Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

Contents lists available at ScienceDirect

Journal of the Taiwan Institute of Chemical Engineers

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

Control of plantwide reactive distillation processes: Hydrolysis, transesterication and two-stage esterication
Shih-Bo Hung a, Jyun-Hong Chen b, Yu-Der Lin b, Hsiao-Ping Huang b, Ming-Jer Lee a, Jeffrey D. Ward b,*, Cheng-Ching Yu b
a b

Department of Chem. Eng., National Taiwan University of Science and Technology, Taipei 10607, Taiwan Department of Chem. Eng., National Taiwan University, Taipei 10617, Taiwan

A R T I C L E I N F O

A B S T R A C T

Article history: Received 8 January 2010 Received in revised form 25 March 2010 Accepted 26 March 2010 The authors dedicate this manuscript to C.C. Yu, who was the principal investigator on this project. Keywords: Reactive distillation Plantwide control Hydrolysis Transesterication Adipic acid

General design principles for controlling plantwide reactive distillation processes are developed in this work. Three cases of plantwide RD processes with large recycle ow rates are used to illustrate the application of the design principals. The cases are hydrolysis of methyl acetate, transesterication of methyl acetate and esterication of adipic acid. The design principles produce workable control structures for the three systems. Temperature and temperature-composition control structures can deal with the two major disturbances: changes in the feed ow rates and reactant composition. Results show that the control structures offer good robustness and acceptable dynamic performances. 2010 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction Reactive distillation (RD) combines reaction and separation in a single unit which provides substantial economic benet for some chemical processes. The coupling of both functions in a single unit makes the system highly nonlinear. Therefore, the operation and control of reactive distillation columns must be addressed at the design stage for successful implementation of the technology. A robust control strategy that ensures safe, stable and economic operation in the presence of disturbances is essential. Many articles have appeared in the literature on the control of RD columns in last decade. Most RD design is based on neat operation, with no excess reactant fed to the RD column, because this conguration usually has greater economic potential. However even in this case, the control issue is more complex than for a traditional distillation column. A main issue is that reactants must be fed in the exact amount to satisfy the stoichiometry down to the last molecule. Luyben and co-workers propose eight control structures for neat reactive distillation with a reversible reaction with two reactants and two products: (CS1CS6 in Al-Arfaj and Luyben (2000); CS7 in Al-Arfaj and Luyben (2002); CS7CS8 in

* Corresponding author. Fax: +886 2 3366 3037. E-mail address: jeffward@ntu.edu.tw (J.D. Ward).

Kaymak and Luyben (2004, 2005)). In general, basic concepts for a RDC with a second order reversible reaction (i.e., A + B $ C + D) are investigated. Other authors have also studied the control of RD columns. Sneesby et al. (1999) study the two-point control of an ethylene glycol column. Linear and nonlinear control of semi-batch reactive distillation for ethyl acetate production has been explored by Engell and Fernholz (2003). The quantication of nonlinearity measurements and the relationships between nonlinearity and control have been investigated by Hung et al. (2006b). Neat operation is not suitable for all RD system designs. Sometimes excess reactant is needed to achieve high conversion of another species and avoid byproducts, for example, Fuchigami (1990) and Subawalla and Fair (1999). Thus control systems sometimes must be developed for an entire plant consisting of several interconnected unit operations including RD. Studies of the plantwide control of processes with reactive distillation columns and separation columns with recycle is less common compared to the control of a reactive distillation column alone or plantwide reactor/ separator systems (Luyben et al., 1998; Wu and Yu, 1996). Al-Arfaj and Luyben (2004) develop an effective control structure for the TAME process (one RD with two columns) using temperature control. Tang et al. (2005) presented control schemes that are workable for the ethyl acetate RD owsheet but the two reactant feeds are still close to neat in his work. Recently, Hung et al. (2008) discussed the steady-state design of a reactive distillation owsheet

1876-1070/$ see front matter 2010 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.jtice.2010.03.021

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

383

Nomenclature AA adipic acid bottom ow rates in ith column Bi BuAc butyl acetate Di top ow rates in ith column DMA dimethyl adipate EA esterication of adipic acid F feed ow rates FR feed ratio in reactive distillation column HAc acetic acid HM hydrolysis of methyl acetate Kc controller gain equilibrium constant Keq Ku ultimate gain mcat catalyst weight MeAc methyl acetate MeOH methanol MMA monomethyl adipate Nr number of rectifying trays Nrxn number of reaction trays Ns number of stripping trays NFAzeotrope azeotrope feed tray location NFBuOH butanol feed tray location NFRecycle recycle stream feed tray location NT total number of trays ultimate period Pu QRi heat duty in ith column RDC reactive distillation column RF recycle ow rate RRi reux ratio in ith column Ti,j temperature of component j in ith column TAC total annual cost TM transesterication of methyl acetate XBi,j bottom composition of component j in ith column XDi,j top composition of component j in ith column tI controller reset time
Table 1 Kinetic equations for three esterication systems.a.
System HM Kinetic model (catalyst) Adsorption-based model (Amberlyst 15) k f a0MeAc a0H O kr a0HAc a0MeOH Ka 2 R mcat ; a0i i i Mi a0MeAc a0H O a0HAc a0MeOH 2
2

with large excess recycle stream for esterication of adipic acid, but the authors did not discuss the control aspects. Lin et al. (2008) gave the optimal reactive distillation plant for hydrolysis of methyl acetate and gave a workable temperature control conguration. In this work, we investigate a new RD owsheet for transesterication of methyl acetate rst. Then we develop control structures for the three RD owsheets: the new transesterication owsheet, the esterication of adipic acid by Hung et al. (2008) and the hydrolysis of methyl acetate by Lin et al. (2008). The objective of this study is to examine and compare the control of three plantwide RD systems and also try to provide generic and effective control structure design procedures for such processes. The transesterication and hydrolysis processes are especially useful for the recovery of methyl acetate from a poly-vinyl alcohol (PVA) plant. The PVA process produces a byproduct stream which is a mixture of methyl acetate and methanol near the azeotropic composition. Methyl acetate has a low economic value, and therefore it is desirable to upgrade the methyl acetate to a more valuable product while at the same time separating the methanol. The transesterication process converts methyl acetate into butyl acetate, and the hydrolysis process converts methyl acetate into acetic acid. In Section 2, the characteristics of the three systems are introduced and the process owsheets are shown. Also, the detailed steady-state design results for the TM owsheet are shown in this section. The principals of control structure design developed in this work are given in Section 3. Because on-line temperature measurement is easy, temperature control structures will be applied to the three systems rst. Later we discuss modied control structures for the case where composition measurement is possible. Modied control structures are also discussed if high purity products are required. Conclusions are given in the last section. In this work, dynamic simulation is carried out using AspenPlus1 and AspenDynamics1. 2. System characteristics Reaction kinetics and thermodynamic properties are important issues for the design of a RD process. If the system characteristics are not known, feasible and economic design is much more difcult.

Equilibrium constant (T = 323 K) Keq = 0.01

k f 6:127 105 exp63730=RT

TM

K MeAc 4:15; K H2 O 5:24; K HAc 3:15; K MeOH 5:64 kr 8:498 106 exp60470=RT Pseudo-homogeneous model (Amberlyst 15)   C MeOH C BuAc r k f C MeAc C BuOH K1 eq k f 2:7 106 exp71960=RT kr 3:64 106 exp72670=RT Quasi-homogeneous model (Amberlyst 35) dC AA k1 C AA C MeOH k1 C MMA C H2 O dt dC MeOH k1 C AA C MeOH k1 C MMA C H2 O k2 C MMA C MeOH k2 C DMA C H2 O dt dC MMA 6 k1 C AA C MeOH k1 C MMA C H2 O k2 C MMA C MeOH k2 C DMA C H2 O k1 5:857 10 exp4097:8=T dt dC H2 O k1 C AA C MeOH k1 C MMA C H2 O k2 C MMA C MeOH k2 C DMA C H2 O dt dC DMA k2 C MMA C MeOH k2 C DMA C H2 O dt k2 2:024 106 exp4201:1=T K 1eq k1 =k1 K 2eq k2 =k2

Keq = 0.97

EA

K1eq = 0.997b, K2eq = 2.56b

a b

pken et al. (2000), (ii) Jimenez and Costa-Lopez (2002) and (iii) Chan et al. (2010). R = 8.314 [kJ/kmol/K], T [K], r [kmol/s], mcat [kgcat], Ci [kmol/m3], Mi [kg/kmol]. (i) Po Equilibrium constants are temperature independent.

384

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

2.1. Reaction kinetics Both the hydrolysis and transesterication of MeAc are onestage reversible reactions. Reactive distillation gives clear advantages for such systems (Kaymak and Luyben, 2004). The hydrolysis of MeAc produces methanol (MeOH) and acetic acid (HAc) and can be described by the following expression (Lin et al., 2008): MeAc H2 O $ HAc MeOH (1)

The transesterication reaction studied in this work occurs between MeAc and butanol (BuOH) to form butyl acetate (BuAc) and methanol (MeOH). The reaction can be written: MeAc BuOH $ BuAc MeOH (2)

point in both ternary residue curve maps. Therefore, it is difcult to consume all of the reactants MeAc and BuOH using neat operation. Excess H2O and MeAc are required in the HM and TM systems before the reaction zone in RDC. Thus, the recycle streams make the operation much easier for the HM and TM systems. Another feature of the EA system is revealed in Fig. 1(c), which shows the vapor-liquid equilibrium of the two reactants (AA and MeOH). The relative volatility difference is quite large and MeOH is more volatile. Since the reaction is carried out in the liquid phase, a very large excess of MeOH is need. Also additional units are needed

The third process is somewhat different in that a two-stage reaction occurs, that is the esterication of adipic acid (AA) and methanol (MeOH) proceeds by two reversible cascade reactions with adipic acid monomethyl ester (monomethyl adipate, MMA) as an intermediate (Hung et al., 2008): AA MeOH $ MMA H2 O MMA MeOH $ DMA H2 O (3)

The catalysts used are solid acidic ion-exchange resins so that the reactive zone can be placed in different sections of the reactive distillation columns. In applying the reaction kinetics to process design, it is assumed that the solid catalyst occupies 50% of the tray holdup volume and the catalyst density (Amberlyst 15: 770 kg/m3, Amberlyst 35: 800 kg/m3) is used to convert the volume into catalyst weight (mcat). All the kinetic units are converted to those that are acceptable to AspenPlus1. Table 1 lists the reaction kinetics used in this study. If we calculate the equilibrium constants at 323 K for each reaction, note that the equilibrium constants in the EA process are independent of temperature, we nd that for the hydrolysis reaction (HM) Keq = 0.01, for the transesterication(TM) Keq = 0.97 and for the two-stage esterication reaction (EA) are K1eq = 0.997 and K2eq = 2.56. The hydrolysis of MeAc is difcult because the equilibrium constant is low. Neat operation for this kind of reaction is unfavorable and water should be the excess reactant based on the adsorption-based kinetics (Table 1). The other two systems seem to be suitable for neat operation based on the equilibrium constants, but the phase equilibrium behaviors should also be considered. 2.2. Phase equilibria To account for non-ideal vapor-liquid equilibrium (VLE) and possible vapor-liquid-liquid equilibrium (VLLE) for these systems, the NRTL (Renon and Prausnitz, 1968) model or UNIQUAC (Abrams and Prausnitz, 1975) model is used to calculate the activity coefcients. We also take the vapor phase dimerization (for acetic acid) into account using the Hayden-OConell second virial coefcient (Hayden and OConnell, 1975) model with the Aspen Plus built-in model parameters. The UNIQUAC model parameters pken et al. (2000) for the HM system; the NRTL are taken from Po model parameters are taken from Aspen Plus built-in values and Hung et al. (2008) for the TM and EA systems, respectively. Table 2 lists the azeotropes and the ranking of azeotropic temperatures (including pure component NBP temperatures). In theory, for the HM system, if all the light reactant (MeAc) is consumed, the lightest remaining pure component (MeOH) can be obtained (Tung and Yu, 2007). However, the product, MeOH, forms an azeotrope MeAc/MeOH at 53.7 8C (Table 2). Thus, MeAc may be brought out in the light ends before it consumed. Similarly, for the TM system the product, BuAc, forms an azeotrope BuOH/BuAc at 116.9 8C (Table 2). Fig. 1(a) and (b) shows that the azeotrope is a saddle

Fig. 1. Residue curve maps and phase diagrams for the three systems.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402 Table 2 Ranking of azeotropic temperatures and pure component normal boiling point temperatures.a. FS1 MeOH/MeAc (0.3407, 0.6593) 53.65 8C FS2 MeOH/MeAc (0.3407, 0.6593) 53.65 8C MeAc 57.05 8C FS3 MeOH 64.53 8C

385

MeAc/H2O (0.8904, 0.1096) 56.43 8C

DMA/H2O (0.0105,0.9895) 99.77 8C H2O 100.02 8C DMA 235.68 8C

MeAc 57.05 8C MeOH 64.53 8C

MeOH 64.53 8C BuOH/BuAc (0.7780, 0.222) 116.9 8C BuOH 117.68 8C BuAc 126.01 8C

H2O 100.02 8C HAc 118.01 8C

a

MMA 261.67 8C AA 337.47 8C

Composition: mole fraction.

for recovery of unreacted MeOH and separation of the products (DMA). 2.3. Process owsheets The plantwide owsheets discussed in this work were optimized to minimize total annual cost (TAC). All the related design parameters and manipulated variables including tray numbers, feed tray locations, recycle ow rates and heat inputs are set to the optimal values. Design of the EA and HM systems has published by Hung et al. (2008) and Lin et al. (2008), respectively. The detailed steady-state design of the TM system is discussed next. The remainder of this section describes the other two systems briey. 2.3.1. Design of TM owsheet The entire process consists of one reactive distillation column and one distillation column with one recycle stream (Fig. 2). The transesterication reaction takes place in the RD column with

partial reux operation. There are two streams fed into the RD column: fresh butanol (50 kmol/h) and a nearly azeotropic composition of methanol and methyl acetate stream from the top of the distillation column. The butanol (heavy key) stream is fed into the top tray of the reactive zone while the other stream (light key) is fed into the lowest tray in the reactive zone. The distillation column also has two feed streams: a fresh feed with a composition close to the binary azeotrope, i.e., 60 mol% methyl acetate and 40 mol% methanol with ow rate 83.33 kmol/h and a recycle stream from the overhead in the RD column. The methyl acetate/methanol feed stream is representative of a byproduct stream from a PVA process as discussed in the introduction. The following design specications are made for the reactive distillation column: a 5-min residence time is assumed in the liquid receiver and half of the holdup volume is packed with catalyst. For reactive trays, we assume that the catalyst occupies half of the tray holdup volume. The tray holdup is determined by the column diameter which is sized using the Tray Sizing Utility in Aspen Plus by assuming a weir height of 10 cm. The concentration of butyl acetate is set to 99 mol% by adjusting the reboiler duty. The top stream of the RD column, mostly methanol and methyl acetate, is fed to the distillation column. The distillation column bottom product is specied to be 99% methanol. The concentration of butyl acetate at the top of RD column is set to be less than 0.1 mol%. Once the conceptual design is completed and specications are given, we can proceed with the preliminary design. The objective is to minimize the total annual cost (TAC) by adjusting the design parameters, e.g., tray numbers in each section, feed location in the column, etc. The TAC is dened as (Douglas, 1988): TAC operating cost capital cost payback period (4)

Here, a payback period of 3 years is used. The operating cost includes the costs of steam, cooling water, and catalyst. The capital cost includes the cost of the column, trays, and heat exchangers. Cost models and corresponding values are the same as those given in Lin et al. (2008) and a catalyst life of 3 months is assumed. In the owsheet, key design parameters are shown in italics in Fig. 2. They are: the number of rectifying, reactive and stripping trays (Nr, Nrxn and Ns), butanol and azeotrope feed tray location (NFBuOH and NFAzeotrope) of the RD column, the total number of trays and feed tray location of the 1st column (NT, NFRecycle and NFMeAc/MeOH). In

Fig. 2. Process owsheet of MeAc transesterication system and design parameters indicated in italics.

386

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

addition to tray number and feed locations in each column, there is still one important design variable (Yi and Luyben, 1997): the recycle ow rate (RF). As the ow rate RF changes, the ow rate from distillation column to the RD column will also change. Thus, the ow rate RF is directly related to the degree of excess methyl acetate ow into the RD column. We have identied 9 design variables above, and a systematic design procedure is devised for the owsheet generation (Chiang et al., 2002; Tang et al., 2005). All the simulations are carried out in Aspen Plus using the RADFRAC module with a FORTRAN module with subroutines for the activity-based reaction kinetics. Given the production rate and product specications, the design steps are (1) Guess the recycle ow rate (RF). (2) Guess a number of trays in the distillation column (NT).

(3) Adjust the recycle feed (NFRecycle) until the TAC is minimized. (4) Adjust the methanol and methyl acetate reactant feed (NFMeAc/ MeOH) until the TAC is minimized. (5) Go back to (2) and change NT until the TAC is minimized. (6) Guess a number of reactive trays (Nrxn). (7) Fix the butanol reactant feed tray (NFBuOH) into the top of reactive zone, and the azeotrope feed tray (NFAzeotrope) in the bottom of reactive zone. (8) Adjust the number of stages in the rectifying zone (Nr) and stripping zone (Ns) until the TAC is minimized. (9) Adjust the feed position (NFAzeotrope) until the TAC is minimized.

Fig. 3. Effects of design variables on Total TAC: (A) number of total trays, (B) feed tray locations of recycle stream, and (C) feed tray locations of MeAc/MeOH stream in 1st column.

Fig. 4. Effects of design variables on Total TAC: (A) number of trays in the stripping section, (B) number of trays in the reactive section, and (C) feed tray locations of azeotrope stream in RD column.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

387

(10) Go back to (6) change a number a reactive trays (Nrxn) until the TAC is minimized. (11) Go back to (1) change the recycle ow rate until the TAC is minimized. The procedure is set up in such a way (i.e., xed specications for all product streams) that the design is decoupled. For example, steps (2)(5), (6)(10) are the design steps for the distillation column and RD column, respectively, given a recycle ow rate. 2.3.2. Design results For the distillation column, Fig. 3 shows the effect of total number of trays and feed tray locations in the separation section (1st column) on the TAC. Because the recycle stream has a higher
Fig. 6. Composition prole in the RD column.

methanol concentration than the azeotropic fresh feed (i.e., 40 mol% methanol and 60 mol% methyl acetate), the recycle stream feed tray location is lower than the fresh feed location. The 1st column has 14 trays with NFRecycle = 5 and NFMeAc/MeOH = 12. For the RD column, Fig. 4 shows it has Nrxn = 35, Nr = 5, Ns = 8 and NFAzeotrope = 12. Note that the tray number is counted from the bottom up. For the recycle ow rate (FR), Fig. 5 shows a minimum in TAC occurs when the recycle ow take the value 112.45 kmol/h (about 45% MeAc in excess). The tradeoff comes from the RD cost and separation column cost. The distillation column cost increases as the recycle ow increases because the column has a higher ow rate which requires more energy (i.e., steam and cooling water). In the RD column, the minimum TAC occurs when there is 55% excess methyl acetate. Fig. 6 presents the composition prole for different
Table 3 Steady-state operating parameters and total annual cost (TAC) for MeAc transesterication process. System Total no. of trays No. of trays in reactive section (Nrxn) No. of trays in rectifying section (Nr) No. of trays in stripping section (Ns) Reactive trays BuOH feed tray (NFBuOH) Azeotrope feed tray (NFAzetrope) MeAc/MeOH feed tray (NFMeAc/MeOH) Recycle ow tray (NFRecycle) Catalyst in each tray/sum (m3) Molar feed ratio (azeotrope/butanol) Top product ow rate (kmol/h) m.f. of methanol m.f. of methyl acetate m.f. of Butanol m.f. of butyl acetate Bottom product ow rate (kmol/h) m.f. of methanol m.f. of methyl acetate m.f. of butanol m.f. of butyl acetate Condenser duty (KW) Reboiler duty (KW) Column diameter (m) Condenser heat transfer area (m2) Reboiler heat transfer area (m2) TAC of RD column ($1000/year) Total capital cost ($1000/year) Column Trays Heat exchanger Total operating cost ($1000/year) Catalyst Energy TAC ($1000/year) (50 kmol/h) RD column 48 35 5 8 943 12 43 1st column 14

12 5 0.1257/4.3995 2.243 112.5 0.7953 0.2020 0.0004 0.0023 49.70 14 10 106 0.0100 0.9900 2929 2943 1.89 155.74 80.34 819.0 688.0 300.5 61.7 325.8 451.5 52.4 399.1 1139.4

112.2 0.3565 0.6435 1015 1012 83.63 0.9900 0.0064 0.0005 0.0031 1780 1728 1.16 114.67 57.16 320.0

Fig. 5. Effects of Excess MeAc (%) on: (A) Total TAC, (B) TAC of 1st column, and (C) TAC of RD column.

388

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

Fig. 7. Optimized process owsheet for the transesterication of methyl acetate(TM) system.

degrees of methyl acetate excess (i.e., 30% and 55%) in the RD column. It is desirable from a kinetic point of view to have the mole fraction of the two reactants as close together as possible (i.e., the ideal case is 50% MeAc and 50% BuOH). From Fig. 6, the greater excess of MeAc is favorable from a kinetics point of view because it

results in the concentration of the two reactants being closer in the reactive zone. But if the RD column has more excess methyl acetate, energy cost will rise gradually. There is tradeoff between energy cost and reaction rate as the excess of methyl acetate is adjusted. In summary, for plantwide design of the transesterication plant, one dominant design variables is identied, FR. The optimized owsheet shown in Fig. 7 can be obtained by carefully adjusting the design parameters. Table 3 summarizes design parameters and corresponding costs. 2.3.3. Discussion The composition proles of the RD column and the distillation column are shown in Figs. 8 and 9, respectively. Column 1 separates methanol (withdrawn from bottom) and methyl acetate (withdrawn from overhead). 99 mol% purity methanol can be obtained in the bottom. In the RD column, butanol and methyl acetate react to produce butyl acetate and methanol and 99 mol% purity butyl acetate can be obtained in the bottom. The optimized design can also be compared with other proposed ow sheets for the transesterication of methyl

Fig. 8. Composition prole in the RD column with: (A) 30% excess MeAc and (B) 55% excess MeAc.

Fig. 9. Composition prole in the 1st column.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

389

Fig. 10. Plantwide reactive distillation process for hydrolysis of methyl acetate (HM).

acetate (see Appendix A). The proposed process is almost 50% less expensive than a conventional process and 36% less expensive than a proposed 3 column process. 2.4. HM owsheet The optimized owsheet for the HM system is shown in Fig. 10. A reactive distillation column is under total reux operation with the reactive zone placed in the upper section of the column, including a reactive reux drum. The entire process consists of one reactive distillation column and two distillation columns with one recycle stream. The hydrolysis reaction takes

place in the RD column with total reux operation. There are three feeds into the RD column: fresh water feed (50 kmol/h), a water-rich recycle stream from bottom of the 3rd column, and the fresh feed with an azeotropic composition which consists of 60 mol% methyl acetate and 40 mol% methanol (83.33 kmol/h). Again, the methyl acetate/methanol feed stream is representative of a byproduct stream from a PVA process as discussed in the introduction. The rst two feeds, mostly the heavy reactant (water), are fed into the reux drum. The third feed stream is the light reactant (MeOH) which is fed to the lower section of the RD column. The acetic acid product has the highest boiling point and is withdrawn form the bottom of the 2nd column. Finally, high

Fig. 11. Plantwide reactive distillation process for production of dimethyl adipate (EA).

390

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

purity methanol product is obtained at the top of methanol product column (the 3rd column). 2.5. EA owsheet The optimized owsheet of the EA system is shown in Fig. 11. The 1st column, a reactive distillation column, has the reactive

zone in the lower section, including a total reactive reboiler. The esterication reaction takes place in the RD column with a xed reux ratio. Two fresh reactant streams (AA: 10 kmol/h, MeOH: 20 kmol/h) are mixed with the recycle stream from the stripper and feed into the bottom of RDC. A large excess of MeOH is needed, thus the ratio of two reactant ow rates (AA and MeOH) is about 15.4. The stripper is used to remove the unreacted methanol so a

Fig. 12. Temperature control structures for (a) TM owsheet, (b) HM owsheet and (c) EA owsheet.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

391

methanol-rich stream from the top is recycled back. Because the remaining organic species and water (MMA/DMA/H2O) have a two liquid phase region (Hung et al., 2006a), a decanter can be used to remove most of the water before the bottom product from the stripper enters the 3rd column. High purity DMA product is obtained from the bottom the 3rd column. 3. Plantwide control structure design In this section, a systematic approach is proposed for the control structure design for these types of reactive distillation owsheets. 3.1. Design principles As pointed out by Al-Arfaj and Luyben (2000), the two fresh feeds cannot be adjusted using simple ratio control. One of the feeds should be under feedback control to maintain stoichiometric balance. Following this principle, workable control structures are developed for the three plants. We use the following steps: (1) Establish control objectives: typical control objectives include: product purity, level and possibly temperature constraints in the reactive distillation column. (2) Determine control degrees of freedom: count the control valves and all possible actuators. (3) Use the limiting reactant fresh feed ow rate to set the production rate.

(4) Ratio the recycle ow of the excess reactant to the limiting reactant fresh feed. This implies the composition of excess reactant in the RD column is maintained as production rate varies. (5) Maintain stoichiometric balance via feedback control by manipulating the fresh feed of the non-limiting reactant or the ratio of the fresh feeds. (6) Start with temperature control. Tune the controllers of the columns in the recycle loop in a sequential manner, starting from the rst column after the RD column, moving down stream away from the RD column, and nally returning to the RD column and tuning its controllers last. As the procedure progresses, individual control loops should be closed (on automatic control) sequentially and one may iterate the steps until the tuning parameters converge. (7) Optimize control objectives. Check the dynamic performance. If necessary, consider alternatives, such as robust control or composition measurement.

Fig. 13. Sensitivity tests for TM owsheet.

Fig. 14. Sensitivity tests for EA owsheet.

392

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

Step (1) and Step (2) are common rules that apply for all chemical processes. Steps (3)(5) refer to control principals of reactor/ separator systems that are mentioned by Luyben et al. (1998). Step (6) is the key point in this work. The purpose of starting with the column down stream from the reactive distillation is to mitigate the impacts of the nonlinear nature of RD columns. If the controllers on the other units are tuned rst, the RDC controllers will be easier to tune because disturbances recycled from the other units will be attenuated by the controllers in those units. Any feeds to RD column may cause unexpected change of the column if the feeds temperature or compositions changed.

3.2. Temperature control structures The benets of temperature control are that temperature can be measured quickly and reliably on-line without an expensive instrument. In this section, control structures based on temperature control are developed using the principals discussed previously. 3.2.1. Step 1. Establish control objectives The control objective is to maintain the temperature prole of each column. Because of the limitation of degrees of freedom, it is

Fig. 15. Temperature control performance for the TM process. (a) 20% production rate changes and (b) 5% (mol%) acetate feed composition changes.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

393

impossible to control every tray temperature. The engineer should choose the most temperature sensitive tray to be controlled. 3.2.2. Step 2. Determine control degrees of freedom I. TM process: The liquid levels in the reux drum and column bottom in the 1st column are regulated by the distillate rate and bottom rate, respectively. The 1st column has two degrees of freedom: reux rate and heat input. For the 2nd column, the bottom liquid level is controlled by manipulating the bottom rate. The degree of freedom is the heat input. The distillate is the

recycled stream and it has to be controlled properly to avoid recycled mass accumulation. Two manipulated variables, reux and distillate rate, also should be set as discussed later in step 5. II. HM process: The 1st column is set to total reux, thus heat input is the only degree of freedom for this column. For the 2nd column, reux rate is controlled by a ratio controller and distillate rate is used to control the liquid level in the reux drum. The bottom heat input can be manipulated, so this column has two degrees of freedom. For the 3rd column, the reux drum level is regulated by the distillate rate, thus the reux rate is one manipulated variable. Another two degrees of

Fig. 16. Temperature control performance for the EA process. (a) 20% production rate changes and (b) 5% and 10% (mol%) alcohol feed composition changes.

394

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

freedom are heat input and bottom product ow rate. Again, the bottom stream is recycled back to 1st column; the bottom liquid level cannot simply be controlled by manipulating the bottom stream ow rate. This ow rate should be set properly as discussed later in step 5. III. EA process: In the 1st column (RDC), there is no outlet from the column bottom and the liquid level in the bottom is controlled by the heat input. There remains one degree of freedom, the reux ratio, that can be used to control one tray temperature in the RDC. The two down stream columns are one stripper and a column with a decanter installed. Heat duty is the only degree of freedom for the 2nd column. The 3rd column is not included in the recycle loop. Heat input is the only degree of freedom for that column as well. The bottom liquid level and two-phase liquid levels in the decanter are all regulated by their exit streams. 3.2.3. Step 3 The limiting reactant fresh feeds are a stream with the (MeAc/ MeOH) azeotropic composition for HM and EM processes and pure adipic acid feed for the EA process. 3.3. Step 4 and 5. Recycle stream and stoichiometric balance Fig. 12(a) shows the control congurations for the TM process. The fresh butanol feed ow rate is as set a ratio of the distillate recycle from the 1st column, and cascade control is implemented to control a tray temperature by manipulating value of the ratio. In order to maintain the proper ratio of reactant streams before entering the RD column, the fresh butanol feed is set as a ratio of

the distillate from the 1st column. Another tray temperature is controlled by manipulating the heat input. Here, the recycle stream ow rate is manipulated to control the reux drum level in the RD column. The reux rate in the RD column is set to be a ratio of the recycle stream rate. Fig. 12(b) shows the complete control conguration for the HM process. The ow rate of the recycle stream from the 3rd column is manipulated to maintain a constant ratio between the limiting reactant fresh feed and the recycle ow rate. The bottom liquid level of the 3rd column is controlled by manipulating the fresh feed of water. Based on the degrees of freedom analysis, tray temperatures in the three columns are controlled by heat input and reux ratio.

Fig. 17. Desired temperature set points under feed ow disturbances for HM system (a) tray 4 in reactive distillation column and (b) tray 5 in 2nd column.

Fig. 18. Desired temperature set points under feed ow disturbances for TM system (a) tray 4 in reactive distillation column, (b) tray 7 in 2nd column and (c) tray 45 in 2nd column.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

395

Fig. 12(c) shows the complete control congurations for the EA process. The 2nd column is a stripper. The recycle ow rate is related to the vapor rate from the stripper. The vapor rate is related to the heat input, so its must be manipulated to control one tray temperature. The outlet composition from the top of the stripper is methanol-rich. The liquid level in the top liquid collection drum is controlled by manipulating the fresh methanol feed. The methanol feed ow rate will affect the methanol vapor rate at the top of the stripper signicantly. The heat input of the 3rd column is used to control one tray temperature. 3.4. Step 6. Selection of temperature control trays and controller design Sensitivity analyses are performed on the three reactive distillation systems. In order to nd the steady-state gains for tray temperature in the linear region, very small step changes (0.1% or less) in the manipulated variables are made. All the sensitivity analyses still start from the column downstream from the RDC. Figs. 13 and 14 show the most sensitive tray in each column. Note that the tray number is counted from bottom (reboiler is 0) to

Table 4 Temperature control tray locations, manipulated variables and tuning parameters for the three plantwide processes.a. Controlled variables HM T1,4 T2,5 T3,15 T1,4 T2,45 T2,7 T1,0 T2,9 T3,0 Manipulated variables QR1 QR2 RR3 QR1 QR2 FR RR1 QR2 QR3 Tuning parameter Kc = 3.35 tI = 0.18 (h) Kc = 14.62 tI = 0.116 (h) Kc = 6.59 tI = 0.192 (h) Kc = 12.7 tI = 0.072 (h) Kc = 2.02 tI = 0.235 (h) Kc = 20.4 tI = 2.104 (h) Kc = 0.983 tI = 0.22 (h) Kc = 5.56 tI = 0.02 (h) Kc = 22.7 tI = 0.145 (h)

TM

EA

a Transmitter span: twice of steady-state value of temperature in 8C. Valve gains: twice of the steady-state value for QR and FR(RR).

top. Each column has one manipulated variable available for tray temperature control thus only SISO control is considered. Fig. 14(a) and (c) show the two most sensitive trays are located in the bottom of the RDC and the 3rd column. In the TM process, the RDC (column 2) has two manipulated variables (Fig. 13 (a) and (b)) and this leads to 2 2 multivariable control. The manipulated variable FR here

Fig. 19. Feedforward control structures for (a) HM owsheet and (b) TM owsheet.

396

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

means the ratio of the butanol feed rate to the total ow rate from the 1st column. The non-square relative gain (NRG; Chang and Yu, 1990) is used for selecting temperature control trays. The relative gain array (RGA) for variable pairing gives the following result:

L 0:04

 QR2

FR  1:04 T 2;7 1:04 0:04 T 2;45

(5)

Therefore QR2 (heat duty in the 2nd column) is paired with T2,45 (Tray 45 in the 2nd column) and FR is paired with T2,7 (Tray 7 in the 2nd column). Once the controlled trays have been determined,

controller settings should be determined next. The principal of step (7) should used for the auto-tuning procedure in order to reduce interacting effects. Relay feedback tests (Shen and Yu, 1994) are performed on the temperature loops to nd the ultimate gain (Ku) and ultimate period (Pu) of each temperature control loop and then the controller tuning parameters are set similar to TyreusLuyben settings (Luyben et al., 1998) as follows: Kc = Ku/3 and tI = 2Pu. As pointed by Lin et al. (2006), relay feedback tests are effective and suitable for such highly nonlinear processes. Table 4 shows the controlled variables, manipulated variable and tuning parameters for the three plantwide processes under temperature control.

Fig. 20. Feedforward control performance for 20% production rate changes for (a) HM process and (b) TM process.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

397

3.5. Step 7. Temperature control performance Performing dynamic simulation using AspenDynamics1, a third-order 0.5 min time lag is assumed for temperature measurement (Luyben et al., 1998). Liquid levels are controlled using proportional-only controllers. Proportional-integral controllers are used for ow, pressure, and temperature controls. Two scenarios are used for testing control performance: one is changing the production rate 20% and the other is changing the feed composition. Lin et al. (2008) report the dynamic performance for the HM owsheet for changes in the production rate and feed composition, respectively. The throughput change within the 1st hour. The feed of fresh water (F H2 O ) also changes dramatically within 1st hour due to the changing bottom level of the 3rd column but it stabilizes within 5 h. Also the temperature controls do well for each column: the heat inputs of all the columns regulate the temperatures quickly. The product compositions (XB2,HAc and XD3,MeOH) have small offsets ($0.5 mol%) when all loops are settled. The results are seen for a change in the feed composition: the process settles within 10 h and the products qualities also have very small offsets. Fig. 15 shows the dynamic performances of the TM process. More oscillations are observed than in HM process but all of the

temperature control loops still settle in 5 h. Again the compositions of products (Fig. 15, XB2,HAc and XD3,MeOH) have small offsets. Fig. 16 shows the dynamic performances for the EA system. The dynamic behavior is similar to that of the HM process. Again all the temperature control loops settle in 5 h and the dynamic responses are symmetric. The controller performances are even better for the case of changing the feed composition. A signicant different is observed for the nal composition of product (dimethyl adipate, XB3,DMA). Unlike the HM and TM processes, the 3rd column in the EA process only separates the main products (H2O and DMA). The azeotrope (DMA/H2O, 0.0105/0.9895) is in the two liquid phase region. Therefore, there is almost no offset for the quality of products. According the above results, the temperature control structure is good enough for the EA process. It has fast response and no steady-state offset. But for the other two systems further improvement in the performance is possible. Next we will discuss improving the control structures. 3.6. Feedforward control The previous results suffer from a common problem associated with temperature control: product purity is no longer on spec. One possible approach to compensate for production rate variation is

Fig. 21. Composition control structures for (a) HM owsheet and (b) TM owsheet.

398

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

adjusting the temperature set point (Huang et al., 2004). The throughput manipulator for the HM and TM owsheets is the methyl acetate feed ow so feedforward compensation is devised by adjusting temperature set points as shown in Figs. 17 and 18. Then we employ the feedforward control structures shown in Fig. 19 for both systems. Fig. 20 shows the dynamic performance for the two processes. Both systems stabilize within 10 h. Compared with the results of the original temperature control, the overshoots of the product compositions are much smaller and the steady-state offsets are also decreased and almost eliminated for the HM system.

3.7. Two-composition one-temperature control Offset-free composition control requires on-line composition measurement. On-line composition measurement instruments are usually expensive and frequently malfunction. Control structures that require composition measurement are considered last. For the TM system, one temperature and two compositions are controlled with three manipulated variables: two heat inputs and one feed ratio (or reux ratio). Four minutes of analyzer dead time is assumed for the composition measurements. As in the temperature control section, 2 2 control is used in the reactive distillation

Fig. 22. Two-composition one-temperature control performance for HM process. (a) 20% production rate changes and (b) 5% (mol%) acetate feed composition changes.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

399

column of the TM system, so again, the relative gain arrays is computed for variable pairings:

Table 5 Controlled composition variables, manipulated variables and tuning parameters for the two plantwide processed.a. Controlled variables HM XB1,MeAc XB2,HAc T3,15 XB1,MeOH T2,45 XB2,BuAc Manipulated variables QR1 QR2 RR3 QR1 QR2 FR Tuning parameter Kc = 22.79 tI = 0.864 (h) Kc = 11.22 tI = 0.888 (h) Kc = 20.08 tI = 0.208 (h) Kc = 3.59 tI = 1.995 (h) Kc = 0.498 tI = 0.91 (h) Kc = 20.4 tI = 1.52 (h)

L 0:03

 QR2

FR  1:03 X B2;BuAc T 2;45 1:03 0:03

(6)

Therefore, in the TM system, the heat input in 1st column is used to control the bottom composition (XB1,MeOH). The BuAc purity (XB2,BuAc) and tray temperature (T2,45) are controlled by the feed ratio and heat input, respectively in the 2nd column. Fig. 21(b) shows the two-composition one-temperature control structure for the TM system. The control structure for the HM system can be congured directly. The bottom compositions in the 1st and 2nd

TM

a Transmitter span: twice of steady-state value of temperature in 8C. Valve gains: twice of the steady-state value for QR and FR(RR).

Fig. 23. Two-composition one-temperature control performance for TM process. (a) 20% Production rate changes and (b) 5% (mol%) acetate feed composition changes.

400

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

columns are controlled by the two heat inputs and the 3rd column still uses temperature control as shown in Fig. 21(a). Table 5 lists the controlled variables, manipulated variable and tuning parameters for two plantwide processes under two-composition onetemperature control. Identical changes in the throughput and feed composition are again used for testing control system performance. Figs. 22(a) and 23(a) show the dynamic performances for 20% feed ow rates changes. All of the product purities return to the original set points (left upper two gures). However the dynamic curves are very asymmetric and have large peak errors. Compared with temperature control, the process takes more time to be settle (HM: 15 h for XB1,MeAc and 10 h for XB2,HAc; TM: 30 h for XB1,MeAc and 15 h for XB2,HAc). For the feed composition disturbances (Figs. 22(b) and 23(b)), both dynamic responses are better than the feed ow rate disturbance responses. The composition controllers accomplish their jobs, but still need about 15 h to reach steady state. Although the dynamic responses are not better than for temperature control, composition control structures allow for offset-free response. 4. Conclusions This work illustrates a systematic approach for synthesizing effective control structures for three cases of plantwide RD owsheets. Design principals are proposed for constructing effective control structures. These steps usually demand a plantwide perspective that often leads to control strategies that differ signicantly from those devised by looking at isolated unit operations. However, basic tools and analysis methods still can be applied, including feed back test and RGA. Simple decentralized PI controllers are used for all the RD plantwide systems and are quite effective. The dynamic results show that the control structures provide robust, stable, safe and economical column operation in the face of disturbances entering the column. Temperature control structures are fast and provide good dynamic responses for all of the systems. Modied temperature control structures using feedforward controls lead to smaller steady-state

offsets and decrease dynamic peak errors. Although composition control offers offset-free control performance, the analyzer dead time may make the dynamic response worse. Analysis of the system characteristics rst, followed by selection of correct control strategies are keys to success for development of an efcient reactive distillation process control structure, regardless of the complexity of the reaction system. As demonstrated in this study, good operability and controllability can be achieved for processes with plantwide complex reactive distillation systems. Acknowledgement This work was supported by the Ministry of Economic Affairs under grant 92-EC-17-A-09-S1-019. Appendix A The transesterication process presented here can compare with other designs published in the literature in the basis of TAC. Details about the method of calculation of the TAC are given in Lin et al. (2008). Luyben proposed two process designs for transestercation (Luyben et al., 2004): one process to produce butyl acetate without using an RD column as shown in Fig. A1 is called the conventional process; the other process utilizes 3 columns as shown in Fig. A2 (one RD column and two distillation columns) and is called RD 3 column. Both butyl acetate purity specications are 99 mol%. We name our process RD 2 column to distinguish from Luyben. In Fig. A3 the TAC of RD 2 column is the base case (100%) to compare with the other process. RD 3 column has 1.5 times the TAC of RD 2 column, and the conventional process has 2 times the TAC of RD 2 column. RD 2 column design can save money compared with previously published designs. Transestercation is one option for the production of a valuable product from a methyl acetate stream. Another option is hydrolysis to produce acetate acid and methanol. Although the product is different, we still compare the TAC of the RD 2 column process for

Fig. A1. Conventional reactor/separator process of the MeAc transesterication system.

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402

401

Fig. A2. Reactive distillation process (3 columns with the high pressure column) of the MeAc transesterication system.

transestercation with a published process for the hydrolysis reaction (Lin et al., 2008). Fig. A4 shows a comparison of these two reaction path in TAC. TAC of transesterication reaction is set as the base (100%) in order to compare other process. Hydrolysis reaction is 1.67 times total TAC of transesterication reaction. Both reaction use impure methyl acetate reactant (i.e., 60 mol% MeAc and 40 mol% MeOH). References
Abrams, D. S. and J. M. Prausnitz, Statistical Thermodynamics of Liquid Mixtures: A New Expression for the Excess Gibbs Energy of Partly or Completely Miscible Systems, AIChE J., 21, 116 (1975). Al-Arfaj, M. A. and W. L. Luyben, Comparison of Alternative Control Structures for an Ideal Two-product Reactive Distillation Column, Ind. Eng. Chem. Res., 39, 3298 (2000). Al-Arfaj, M. A. and W. L. Luyben, Comparative Control Study of Ideal and Methy Acetate Reactive Distillation, Chem. Eng. Sci., 57, 5039 (2002). Al-Arfaj, M. A. and W. L. Luyben, Plant-wide Control for TAME Production Using Reactive Distillation, AIChE J., 50, 1462 (2004). Chang, J. W. and C. C. Yu, The Relative Gain for Nonsquare Multivariable Systems, Chem. Eng. Sci., 45, 1309 (1990). Chan, K. W., Y. T. Tsai, H. M. Lin, and M. J. Lee, Esterication of Adipic Acid with Methanol over Amberlyst 35, J. Tw. Inst. Chem. Eng., in press, doi:10.1016/ j.jtice.2009.12.001. Chiang, S. F., C. L. Kuo, C. C. Yu, and D. S. H. Wong, Design Alternatives for the Amyl Acetate Process: Coupled Reactor/Column and Reactive Distillation, Ind. Eng. Chem. Res., 41, 3233 (2002). Douglas, J. M., Conceptual Design of Chemical Processes, , McGraw-Hill, New York, USA (1988). Engell, S. and G. Fernholz, Control of a Reactive Separation Process, Chem. Eng. Process., 42, 201 (2003). Fuchigami, Y., Hydrolysis of Methyl Acetate in Distillation Column Packed with Reactive Packing of Ion-exchange Resin, J. Chem. Eng. Jpn., 23, 354 (1990). Hayden, J. G. and J. P. OConnell, A Generalized Method for Predicting Second Virial Coefcients, Ind. Eng. Chem. Process. Des. Dev., 14, 209 (1975). Huang, S. G., C. L. Kuo, S. B. Hung, Y. W. Chen, and C. C. Yu, Temperature Control of Heterogeneous Reactive Distillation, AIChE J., 50, 2203 (2004). Hung, S. B., I. K. Lai, H. P. Huang, M. J. Lee, and C. C. Yu, Reactive Distillation for Twostage Reaction Systems: Adipic Acid and Glutaric Acid Esterication, Ind. Eng. Chem. Res., 47, 3076 (2008). Hung, S. B., H. M. Lin, C. C. Yu, H. P. Huang, and M. L. Lee, Liquid-liquid Equilibria of Aqueous Mixtures Containing Selected Dibasic Esters and/or Methanol, Fluid Phase Equilib., 248, 174 (2006a). Hung, S. B., M. J. Lee, Y. T. Tang, Y. W. Chen, I. K. Lai, W. J. Hung, H. P. Huang, and C. C. Yu, Control of Different Reactive Distillation Congurations, AIChE J., 52, 1423 (2006b).

Fig. A3. Comparison of TAC with the different owsheets of the transesterication system.

Fig. A4. Comparison of TAC with the transesterication system and the hydrolysis system.

402

S.-B. Hung et al. / Journal of the Taiwan Institute of Chemical Engineers 41 (2010) 382402 with Methanol and Methyl Acetate Hydrolysis, Ind. Eng. Chem. Res., 39, 2601 (2000). Renon, H. and J. M. Prausnitz, Local Compositions in Thermodynamic Excess Functions for Liquid Mixtures, AIChE J., 14, 135 (1968). Shen, S. H. and C. C. Yu, Use Of Relay-Feedback Test for Automatic Tuning Of Multivariable Systems, AIChE J., 40, 627 (1994). Sneesby, M. G., M. O. Tade, and T. N. Smith, Two-point Control of a Reactive Distillation Column for Composition and Conversion, J. Process Control, 9, 19 (1999). Subawalla, H. and J. R. Fair, Design Guidelines for Solid-catalyzed Reactive Distillation Systems, Ind. Eng. Chem. Res., 38, 3696 (1999). Tang, Y. T., H. P. Huang, and I. L. Chien, Plant-wide Control of a Complete Ethyl Acetate Reactive Distillation Process, J. Chem. Eng. Jpn., 38, 130 (2005). Tung, S. T. and C. C. Yu, Effects of Relative Volatility to the Design of Reactive Distillation Systems, AIChE J., 53, 1278 (2007). Wu, K. L. and C. C. Yu, Reactor/Separator Processes with Recycle. 1. Candidate Control Structure for Operability, Comput. Chem. Eng., 20, 1291 (1996). Yi, C. K. and W. L. Luyben, Design and Control of Coupled Reactor/Column Systems Part 3. A Reactor/Stripper with Two Columns and Recycle, Comput. Chem. Eng., 21, 69 (1997).

Jimenez, L. and J. Costa-Lopez, The Production of Butyl Acetate and Methanol via Reactive and Extractive Distillation. II. Process Modeling, Dynamic Simulation, and Control Strategy, Ind. Eng. Chem. Res., 41, 6735 (2002). Kaymak, D. B. and W. L. Luyben, Quantitative Comparison of Reactive Distillation with Conventional Multiunit Reactor/Column/Recycle Systems for Different Chemical Equilibrium Constants, Ind. Eng. Chem. Res., 43, 2493 (2004). Kaymak, D. B. and W. L. Luyben, Comparison of Two Types of Two-temperature Control Structures for Reactive Distillation Columns, Ind. Eng. Chem. Res., 44, 4625 (2005). Lin, Y. D., J. H. Chen, H. P. Huang, and C. C. Yu, Process Alternatives for Methyl Acetate Conversion Using Reactive Distillation. 1. Hydrolysis, Chem. Eng. Sci., 63, 1668 1682 (2008). Luyben, W. L., B. D. Tyreus, and M. L. Luyben, Plantwide Process Control, , McGraw-Hill, New York (1998). Luyben, W. L., K. M. Pszalgowski, M. R. Schaefer, and C. Siddons, Design and Control of Conventional and Reactive Distillation Processes for the Production of Butyl Acetate, Ind. Eng. Chem. Res., 43, 8014 (2004). pken, T., L. Go tze, and J. Gmehling, Reaction Kinetics and Chemical Equilibrium Po of Homogeneously and Heterogeneously Catalyzed Acetic Acid Esterication