixed alcohols was placed above HAc at equal molar BuOH/AmOH ratio. If the other conjecture was used, the feed location of the mixed alcohols should be determined by the weaker reaction. Table 3 shows that the reaction rate constant cannot be compared due to different units. However, the reaction equilibrium constant of AmOH esterification is smaller than that for BuOH (4.875 times smaller at 402 K). Because AmOH has a higher boiling point than HAc, the mixed alcohols feed location should be advantageously placed above the HAc feed

Table 5. Reaction Kinetics of BuOH/AmOH Mixture System with Equilibrium Constant Exchangea forward reaction rate constant (T ) 402 K)

Keq (T ) 402 K)

(1) BuAc (Pseudo-homogeneous Model)

r ) mcat(k1aHAcaBuOH - k2aBuAcaH2O)

k1 ) 2.23 × 10-3 [kmol/(kgcat · s)]

2

k1 ) 3.3856 × 106 exp(-70660/RT), k2 ) 4.9432 × 106 exp(-74241.7/RT) (2) AmAc (Quasi-homogeneous Model)

r ) mcat(k3CHAcCBuOH - k4CBuAcCH2O)

k3 ) 5.9 × 10-6 [m3/(kmol · kgcat · s)]

9.75

k3 ) 31.1667 exp(-51740/RT) k4 ) 0.46267 exp(-45280/RT) a

R ) 8.314 (kJ/kmol/K), T (K), r (kmol/s), mcat (kgcat), Ci (kmol/m3cat).

Table 6. Reaction Kinetics of BuOH/AmOH Mixture System with Reaction Rate Constant Changea forward reaction rate constant (T ) 402 K)

Keq (T ) 402 K)

(1) BuAc (Pseudo-homogeneous Model)

r ) mcat(k1aHAcaBuOH - k2aBuAcaH2O)

k1 ) 2.23 × 10-4 [kmol/(kgcat · s)]

9.75

k1 ) 3.3856 × 105 exp(-70660/RT), k2 ) 1.0135 × 105 exp(-74241.7/RT) (2) AmAc (Quasi-homogeneous Model)

r ) mcat(k3CHAcCBuOH - k4CBuAcCH2O) k3 ) 311.667 exp(-51740/RT) k4 ) 22.533 exp(-45280/RT) a

R ) 8.314 (kJ/kmol/K), T (K), r (kmol/s), mcat (kgcat), Ci (kmol/m3cat)

k3 ) 5.9 × 10-5 [m3/(kmol · kgcat · s)]

2

7202

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

Table 7. Results of Relative Feed Locations Comparison at Each Reaction Kinetic Condition

location. It is hard to determine whether or not reaction kinetics will affect the relative feed location or the ranking of the mixed alcohols versus HAc. For this reason, the equilibrium constants and reaction rate constants are modified to observe the effects of relative feed location under the optimal condition of minimum TAC. The following simulations are based on the “indirectsequence” configuration with aqueous reflux. The HAc specification is released in the BuAc product stream to make the product specifications of AmAc and BuAc the same (both are set at 99 mol %). Three cases are simulated using original kinetics, a hypothetical case with the equilibrium constant Keq changed and another hypothetical case with the reaction rate constant k changed. Table 5 shows how to change Keq by adjusting the pre-exponential factor. Table 6 displays a new k under the fixing Keq by a decreasing of 10 times of the k1 and k2 pre-exponential factors of the BuAc reaction and also bu an increasing of 10 times of the k3 and k4 preexponential factors of the AmAc reaction. Based on the optimization procedure of the “indirect-sequence” configuration with aqueous reflux, Table 7 shows the overall results with different kinetics. Figure 13 displays the reaction rate profiles of these three cases. Under original kinetic conditions, the mixed alcohols feed location is more advantageously placed above the HAc regardless of the reason of boiling point ranking or equilibrium constant. The first row of Table 7 also reflects the results. If the correct relative feed location is given, the TAC and Qr can decrease 12% and 15%, respectively. The second row of Table 7 displays the condition of Keq changed. Basing the relative feed location

on a comparison of the equilibrium constants increases the TAC and Qr. This result shows that a comparison of the equilibrium constants should not be used as the judgment of the relative feed location. The third row of Table 7 shows that the decision of the relative feed location should be altered when the k1 and k2 pre-exponential factors of the BuAc reaction decrease 10 times and the k3 and k4 pre-exponential factors of the AmAc reaction increase 10 times. However, the TAC and Qr are almost the same even if the relative feed location is set wrong. Comparing the reaction rate profiles of the best relative feed location in these three cases clearly reveals that the reaction rate profiles of BuAc and AmAc are similar for the original kinetics and Keq changed cases. However, Figure 13c shows that the reaction rate profiles are quite different than the profiles in Figure 13 panels a and b. This reflects that the relative reaction rates of the two BuAc and AmAc reactions are changed in comparison with the original kinetic case. The above observations of Figure 13 and Table 7 imply that the equilibrium constant is not a major factor affecting the relative feed location. The relative feed location is more significantly affected by the reaction rate constant k than the equilibrium constant Keq. Another result shown in Table 7 is that the saving of the TAC or Qr depends on both the reason of boiling point ranking and the relative reaction rates of the two reactions. When the two reasons all indicate the same relative feed location as in the original kinetic case, if the relative feed location is guessed wrong, the TAC and Qr will increase significantly. However,

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

7203

5. Conclusion This study presents a feasible design for the esterification of BuOH and AmOH mixtures with HAc, and also finds the rules of relative feed location for an alternating system. Three design configurations are explored in this work: one “separation first” design and two “reaction-first” designs. The best choice between these different configurations is based on TAC savings and energy consumption. Comparing these three cases shows that esterification by RD column directly in conjunction with the “indirect-sequence” design with aqueous reflux is the best configuration. Compared with “separation first” design, this configuration can eliminate one RD column. This configuration can also avoid the remixing effect toward the bottom of the RD columns compared with “direct-sequence” configuration. The heaviest product, AmAc, is obtained at the bottom of the RD columns and the RD overhead is fed to a decanter for water separation. Part of the aqueous phase is refluxed back to the RD column for extracting more BuAc to the top of this column. Organic phase material is fed to another distillation column to further purify the BuAc product at the bottom of this column. This proposed design can save 31% reboiler duty and reduce the TAC by 22% compared with the “separation first” design. Because the boiling point of the acid in this system lies between the two mixed alcohols (BuOH < HAc < AmOH), the relative feed location of this alternating system is also an important issue in this study. Reaction kinetics and equilibrium constants are changed to verify some conjectures. Results show that the boiling point of mixed alcohols and a comparison of the two reaction rate constants are the most important factors to be considered in determining the relative feed location of this alternating system. Acknowledgment This work is supported by the National Science Council of ROC under Grant No. NSC 95-2221-E-002-379-MY3. Supporting Information Available: Vapor pressure model; TAC calculation. This material is available free of charge via the Internet at http://pubs.acs.org. Nomenclature

Figure 13. BuAc and AmAc reaction rate of (a) original kinetics, (b) equilibrium constant exchanged, and (c) changing of the reaction rate constant.

if the boiling point ranking and the reaction kinetics contrarily indicate the relative feed location, either locations offer similar TAC or Qr.

AmAc ) amyl acetate AmOH ) n-amyl alcohol BuAc ) butyl acetate BuOH ) n-butanol HAc ) acetic acid Keq ) equilibrium constant ki ) reaction rate constant (i ) 1, 2, 3, 4) LLE ) liquid-liquid envelope NT ) total number of stages NF ) feed tray location NFHAc ) the HAc feed location NFOH ) the mixed alcohols feed location NFRecycle ) recycle stream feed location Nrxn ) stage numbers of the reactive zone NR ) rectifying stage numbers of RD NS ) stripping stage numbers of RD Qr ) reboiler duty (kW) r ) reaction rate (kmol/s) RCM ) residual curve map RD ) reactive distillation RR ) distillation column reflux ratio SR ) aqueous reflux split ratio

7204

Ind. Eng. Chem. Res., Vol. 48, No. 15, 2009

TAC ) total annual cost xi ) component i mole fraction (i ) HAc, BuOH, AmOH, BuAc, AmAc, H2O)

Literature Cited (1) Steinigeweg, S.; Gmehling, J. n-Butyl Acetate Synthesis via Reactive Distillation: Thermodynamic Aspects, Reaction Kinetics, Pilot-Plant Experiments, and Simulation Studies. Ind. Eng. Chem. Res. 2002, 41, 5483–5490. (2) Gangadwala, J.; Mankar, S.; Mahajani, S.; Kienle, A.; Stein, E. Esterification of Acetic Acid with Butanol in the Presence of Ion-Exchange Resins as Catalysts. Ind. Eng. Chem. Res. 2003, 42, 2146–2155. (3) Gangadwala, J.; Kienle, A.; Stein, E.; Mahajani, S. Production of Butyl Acetate by Catalytic Distillation: Process Design Studies. Ind. Eng. Chem. Res. 2004, 43, 136–143. (4) Singh, A.; Hiwale, R.; Mahajani, S. M.; Gudi, R. D.; Gangadwala, J.; Kienle, A. Production of Butyl Acetate by Catalytic Distillation. Theoretical and Experimental Studies. Ind. Eng. Chem. Res. 2005, 44, 3042– 3052. (5) Gangadwala, J.; Kienle, A. MINLP Optimization of Butyl Acetate Synthesis. Chem. Eng. Proc. 2007, 46, 107–118. (6) Chiang, S. F.; Kuo, C. L.; Yu, C. C.; Wong, D. S. H. Design Alternatives for the Amyl Acetate Process: Coupled Reactor/Column and Reactive Distillation. Ind. Eng. Chem. Res. 2002, 41, 3233–3246. (7) Tang, Y. T.; Chen, Y. W.; Huang, H. P.; Yu, C. C.; Hung, S. B.; Lee, M. J. Design of Reactive Distillations for Acetic Acid Esterification. AIChE J. 2005, 51, 1683–1699. (8) Thurman, L. R.; Harris, J. B.; McAtee, M. R. Co-production of Low Molecular Weight Esters of Alkanols, Patent No. EP 0260572 A1, 1987, BASF Corporation. (9) Cooke, N. C.; Yeomans, B. Co-production of Ethyl Acetate and n-Butyl Acetate. Patent No. 1742/73, 1975, BP Chemicals Limited. (10) Van Acker, P. E.; Mathieu, O.; Milner, R. J.; Pacynko, W. F., Ester Co-production. Patent No. WO 98/42652, 1998, BP Chemicals Limited. (11) Schulz, E.; Baucer, H.; Merscher, D. Coupled Production of Two Esters. Patent No. US 7,115,773 B2, 2006, Celanese Chemicals Europe GmbH. (12) Venimadhavan, G.; Malone, M. F.; Doherty, M. F. A Novel Distillate Policy for Batch Reactive Distillation with Application to the Production of Butyl Acetate. Ind. Eng. Chem. Res. 1999, 38, 714. (13) Hayden, J. G.; O’Connell, J. P. A Generalized Method for Predicting Second Virial Coefficients. Ind. Eng. Chem. Process Des. DeV. 1975, 14, 209–216.

(14) Gmehling, J.; and Onken U.; Arlt, W. Aqueous-Organic System (Supplement 1). Vapor-Liquid Equilibrium Data Collection; Chemistry Data Series, Vol. 1; DECHEMA: Frankfurt/Main, Germany, 1981; Part 1a, p 383. (15) Gmehling, J.; and Onken U.; Arlt, W. Organic Hydroxy Compounds: Alcohols and Phenols. Vapor-Liquid Equilibrium Data Collection; Chemistry Data Series, Vol. 1; DECHEMA: Frankfurt/Main, Germany, 1990; Part 2b, pp 138, 194-197. (16) Gmehling, J.; and Onken U.; Weidlich, U. Organic Hydroxy Compounds: Alcohols and Phenols (Supplement 2). Vapor-Liquid Equilibrium Data Collection; Chemistry Data Series, Vol. 1; DECHEMA: Frankfurt/Main, Germany, 1982; Part 2d, p 432. (17) Gmehling, J.; and Onken U.; Grenzheuser, P. Carboxylic Acids, Anhydrides, Esters. Vapor-Liquid Equilibrium Data Collection; Chemistry Data Series, Vol. 1; DECHEMA: Frankfurt/Main, Germany, 1982; Part 5, pp 147-151. (18) Gmehling, J.; Menke, J.; Krafczyk, J.; Fisher K. Azeotropic Data, 2nd ed.; Wiley-VCH: Weinheim, Germany, 2004. (19) Koichi, I.; Hitoshi, K. Isobaric Vapor-Liquid-Liquid Equilibria with A Newly Developed Still. Fluid Phase Equilib. 2001, 192, 171–186. (20) Cho, T. H.; Ochi, K.; Kojima, K. Measurement of Vapor-Liquid Equilibrium for System with Limited Miscibility. Fluid Phase Equilib. 1983, 11, 137–152. (21) Ruiz Bevia, F.; Prats Rico, D.; Gomis Yagu¨es, V.; Varo Galvan˜, P. Quaternary Liquid-Liquid Equlibrium: Water-Acetic Acid-1-Butanoln-Butyl Acetate at 25°C. Fluid Phase Equilib. 1984, 18, 171–183. (22) Tan, T. C.; Aravinth, S. Liquid-Liquid Equilibria of Water/Acetic Acid/1-Butanol SystemsEffects of Sodium (Potassium) Chloride and Correlations. Fluid Phase Equilib. 1999, 163, 243–257. (23) Esquı´el, M. M.; Bernardo-Gil, M. G. Liquid-Liquid Equilibria for Systems: Water/1-Pentanol/Acetic Acid and Water/1-Hexanol/Acetic Acid. Fluid Phase Equilib. 1991, 62, 97–107. (24) Lee, M. J.; Wu, H. T.; Kang, C. H.; Lin, H. M. Kinetic Behavior of Amyl Acetate Synthesis Catalyzed by Acidic Cation Exchange Resin. J. Chin. Inst. Chem. Eng. 1999, 30, 117–122. (25) Douglas, J. M. Conceptual Design of Chemical Process; McGrawHill: New York, 1988. (26) Tang, Y. T.; Huang, H. P.; Chien, I. L. Design of a Complete Ethyl Acetate Reactive Distillation System. J. Chem. Eng. Jpn. 2003, 36, 1352– 1363.

ReceiVed for reView December 9, 2008 ReVised manuscript receiVed June 1, 2009 Accepted June 12, 2009 IE801891Q