Recovery Column Design for Isopropyl

Ind. Eng. Chem. Res. 2007, 46, 2535-2543

2535

Combined Preconcentrator/Recovery Column Design for Isopropyl Alcohol Dehydration Process Saiful Arifin and I-Lung Chien* Department of Chemical Engineering, National Taiwan UniVersity of Science and Technology, Taipei 106, Taiwan

The design and control of a complete heterogeneous azeotropic distillation column system using cyclohexane (CyH) as entrainer has been presented by Chien et al. (Ind. Eng. Chem. Res. 2004, 43, 2160). In that study, the feed composition to the heterogeneous azeotropic column is assumed to be near the isopropyl alcohol (IPA) and water azeotropic composition (water at 31 mol % in that paper). However, for a typical waste IPA stream in the semiconductor industry where IPA is used as a cleaning agent, this waste stream is more diluted. Taking a typical waste IPA stream with equal molar of IPA and water as an example, design and control of the overall isopropyl alcohol dehydration process will be investigated. The resulting proposed design flow sheet is a two-column system which combines preconcentrator column and recovery column into a single column. This design adapted from Ryan and Doherty (AIChE J. 1989, 35, 1592) for their ethanol dehydration process not only minimizes the total annual cost but also is very robust when operating in the face of various fresh feed disturbances. Only one tray temperature control loop in the heterogeneous azeotropic column and another one in the combined preconcentrator/recovery column are needed in the proposed overall control strategy. 1. Introduction Because of the performance of natural liquid-liquid separation with no energy requirement at the decanter, heterogeneous azeotropic distillation is commonly used in industry to separate mixtures of close relative volatility and breaking azeotropes. Widely used industrial applications include organic alcohol and acetic acid dehydration systems. A nice review paper by Widagdo and Seider1 showed that parametric sensitivity, multiple steady states, and long transient and nonlinear dynamics were found by many authors using theoretical models and computer simulation. These heterogeneous azeotropic distillation columns are known to be difficult to operate and control. Ethanol and isopropyl alcohol dehydration are the two most commonly seen industrial examples. For ethanol dehydration, Kovach and Seider2 performed step tests on feed rate and operating variables for an industrial column for dehydrating ethanol using di-sec-butyl ether (DSBE) as entrainer and found erratic behavior which was attributed to parametric sensitivity. Bozenhardt3 proposed control strategy involving average temperature control, online break point position control, and five feedforward control loops for the ethanol + ether + water system. Rovaglio et al.4-6 proposed average temperature control and two feedforward control loops for the ethanol + benzene + water system. Mu¨ller et al.7 were able to fit data obtained by a laboratory tray column with an equilibrium stage model for an ethanol + cyclohexane + water system and later Mu¨ller and Marquardt8 experimentally reported evidence of multiple steady states of this system. Recently, Luyben9 proposed a plantwide control strategy for ethanol + benzene + water three-column system showing counterintuitive behavior in the aqueous level control loop. For the isopropyl alcohol (IPA) dehydration, previous studies10-13 of this system by using cyclohexane (CyH) as an entrainer showed that the optimum operation point should be * To whom correspondence should be addressed. Tel.: +886-22737-6652. Fax: +886-2-2737-6644. E-mail: [email protected].

located at a critical reflux, a transition point at which the distillation path switches from a route that passes through IPA + H2O azeotrope to one that passes through IPA + CyH azeotrope. At this critical reflux, a very high purity IPA product can be obtained with minimum energy consumption and maximum product recovery. However, from bifurcation analysis, this critical reflux point is right at the edge of the top stable branch and the middle unstable branch of steady-state operating conditions; thus, this steady state is extremely sensitive to any small perturbations. An inverse double loop control strategy was recommended to maintain a steady column temperature profile. This control strategy was verified via dynamic simulation and experimental test. Chien et al.14 extended the previous study of a single heterogeneous azeotropic distillation column to include a recovery column. Design and control of this two-column system are proposed in that paper. Using a different separation method, Luyben15 proposed plantwide control strategy for a three-column system via extractive distillation with ethylene glycol as the extractive agent. In Chien et al.,14 the feed composition to the heterogeneous azeotropic column is assumed to be near the isopropyl alcohol (IPA) and water azeotropic composition (water at 31 mol % in that paper). However, for a typical waste IPA stream in the semiconductor industry where IPA is used as a cleaning agent, this waste stream is more diluted. Taking a typical waste IPA stream with equal molar of IPA and water as an example, design and control of the overall isopropyl alcohol dehydration process will be investigated. The organization of this paper is as follows. Section 2 will present three design alternatives for the separation of IPA and water with equal molar feed composition. The base case design flow sheet of each design alternative is determined by minimization of total annual cost (TAC) of the overall process. A twocolumn system which combines preconcentrator column and recovery column into a single column is found to obtain minimum TAC. Section 3 will discuss the overall control strategy of this proposed two-column system. Only tray temperature(s) will be used in the proposed control strategy to hold

10.1021/ie061446c CCC: $37.00 © 2007 American Chemical Society Published on Web 03/21/2007

2536

Ind. Eng. Chem. Res., Vol. 46, No. 8, 2007

Figure 1. Process flow sheet of design alternative #1.

Figure 2. Material balance lines for design alternative #1.

the final product purity specification in the face of fresh feed composition and flow rate disturbances. Some conclusion remarks will be given in section 4. 2. Steady-State Design of the Overall Process

Figure 3. Sequential iterative optimization procedure for this system.

The feed composition in this study is assumed to be equal molar of IPA and water and the feed rate of the fresh feed stream is chosen to be with industrial scale of 100 kmol/h. With more dilute feed stream as compared to the ones in Chien et al.,14 the steady-state design of the overall process should be re-investigated. The first obvious choice is to see if the original design in Chien et al.14 can handle the feed composition variations. 2.1. Design Alternative #1 (Two-Column System as in Chien et al.14). The process flow sheet for this design alternative can be seen in Figure 1 with the conceptual material balance lines as in Figure 2. In this figure, the fresh feed (FF) combined with the distillate from the recovery column (D2) to become the point (FF + D2) in Figure 2. This point combined with the organic reflux and then separated into pure IPA at the bottom of the heterogeneous azeotropic column and the top vapor close to the ternary azeotrope. This top vapor stream after subcooling to 40 °C will naturally split into two liquid phases. The organic phase, which contains mostly entrainer (CyH), is designed to be totally refluxed back to the heterogeneous azeotropic column while the aqueous phase, still containing a significant amount of IPA, is fed into the recovery column. The bottom stream of the recovery column (B2) is pure water while the distillate (D2)

is recycled back to the heterogeneous azeotropic column. From the figure, one would immediately find out that because the fresh feed is more diluted, the material balance line drawn from the point of FF + D2 to the point of organic reflux is very close to the point of ternary azeotrope. From the level rule, the top vapor flow rate will be much greater than the flow rate of the bottom IPA product, thus requiring large recycle flow rate in the overall system. The optimal process flow sheet was obtained by minimization of the total annual cost (TAC) with five design variables: IPA distillate composition (XD2) of the recovery column, total number of stages for the heterogeneous azeotropic column and the recovery column (N1 and N2), and the two feed stages (NF1 and NF2). The product specification of IPA is set to be ultrapure (99.9999 mol %) to be used in the semiconductor industry. The product specification of water is set to be the same as that in Chien et al.14 (99.9 mol %). In each simulation run, the IPA product specification is achieved by varying the reboiler duty of the heterogeneous azeotropic column and the water product specification is achieved by varying the reboiler duty of the recovery column. The entrainer makeup flow rate will be very small to balance the entrainer loss from the two bottom streams.

Ind. Eng. Chem. Res., Vol. 46, No. 8, 2007 2537

Figure 5. Optimal design and operating conditions for design alternative #1.


Figure 4. TAC plots for design alternative #1.

The TAC of the overall system includes the annualized capital costs and the operating costs. The annualized capital costs include column shells of the two columns, internal trays of the two columns, and also the reboilers and condensers of the two columns. The payback period was assumed to be 3 years in the calculation. The operating costs include the steam and cooling water required for operating these two columns and also very small entrainer makeup cost. A sequential iterative procedure was used to obtain the optimal design flow sheet with the procedure summarized in Figure 3. The optimization method is intuitive; however, it is

rather time-consuming when the number of design variables becomes large in a plantwide optimization. Figure 4 shows some of the TAC plots within the optimization search. The top plot is a summarized plot at each XD2 and shows that the optimal distillation composition is roughly at 0.62 (or can be said to be between 0.615 and 0.625). Of course, more precise optimal XD2 can be calculated which requires more simulation times. However, because the plot is flat enough at low TAC, not much can be gained in being more precise. This IPA distillate composition represents tradeoff in the overall process design. The costs associated with the recovery column will be increased when this distillate composition is purer in IPA. However, setting XD2 to be more pure in IPA will make the material balance line connecting points FF + D2 and organic reflux (see Figure 2) further away from the ternary azeotropic point, thus reducing recycle flow rate in the overall system. The optimal design and operating condition can be seen in Figure 5. Notice that the ratio of aqueous outlet flow (2653.24 mol/min) to the bottom IPA product flow (832.50 mol/min) is at 3.19; thus, large flow rate is going from the heterogeneous azeotropic column to the recovery column. The TAC and the operating cost will be compared to two other design alternatives later. 2.2. Design Alternative #2 (Three-Column System with a Preconcentrator Column). As explained above, the above twocolumn system may not be suitable for fresh feed composition with dilute IPA; thus, intuitive thinking says to add a preconcentrator column so that the feed to the heterogeneous azeotropic column can be placed closer to the IPA + water azeotrope. The process flow sheet for this design alternative can be seen in Figure 6 with the conceptual material balance lines as in Figure 7. From Figure 7 the decision for the placement of point D1 is a tradeoff between less flow rate from the heterogeneous azeotropic column to the recovery column and the TAC of the preconcentrator column. Placing D1 closer to the IPA + water

2538


Figure 7. Material balance lines for design alternative #2.

azeotrope will increase the TAC of the preconcentrator column. However, the benefit is that the top vapor flow rate of the heterogeneous azeotropic column can be lower, thus reducing the organic reflux flow rate back to the same column and also reducing the aqueous outlet flow rate to the recovery column. The placement of point D1 is an additional design variable in this design alternative #2. Since there is no recycle flow back to the first preconcentrator column, thus this stand-alone column can be optimized first with two product specifications and two design variables of total number of stages and feed tray location. One of the product specifications is the estimate of the XD1 composition and the other product specification is the setting of the bottom water product composition at 0.999 by varying reboiler duty. For the first column, we can obtain the optimized result for each estimate of the XD1 composition. With each result, we can go through the same procedure as outlined previously and obtain the optimal design and operating conditions for this overall three-column system. The optimized results can be seen in Figure 8. Notice that the ratio of aqueous outlet flow (2056.01 mol/min) to the bottom IPA product flow (832.50 mol/min) has been reduced from 3.19 to 2.47. The comparison of the TAC and the operating cost to the other two design alternatives will be shown later.

Figure 8. Optimal design and operating conditions for design alternative #2.

2.3. Design Alternative #3 (Two-Column System with Combined Preconcentrator/Recovery Column). From previous Figure 7, the material balance lines of the preconcentrator column and the recovery column are quite close together; thus, it should be possible to combine these two columns together into one column. Ryan and Doherty16 in 1989 proposed an ingenious design flow sheet for their ethanol dehydration process with benzene as the entrainer. In the following, we will apply this design to the isopropyl alcohol dehydration process and we will also examine the overall control strategy for this design alternative in section 3. The process flow sheet for this design alternative with only two columns can be seen in Figure 9 with the conceptual material balance lines as in Figure 10. In this design flow sheet, fresh feed and the aqueous outlet flow are combined and enter into the C-2 column which acts as the preconcentrator and also as the recovery column. The bottom of this column is designed to have water purity of 99.9 mol % while the top will feed into the heterogeneous azeotropic column (C-1). The top composition of the C-2 column (XD2), similar to previous design alternatives, is a design variable which will be determined to minimize the TAC of the overall system. The bottom IPA product purity of the heterogeneous azeotropic column is again set at a very high purity of 99.9999 mol % for semiconductor industrial usage. The optimization procedure to be followed is exactly the same as the one for design alternative #1 with the total of five design variables (XD2, total number of stages of C-1 and C-2, and feed tray locations of C-1 and C-2). The procedure for the iterative sequential optimization search was outlined previously in section 2.1 with Figure 11 summarizing the optimization result at each estimate of XD2. From the figure the optimized XD2 is at 0.620. With this XD2, the optimized total number of stages for C-1 is 18 (including reboiler) with feed tray location at tray #2 (counting from the top) and the optimized total number of stages for C-2 is 10 (including reboiler and condenser) with feed tray location at tray #6. The optimal design and operating conditions for this two-column system can be seen in Figure 12. Notice that the ratio of aqueous outlet flow (1992.54 mol/ min) to the bottom IPA product flow (832.50 mol/min) is again reduced from that of design alternative #1. 2.4. Comparison of the Three Design Alternatives. The comparison of all the costs calculated for the three design alternatives is summarized in Table 1. Notice that the TAC for



Figure 12. Optimal design and operating conditions for design alternative #3. Table 1. Cost Comparison of the Three Design Alternatives

costs C-1 column annualized equipment cost ($) C-1 column steam cost ($) C-1 column cooling water cost ($) C-2 column annualized equipment cost ($) C-2 column steam cost ($) C-2 column cooling water cost ($)

design alternative #1



576244.47

136819.22

464008.31

295210.38 14720.18

110895.98 3753.55

211920.85 11023.43

208802.94

471680.33

242410.46

178006.00 6481.11

216942.67 11325.27

251794.40 8956.09

C-3 column annualized equipment cost ($) C-3 column steam cost ($) C-3 column cooling water cost ($) Figure 10. Material balance lines for design alternative #3.

makeup entrainer cost ($) TAC

165263.05 132398.07 4821.39 129.91

127.17

266.60

1279594.99

1254026.70

1190380.14

This result is somewhat different than the study by Ryan and Doherty16 for their ethanol dehydration process. They concluded that the two-column system of combining preconcentrator column and recovery column has lower capital costs but higher operating costs than the three-column system so that the total annualized cost is about the same for both systems. However, they also cautioned that no generalizations should be made from this one example about other systems. The control of the overall process is also very important in determining which design should be used for this isopropyl alcohol dehydration process. If the design with the lowest TAC were not able to properly maintain the product purity in the face of typical fresh feed disturbances, this process design would not be recommended. Figure 11. TAC plots for design alternative #3.

3. Overall Control Strategy Development design alternative #3 is the lowest with 7.0% savings compared to design alternative #1 and 5.1% savings compared to design alternative #2. The operating cost for design alternative #3 is slightly more than that of design alternative #2 (0.77% more); however, it is less than the cost of the other two-column system of design alternative #1 (2.1% less). It is decided that design alternative #3, although with slightly more operating cost, would be the best design flow sheet for the isopropyl alcohol dehydration process because the TAC is the smallest and also there are fewer pieces of process equipment in the overall design flow sheet.

In the following, we will investigate the proper overall control strategy for the process design with the lowest TAC (design alternative #3) first. Only tray temperature control loop(s) will be used in the overall control strategy for wider industrial applications. The Aspen Plus steady-state simulation in the last section is exported to the dynamic simulation of Aspen Dynamics. The tray sizing option in Aspen Plus is utilized to calculate the column diameter of both columns with the assumed tray spacing of 0.6096 m. The diameter of the C-1 and the C-2 columns are

2540


Figure 14. Proposed overall control strategy.

Figure 13. Open-loop sensitivity plots.

calculated to be 1.61 and 1.25 m, respectively, and the weir heights of both columns are assumed to be 0.0508 m. Other equipment sizing followed the recommendation by Luyben.17 The volumes of the column base for the C-1 column and also for the C-2 column and the reflux drum of the C-2 column are all sized to give 10 min holdup with 50% liquid level. The decanter is sized to be bigger to allow for two liquid phases to separate. The holdup time of 20 min for the decanter is used in the dynamic simulation. Pressure-driven simulation in Aspen Dynamics is used with the top stage pressures of the C-1 column set at 1.1 atm to allow for some pressure drop in the condenser and decanter. The top pressure of the C-2 column is set to be at atmospheric pressure. The pressure drops inside both columns are automatically calculated in Aspen Dynamics to account for liquid hydraulics and vapor traffic. The bottom pressure for the C-1 and the C-2 columns are calculated to be 1.200 and 1.056 atm, respectively. 3.1. Inventory Control Loops. The inventory control loops according to our past experience are designed as follows. For the heterogeneous azeotropic column with the decanter, the aqueous phase level is controlled by manipulating the aqueous outlet flow; the column bottom level is controlled by manipulating the bottom IPA product flow; and the column top pressure is controlled by manipulating the top vapor flow. One important loop pairing from past experience (c.f. Chien et al.14) is that the organic phase level should be controlled by entrainer makeup flow, not by an internal recycling flow of organic reflux, so that the snowballing effect can be avoided. For the combined preconcentrator/recovery column, the reflux drum level is

controlled by the distillate flow (feed to heterogeneous azeotropic column); the column bottom level is controlled by the bottom water product flow; and the column pressure is controlled by the condenser duty. In all the closed-loop simulation runs, the P-only controller is used in all level loops for both columns. The reason for using a P-only controller is to provide maximum flow smoothing and also because maintaining a liquid level at setpoint value is often not necessary. Kc ) 2 as suggested in Luyben17 is used in most of the level loops. For the organic phase level loop to manipulate the entrainer makeup flow, Kc ) 10 is used to simulate very fast control behavior favorable in this control loop for quickly sending more entrainer or quickly stop feeding makeup entrainer into this system. For the top pressure control loops of both columns, tight PI controller tuning parameters of Kc ) 20 and τI ) 12 min are used. The remaining manipulated variables for the heterogeneous azeotropic column are the organic reflux flow and the reboiler duty, and the remaining manipulated variables for the preconcentrator/recovery column are the reflux flow and the reboiler duty. We will investigate the simplest overall control strategy first with only one tray temperature control loop at each column. Notice in Chien et al.14 it is not possible to properly reject feed disturbances with only one tray temperature control loop in the heterogeneous azeotropic column. The reason for the difficulty in operating and controlling this very sensitive heterogeneous azeotropic column was explained in detail in previous studies (cf. Chien et al.11-14). 3.2. Tray Temperature Control Point(s). Open-loop sensitivity analysis is used to determine the tray temperature control point for both the heterogeneous azeotropic column and the preconcentrator/recovery column. Figure 13 shows the openloop sensitivity plots for both columns. The one on the top is for the heterogeneous azeotropic column with (1% changes in its reboiler duty, and the one on the bottom is for the preconcentrator/recovery column with (1% changes in its reboiler duty. Tray #7 of the heterogeneous azeotropic column and tray #9 of the preconcentrator/recovery column are selected as the temperature control points because of high sensitivity and nearly linear behavior. PI control form is used in these two temperature control loops. The tuning constants are determined using the IMC-PID tuning rules of Chien and Fruehauf.18 From the open-loop step responses, the controlled temperature at C-1 column is suitable to assume to behave like an integrating plus deadtime model


Figure 15. Closed-loop responses with (20% feed flow rate changes.

and the controlled temperature at C-2 column is assumed to behave like a first-order plus deadtime model. The closed-loop time constants of these two loops are set to be twice their model apparent dead times. The resulting PI tuning constants for the tray #7 temperature loop of the heterogeneous azeotropic column are Kc ) 3.58 and τI ) 6.25 min and the tuning constants for the tray #9 temperature loop of the preconcentrator/recovery column are Kc ) 0.40 and τI ) 12.8 min. The other two manipulated variables not used in the two tray temperature control loops are the organic reflux flow for the heterogeneous azeotropic column and the reflux flow for the preconcentrator/recovery column. Ratio control scheme is designed so that these two manipulated variables can be adjusted in the face of disturbances. The organic reflux flow is set to maintain a constant ratio to the feed flow of the heterogeneous azeotropic column and the reflux ratio of the preconcentrator/

recovery column is also maintained. The overall proposed control strategy is summarized in Figure 14. 3.3. Simulation Results. Two types of disturbances will be used to test the proposed control strategy. The first disturbance change that is considered as a known disturbance is (20% changes in the fresh feed flow rate. With these changes, the organic reflux flow rate will be adjusted to hold the (organic reflux)/(feed to heterogeneous azeotropic column) ratio and the reflux flow of the preconcentrator/recovery column will also be adjusted to hold the reflux ratio at a constant value. Figure 15 shows the dynamic responses for some of the important variables in the process. The top left and right plots show that the two tray temperatures are smoothly controlled back to their setpoint values. The middle left and right plots show the feed composition and feed flow rate entering into the heterogeneous azeotropic column. It is noticed that there is only

2542


Figure 16. Closed-loop responses with (20% feed IPA composition changes.

slight changes in the feed composition to this “more sensitive” column and the feed rate changes (assumed measurable) are also quite smooth. There is no snowballing effect in this overall system even though no recycling flow rate is fixed in the proposed control strategy. The bottom two plots show that the two product compositions are still very close to their purity specifications. The IPA product stream is still in the ultrapure region (>99.99986 mol %) and the water product stream is back to 99.9 mol %. With this (20% changes in the fresh feed flow rate, the IPA product flow rate and water product flow rate also increase/decrease correspondingly to their new values. For example, with +20% in the fresh feed flow rate, the IPA product flow rate changes from 832.50 to 998.62 mol/min (also a +20% increase) and the water product flow rate changes from 834.17 to 1000.83 mol/min (also a +20% increase). The second disturbance tested is the fresh feed composition changes, which are often considered a tougher disturbance to

reject because they are unmeasured. Large variations of (20% changes in the fresh feed IPA composition are tested with the closed-loop results shown in Figure 16. Again the two tray temperatures are smoothly controlled back to their setpoint values. More importantly, for the more sensitive heterogeneous azeotropic column its feed composition does not vary much (only change from 0.620 to 0.625 or from 0.620 to 0.615). The final IPA product composition is still kept in the ultrapure region (>99.9997 mol %) and the water product composition is returned back to 99.9 mol %. This desirable result is mainly due to the combined preconcentrator/recovery column dampening the large disturbance from fresh feed, thus making the feed variations to the more sensitive C-1 column small and easier to handle. The three-column system should also have this desirable feature. However, it requires more process, instrumental, and control equipment and the TAC is also higher than the proposed design.


4. Conclusions In this paper, design and control of a complete heterogeneous azeotropic distillation column system with more dilute fresh feed stream has been studied. Total annual cost and operating cost of three design alternatives are compared. The first design alternative having fresh feed directly entering into the heterogeneous azeotropic column gives the highest TAC and operating cost. The operating costs of the second design alternative (threecolumn system) and the proposed design alternative (twocolumn system) are comparable with the proposed two-column system, giving a 5% savings in TAC. The proposed design combining preconcentrator column and recovery column into a single column is very robust in the face of wide variations in the fresh feed flow rate and feed composition. Only one tray temperature control loop in the heterogeneous azeotropic column and another one in the combined preconcentrator/recovery column are needed in the proposed overall control strategy. The more complicated inverse double-loop control strategy as in previous papers (c.f. Chien et al.11-14) is not needed because the combined preconcentrator/ recovery column “buffered” the disturbances from directly entering into the heterogeneous azeotropic column. The benefit of dampening the fresh feed disturbances by a preconcentrator column as in three-column system is retained without extra process, instrumental, and control equipment. Literature Cited (1) Widagdo, S.; Seider, W. D. Azeotropic distillation. AIChE J. 1996, 42, 96. (2) Kovach, J. W., III; Seider, W. D. Heterogeneous Azeotropic Distillation: Experimental and Simulation Results. AIChE J. 1987, 33, 1300. (3) Bozenhardt, H. F. Modern Control Tricks Solve Distillation Problems. Hydrocarbon Process. 1988, June, 47. (4) Rovaglio, M.; Faravelli, T.; Biardi, G.; Gaffuri, P.; Soccol, S. Precise Composition Control of Heterogeneous Azeotropic Distillation Towers. Comput. Chem. Eng. 1992, 16 (Suppl.), S181. (5) Rovaglio, M.; Faravelli, T.; Biardi, G.; Gaffuri, P.; Soccol, S. Key Role of Entrainer Inventory for Operation and Control of Heterogeneous Azeotropic Distillation Towers. Comput. Chem. Eng. 1993, 17, 535.

(6) Rovaglio, M.; Faravelli, T.; Gaffuri, P.; Di Palo, C.; Dorigo, A. Controllability and Operability of Azeotropic Heterogeneous Distillation Systems. Comput. Chem. Eng. 1995, 19 (Suppl.), S525. (7) Mu¨ller, D.; Marquardt, W.; Hauschild, T.; Ronge, G.; Steude, H. Experimental Validation of an Equilibrium Stage Model for Three-phase Distillation. Inst. Chem. Eng. Symp. Ser. 1997, 142, 149. (8) Mu¨ller, D.; Marquardt, W. Experimental Verification of Multiple Steady States in Heterogeneous Azeotropic Distillation. Ind. Eng. Chem. Res. 1997, 36, 5410. (9) Luyben, W. L. Control of a Multiunit Heterogeneous Azeotropic Distillation Process. AIChE J. 2006, 52, 623. (10) 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. (11) 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. (12) Chien, I. L.; Wang, C. J.; Wong, D. S. H.; Lee, C. H.; Cheng, S. H.; Shih, R. F.; Liu, W. T.; Tsai, C. S. Experimental Investigation of Conventional Control Strategies for a Heterogeneous Azeotropic Distillation Column. J. Process Control 2000, 10, 333. (13) Chien, I. L.; Chen, W. H.; Chang, T. S. Operation and Decoupling Control of a Heterogeneous Azeotropic Distillation Column. Comput. Chem. Eng. 2000, 24, 893. (14) 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. (15) Luyben, W. L. Plantwide Control of an Isopropyl Alcohol Dehydration Process. AIChE J. 2006, 52, 2290. (16) Ryan, P. J.; Doherty, M. F. Design/Optimization of Ternary Heterogeneous Azeotropic Distillation Sequences. AIChE J. 1989, 35, 1592. (17) Luyben, W. L. Plantwide Dynamic Simulators in Chemical Processing and Control; Marcel Dekker; New York, 2002. (18) Chien, I. L.; Fruehauf, P. S. Consider IMC Tuning to Improve Controller Performance. Chem. Eng. Prog. 1990, 86 (10), 33.

ReceiVed for reView November 12, 2006 ReVised manuscript receiVed February 13, 2007 Accepted February 19, 2007 IE061446C

Recovery Column Design for Isopropyl

Recommend Documents