Evolutional Design and Control of the Equilibrium-Limited Ethyl

---

Article pubs.acs.org/IECR

Evolutional Design and Control of the Equilibrium-Limited Ethyl Acetate Process via Reactive Distillation−Pervaporation Hybrid Configuration Hao-Yeh Lee,‡ Sheng-Yu Li,† and Cheng-Liang Chen*,† †

Department of Chemical Engineering, National Taiwan University, Taipei 10617, Taiwan Department of Chemical Engineering, National Taiwan University of Science and Technology, Taipei 10607, Taiwan

‡

S Supporting Information *

ABSTRACT: Ethyl acetate (EtAc) is an important chemical and is normally produced via esterification reaction of acetic acid with ethanol. Due to the thermodynamic limitation of the EtAc esterification system, an energy efficient two-column process which is composed of a reactive-distillation (RD) column and a stripper has been proposed recently1,2 to remove water for yielding high purity EtAc product. Considering recent phenomenal advancement of membrane technology, this study raises a RD−pervaporation (PV) hybrid configuration for producing high purity EtAc with further decreased energy consumption. A side stream with high EtAc purity is drawn from the rectifying section of the RD column, and pervaporation units are applied to remove water for producing high spec EtAc product. The optimized design achieves 13% energy savings compared to the two-column process.2 Based on the optimized RD−PV hybrid process, two different control strategiesthe single-point temperature control and the dual-points temperature controlare proposed to examine the performance of rejecting the main disturbances to the process. A lot of simulation tests show that the single-point temperature control structure can handle throughput disturbance but fails to maintain adequate product specification should the acid purity in the feed be lower than a specific value. However, the dual-points temperature control structure can successfully deal with both throughput and feed composition disturbances.

1. INTRODUCTION In chemical industries, ethyl acetate (EtAc) is usually used as a solvent to produce inks and adhesives. The main method to manufacture EtAc is through the esterification reaction of acetic acid (HAc) and ethanol (EtOH) under acidic conditions. However, the esterification reaction has an equilibrium limitation which brings about high operating costs, due to the heavy burden of separation and recovery of unreacted reactants. Reactive distillation (RD) can be used to break this limitation. The RD configuration is a technique which combines reaction and separation tasks into a single unit; thus, it can highly reduce the total capital cost. Meanwhile, high conversion and selectivity can be achieved by shifting the chemical equilibrium boundaries, which brings a large amount of energy savings. In 1932, Keyes reported a novel ethyl acetate (EtAc) process which consisted of a pre-esterification reactor, a RD column, two recovery columns, and a decanter.3 Since then, more and more researchers started to study various processes which related to the RD unit: either its steady state design or its dynamic control. In the 1980s, the production of methyl acetate with a RD column was commercialized by Eastman Chemical Co.4 In recent years, a significant number of researchers have © XXXX American Chemical Society

discussed the characteristics of the RD and not only raised several conceptual designs for the RD process but also summarized the potential of RD in the industry. Tang et al.1 provided generalized RD configurations for esterification reaction of different alcohols with acetic acid. They classified these different alcohols, ranging from C1 to C5, into three types. The production of EtAc was classified as type II. The configuration of this type is much simplified with the typical RD processes. Instead of using pure feed, Lai et al.2 reported an alternative design which used impure feeds in the process. Thus, without advanced purification of these raw materials, significant cost reduction can be achieved. Nowadays, usage of energy in industry has captured more and more attention, which means further improvement of many existing processes is necessary. Therefore, an increasing amount of researchers are focusing on the development of new process design. Membrane technology has been considered to be an Received: April 11, 2016 Revised: July 9, 2016 Accepted: July 20, 2016

A

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

show that EtAc product purity is remarkably enhanced. For the dynamic control of the pervaporation hybrid process, Luyben16 studied a control structure of the hybrid system for separating the ethanol/water azeotrope, providing effective disturbance rejection for both production rate and feed composition changes. In this work, a new RD-PV hybrid configuration is raised for production of high spec EtAc. In the process, the RD unit is used for reaction and preliminary separation, and it is found that under suitable arrangement of the RD column and the decanter, a side stream with extremely high acetate concentration can be drawn from the middle stage of the RD column. The PV units are applied for further purification to satisfy the product specification. Variables of the RD column and the PV units in the process are subsequently investigated to find the optimal design, which corresponds to the lowest total annual cost (TAC) while meeting product specifications. After finishing the steady state design, feasible control structures are then developed for this hybrid system to reject the main operating disturbances, such as throughput change and feed composition variations. The commercial simulators Aspen Plus and Aspen Plus Dynamics are used for simulation.

effective means of separation because it is not limited by the volatility of the components. In addition, the membranes show advantages including low energy consumption, high selectivity and compact and modular design. These merits have made membrane technologies attract increased attention in recent years. There are several types of membrane systems; among them, pervaporation (PV) is one of the most promising alternatives. The word pervaporation was first studied in 1917 by Kober.5 It refers to a process in which one or more components of a fluid mixture selectively permeates a dense membrane. Although PV seems to have a lot of advantages, it still has some limitations in practical application due to its high capital cost and low capacity. The most common way to use this technology is to combine PV with a conventional unit, called PV-based hybrid process, thus overcoming the disadvantages of each. A hybrid process offers significant benefits over the conventional processes. In 1999, Lipnizki et al.6 focused on the PV-based hybrid processes and gave an overview of their applications, process designs, and economics. In 2014, Van der Bruggen and Luis7 also discussed the application of the pervaporation unit, including organophilic separation, membrane reactors, and bioethanol upgrading. Among these hybrid processes, the integration of pervaporation with the conventional esterification process has great potential for further development. Via continuously removing water from the reactor, the chemical equilibrium can be shifted; thus, a higher conversion can be achieved without intensive energy consumption. Focused on the membrane reactors, Van der Bruggen8 made a complete introduction about this process. For the ethyl acetate esterification process, Waldburger and Widmer9 first reported the observation of the continuous tube membrane reactor. Comparing this novel design to the traditional reactor and distillation configuration, 75% energy and 50% total cost can be saved. de la Iglesia et al.10 achieve high conversion (90%) by using the mordenite membranes. Jyoti et al.11 gave a detailed review of the esterification− pervaporation process. Different operating conditions of pervaporation−integrated processes, such as temperature, catalyst concentration, etc., are discussed in this study.11 It is noted that a distillation unit is still necessary to guarantee manufacture of high spec products. Another consideration of using a reaction pervaporation process is the acid factor. Due to the existence of acid, ceramic membranes are recommended to separate water. However, ceramic membranes are still very expensive compared to the polymeric membranes. Therefore, the RD-PV configuration is a more practical method on the industrial scale. In the future, once the membrane technology is improved furthermore, it can replace the RD-PV process definitely. For the discussion of the RD-PV hybrid process, Bausa and Marquardt12 provided a shortcut design for hybrid membrane/distillation processes to estimate the minimum membrane area; thus, the investment cost can be analyzed. Aiouache and Goto13 also reported a tert-amyl alcohol RD-PV hybrid process and investigated the effect of various membrane properties. The results showed that by continuous removal of generated water, product yield can be enhanced by 10%. Buchaly et al.14 presented a new n-propyl propionate RD process with pervaporation. The membrane module is placed on the top of the RD column in order to selectively remove the produced water without use of entrainers. Lv et al.15 investigated a new EtAc hybrid process, where the membrane is located in the bottom stream in order to selectively remove water from the reboiler and recycle acetic acid. The results

2. MODEL BUILDING 2.1. Thermodynamic Property. The EtOH−HAc−EtAc− H2O quaternary system exhibits a nonideal phase behavior. To

Figure 1. RCM and liquid−liquid envelope for the EtAc system at 1 atm and 40 °C.

Table 1. Pervaporation Parameters of Each Species in the EtAc System Species i

Di,0 [m·h−1]

Ei/R [K]

EtOH H2O EtAc HAc

0.01176 1.32441 0.06158 0.12321

2732.58 1959.21 2987.62 2021.89

predict the phase equilibrium of these four components, including the vapor−liquid equilibrium (VLE) and its possible vapor−liquid−liquid equilibrium (VLLE) phase behaviors in this quaternary system, the nonrandom two liquids (NRTL) activity coefficient model is used in the following. Furthermore, B

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 2. Process flowsheet: (A) Two column process;2 (B) Proposed RD−PV hybrid process (optimal configuration).

considering the dimerization of acetic acid in the vapor phase, the second viral coefficients of Hayden−O’Connell (HOC)17 are also adopted. All binary parameters for this EtAc system can be found in ref 18. Detailed thermodynamic information and the following kinetic models are available in the Supporting Information.

Figure 1 shows that the NRTL-HOC model predicts four azeotropes in this quaternary system, including three binary azeotropes and one ternary azeotrope. Also, there is a significant liquid−liquid envelope. 2.2. Reaction Kinetics. For the EtAc production, HAc and EtOH undergo an esterification reaction to yield EtAc and C

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

the dynamic balances of all compositions and energy in each cell n.

water, as shown in eq 1, where the acidic ion-exchange resin Purolite CT179 is used to catalyze the reaction: k1

EtOH + HAc XooY EtAc + H2O

Total mass balance:

(1)

k −1

Wi A·t

α=

yi /(1 − yi )

Component balance: dz R,n, i MR = FR,n − 1z R,n − 1, i − FR,nz R,n, i − FP,nz P,n, i dt

(7)

Energy balance: dhR,n = FR,n − 1hR,n − 1 − FR,nhR,n − FP,nHP,n MR dt

(8)

In these equations, variable F with subscript n denotes the molar flow rate from cell n, subscripts R and P represent the retentate or the permeate sides, respectively, zi is the molar fraction of species i, h is the molar enthalpy of liquid in retentate side, and H is the molar enthalpy of vapor in the permeate side.

(2)

xi /(1 − xi)

dMR = FR,n − 1 − FR,n − FP,n dt (6)

The reaction rate expression for this reaction can be found in ref 19. To implement this rate equation into Aspen Plus for RD simulation, some parameters are clarified here: (1) The Weir height of the RD column is 0.1016 m; (2) The downcomer area occupies 10% of the tray area; (3) Half of the liquid holdup is filled with catalyst in the reactive section; (4) The density of catalyst is assumed as 770 kg/m3; (5) The column base holdup is taken as 10 times the tray holdup. 2.3. Pervaporation Model. In this study, we use flux (Ji, kg·m−2·h−1) and the separation factor (α) to describe the membrane performance. Ji =

0=

3. STEADY STATE DESIGN Figure 1 illustrates the residue curve map (RCM) and the liquid−liquid envelope for the EtAc system at 1 atm and 40 °C.

(3)

where Wi is the weight of component i in the permeate (kg), t is the permeation time interval (h−1), and yi and xi are the weight fractions of component i in the permeate and feed, respectively. In a pervaporation membrane with an effective membrane area A (m2) for mass transfer, the diffusion rate of species i (JiA or FP,i, kg·h−1) is determined by the concentration difference between the retentate and the permeate sides (CR,i and CP,i, kg·m−3), Ji A = FP, i = ADi (CR, i − CP, i)

(4) −1

The diffusion coefficient (Di, m·h ) is usually temperature dependent and can be represented as the Arrhenius form, Di = Di,0 e−Ei /RT

(5)

The membrane model parameters used in this study, Di,0 and Ei/R, are regressed from experimental data published by Lv et al. (2012),15 which reports a PV/ceramic composite membrane. In the literature, the PV dehydration performance in a quaternary HAc/H2O/EtAc/EtOH (80/12/6/2 wt %) solution was tested at different temperatures. As the temperature increased from 50 to 100 °C, the total flux increased from 0.35 to 1.25 (kg·h−1·m−2), and the separation factor of water decreased from 15.4 to 12.6. These data are used to find the membrane model parameters, Di,0 and Ei/R, such as given in Table 1. The detailed regression procedure and results of model validation between the literature and the model prediction are available in the Supporting Information. Considering a possible concentration gradient inside the membrane, a simple lumped model16 is used for describing the dynamic behavior of the pervaporation unit. In the lumped model, the membrane unit is equally divided into 10 lumps to ensure acceptable error between the simplified lumped model and the perfect plug flow.20 The molar holdup in each cell MR is assumed to be constant, and the size of the membrane holdup is based on the membrane area, 0.003 m3/m2.21 The following ordinary differential equations are used to describe

Figure 3. Liquid composition profile in the RD column: (A) Two column process;2 (B) Proposed RD−PV hybrid process.

Therein, when most HAc in this quaternary system is consumed in the reactive section of the RD column, the lowest temperature in the whole system is the homogeneous ternary azeotrope, and the second lower one is EtAc and the H2O heterogeneous binary azeotrope. This means, under nominal D

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

feed flow rate, 50.8 kmol/h, which is the same as Lai et al.’s work for comparison.2 The RD column in this process is taken as a reactor and also a preliminary separation unit, which only consists of a reactive section and a rectifying section with nearly zero bottom flow rate (0.01 kmol/h in this case). Two inlet streams are fed into the column from the bottom, as suggested by Tang et al.1 When most acid is consumed, a mixture of EtOH/H2O/EtAc ternary azeotrope and EtAc/H2O binary azeotrope appears in the top of the RD column, and its composition can be separated into two liquid phases via the decanter (see Figure 1) operated under atmosphere and 40 °C. The aqueous phase liquid with high water purity is directly removed from the decanter, and the organic phase liquid, consisting of mostly EtAc, is totally refluxed back to the RD column. Figure 3 shows the liquid concentration profile in the RD column of these two processes. It can be found that a stream with high EtAc concentration is drawn from the middle of the rectifying section in the RD column. In order to enhance the EtAc purity in the side stream without consuming too much energy, more rectifying stages are applied in this RD−PV hybrid process. Although the EtAc purity in the side stream drawn from the RD column is quite high, it still does not meet the desired product specification. Thus, membrane modules are used to enhance the product purity to the required specification. For the design of the pervaporation units, there is a maximum membrane area in a single module.23 The membrane module thus must be divided into several smaller modules to make sure it does not exceed the limitation. In this case, the flux through the membrane is an important consideration. To enhance the flux, inlet temperature is set at 100 °C,15 and series arrangement is adopted.20 However, the side stream from the column would vaporize under this temperature. To prevent this phenomenon, a feed pump is placed before the heater to increase its boiling point. Due to the characteristics of the PV units, loss of product from the permeate side is inevitable. Therefore, treatment of the permeate stream is important. In this study, the permeate stream is sent back to the decanter to recover EtAc residue. Since the water purity in the permeate side is high, mixing this stream with the top stream from the RD column can improve the performance of the decanter and the water purity in the aqueous phase can be enhanced as a result. 3.2. Process Optimization. The total annual cost (TAC) analysis is used to find the optimal design. The calculation of TAC is the combination of the annual operating cost (AOC) and the total capital cost (TCC) as shown in eq 9.

Figure 4. Optimization procedure.

operation, the vapor composition out of the RD column is in between the EtAc/H2O binary azeotrope and the ternary azeotrope. Also, its composition will locate in the liquid−liquid envelope because the ternary azeotrope is quite close to the two liquid phases region (see Figure 1). From Figure 1, the tie line end point of the aqueous phase in the liquid−liquid envelope is always toward the relatively pure water. This implies that a decanter on the top of the column is a good choice to remove water generated from the esterification reaction of the system. For handling the organic phase, most previous researchers selected to make it partially reflux to the RD column, and send the remaining to an additional stripper to obtain high purity EtAc product with the top recycled back to the decanter. This configuration was presented by Burkett and Rossiter22 and Tang et al.1,18 In 2007, Lai et al.2 proposed a RD process with an azeotropic feed instead of using a pure feed which brings a significant raw material cost savings. In this study, a modified process flowsheet is proposed, with total reflux of the organic phase from the decanter, and a stream with extremely high concentration EtAc is drawn from the middle stage of the rectifying section in the RD column. Detailed discussion of this configuration is given in the following subsections. 3.1. Process Flowsheet. Figure 2 shows the optimal twocolumn design presented by Lai et al.2 and also the proposed EtAc hybrid process flowsheet. The total reduction of the reboiler duty is 13%, from 7,092 kW of the conventional twocolumn design to 6,192 kW of the proposed RD-PV hybrid process. The detailed optimization procedure of the proposed hybrid process will be elucidated in subsection 3.2. The hybrid process is targeting to synthesize and purify EtAc to meet the 99 mol % product specification. Also, the HAc mole purity should be lower than 100 ppm in the product stream. In this research, the feed streams are set to 87 mol % for the EtOH stream and 95 mol % for the HAc stream, where the impurity is water, and the flow rate of the HAc feed stream is fixed at 50.8 kmol/h, which is related to the product capacity.2 The EtOH feed flow rate is used to meet the specification that the HAc concentration in the product stream is lower than 100 ppm. The throughput of the EtAc product is only related to the HAc

TAC = AOC +

TCC payback period

(9)

In this study, the payback period is set as 3 years. The calculation for column and heat exchanger is based on Douglas,24 as given in the Appendix. The cost of piping and pumps is ignored during the calculation. The membrane price is taken from Szitkai et al.,25 and the lifetime of the membrane is assumed to be 3 years.20 In this study, the membrane price is divided into two parts: the membrane material with cost per unit area and the membrane modules with cost per unit.20 The capital cost and the operating cost of the vacuum pump are based on Woods26 and Oliveira et al.,27 respectively. In the following subsections, all variables are divided into two parts for subsequent discussion. E

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 5. Effect of the main design variables of the system (based on TAC).

(5) Side stream flow rate is used as the manipulated variable to control the EtAc purity in the side stream. The remaining variables to be determined in the RD section include the side stream drawn stage, the number of trays in the reactive (Nrxn) and rectifying (NR) sections, and the EtAc purity in the side stream. The EtAc purity is an important consideration during the optimization. Higher purity would reduce the use of the membrane area, but the operating cost would be higher in the RD column. 3.2.2. Variables in the Membrane Section. As for variables in the membrane section, the membrane area is determined to meet the final product specification. The inlet temperature is fixed at 100 °C for higher flux.15 The permeate side pressure is kept at 0.16 bar.20 Since the retentate side is in the liquid phase and the permeate side in the vapor phase, the variation of inlet pressure is not considered in this study because it does not have a significant effect.28 In this study, 5 bar is set to avoid vaporization of the side stream at 100 °C.

3.2.1. Variables in the RD Column. There are several variables that need to be clarified for optimizing the RD column, including column pressure, reactants feed stages, reactants feed ratio, number of trays in reactive and rectifying sections, reboiler duty, side stream drawn stage, and side stream flow rate. The optimization is also based on the following assumptions: (1) The column is operated under normal pressure while cooling water can be used. (2) For the reactants feed stage, it was already discussed by Tang et al.1 In this study, the condition is kept as the same design to simplify the optimization procedure. (3) The HAc feed flow rate is kept at 50.8 kmol/h,2 which is related to the product capacity. The EtOH feed flow rate is 56.96 kmol/h to keep the HAc concentration at the side stream to meet its specification. (4) Reboiler duty is used to maintain the column sump level and a very small bottom flow rate is fixed as 0.01 kmol/h in this study. F

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research Table 2. Detailed Cost Distribution of the Two Column EtAc Process2 and the Proposed RD-PV Hybrid Configuration TwoColumn Design (Laiet al.2007) RD-PV Hybrid Process (this work) RD Column Capital Cost ($1000/year) 167.20 Column 253.13 31.74 Trays 54.39 112.68 Reboiler 125.61 133.83 Condenser 168.50 69.24 Sub-Cooler 73.47 RD Column Operating Cost ($1000/year) Catalyst 71.25 Catalyst 81.00 Steam 436.58 Steam 528.66 Cooling Water 18.10 Cooling Water 21.39 Purification Section Capital Cost ($1000/year) Stripper Membrane Section Column 59.96 Membrane Material 188.58 Trays 8.09 Membrane Module 173.49 Reboiler 39.94 Vacuum Pump 2.03 Condenser 67.58 Heat Exchanger 8.59 Sub-Cooler 34.76 Feed Pump 1.49 Purification Section Operating Cost ($1000/year) Stripper Membrane Section Steam 168.95 Steam 6.20 Cooling Water 6.30 Electricity 10.03 Chilled Water 0.66 Total Annual Cost (TAC) ($1000/year) 1426.20 1697.22 Raw Material Cost ($1000/year) HAc 17,455.34 HAc 17,455.34 EtOH 24,229.26 EtOH 24,014.24 Total Annual Cost Including Raw Material ($1000/year) 43,110.80 43,166.80 Column Trays Reboiler Condenser Sub-Cooler

Figure 6. Influence of membrane price.

specification, Figure 4 shows all the steps to obtain a nearly optimal steady state design. 3.2.4. Results and Discussion. Figure 5 shows the optimization results which follow the optimization procedure, and the results indicate the optimal operating configuration, which is also noted as a red line with a solid circle. From this figure, an appropriate side stream drawn stage of each configuration can be observed, and its value is always related to the numbers of rectifying stage (NR) or reactive stage (Nrxn). As Nrxn or NR increases, the suitable drawn stage number would move down with fixed other stage. It is interesting to notice that the optimal operating condition is to keep the side stream EtAc purity at around 96 mol % based on the TAC analysis. These phenomena can be explained by the expensive membrane price and the weak membrane performance. Thus, the annualized capital cost in the membrane separation section will be very high (around 22% TAC of this system; see Table 2). However, if the side stream EtAc purity is higher than 96 mol %, the reboiler duty in the RD column will increase significantly (see Table 3). Table 2 lists a detailed cost distribution of the hybrid process and the two column design. It can be found that the TAC of the former design is a little higher than the latter case but there is a significant savings in raw materials and reboiler utility. As

3.2.3. Optimization Strategy. The design procedure follows the direct search method proposed by Hooke and Jeeves in 1961.29 During the optimization, the sensitivity test for the design and control variables can be recorded, making the researchers easily understand the effect of these variables of the process. The direct search method can be applied to most processes which are simulated by some commercial simulators such as Aspen Plus. In this study, with a given product

Table 3. Sensitivity of EtAc Purity in the Side Stream to Total Cost of Steam and Membrane under Three Membrane Material Prices $387.5/m2

$271/m2

$500/m2

EtAc Purity in side stream

Steam

Membrane Module

Membrane Material

Total (Steam + Membrane)

Membrane Material

Total (Steam + Membrane)

Membrane Material

Total (Steam + Membrane)

0.900 0.910 0.920 0.930 0.940 0.950 0.955 0.960 0.965 0.970 0.975 0.981

460.078 466.297 473.106 480.834 490.095 501.828 511.461 528.659 566.211 633.671 733.808 938.051

1299.751 732.123 514.331 388.396 303.154 251.988 212.148 173.492 165.196 127.862 120.311 80.976

1409.166 782.725 543.789 417.247 328.368 254.071 219.067 188.583 160.167 133.423 108.920 74.958

3168.995 1981.145 1531.227 1286.477 1121.617 1007.887 942.675 890.735 891.574 894.956 963.039 1093.984

985.5069 547.4028 380.3015 291.8038 229.6454 177.6857 153.2053 131.8867 112.0133 93.31018 76.17384 52.42251

2745.336 1745.822 1367.739 1161.034 1022.894 931.502 876.814 834.038 843.420 854.843 930.293 1071.449

1818.278 1009.968 701.663 538.383 423.700 327.833 282.667 243.333 206.667 172.159 140.542 96.721

3578.107 2208.388 1689.101 1407.613 1216.949 1081.649 1006.275 945.485 938.074 933.692 994.661 1115.747

G

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 8. Influence of different membrane performances.

module. In the discussion of membrane cost, the material price is studied, which means, during the sensitivity test, the price of the membrane module remains the same. Table 3 shows the cost distribution of different material prices. It can be observed that the best trade-off is located at 96% EtAc purity for both $387.5/m2 and $271/m2 (See Figure 7 and Table 3). The reason is that an additional membrane module is required, and the cost of the membrane module will increase from 173 to 212 ($1000/y) when the purity decreases from 96% to 95.5%. However, the savings in the steam consumption is limited. Thus, the optimal operating condition remains the same. After further decreasing the membrane price, the best trade-off is still located at the same point because the savings in steam consumption cannot compensate for the cost of an additional membrane module. It is noted that the best trade-off would move to higher purity when the membrane price becomes very expensive where the membrane dominates the total cost. Taking membrane material cost $500/m2 as an example, the best trade-off will locate at 97% purity. In addition to the membrane price, the membrane performance would also affect the competitiveness of the proposed hybrid process. In this configuration, the feed stream to the pervaporation section contains about 3.7 mol % water, 96 mol % EtAc, and 0.3 mol % EtOH, and different membrane performance would affect the required membrane area to purify the product. An advantage of the process simulation is that the virtual membrane can be built and, hence, make discussion easier. Figure 8 shows the required membrane areas of different membrane fluxes and separation factors to purify EtAc from 96 mol % to 99 mol %. The dashed line in the figure represents the area 1,460 m2, which is the original value in the steady state design. If a better membrane can be applied to the hybrid process, the required membrane area can be further reduced. For example, if a membrane has a 150 separation factor, its flux should be higher than the current value (30.5 g·h−1·m−2) to reduce the required area. In this study, three different membrane factors have been discussed, including membrane price, membrane separation factor, and membrane flux. The original membrane price is $387.5/m2 with about 150 separation factor and 30.5 (g·h−1· m−2) flux. With the same operating condition of the RD column, to make the hybrid process more competitive than the traditional two column design, besides reducing the membrane price, reducing the membrane area is also a practical way. After the TAC analysis, it is found that once the membrane area is lower than 1,127 m2 (with the same membrane price $387.5/ m2), the hybrid process will be more economical. To reduce the membrane area, the membrane flux or separation factor should

Figure 7. Influences of EtAc purity in the side stream on the total annual cost under various membrane costs: (A) $387.5/m2; (B) $271/ m2; and (C) $500/m2.

discussed before, the membrane price, taken as $387.5/m2 during the optimization, is a dominant cost factor. The membrane price is expected to decline with the improvement of membrane manufacturing technology. Figure 6 shows the influence of membrane price on the cost of the hybrid process.2 It can be found that once the membrane price is less than $271/m2, the hybrid process would become more competitive over the conventional two column design. The cost calculation in the membrane section includes two different parts: the membrane material and the membrane H

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 9. Open loop sensitivity test for (A) feed ratio and (B) side stream flow rate.

In the control of the pervaporation process, a permeate drum is installed to deal with the permeate stream from the membrane units and also to make sure that the feed to the decanter is always in the liquid phase.16,20 For convenience, the operating condition (i.e., temperature and pressure) is kept the same for each membrane system. The whole control loops can be divided into two parts, including inventory control and quality control. The inventory control is used to keep material balance in the system, and the function of quality control is to maintain the product purity. In this process, the RD column pressure is controlled by the coolant flow rate to the condenser and the sump level is controlled by the reboiler duty since its bottom flow rate is nearly zero. The organic and aqueous levels in the decanter are controlled by the organic and aqueous outlet flow rates, respectively. The decanter temperature is maintained by manipulating the coolant flow rate of the subcooler. The pressure and liquid level of the permeate drum are controlled by the coolant flow rate of the permeate condenser and its liquid outlet flow rate. Two feed streams into the process are flow-controlled, and a ratio controller is installed. The throughput manipulator is the EtOH feed flow rate in this study.2 After all of the inventory loops are set, the remaining control degrees of freedom in the process are the feed ratio FHAc/FEtOH, the side stream flow rate, and the membrane inlet temperature. There are two manipulated variables, the feed ratio and the side stream flow rate, in the RD column. An open-loop sensitivity test is conducted to find the measured point of the column (see Figure 9). A close-loop sensitivity test in the Aspen Plus is conducted to determine the membrane inlet temperature. The results are listed in Table 4, where the steady state value represents the status before any disturbance. The other rows mean the deviation of the variables with different disturbances under perfect control. As the side stream flow rate is different from the steady state value, the membrane inlet temperature should be adjusted accordingly. From Table 4, one can observe that when the side stream flow rate increases 10%, the membrane inlet temperature should also be adjusted to 9 °C higher than its steady state value and vice versa. Thus, a membrane inlet temperature can be determined according to the following empirical relation, which is represented as f(x) in Figure 10.

be higher than the original value. Thus, three ways to reduce the total cost of the proposed RD-PV hybrid design are highlighted in the following: (1) If the membrane flux and separation factor remain the same, the price should be lower than $271/m2; (2) If the membrane flux and price remain the same, the separation factor should be higher than 300; (3) If the membrane separation factor and price remain the same, the flux should be higher than 44.6 (g·h−1·m−2).

4. PROCESS CONTROL After the economically optimal configuration is designed, the control structure of this process is investigated in this section. Table 4. Closed Loop Sensitivity Analysis for Throughput and Composition Change EtOH feed flow rate (Disturbance)

Side stream flow ratea

Ratio of side stream flow rate to EtOH feed flow rate

Membrane inlet temperature (°C)

Steady state 56.96 49.60 0.8708 100.0 Disturbance from throughput change (EtOH feed flow rate) +20% 68.35 59.55 0.8712 118.0 +10% 62.66 54.57 0.8710 109.0 −10% 51.26 44.64 0.8708 091.5 −20% 45.57 39.66 0.8704 082.7 Disturbance from feed composition change 90% HAc 56.96 49.55 0.8676 099.9 100% HAc 56.96 49.64 0.8691 100.0 82% EtOH 56.96 46.67 0.8173 095.6 77% EtOH 56.96 43.74 0.7660 091.4 Disturbances from FEtOH composition and throughput change simultaneously 82% EtOH 68.35 (+20%) 56.0160 0.8195 107.1 82% EtOH 62.66 (+10%) 51.3599 0.8197 101.9 82% EtOH 51.26 (−10%) 42.0658 0.8206 087.4 82% EtOH 45.57 (−20%) 37.3537 0.8197 079.8 77% EtOH 68.35 (+20%) 52.2859 0.7650 103.7 77% EtOH 62.66 (+10%) 47.9593 0.7654 097.6 77% EtOH 51.26 (−10%) 39.2952 0.7666 084.4 77% EtOH 45.57 (−20%) 34.8980 0.7658 079.9

Adjust the side stream flow rate (manipulated variable) to maintain the same EtAc purity in the side stream. a

The objective of the control strategies is to keep the product purity at 99 mol % and the acid concentration in the side stream less than 100 ppm (mole basis) under throughput and raw material composition disturbances.

Tmsp = Tm , ss + 90 × I

(Fside) − (Fside)ss (Fside)ss

(10)

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 10. (A) Single-point temperature control structure; (B) Dual-points temperature control structure.

where Tsp m represents the set point of the membrane inlet temperature, Fside is the side stream flow rate, and subscript “ss” means the value under steady state. In the following subsections, two control strategies are studied based on the discussion above, including the singlepoint and the dual-points temperature control strategies. In

these structures, 5 and 10 min dead times are set for the measurement of temperature and composition, respectively. 4.1. Single-Point Temperature Control Structure. Figure 10(A) shows the single-point temperature control structure. The feed ratio is used to control the 15th tray temperature. The membrane inlet temperature is varied with J

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research Table 5. Controller Settings for Single-Point and Dual-Point Temperature Control Structures PID Settings Controller LC

FC

P

PI

PC

PI

TC

PI

Controlled Variable

Manipulated Variable

Aqueous phase level in decanter Organic phase level in decanter Permeate drum level Sump level EtOH feed flow rate HAc feed flow rate Side stream flow rate Column pressure Permeate drum pressure Decanter temperature Membrane inlet temperature 15th tray temperature in RD column

Valve position in aqueous outlet Valve position in organic outlet Valve position in liquid outlet Steam flow rate in reboiler

5th tray temperature in RD column

Side stream flow rate

Kc = 2

Kc = 20 τI = 12 Kc = 1 τI = 20 Kc = 18.3 τI = 14.5

Kc = 18.8 τI = 48.5 Kc = 70.0 τI = 30.3

demonstrated product purity oscillation for the +20% throughput change with different controller gains. A benefit can be observed from Kc = 2 to Kc = 20, but there is only slight improvement for further increase of the Kc value. The following dual-point temperature control structure is expected to enhance the performance under large throughput disturbance. 4.2. Dual-Points Temperature Control Structure. Figure 10(B) shows the dual-points temperature control structure. According to the results of open-loop sensitivity analysis (see Figure 9), the proper controlled trays are the 5th and the 15th. The feed ratio (FR) and the side stream flow rate (Fside) are used to control these tray temperatures. Relative Gain Analysis (RGA)31 is used to find proper pairing of the controlled and the manipulated variables. Equations 14 and 15 depict the steady gain and RGA paring of the process.

(11)

(12)

where xEtOH, Feed is the EtOH composition of FEtOH. Therefore, the side stream can be determined by the EtOH feed flow rate and further modified by a composition analyzer to detect the composition variation of EtOH in the feed stream. Combination of eqs 11 and 12 provides an estimate of the side stream flow rate, such as, Fside ≅ FEtOH × xEtOH , Feed

Kc = 20 Kc = 0.5 τI = 0.3

Coolant flow rate in condenser Coolant flow rate in condenser Coolant flow rate in subcooler Steam flow rate in heater Feed ratio

where FEtOH is the fresh EtOH feed flow rate, and 0.87 is the steady state feed EtOH composition. However, when the feed EtOH composition varies, the relation will be inaccurate. The error between eq 11 and perfect control is always related to the feed composition, which can be written as Error ≅ FEtOH × (xEtOH , Feed − 0.87)

Dual-points Kc = 2

Valve position

the side stream flow rate according to eq 10. So the remaining variable is only the side stream flow rate. Based on the results of close-loop sensitivity analysis (see Table 4), it is found that the ratio of the side stream flow rate to the feed flow rate is nearly fixed over throughput changes, such as follows:

Fside ≅ FEtOH × 0.87

Single-point

⎡T5 ⎤ ⎡−4.67 571.86 ⎤⎡ F ⎤ side ⎢ ⎥=⎢ ⎥⎢ ⎥ ⎢⎣T15 ⎥⎦ ⎣ 34.26 4553.94 ⎦⎣ FR ⎦

Fside

(13)

FR

Λ = ⎡ 0.520482 0.479518 ⎤T 5 ⎢ ⎥ ⎣ 0.479518 0.520482 ⎦T15

To measure the EtOH composition of FEtOH, a composition analyzer is installed which is represented as “AT” in Figure 10(A). Parameters for all controllers are listed in Table 5. The temperature controller of the RD column is tuned from the closed-loop ATV method,30 and then the Tyreus−Luyben tuning rule is used to set the controller parameters. Figures 11(A)−(C) show the dynamic responses under different disturbances. It can be noticed that, for large throughput change, there would be a large oscillation in product purity. In this single-point temperature control structure, the side stream flow rate is determined by the feed flow rate. As the feed flow rate changes, the side stream flow rate would also change immediately. But the change of the reboiler duty is not as fast as the side stream flow rate. Hence, the product purity of this case would have a large oscillation, especially for the positive throughput change. This phenomenon can be improved by making the sump level controller more aggressive. Here, larger controller gain is selected for the sump level controller and the other level controllers which related to the recycle to make the response more quick. Figure 13

(14)

(15)

The result indicates that the side stream flow rate should be used to control the 5th tray temperature, and the feed ratio controls the 15th tray temperature. This relation is consistent with the heuristic pairing rule, in which the feed ratio (lower manipulated variable) is used to control the lower stage tray temperature, and the side stream flow rate (higher manipulated variable) is used to control the higher tray stage temperature. Controller parameters are also given in Table 5; the temperature controllers’ settings of the RD column are tuned by using the closed-loop ATV method and the Tyreus−Luyben tuning relation. Figures 12(A)−(C) show the dynamic responses under throughput and composition disturbances. Although the response time is slightly slower than the single-point temperature control structure, the deviation of the product purity is much smaller in the dual-points temperature control, especially for the disturbance from HAc composition (Figure 13). The reason is that, for the single-point temperature control K

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 11. Dynamic responses of single-point temperature control, (A) throughput disturbance; (B) HAc disturbance; (C) EtOH disturbance.

structure, the side stream flow rate is estimated by the EtOH feed flow rate. From the close-loop sensitivity analysis in Table

4, while the HAc composition is different from its steady state value, the set point of the side stream flow rate is still kept at L

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 12. Dynamic responses of dual-point temperature control: (A) throughput disturbance; (B) HAc disturbance; and (C) EtOH disturbance. M

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Industrial & Engineering Chemistry Research

Article

■

APPENDIX In this study, the unit costs are based on Douglas’ book, Conceptual Design of Chemical Processes,24 and a payback period of three years is used for the exchanger and column, and the membrane lifetime is assumed as three years during the calculation. The M&S index is taken as 1448.3 (2010, 1st quarter). A. Height of Column

LC [ft] = 2.3 × (NT − 1)

where NT is the total number of trays in the column B. Cost of Column Column Cost [$] =

M&S 0.802 × 101.9 × D1.066 × (2.18 + 3.67) C LC 280

where DC [ft] is the diameter of the column.

Figure 13. Product purity response with different Kc settings under +20% throughput change.

C. Cost of Tray Tray Cost [$] =

the same value. Thus, there will be product deviation in the single-point temperature control in this kind of disturbance. However, there is still a disadvantage in the dual-points temperature control structure. When the throughput changes, the side stream flow rate also oscillates. Meanwhile, the membrane inlet temperature also oscillates with the side stream flow rate due to the relation of eq 10. There would be a maximum temperature limitation which the membrane can afford in the real application of the membrane module. When the oscillation of the side stream flow rate is too large, the membrane may exceed its temperature limitation and further damage the membrane. To avoid this situation, a low selector is necessary before the set point signal is sent to the membrane heater.

M&S × 4.7 × D1.55 C LC × (1 + 1.8 + 1.7) 280

D. Reboiler Heat Transfer Area

AR [ft2] =

QR UR ΔTR

where UR is 250 [BTU/(h ft2)]. E. Condenser Heat Transfer Area

A C [ft2] =

QC UCΔTC

where UR is 150 [BTU/(h ft2)]. F. Cost of Heat Exchanger

Heat Exchanger Cost [$] =

5. CONCLUSION This paper proposes a novel EtAc production process consisting of the reactive distillation (RD)−pervaporation (PV) hybrid configuration. The RD−PV hybrid process is targeted to synthesize 99 mol % EtAc with 87 mol % EtOH and 95 mol % HAc fed into the bottom of the RD column. A decanter located at the top of the RD column removes water from the hybrid process. A side stream with around 96 mol % EtAc is drawn from the rectifying section of the RD column and is fed to the membrane modules to enhance the EtAc purity to the desired specification (99 mol %). About 13% energy savings can be achieved when compared to the conventional two-column process. Total annual cost (TAC) analysis is used to find the optimal design. The proposed hybrid process would become more competitive over the conventional two-column design once the membrane price is less than $271/ m2. Two control strategies, including the single-point temperature control structure and the dual-points temperature control structure, are then investigated for studying rejection of the possible main disturbances to this process, such as throughput change and EtOH and HAc feed compositions. Simulation tests show that both control structures are able to maintain high product purity and give acceptable transient responses to those major disturbances. However, the single-point temperature control structure requires an additional composition analyzer on the EtOH feed stream to adjust the side stream flow rate to the membrane modules.

M&S × A0.65(2.29 + FC) 280

where FC is 5.0625 for the reboiler and 3.75 for the condenser. G. Cost of Membrane Material20

Membrane Material Cost [$] = 387.5/m 2 H. Cost of Membrane Module20

Membrane Module Cost [$] = 125550 ×

⎛ A ⎞0.3 ⎜ ⎟ ⎝ 324 ⎠

where A [m2] is the membrane area for each module. I. Cost of Vacuum Pump26

Vacuum Pump Cost [$] ⎛ 60 × Fp × 8.314 × 273.15 ⎞0.55 = 4200 × ⎜ ⎟ 3600 × 101.325 ⎠ ⎝

where Fp is the total permeate rate through the membrane [kmol/h]. J. Cost of Feed Pump26

Feed Pump Cost [$] = 26700 ×

⎛ 24 × 3600 × FF ⎞0.53 ⎜ ⎟ ⎝ ⎠ 50000

where FF is the total feed rate to the pump [m3/s]. N

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research K. Steam and Cooling Water Cost

Annual Cost of Steam =

(2) Lai, I. K.; Hung, S. B.; Hung, W. J.; Yu, C. C.; Lee, M. J.; Huang, H. P. Design and control of reactive distillation for ethyl and isopropyl acetates production with azeotropic feeds. Chem. Eng. Sci. 2007, 62, 878−898. (3) Keyes, D. Esterification processes and equipment. Ind. Eng. Chem. 1932, 24, 1096−1103. (4) Agreda, V. H.; Partin, L. R. Reactive distillation process for the production of methyl acetate. U.S. Patent 4,435,595, April 26, 1982. (5) Kober, P. A. Pervaporation, perstillation and percrystallization. J. Membr. Sci. 1995, 100, 61−64. (6) Lipnizki, F.; Field, R. W.; Ten, P.-K. Pervaporation-based hybrid process: a review of process design, applications and economics. J. Membr. Sci. 1999, 153, 183−210. (7) Van der Bruggen, B.; Luis, P. Pervaporation as a tool in chemical engineering: a new era? Curr. Opin. Chem. Eng. 2014, 4, 47−53. (8) Van der Bruggen, B. 3.06-Pervaporation Membrane Reactors. In Comprehensive Membrane Science and Engineering; Drioli, E., Giorno, L., Eds.; Elsevier: Oxford, 2010. (9) Waldburger, R. M.; Widmer, F. Membrane reactors in chemical production processes and the application to the pervaporation-assisted esterification. Chem. Eng. Technol. 1996, 19, 117−126. (10) de la Iglesia, Ó .; Mallada, R.; Menéndez, M.; Coronas, J. Continuous zeolite membrane reactor for esterification of ethanol and acetic acid. Chem. Eng. J. 2007, 131, 35−39. (11) Jyoti, G.; Keshav, A.; Anandkumar, J. Review on Pervaporation: Theory, Membrane Performance, and Application to Intensification of Esterification Reaction. J. Eng. 2015, 2015, 927068. (12) Bausa, J.; Marquardt, W. Shortcut design methods for hybrid membrane/distillation processes for the separation of nonideal multicomponent mixtures. Ind. Eng. Chem. Res. 2000, 39, 1658−1672. (13) Aiouache, F.; Goto, S. Reactive distillation−pervaporation hybrid column for tert-amyl alcohol etherification with ethanol. Chem. Eng. Sci. 2003, 58, 2465−2477. (14) Buchaly, C.; Kreis, P.; Górak, A. Hybrid separation processes Combination of reactive distillation with membrane separation. Chem. Eng. Process. 2007, 46, 790−799. (15) Lv, B.; Liu, G.; Dong, X.; Wei, W.; Jin, W. Novel reactive distillation−pervaporation coupled process for ethyl acetate production with water removal from reboiler and acetic acid recycle. Ind. Eng. Chem. Res. 2012, 51, 8079−8086. (16) Luyben, W. L. Control of a Column/Pervaporation Process for Separating the Ethanol/Water Azeotrope. Ind. Eng. Chem. Res. 2009, 48, 3484−3495. (17) 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. (18) 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. (19) Hangx, G.; Kwant, G.; Maessen, H.; Markusse, P.; Urseanu, I. Reaction kinetics of the esterification of ethanol and acetic acid towards ethyl acetate. Deliverable 22, Intelligent Column Internals for Reactive Separations (INTINT); Technical Report to the European Commission; 2001. (20) Santoso, A. Design and Control of Hybrid DistillationMembrane Systems for Separating Azeotropic Mixtures. M.S. Thesis, National Taiwan University, Taiwan, 2010. (21) Geankoplis, C. J. Transport processes and separation process principles; Prentice Hall Professional Technical Reference: New York, 1993. (22) Burkett, R. J. In Choosing the right control structure-examples from the petrochemical industry; Choosing the Right Control Structure for Your Process (Digest No. 1998/280), IEE Colloquium on; 1998; IET: 1998; pp 8/1−8/3. (23) Seider, W. D.; Seader, J. D.; Lewin, D. R. Product and process design principles: synthesis, analysis, and evaluation; Wiley: New York, 2004. (24) Douglas, J. M. Conceptual design of chemical processes; McGrawHill: New York, 1988.

Q $Cs × H × (8150) 1000 912

Annual Cost of Cooling Water =

Q 0.03 × C × (8150) 8340 30

L. Power of Vacuum Pump27

Annual Cost of Power ⎧ ⎪⎛ Fp ·8.314· T ⎞⎛ k r ⎞ = 8150 × 0.04 × ⎨⎜ ⎟ ⎟⎜ ⎪⎝ 3600 ⎠⎝ k r − 1 ⎠ ⎩ ⎡⎛ ⎤⎫ ⎞(k r /k r − 1) ⎢⎜ 1.013 ⎟ ⎥⎪ − 1⎥⎬ ⎢⎜ P ⎟ ⎣⎝ op ⎠ ⎦⎪ ⎭

where kr is the heat capacity ratio. Pop is the pressure on the permeate side. M. Chilled Water Cost

In this study, the permeate side is under 0.16 bar. In order to cool it to the liquid, the usage of chilled water is necessary, which is $4.43/GJ. N. Raw Material Cost32

Acetic Acid: 33 [US CTS/LB] Ethanol: 62.5 [GBP/HL] O. Catalyst Cost (assuming a catalyst life of 3 months)

Catalyst Cost [$] = catalyst loading [kg] × 7.7162

■

$ kg

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.iecr.6b01358. Table S1, NRTL model parameters; Table S2, HOC parameters; Table S3, composition and temperature of the azeotropes for the EtAc system; Table S4, kinetic model for the EtAc esterification reaction; Table S5, pervaporation parameters of each species in the EtAc system; Figure S1, effect of operation temperature on flux; Figure S2, effect of operation temperature on permeate concentration (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel: +886-2-33663039. Fax: +8862-23623040. Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS This work is supported by the Ministry of Science and Technology Taiwan under grant MOST 104-2218-E-002-006. REFERENCES

(1) 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. O

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research (25) Szitkai, Z.; Lelkes, Z.; Rev, E.; Fonyo, Z. Optimization of hybrid ethanol dehydration systems. Chem. Eng. Process. 2002, 41, 631−646. (26) Woods, D. R. Cost Estimation for the Process Industries; McMaster University: Hamilton, Ontario, Canada, 1983. (27) Oliveira, T. A.; Cocchini, U.; Scarpello, J. T.; Livingston, A. G. Pervaporation mass transfer with liquid flow in the transition regime. J. Membr. Sci. 2001, 183, 119−133. (28) Tusel, G.; Brüschke, H. Use of pervaporation systems in the chemical industry. Desalination 1985, 53, 327−338. (29) Hooke, R.; Jeeves, T. A. ̀̀ Direct Search’’ Solution of Numerical and Statistical Problems. J. Assoc. Comput. Mach. 1961, 8, 212−229. (30) Åström, K. J.; Hägglund, T. Automatic tuning of simple regulators with specifications on phase and amplitude margins. Automatica. 1984, 20, 645−651. (31) Bristol, E. On a new measure of interaction for multivariable process control. IEEE Trans. Autom. Control 1966, 11, 133−134. (32) ICIS Home Page. http://www.icis.com/ (accessed Oct 12, 2015).

P

DOI: 10.1021/acs.iecr.6b01358 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX