Design and Control of an Isopropyl Alcohol Dehydration Process via

790

Ind. Eng. Chem. Res. 2008, 47, 790-803

Design and Control of an Isopropyl Alcohol Dehydration Process via Extractive Distillation Using Dimethyl Sulfoxide as an Entrainer Saiful Arifin and I-Lung Chien* Department of Chemical Engineering, National Taiwan UniVersity of Science and Technology, Taipei 106, Taiwan

In this paper, design and control of an isopropyl alcohol (IPA) dehydration process via extractive distillation have been investigated. The heavy-boiling entrainer used to aid the separation is dimethyl sulfoxide (DMSO). The design flowsheet includes an extractive distillation column and an entrainer recovery column with the top product of the extractive distillation column to be IPA and the top product of the entrainer recovery column to be water. The bottom product of the entrainer recovery column is the recovered DMSO which is recycled back to the extractive distillation column. The optimal design flowsheet of this complete process has been established showing that the total annual cost and the needed steam cost of this design flowsheet is significantly less than a competing design flowsheet via heterogeneous azeotropic distillation. A very simple overall control strategy has also been proposed which requires only one tray temperature control loop in each column to hold the high-purity specifications of the two products. Dynamic simulations reveal that fixing of the reflux ratio is not a suitable control strategy. Instead, the strategy to fix the two reflux flow rates should be used to reject feed disturbances. 1. Introduction

Table 1. NRTL Model Parameters of This System

Isopropyl alcohol (IPA) is widely used in the semiconductor industry as a cleaning agent, thus the recovery of this cleaning agent from the waste solvent stream is an important topic worthy of detailed study. This waste solvent stream contains mainly isopropyl alcohol and water which contains a minimum-boiling azeotrope and, thus, is difficult to separate. The usual practice in industry for the separation of a mixture with these two components is to add cyclohexane (CyH) as an entrainer via heterogeneous azeotropic distillation.1-3 However, it is wellknown that heterogeneous azeotropic distillation can exhibit hign parametric sensitivity, multiple steady-states, long transients, and nonlinear dynamics,4 which can limit the operating range of this IPA dehydration system under feed disturbances. An alternative way for the separation of this mixture is to use extractive distillation by adding a heavy-boiling entrainer into this system. The entrainer acts to increase the relative volatility of the IPA and water mixture far away from unity without inducing liquid-liquid phase separation and also does not form additional azeotropes with any of the other components. Entrainer screening is an important step in extractive distillation before designing a distillation sequence. Several rules should be followed to get a good entrainer candidate for azeotropic mixtures.5-8 The presence of the entrainer is the key difference between extractive distillation and simple distillation; however, it presents more degrees of freedom in the overall system to be designed. The proper amount of the entrainer feed should be designed for economical purposes which complicates the design and optimization of extractive distillation. Quite a few papers9-19 in the open literature have studied the feasibility of the design and synthesis of this type of processes. A residue curve map (RCM) can be used as a simple method for designing and distinguishing between feasible and infeasible sequences in extractive distillation.20 Thermal integration in extractive distillation has also been studied in the literature.21 In extractive distillation, infinite reflux does not always imply maximum separation. Separations which are infeasible at infinite * To whom correspondence should be addressed. Tel.:+886-22737-6652. Fax: +886-2-2737-6644. E-mail: [email protected].

comp,i comp,j aij aji bij bji cij

IPA H2O 0 0 185.4 777.3 0.50

IPA DMSO 0 0 115.2787 -25.0123 0.30

Aspen Plus NRTL:

∑ xτ G j ji

ln γi )

ji

j

∑

+ xkGki

∑ j

k

where Gij ) exp(-Rijτij)

xjGij

∑

xkGkj

k

τij ) aij +

H2O DMSO -1.2449 1.7524 586.801 -1130.215 0.30

[ ] ∑x τ

m mjGmj

τij -

m

∑xG k

kj

k

bij T

Rij ) cij, τii ) 0, Gii ) 1.

reflux may be feasible at finite reflux. This is an important property of extractive distillation and plays an important role in the entrainer selection procedure.22 Bifurcation/multiple steady states are well-known in extractive distillation. It has great effect upon the flexibility, operability, and controllability of the column.23-25 Steady state design also effects the complexity of the dynamic and control problem of an extractive column.26 Several authors27-30 have studied the dynamic and control of extractive distillation. In a recent paper by Luyben,31 the plantwide control strategy of an isopropyl alcohol process using ethylene glycol as the entrainer has been studied. In this paper, instead of using ethylene glycol as the entrainer, dimethyl sulfoxide (DMSO) will be used for isopropyl alcohol dehydration. Gmehling and Mo¨llmann32 studied five entrainers for the IPA-water system and found dimethyl sulfoxide (DMSO) to be the best candidate which has the largest separation factor and introduces no further azeotrope in the system. The optimal design and overall control strategy of this isopropyl alcohol dehydration process will be studied in this paper. The fresh feed and product specifications will be chosen to be exactly the same as an earlier paper by Arifin and Chien3 so that direct comparison of the two design flowsheets can be made. Section 2 shows the thermodynamic model

10.1021/ie070996n CCC: $40.75 © 2008 American Chemical Society Published on Web 01/08/2008

Ind. Eng. Chem. Res., Vol. 47, No. 3, 2008 791

Figure 1. T-x-y and x-y experiment and predicted plots at 1 atm for the IPA-H2O-DMSO system.

used in this investigation. Vapor-liquid equilibrium from previous experimental data and azeotropic composition and temperature for this system in the literature will be compared with our model to demonstrate the reliability of this model. Section 3 shows the optimal design flowsheet of the overall process. The optimal design flowsheet via extractive

distillation will be compared to the design flowsheet via heterogeneous azeotropic distillation. Section 4 will present the study on the overall control strategy. Only the tray temperature control strategy will be used for wider industrial applications. Some concluding remarks will be given in Section 5.

792

Ind. Eng. Chem. Res., Vol. 47, No. 3, 2008

Figure 2. RCM and equivolatility plots at 1 atm for the IPA-H2O-DMSO system.

Figure 3. Proposed flowsheet of the IPA dehydration process.

2. Thermodynamic Model Used in the Simulation According to Gmehling,33 isopropy alcohol (normal bp 82.35 °C) forms a minimum-boiling azeotrope with water (normal bp 100 °C) at 1 atm with an azeotropic temperature of 80.10 °C

and azeotropic composition of 68.70 mol % IPA. Thus, these two components cannot be separated in a single column. In this paper, dimethyl sulfoxide (DMSO) will be used as the entrainer to aid the separation. The reason to choose DMSO, according


Figure 5. Optimization results for case 3 with N2 and NF2 as the design variables. Table 2. Breakdown of TAC (TAC dollars × 1000) case 1

case 2

case 3

first column reboiler cost 55.84 condenser cost 29.72 column cost 154.42 tray cost 14.65 steam cost 158.99 cooling water cost in condenser 2.50 416.12

55.42 45.84 185.41 19.19 157.18 4.86 467.90

55.67 32.51 153.53 14.52 158.26 2.87 417.36

45.20 25.47 99.34 8.55 168.94 2.81 350.31 0.27

44.47 25.07 101.77 8.79 164.73 2.74 347.57 0.29 33.67 2.01 800.90

second column reboiler cost 44.59 condenser cost 25.19 column cost 94.65 tray cost 8.03 steam cost 165.41 cooling water cost in condenser 2.76 340.63 entrainer cost 0.70 cooler cost 53.96 cooling water cost in cooler 2.40 TAC 813.81

818.48

to Table 6 of the paper by Gmehling and Mollmann32 to compare five entrainers, is that DMSO has the largest separation factor (R∞1,2 ) 5.108) and also introduces no further azeotrope in the system. In comparison, the separation factor for ethylene glycol as entrainer for this system is lower at 4.234. For this overall three-component system, the NRTL model will be used to describe the nonideality of the liquid phase while the vapor phase is assumed to be ideal. The NTRL model parameters of the IPA-H2O pair are taken from the work of Wang et al.,34 and the model parameters of the other two pairs are taken from Aspen Plus. The complete NRTL model parameters are shown in Table 1. All other physical property model parameters are taken from the built-in values in Aspen Plus.

Figure 4. Typical optimization results for case 3.

In order to verify the validity of the thermodynamic model, the y-x and T-xy vapor-liquid equilibrium plots predicted by the above model will be compared with the experimental data from Gmehling and Onken.35 Figure 1 demonstrates that the model fits the experimental date very well (note that no experimental data was found for the IPA-DMSO pair). The predicted azeotropic temperature and azeotropic composition

794


Figure 6. Optimal process flowsheet for this IPA dehydration process.

Figure 7. Material balance lines for this IPA dehydration process.

(at 1 atm) from this model are at 80.00 °C and 68.88 mol % IPA, respectively. Again, very good agreement has been achieved. The RCM and equivolatility curves for this three-component system are shown in Figure 2. From the RCM, we can see that the IPA-H2O azeotrope is the unstable node, DMSO is the stable node, and both IPA and water are the saddles. For the equivolatility curves, the relative volatility between IPA and water was calculated in the presence of DMSO and the equal values of the relative volatility at different compositions were collected. It is noticed that the IPA/water mixture can be separated easier in the presence of DMSO because of higher values of the relative volatility. 3. Steady State Design and Economic Analysis 3.1. Design Flowsheet via Extractive Distillation. The proposed flowsheet for the IPA-water separation with DMSO as the entrainer can be seen in Figure 3. The IPA-water mixture

and the entrainer are fed into the extractive distillation column with the stages above the entrainer feed as the rectifying section, the stages between the feed stages of the entrainer and the IPAwater mixture as the extractive section, and the stages below the feed stage of the IPA-water mixture as the stripping section. The presence of DMSO alters the relative volatility between IPA and water causing IPA to move toward the top part and water to move toward the bottom part of the column. In the rectifying section, there is essentially no water, thus the simple separation between IPA and DMSO is performed with pure IPA going to the distillate and DMSO returning to the extractive section for entrainer usage. In the stripping section, IPA as the lightest component is stripping toward the extractive section of the column resulting in only negligible IPA in the column bottoms stream. This column bottoms stream is fed into the entrainer recovery column to produce almost pure water in the distillate and almost pure DMSO in the column bottoms. DMSO as a heavy entrainer will be recycled back to the extractive


Figure 8. Liquid composition profiles of the two columns.

Figure 9. Optimal design flowsheet for the IPA dehydration via heterogeneous azeotropic distillation.

distillation column. To balance tiny entrainer losses in both D1 and D2 streams, small makeup stream of entrainer should be added. The product specifications are set to be exactly the same as in the work of Chien et al.2 and Arifin and Chien.3 They are the following: 99.9999 mol % of IPA in D1 and 99.9 mol % of water in D2. The reason for the ultrahigh purity in IPA is for semiconductor industry usage. The feed composition as in a typical waste IPA stream in the semiconductor industry contains equal moles of IPA and water. With this feed composition not far from the azeotropic composition, the flowsheet arrangement with a preconcentrator column was not considered for the sake of having a simpler process as well as simpler instrumental and control equipment. The feed flow rate is assumed to be 100 kmol/h () 1666.67 mol/min), also exactly the same as in the work of Arifin and Chien.3 For easy operation purposes, reflux drums of both columns are operated at atmospheric pressure with the column pressure drop automatically calculated in the Aspen simulation. The design variables to be determined in the flowsheet include the following: the feed ratio (FE/FF), total stages of the extractive distillation column (N1), entrainer and fresh feed tray locations (NFE and NFF), total stages of the entrainer recovery

column (N2), and feed tray location of the entrainer recovery column (NF2). As can be seen in the work of Knight and Doherty36 and also summarized in Chapter 5 of the book by Doherty and Malone,37 entrainer feed temperature can also be considered as another design variable, thus a cooler is included in Figure 3. There are too many design variables that need to be determined; thus for simplification purposes, the more important extractive distillation column is optimized first. We considered three different cases for the entrainer feed temperature, they are the following: Case 1: Subcooling the entrainer feed temperature (from B2) to 40 °C. Case 2: Operating the entrainer feed temperature the same as the temperature in the B2 stream (196.6 °C) with no cooler. Case 3: Operating the entrainer feed temperature 5-15 °C below the top temperature of the extractive distillation column (as suggested by Knight and Doherty36). We used 72 °C in the following simulation. In each case, there are four design variables that need to be determined (FE/FF feed ratio, N1, NFE, and NFF). A sequential iterative optimization search is used to find the optimal design with feed ratio as the outer iterative loop, N1 as the middle loop,

796


Figure 10. Conceptual design and material balance lines for the IPA dehydration via heterogeneous azeotropic distillation.

Figure 11. Open-loop sensitivity plots for (0.1% changes in Q1 or Q2.

and NFE and NFF as the inner iterative loop. The two design specifications for all the Aspen simulations are the following: setting the top composition at 99.9999 mol % IPA and setting the ratio of IPA to water in the bottom stream to be 0.001. The

reason for this bottom specification is to set the IPA loss through the column bottoms. The above two design specifications can be met by varying the remaining two degrees of freedom in this column (e.g., reboiler duty and reflux flow rate).


Figure 12. Overall control strategy with reflux ratio fixed.

Total annual cost (TAC) is used as the objective function to be minimized which includes annualized capital costs and operating costs. The capital costs include the column shell, trays, reboiler, and condenser, and a payback period of 3 years is assumed. The calculation formula for the above equipment can be found in the work of Douglas.38 The operating costs include the steam and cooling water for the operation of the reboiler and condenser. The optimization procedure to minimize the TAC by varying the four design variables is summarized below: (1) Guess the FE/FF feed ratio. (2) Guess the total stages of the extractive distillation column (N1). (3) Guess the entrainer feed location (NFE) and the fresh feed location (NFF). (4) Change the reboiler duty and the reflux flow rate until the two design specifications can be met. (5) Go back to step 3 and change NFE and NFF until TAC is minimized. (6) Go back to step 2 and change N1 until TAC is minimized. (7) Go back to step 1 and change the FE/FF feed ratio until TAC is minimized. Figure 4 shows some typical results for case 3. The top shows the results by varying NFE and NFF with N1 fixed at 41 and the FE/FF feed ratio fixed at 1.025. From this top plot, the best NFE is at the 7th stage and the best NFF is at the 35th stage (stages are counted from top to bottom with the condenser as the first stage and the reboiler as the last stage). The middle plot is the collection of all the similar top plots at various N1 values and fixing the feed ratio at 1.025. It is noticed that N1 ) 41 is the best. The bottom plot is the summary of all the simulation results at various feed ratios with the best feed ratio as 1.025. For case 3, the final optimal design variables are as follows: FE/FF feed ratio at 1.025, N1 ) 41, NFE ) 7, and NFF ) 35. A similar study can be done for cases 1 and 2, and the resulting optimal design variables for case 1 are as follows: FE/ FF feed ratio at 1.00, N1 ) 41, NFE ) 8, and NFF ) 35. The resulting optimal design variables for case 2 are as follows: FE/ FF feed ratio at 0.95, N1 ) 40, NFE ) 6, and NFF ) 34. After the optimal design variables for the extractive distillation column are determined, the total TAC can be calculated with the entrainer recovery column and the recycle stream included. Additional costs in the total TAC include the annualized capital

cost for the entrainer recovery column, the costs associated with the cooler from B2 to the entrainer feed, the operating costs of the steam and cooling water to operate the entrainer recovery column, and the entrainer makeup cost. As an example, Figure 5 shows the results of the optimization runs for case 3 with N2 and NF2 as the design variables where the y-axis is the total TAC of the complete flowsheet. From the figure, N2 should be 24 and NF2 should be at the ninth stage. Table 2 shows the minimized TAC results for cases 1, 2, and 3. It is noticed that the TAC of case 3 is the lowest which verifies the recommendation by Knight and Doherty.36 The subcooling of the entrainer recycled temperature makes the TAC of the extractive distillation column lower; however, the additional cooler cost is imposed. Therefore, a tradeoff can be made for the degree of subcooling resulting in case 3 to be better than case 1. Comparing to the case with no cooling at all (case 2), there is only a slight reduction of the TAC by 2.1% for the best case (Case 3). The effect of the feed ratio is also surprisingly small if for each case the optimal design is carefully determined. For example, the bottom plot of Figure 4 shows the summary effect of the changing feed ratio to the TAC of the extractive distillation column. In a feed ratio range from 0.9 to 1.1, the best case (FE/FF ) 1.025) only reduces the TAC from the worst case (FE/FF ) 0.90) by 0.27%. On the contrary, the determination of the entrainer feed location is relatively more important. With the feed ratio, N1, and NFF at optimal values, the reduction of the TAC from changing NFE from the sixth stage to the seventh stage can be as big as 7.3%. This reveals that a sufficient number of stages is needed in the rectifying section to satisfy the high-purity specification. In this study, heat integration was not considered. Presumably feed-effluent heat exchangers can be placed for preheating of B1 into F2 by cooling of B2 or for preheating of fresh feed into extractive distillation column by cooling of B2. However, additional heat exchange equipment will be needed; thus, detailed calculations need to be made to see if it is worthwhile. The final optimal flowsheet for this system is shown in Figure 6, and the material balance lines for this separation process are demonstrated in Figure 7. The liquid composition profiles of the two columns for the flowsheet in Figure 6 are shown in

798


Figure 13. Closed-loop responses for (20% changes in feed IPA composition using overall control strategy in Figure 12.

Figure 8. From top plot of Figure 8, it is noticed that the goal of the extractive section is fulfilled in eliminating water going into the rectifying section. Because the product specification is ultrapure, thus, quite a few stages are needed to serve this purpose. We will compare this flowsheet to the optimized design flowsheet via heterogeneous azeotropic distillation next. 3.2. Comparison to the Design Flowsheet Using Heterogeneous Azeotropic Distillation. Arifin and Chien3 developed an overall design flowsheet using cyclohexane as an entrainer via heterogeneous azeotropic distillation. Two columns were also used in their overall design flowsheet with one column serving as the combined preconcentrator/recovery column and another one as the heterogeneous azeotropic column with a

decanter connected at the column’s top. The optimal design flowsheet of their two-column system is shown in Figure 9. Because the feed rate, feed composition, and the two product specifications are exactly the same, direct comparison can be made. Comparing the optimal flowsheet developed in this paper via extractive distillation to that of Figure 9 via heterogeneous azeotropic distillation, it is found that the extractive distillation column is much taller (NT ) 40 if not including a condenser) than the heterogeneous azeotropic column (NT ) 18). However, the extractive distillation column is much thinner (column diameter of 0.78 m) than the heterogeneous azeotropic column (column diameter of 1.61 m). The total annual cost of the overall process via extractive distillation is $8.009 × 105 (see Table


Figure 14. Overall control strategy with reflux flow fixed.

2). This can be directly compared to the TAC of the overall process via heterogeneous azeotropic distillation ($1.190 × 106; see Table 1 in Arifin and Chien3). This shows that this separation via extractive distillation reduces the TAC by as much as 32.7%. The significant reduction is also in the needed steam cost for the separation. The total steam cost for the overall process via extractive distillation from Table 2 is $3.230 × 105, while via heterogeneous azeotropic distillation from Table 1 of the paper by Arifin and Chien3 it is $4.637 × 105. A reduction of 30.3% of the needed steam cost can be realized via extractive distillation. The main reason for the high TAC and energy requirement can be explained by the particular residue curve map (RCM) and material balance lines for the feasible separation via heterogeneous azeotropic distillation for this chemical system. A conceptual diagram of the separation via heterogeneous azeotropic distillation can be seen in Figure 10 with the dashedline showing the material balance around the heterogeneous azeoptropic distillation column. The material balance lines in the bottom plot of Figure 10 represent the ideal case when column bottoms reach the composition of pure IPA, the column top vapor approaches a ternary azeotrope, and D2 is right at the distillation boundary. With this ideal case, we can easily estimate top vapor flow using the formula noted in Figure 10. In this calculation of the top vapor flow rate, B1 can be estimated to be half of the fresh feed flow rate and “b” and “a” can easily be determined by a line between D2 and organic reflux and another line between B1 and V1. This means that the top vapor flow rate will be quite large resulting in a high organic reflux flow rate back into the column and also high aqueous outlet flow rate into the C2 column. This in turn requires more reboiler duty for the two columns to generate the vapor rate going up the column and larger diameters for the two columns. On the other hand, there is no distillation boundary in the RCM for the separation via extractive distillation. In fact, the reflux flow rate for the extractive distillation column is only 527.81 mol/min (comparing to organic reflux flow rate of 3053.21 mol/min in the heterogeneous azeotropic distillation system). The two only possible drawbacks for the overall process via extractive distillation are that the extractive distillation column needs to be very high; this may cause some difficulty for the construction of this column, and the two columns need

to be operated at higher temperatures. Thus, steam with higher pressure will be needed for the two reboilers. 4. Overall Control Strategy Development The steady-state economics of the overall process via extractive distillation is much better than the overall process via heterogeneous azeotropic distillation. In the following, we will investigate the control strategy of this complete process in Figure 6. We would like to restrict ourselves to use tray temperature control loop(s) to indirectly hold the product purity for wider industrial application purpose. Large feed disturbances ((20% changes in the feed composition and also (20% changes in the fresh feed flow) will be used to test the proposed overall control strategy. Pressure-driven simulation in Aspen Dynamics is used in the control strategy development. Before converting the Aspen Plus result to Aspen Dynamics, sizing of all equipment is needed. The tray sizing tool in Aspen Plus is used to calculate the column diameters of both columns to be 0.78 and 0.82 m for the first and the second column, respectively. Tray spacing and weir height of both columns are assumed to be 0.6096 and 0.0508 m, respectively. A 10-min holdup time with a 50% liquid level is used to calculate volume of each column base and reflux drum of each column. The top pressures of the first and the second column are set at atmospheric pressure. The tray rating tool in Aspen Plus is used to automatically calculate the pressure drop along both columns. Some simple regulatory control loops are determined first: the levels of the reflux drums for both columns are controlled by manipulating the distillate; top pressures of both columns are controlled by the condenser duty; the bottom level of the extractive distillation column is controlled by manipulating the bottom flow; and the entrainer feed temperature is controlled at 72 °C by manipulating the cooler duty. For the two reflux flows, an immediate thinking is to keep the two reflux ratios at their nominal values during disturbances. This was also used in the overall control structure by Luyben.31 An important inventory control loop in this overall process is the bottom level of the entrainer recovery column. The control of this level was suggested by Grassi39 and Luyben31 to be held by the entrainer makeup flow. However, because this flow is very small, the bottom level essentially floats as changes in the

800


Figure 15. Closed-loop responses for (20% changes in feed IPA composition using overall control strategy in Figure 14.

entrainer flow rate occur. With this control pairing, the entrainer feed to the first column is flow-controlled by Grassi39 and modified by Luyben31 by adding a first-order lag, a high selector, and a low selector to determine the entrainer feed and fresh feed flow rates. We adapted this control pairing for the overall control strategy in our study by using entrainer makeup flow to control the bottom level of the entrainer recovery column and using a flow-controlled entrainer feed. The above control strategy leaves two reboiler duties which can be used by a tray temperature control loop in each column. Figure 11 shows the results of an open-loop sensitivity analysis with (0.1% changes in reboiler duty of the extractive distillation column or the entrainer recovery column. The temperature control point will be selected at a particular stage with high

sensitivity and also with near-linear behavior. For the extractive distillation column, the temperature at the 38th stage (location shown with a perpendicular line in the top plot in Figure 11) is chosen as the control point, and for the entrainer recovery column, the temperature at the 12th stage (location also shown with a perpendicular line in the bottom plot in Figure 11) is chosen as the control point. There are two temperature breaks for the entrainer recovery column; however, the temperature control point in the rectifying section (e.g., the sixth stage) was not chosen because of the inferior open-loop dynamic response due to larger deadtime. The control strategy of the overall process is summarized in Figure 12. Proportional (P) only controllers with Kc ) 2 are used for both reflux drum levels as suggested in the work of Luyben40


Figure 16. Closed-loop responses for (20% changes in fresh feed flow rate using overall control strategy in Figure 14.

and Kc ) 10 is used for both base bottom levels for faster dynamics of the internal flow of the overall process and also for faster increase or decrease of entrainer makeup into the system. The proportional and integral (PI) settings of the top pressure control loops for both columns are set at Kc ) 20 and τI ) 12 min. For the two crucial tray temperature loops, openloop tests were performed for determining the PI tuning constants following the IMC-PI tuning rule of Chien and Fruehauf41 with the assumption of integrating plus the deadtime model form for the initial dynamic response. The results of those calculations are the following: Kc ) 1.54 and τI ) 7.5 min for the tray temperature loop in the extractive distillation column;

and Kc ) 1.72 and τI ) 13.75 min for the tray temperature loop in the entrainer recovery column. The unmeasured feed IPA composition disturbances will be used to test this control strategy with (20% changes in the IPA in the fresh feed composition at time ) 10 min. Figure 13 shows the closed-loop results for this unmeasured disturbance. From the top two plots, it is found that the two tray temperature control loops perform well in bringing the temperatures back to their setpoints. The middle two plots show that the IPA product composition is still maintained at high purity for the first 500 min but becomes less pure after that time. The purity of the water product is dropped to an unacceptable level between 100

802


and 500 min during +20% changes in the feed IPA composition. The bottom plot also shows the problem with this overall control strategy. The bottom level of the entrainer recovery column was continuously dropping during +20% changes of the feed IPA composition until it was almost empty. At that time, the entrainer feed cannot be maintained at its setpoint and the valve is fully opened. For the +20% changes in the feed IPA composition, the water product flow rate should desirably be decreased. By holding the reflux ratio at a nominal constant value, the reflux flow rate was also decreased resulting in less material going down the column and also a drop in the purity of the water product. A modified overall control strategy is to hold both reflux flow rates at constant values and only change with the fresh feed flow rate. This modified overall control strategy was recommended by Grassi39 and is shown in Figure 14. It is noticed that the temperature control points for both columns are the same from a similar open-loop sensitivity analysis as shown in Figure 11. Figure 15 shows the closed-loop responses under this overall control strategy with the unmeasured feed composition disturbance the same as that shown in Figure 13. It is noticed that the two product compositions are maintained much closer to high purity than in Figure 13. The bottom level of the entrainer recovery column was steadily maintained at new values under feed composition disturbances without the problem exhibited under the reflux ratio scheme. Another feed disturbance to test the overall control strategy is the throughput changes. These changes are made by increasing/decreasing of the fresh feed flow rate. Figure 16 shows the closed-loop responses using the overall control strategy in Figure 14 with (20% changes of the fresh feed flow rate. With this control strategy, the two reflux flow rates and the entrainer feed flow rate will be changed in accordance with the fresh feed flow rate. From Figure 16, it is found that the overall control strategy performed nicely in maintaining high purity of the two products. Although not shown in this figure, both product flow rates are desirably increased/decreased to new values. 5. Conclusion Design and control of an isopropyl alcohol dehydration process via extractive distillation is thoroughly investigated in this paper. Using total annual cost (TAC) as the objection function, the optimal design flowsheet of the process of a twocolumn system is developed. This optimal design flowsheet can be compared directly to an earlier study via heterogeneous azeotropic distillation. From the simulation results, it is found that the process via extractive distillation is much more competitive than the one via heterogeneous azeotropic distillation. The TAC is reduced by as much as 32.7% using the process via extractive distillation, and the required steam cost was also cut by 30.3%. The two only possible drawbacks for the process via extractive distillation are that the extractive distillation column needs to be very high; this may cause some difficulty for the construction of this column. The two columns also need to be operated at higher temperatures; thus, steam with higher pressure will be needed in the two reboilers. As for the overall control strategy of the process, it is found that fixing the two reflux ratios does not work during +20% changes of the feed IPA composition. The recommended overall control strategy is the one which fixed the reflux flow rates of the two columns and only varied with the fresh feed flow rate. The high purity of the two products can be maintained despite very wide feed disturbance variations.

Acknowledgment This work is supported by the National Science Council of the R.O.C. under grant NSC 95-2221-E-011-131. Literature Cited (1) Chien, I. L.; Wang, C. J.; Wong, D. S. H. Dynamics and Control of a Heterogeneous Azeotropic Distillation Column: Conventional Control Approach. Ind. Eng. Chem. Res. 1999, 38, 468-478. (2) Chien, I. L.; Zeng, K. L.; Chao, H. Y. Design and Control of a Complete Heterogeneous Azeotropic Distillation Column System. Ind. Eng. Chem. Res. 2004, 43, 2160-2174. (3) Arifin, S.; Chien, I. L. Combined Preconcentrator/Recovery Column Design for Isopropyl Alcohol Dehydration Process. Ind. Eng. Chem. Res. 2007, 46, 2535-2543. (4) Widagdo, S.; Seider, W. D. Azeotropic Distillation. AIChE J. 1996, 42, 96-130. (5) Foucher, E. R.; Doherty, M. F.; Malone, M. F. Automatic Screening of Entrainers for Homogeneous Azeotropic Distillation. Ind. Eng. Chem. Res. 1991, 30, 760-772. (6) Laroche, L.; Bekiaris, N.; Andersen, H. W.; Morari, M. Homogeneous Azeotropic Distillation. Comparing Entrainers. Can. J. Chem. Eng. 1992, 69, 1302-1319. (7) Rodriguez-Donis, I.; Gerbaud, V.; Joulia, X. Entrainer Selection Rules for the Separation of Azeotropic and Close-boiling-temperature Mixtures by Homogeneous Batch Distillation Process. Ind. Eng. Chem. Res. 2001, 40, 2729-2741. (8) Chen, B.; Lei, Z.; Li, Q.; Li, C. Application of CAMD in Separating Hydrocarbons by Extractive Distillation. AIChE J. 2005, 51, 3114-3121. (9) Van Dongen, D. B.; Doherty, M. F. Design and Synthesis of Homogeneous Azeotropic Distillations. 1. Problem Formulation for a Single Column. Ind. Eng. Chem. Fundam. 1985, 24, 454-463. (10) Levy, S. G.; Van Dongen, D. B.; Doherty, M. F. Design and Synthesis of Homogeneous Azeotropic Distillations. 2. Minimum Reflux Calculations for Nonideal and Azeotropic Columns. Ind. Eng. Chem. Fundam. 1985, 24, 463-474. (11) Levy, S. G.; Doherty, M. F. Design and Synthesis of Homogeneous Azeotropic Distillations. 4. Minimum Reflux Calculations for Multiplefeed Columns. Ind. Eng. Chem. Fundam. 1986, 25, 269-279. (12) Knight, J. R.; Doherty, M. F. Design and Synthesis of Homogeneous Azeotropic Distillations. 5. Columns with Nonnegligible Heat Effects. Ind. Eng. Chem. Fundam. 1986, 25, 279-289. (13) Laroche, L.; Bekiaris, N.; Andersen, H. W.; Morari, M. Homogeneous Azeotropic Distillation: Separability and Flowsheet Synthesis. Ind. Eng. Chem. Res. 1992, 31, 2190-2209. (14) Thong, D. Y. C.; Jobson, M. Multicomponent Homogeneous Azeotropic Distillation 1. Assessing Product Feasibility. Chem. Eng. Sci. 2001, 56, 4369-4391. (15) Thong, D. Y. C.; Jobson, M. Multicomponent Homogeneous Azeotropic Distillation 2. Column Design. Chem. Eng. Sci. 2001, 56, 43934416. (16) Thong, D Y. C.; Jobson, M. Multicomponent Homogeneous Azeotropic Distillation 3. Column Sequence Synthesis. Chem. Eng. Sci. 2001, 56, 4417-4432. (17) Hilal, N.; Yousef, G.; Langston, P. The Reduction of Extractive Agent in Extractive Distillation and Auto-Extractive Distillation. Chem. Eng. Proc. 2002, 41, 673-679. (18) Bru¨ggemann, S.; Marquardt, W. Shortcut Methods for Nonideal Multicomponent Distillation: 3. Extractive Distillation Columns. AIChE J. 2004, 50, 1129-1149. (19) Liu, G.; Jobson, M.; Smith, R.; Wahnschafft, O. M. Recycle Selection for Homogeneous Azeotropic Distillation Sequences. Ind. Eng. Chem. Res. 2005, 44, 4641-4655. (20) Doherty, M. F.; Caldarola, G. A. Design and Synthesis of Homogeneous Azeotropic Distillations. 3. The sequencing of Columns for Azeotropic and Extractive Distillations. Ind. Eng. Chem. Fundam. 1985, 24, 474-485. (21) Knapp, J. P.; Doherty, M. F. Thermal Integration of Homogeneous Azeotropic Distillation Sequences. AIChE J. 1990, 36, 969-984. (22) Laroche, L.; Bekiaris, N.; Andersen, H. W.; Morari, M. The Curious Behavior of Homogeneous Azeotropic DistillationsImplication for Entrainer Selection. AIChE J. 1992, 38, 1309-1328. (23) Bekiaris, N.; Meski, G. A.; Radu, C. M.; Morari, M. Multiple Steady States in Homogeneous Azeotropic Distillation. Ind. Eng. Chem. Res. 1993, 32, 2023-2038.

Ind. Eng. Chem. Res., Vol. 47, No. 3, 2008 803 (24) Knapp, J. P.; Doherty, M. F. Minimum Entrainer Flows for Extractive Distillation: A Bifurcation Theoretic Approach. AIChE J. 1994, 40, 243-268. (25) Dorn, C.; Morari, M. Qualitative Analysis for Homogeneous Azeotropic Distillation. 2. Bifurcation Analysis. Ind. Eng. Chem. Res. 2002, 41, 3943-3962. (26) Andersen, H. W.; Laroche, L.; Morari, M. Effect of Design on the Operation of Homogeneous Azeotropic Distillation. Comput. Chem. Eng. 1995, 19, 105-122. (27) Abu-Eishah, S. I.; Luyben, W. L. Design and Control of a Twocolumn Azeotropic Distillation System. Ind. Eng. Chem. Process Des. DeV. 1985, 24, 132-140. (28) Andersen, H. W.; Laroche, L.; Morari, M. Dynamics of Homogeneous Azeotropic Distillation Columns. Ind. Eng. Chem. Res. 1991, 30, 1846-1855. (29) Jacobsen, E. W.; Laroche, L.; Morari, M.; Skogestad, S.; Andersen, H. W. Robust Control of Homogeneous Azeotropic Distillation Columns. AIChE J. 1991, 37, 1810-1824. (30) Ulrich, J.; Morari, M. Operation of Homogeneous Azeotropic Distillation Column Sequences. Ind. Eng. Chem. Res. 2003, 42, 45124534. (31) Luyben, W. L. Plantwide Control of an Isopropyl Alcohol Dehydration Process. AIChE J. 2006, 52, 2290-2296. (32) Gmehling, J.; Mo¨llmann, C. Synthesis of Distillation Process Using Thermodynamic Models and the Dortmund Data Bank. Ind. Eng. Chem. Res. 1998, 37, 3112-3123. (33) Gmehling, J. Azeotropic Data, 2nd ed.; Wiley-VCH: Weinheim, 2004.

(34) Wang, C. J.; Wong, D. S. H.; Chien, I. L.; Shih, R. F.; Liu, W. T.; Tsai, C. S. Critical Reflux, Parametric Sensitivity, and Hysteresis in Azeotropic Distillation of Isopropyl Alcohol + Water + Cyclohexane. Ind. Eng. Chem. Res. 1998, 37, 2835-2843. (35) Gmehling, J.; Onken, U. Vapor-Liquid Equilibrium Data Collection; Dechema: Frankfurt, Germany, 1977. (36) Knight, J. R.; Doherty, M. F. Optimal Design and Synthesis of Homogeneous Azeotropic Distillation Sequences. Ind. Eng. Chem. Res. 1989, 28, 564-572. (37) Doherty, M. F.; Malone, M. F. Conceptual Design of Distillation Systems; McGraw-Hill: New York, 2001. (38) Douglas, J. M. Conceptual Design of Chemical Processes; McGrawHill: New York, 1988. (39) Grassi, V. G. Process design and control of extractive distillation. In Practical Distillation Control; Van Nostand Reinhold Press: New York, 1992: pp 370-404. (40) Luyben, W. L. Plantwide Dynamic Simulators in Chemical Processing and Control; Marcel Dekker: New York, 2002. (41) Chien, I. L.; Fruehauf. P. S. Consider IMC Tuning to Improve Controller Performance. Chem. Eng. Prog. 1990, 86, 33-41.

ReceiVed for reView July 23, 2007 ReVised manuscript receiVed October 15, 2007 Accepted October 19, 2007 IE070996N

Design and Control of an Isopropyl Alcohol Dehydration Process via

Recommend Documents