Comparison of Pressure-Swing and Extractive-Distillation Methods for

Jun 8, 2005 - Comparison of Pressure-Swing and Extractive-Distillation Methods for Methanol-Recovery Systems in the TAME Reactive-Distillation Process...
0 downloads 7 Views 415KB Size
Ind. Eng. Chem. Res. 2005, 44, 5715-5725

5715

Comparison of Pressure-Swing and Extractive-Distillation Methods for Methanol-Recovery Systems in the TAME Reactive-Distillation Process William L. Luyben† Process Modeling and Control Center, Department of Chemical Engineering, Lehigh University, Bethlehem, Pennsylvania 18015

The process to produce tert-amyl methyl ether via reactive distillation requires a methanolrecovery section because the presence of C5-methanol azeotropes means that a significant amount of methanol is present in the distillate from the reactive column. The use of pressureswing azeotropic distillation was studied in a previous paper [Al-Arfaj, M. A.; Luyben, W. L. Plantwide control for TAME production using reactive distillation. AIChE J. 2004, 50 (7), 1462] in which both the steady-state design and the plantwide control of the entire process were developed. This paper presents a quantitative steady-state and dynamic comparison of the pressure-swing process with an extractive-distillation process. Water is the extractive agent. The extractive-distillation process is found to be much more economical (40% lower capital investment and 60% lower energy cost). The plantwide dynamic controllability performances of the two systems are essentially equivalent. 1. Introduction Improved designs of chemical processes have produced significant reductions in capital investment, energy costs, environmental emissions, and safety hazards. Consideration of both steady-state economics and dynamic controllability is vital in developing these highperformance designs. Reactive distillation and coupled reactor/distillation column systems are two of the technologies that have been applied to achieve these objectives. These methods also provide process intensification improvement in some chemical systems. Subawalla and Fair2 studied a reactive-distillation system for the production of tert-amyl methyl ether (TAME). The steady-state designs of a prereactor and a reactive-distillation column were presented. Al-Arfaj and Luyben1 extended this work to design a complete plant with methanol recovery using a pressure-swing azeotropic separation method. They also developed a plantwide control structure for the three-column process with two recycles and two fresh feed streams. The C5 feed stream to the TAME process contains about 24 mol % reactive isoamylenes: 2-methyl-1butene (2M1B) and 2-methyl-2-butene (2M2B). The remaining components are pentanes and pentenes (largely isopentane, iC5), which are inert in the TAME reaction. TAME is the highest boiling component, so it leaves in the bottoms stream from the reactive-distillation column. The lighter C5’s leave in the distillate stream along with a significant amount of methanol. Methanol forms minimum-boiling azeotropes with many of the C5’s. The reactive column operates at 4 bar (the optimum pressure2 that balances the temperature requirements for reaction with those for vapor-liquid separation). At this pressure, isopentane and methanol form an azeotrope at 339 K that contains 26 mol % methanol. Therefore, the distillate from the reactive †

Tel.: (610) 758-4256. E-mail: [email protected].

column contains a significant amount of this reactant, which must be recovered. Because the iC5-methanol azeotrope is pressuresensitive (79 mol % iC5 at 10 bar and 67 mol % iC5 at 4 bar), it is possible to use two distillation columns, operating at two different pressures, to separate methanol from the other C5 components. This pressure-swing process is the one studied in an earlier paper.1 An alternative separation process for this system is extractive distillation, which is mentioned in Stichlmair and Fair.3 These authors provide no details of either the steady-state design or the dynamic control. This paper provides a quantitative comparison of the pressure-swing and extractive-distillation processes. 2. Chemical Kinetics and Phase Equilibrium The liquid-phase reversible reactions considered are

2M1B + MeOH T TAME 2M2B + MeOH T TAME 2M1B T 2M2B The kinetics for the forward and reverse reactions are given by Al-Arfaj and Luyben1 (Table 1 in their paper). Their reaction rates are given in units of kmol/s/kgcat and were converted to the Aspen-required units of kmol/s/m3 by using a catalyst bulk density of 900 kg/ m3. The reactive stages in the column each contain 1100 kg of catalyst. This corresponds to 1.22 m3 on each tray, which gives a weir height of 0.055 m for a reactive column with a diameter of 5.5 m. The phase equilibrium of this system is complex because of the existence of azeotropes. The UNIFAC physical property package in Aspen Plus is used to model the vapor-liquid equilibrium in all units except the methanol-water column in the extractive-distillation process. The van Laar equations are used for the

10.1021/ie058006q CCC: $30.25 © 2005 American Chemical Society Published on Web 06/08/2005

5716

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 1. TAME process with pressure-swing methanol recovery.

Table 1. Stream Information for Reactor and Column C1, for Columns C2 and C3 in the Pressure-Swing Process, and for Columns C2 and C3 in the Extractive-Distillation Process A. Reactor and Column C1 fresh methanol (kmol/h) MeOH 2M1B 2M2B TAME nC5 iC5 1-pentene 2-pentene total T (K) P (atm)

feed (kmol/h)

230

MeOH reactor (kmol/h)

reactor effluent (kmol/h)

MeOH C1 (kmol/h)

313

188 13.7 112 125 88.4 501 38.1 162 1228 355 6

235

85.6 165

230 325 17

88.4 501 38.1 162 1040 343 10

313 351 7

B1 (kmol/h)

235 351 4

D1 (kmol/h) 316 3.31 14.1

0.01 0.60 232 0.22 0.17 0.18 0.76 234 415 4.29

88.2 501 37.9 161 1122 341 3.95

B. Columns C2 and C3 in the Pressure-Swing Process B2 (kmol/h) MeOH 2M1B 2M2B TAME nC5 iC5 1-pentene 2-pentene total T (K) P (atm)

D2 (kmol/h)

230

B3 (kmol/h)

D3 (kmol/h)

313

188 13.7 112 125 88.4 501 38.1 162 1228 355 6

85.6 165 88.4 501 38.1 162 1040 343 10

230 325 17

313 351 7

C. Columns C2 and C3 in the Extractive-Distillation Process

MeOH 2M1B 2M2B TAME nC5 iC5 1-pentene 2-pentene water total T (K) P (atm)

B2 (kmol/h)

D2 (kmol/h)

B3 (kmol/h)

D3 (kmol/h)

316

0.03 3.31 14.1

1.03

315

1026 1027 379 1.2

0.32 315 338 1.0

water makeup (kmol/h)

extract water to C2 (kmol/h)

24 24 325 7

1050 1050 322 2.5

88.4 501 38.0 161 1025 1341 325 17

830 326 2.5

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5717

Figure 2. (A) Composition profiles in a reactive column. (B) Temperature profile in a reactive column.

methanol-water separation because of their known ability to accurately match experimental data. 3. Pressure-Swing Process 3.1. Steady-State Design. The pressure-swing process is based on that presented in a previous paper1 with some differences in recycle flow rates, as discussed below. The process is shown in Figure 1. Table 1 gives information for the important streams. A. Prereactor. The prereactor is a cooled tubular reactor containing 9544 kg of catalyst. The C5 fresh feed (1040 kmol/h) and 313 kmol/h of methanol are fed to the reactor. B. Reactive Column C1. The reactor effluent is fed into a 36-stage reactive distillation column (C1) at stage

28. We use Aspen notation of numbering stages from the top, with stage 1 being the reflux drum. The Subawalla and Fair2 design has a catalyst present at stages 7-23. The reactor effluent is fed five trays below the reactive zone. A methanol stream is fed at the bottom of the reactive zone (stage 23). The flow rate of the methanol fed to the reactive column is 235 kmol/h in this paper. In our previous paper, it was given as 190 kmol/h. The only reason we can offer for this difference is the use of two different versions of Aspen Plus. In the previous paper, version 10 was used. In the current paper, version 12 is used. As mentioned later, the other difference is the flow rate of distillate recycle (D3) from the high-pressure column back to the low-pressure column. In the previous paper, this flow

5718

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Table 2. Design Parameters and Economic Results pressureswing process C2

C3

4.2 10 24.5 2065 19.5 985

extractivedistillation process

diameter (m) stages QC (MW) AC (m2) QR (MW) AR (m2) QHX (MW) AHX (m2) shell cost (106 $) HX cost (106 $) energy cost (106 $/year) capital (106 $) diameter (m) stages QC (MW) AC (m2) QR (MW) AR (m2) shell cost (106 $) HX cost (106 $) energy cost (106 $/year) capital (106 $)

0.339 1.69 2.89 2.02 6.2 10 20.2 1703 27.2 1374 0.508 0.304 4.03 2.23

1.74 12 5.20 438 5.96 301 1.24 104 0.158 0.828 0.985 1.21 2.2 32 9.53 803 9.05 457 0.496 0.954 1.34 1.45

total capital (106 $) total energy (106 $/year) TAC (106 $/year)

4.25 6.92 8.34

2.44 2.22 3.04

Table 3. Basis of Economics condensers heat-transfer coefficient ) 0.852 kW/K‚m2 differential temperature ) 13.9 K capital cost ) 7296Area0.65 where Area is in m2 reboilers heat-transfer coefficient ) 0.568 kW/K‚m2 differential temperature ) 34.8 K capital cost ) 7296Area0.65 where Area is in m2 column vessel capital cost ) 17640D1.066 L0.802 where D and L are in m energy cost ) $4.7/106 kJ TAC ) capital cost/payback period + energy cost payback period ) 3 years

rate was 1421 kmol/h. In the current paper, it is 1616 kmol/h. Figure 2 gives composition and temperature profiles in the reactive column C1. The reflux ratio is 4, which gives a bottoms purity of 99.2 mol % TAME and a distillate impurity of 0.1 ppm of TAME. Reboiler heat input and condenser heat removal are 38.2 and 39 MW, respectively. The operating pressure is 4 bar. The column diameter is 5.5 m. The overall conversion of 2M1B and 2M2B in the C5 fresh feed is 92.4%. The distillate D1 has a composition of methanol (28 mol % methanol) that is near the azeotrope at 4 bar. It is fed at a rate of 1122 kmol/h to the first of the twocolumn pressure-swing methanol-recovery columns. C. Methanol-Recovery Column C2. This column operates at a pressure of 2 bar so that the reflux-drum temperature (318 K) is high enough for the use of cooling water in the condenser. The separation is a fairly easy one, so using only 10 stages and a reflux ratio of 0.4 yields a bottoms purity of 99.9 mol % methanol. The

bottoms (318 kmol/h) is mixed with a fresh feed of pure methanol (230 kmol/h), and the total is split between the methanol fed to the reactor and the methanol fed to the reactive column. The distillate D2 is fed to the high-pressure column. Its composition (22.8 mol % methanol) is near the azeotropic composition at the 2 bar pressure. Reboiler heat input and condenser heat removal are 19.5 and 24.5 MW, respectively. The column diameter is 4.2 m. D. C5-Recovery Column C3. The pressure is set at 10 bar, which shifts the azeotropic composition so that the distillate stream from this column D3 has a composition of 34.2 mol % methanol. Higher and lower pressures were explored to see their effects on the economics. The 10 bar pressure seems to be about the optimum because going above this pressure does not shift the azeotrope significantly and raises the base temperature, which would require a higher temperature energy input. The separation is a fairly easy one, so using only 10 stages and a reflux ratio of 1 yields a bottoms impurity of 0.01 mol % methanol. This bottoms stream B3 is the C5 product stream. The reflux-drum temperature is 373 K at this high pressure, which means that some heat integration between C2 and C3 may be economical (the base of C2 is at 356 K, and the reflux drum of C3 is at 373 K). This possibility is not considered in this paper. The distillate is recycled back to C2 at a flow rate of 1616 kmol/h. Reboiler heat input and condenser heat removal are 27.2 and 20.2 MW, respectively. The column diameter is 6.2 m. E. Flowsheet Convergence of Recycles. As any user of flowsheet simulators knows, the convergence of steady-state simulators when recycle streams are present can be very difficult. Such is the case with both of these processes because they both involve two recycle streams. An alternative approach is to use a dynamic simulation to converge the process flowsheet to a steady state. The process conditions shown in Figures 1 and 5 are obtained in this way. Luyben4 discusses this procedure in detail. 3.2. Economics. The design parameters, the capital investments, and the energy costs of the two columns C2 and C3 in the pressure-swing process are given in Table 2. Table 3 gives the economic basis used. Because the reactor and reactive-distillation column C1 are identical in both the pressure-swing and extractivedistillation processes, only the economics of the separation sections are compared. The total capital investment in the two columns, with their associated reboilers and condensers, is $4 250 000. The energy cost for the two reboiler heat inputs is $6.92 × 106/year. The total annual cost (TAC), using a payback period of 3 years for capital investment, is $8.33 × 106/year. These numbers are compared to the extractive-distillation recovery process in the following section of this paper.

Table 4. Controller Tuning Constants column C1 pressure swing extractive distillation

controller

KC

τI (min)

transmitter range

set point

TC1 CC TC2 TC3 TC2 TC3

0.43 0.2 0.41 1.8 0.16 2.4

9 15 11 8.4 9 9.6

350-450 K 0-0.5 mf MeOH 300-400 K 350-450 K 300-400 K 300-400 K

384 K 0.20 mf MeOH 346 K 387 K 361 K 350 K

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5719

Figure 3. (A) Control structure for the pressure-swing process: reactors and column C1. (B) Control structure for the pressure-swing process: columns C2 and C3.

5720

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 4. (A) Feed rate disturbances for the pressure-swing process. (B) Feed composition disturbances for the pressure-swing process.

3.3. Dynamic Plantwide Control. In preparation for exporting the steady-state flowsheet into Aspen Dynamics, all equipment is sized. Column diameters are

calculated by Aspen Tray Sizing. Reflux drums and column bases are sized to provide 5 min of holdup when 50% full, based on the total liquid entering the surge

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5721

Figure 5. TAME process with extractive-distillation methanol recovery.

capacity. Pumps and control valves are specified to give adequate dynamic rangeability. Typical valve pressure drops are 2 atm. When the flowsheet with a tubular reactor was exported into Aspen Dynamics, the program would not run. A liquid-filled plug-flow reactor will not run in version 12 of Aspen Dynamics. To work around this limitation, the tubular reactor was replaced by two continuous stirred tank reactors in series. Operating temperatures in both reactors were set at 355 K and volumes at 10 m3. This design gave the same reactor effluent as the tubular reactor. The plantwide control structure presented in the previous paper1 is used and is shown in Figure 3. A tray temperature is controlled in each column by manipulating the reboiler heat input. The trays are selected by finding the location where the temperature profile is steep: stage 31 in column C1 (see Figure 2B), stage 9 in column C2, and stage 7 in column C3. In addition, an internal composition in column C1 is controlled by manipulating the flow rate of methanol to the column. Stage 18 is selected (see Figure 2A). The flow rate of methanol to the reactor is ratioed to the feed flow rate. All temperature and composition controllers have 1-min deadtimes. The proportional-integral controllers are tuned by running a relay-feedback test and using the Tyreus-Luyben settings. Table 4 gives controller constants. All liquid levels are controlled by proportional controllers with gains of 2 for all level loops except the two reactors, which have gains of 10. The liquid levels in the reflux drums are controlled by manipulating the distillate flow rates. The reflux ratios in all columns are controlled by manipulating the reflux. Column pressure

controllers use default controller settings and manipulate condenser heat removal (not shown in Figure 3). The liquid levels in the base of columns C1 and C3 are controlled by manipulating the bottoms flow rates. In column C2, the base level is controlled by manipulating the fresh feed of methanol, as shown in Figure 3B. This exactly balances the stoichiometry by appropriately adjusting the addition of fresh methanol as it is consumed in the TAME reaction. Remember that the downstream methanol is set by two flow controllers. This means that increasing the methanol fresh feed flow produces an immediate decrease in the flow rate of bottoms B2 from C2. Thus, there is an instantaneous effect on the base level. As demonstrated in the earlier paper, this control structure effectively handles quite large disturbances. Figure 4A gives the responses to positive and negative 20% changes in the feed to the process occurring at 0.2 h. The control structure maintains TAME purity very close to its specification. The system reaches a new steady state in about 3 h. Feed composition disturbances are also tested. The responses of the system to two types of composition disturbances are shown in Figure 4B. The curves labeled +∆z correspond to 20% increases in the molar flow rates of the two reactants in the feed (2M1B and 2M2B), with an appropriate reduction in the isopentane molar flow rate such that the total feed flow rate remains constant. The curves labeled -∆z correspond to 20% decreases in the molar flow rates of the two reactants, with an appropriate reduction in the isopentane molar flow rate. The control structure provides effective control of the process.

5722

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 6. Control structure for the extractive-distillation process: columns C2 and C3.

4. Extractive-Distillation Process The alternative methanol recovery section studied in this paper uses water as an extractive agent in an extractive-distillation column to remove methanol from the distillate stream coming from the reactive-distillation column. A second column separates the methanolwater mixture coming from the base of the extraction column and recycles both methanol and water back to upstream units in the process. 4.1. Steady-State Design. Figure 5 gives the steadystate flowsheet of the process. The reactor and column C1 are identical with those used in the pressure-swing process. The methanol-containing distillate D1 from the top of the reactive column is fed to stage 6 of a 12-stage extraction column. Water is fed on the top tray at a rate of 1050 kmol/h and a temperature of 322 K, which is achieved by using a cooler (heat removal 1.24 MW). The column is a simple stripper with no reflux. The column operates at 2.5 atm so that cooling water can be used in the condenser (the reflux-drum temperature is 326 K). The reboiler heat input is 5.96 MW. The overhead vapor is condensed and is the C5 product stream. This column is designed by specifying a very small loss of methanol in the overhead vapor (0.01% of methanol fed to the column) and finding the minimum flow rate of extraction water that achieves this specification. Adding more than 10 trays or using reflux did not affect the recovery of methanol. The bottoms is essentially a binary methanol-water (23.5 mol % methanol), which is fed to stage 12 of a 32stage column operating at atmospheric pressure. The

number of trays in the second column was optimized by determining the TAC of the columns over a range of tray numbers. The reboiler heat input and condenser heat removal are 8.89 and 9.53 MW, respectively. The column diameter is 2.24 m. A reflux ratio of 2.1 produces 316 kmol/h of highpurity methanol in the distillate (99.9 mol %) and 1026 kmol/h of high-purity water in the bottoms (99.9 mol %). The methanol is combined with 230 kmol/h of a fresh methanol feed, and the total is split between the methanol feed streams to the reactor and to the reactive column. The water is combined with a small water makeup stream, cooled, and recycled back to the extractive column C2. Some makeup water is needed because a small amount of water goes overhead in the vapor from C2 (2.9 mol % water). The solubility of water in pentanes is quite small, so the reflux drum of column C2 would form two liquid phases (not shown in Figure 5). The aqueous phase would be 19.9 kmol/h and 99.9 mol % water. The organic phase would be 809 kmol/h and 0.5 mol % water. Table 1C gives the molar flow rates and conditions for the important streams in the separation section of this flowsheet. 4.2. Economics. Design parameters, capital costs, and energy costs for the separation sections of both processes are given in Table 2. The reboiler heat inputs in the two columns of the extractive-distillation process are about 30% of those in the pressure-swing process. This reduces the column diameters and heat-exchanger

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5723

Figure 7. (A) Feed rate disturbance for the extractive-distillation process. (B) Feed composition disturbances for the extractive-distillation process.

areas, so the capital cost is also much smaller (about 40% lower). These economics clearly indicate a large advantage for the extractive-distillation process over the pressureswing process from the perspective of steady-state economics. The issue of dynamic controllability must also be considered. 4.3. Plantwide Dynamic Control. The control scheme for the reactor and reactive column is identical

with that used in the pressure-swing process. Figure 6 gives the control structure developed for columns C2 and C3 in the extractive-distillation separation section. The flow rate of extraction water fed to the top of C2 is ratioed to the feed to this column D1 by using a multiplier and a remote-set flow controller. The temperature of the extraction water is controlled by manipulating cooling water to the cooler. The temperature at stage 7 is controlled by the reboiler heat input. The

5724

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 8. (A) Comparison for the feed rate change. (B) Comparison for the feed composition change.

base level is controlled by manipulating the bottoms, and the reflux-drum level is controlled by manipulating the distillate. There is no reflux. The binary methanol-water mixture from the bottoms of C2 is fed to C3. A constant reflux ratio is maintained in this column by adjusting the reflux flow rate. The temperature at stage 7 is controlled by the reboiler heat input. There are two key plantwide material balance loops associated with column C3. The level in the reflux drum provides a good indication of the inventory of methanol in the system. If this level is going down, more methanol

is being consumed in the reaction than is being fed into the process. Therefore, the control structure maintains the reflux-drum level in C3 by manipulating the methanol fresh feed. Note that the flow rate of the total methanol (D3 plus fresh methanol feed) is fixed by the two downstream flow controllers setting the flow rates to the reactor and to column C1. This means there is an immediate effect of the fresh feed flow rate on the reflux-drum level. The distillate flow D3 changes inversely with the fresh feed flow because the downstream flow rate is fixed. Thus,

Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5725

the reflux-drum level sees the change in the methanol fresh feed instantaneously. At the other end of the column, the base level provides a good indication of the inventory of water in the system. Ideally, there should be no loss of water because it just circulates around between the extractive column and the recovery column. However, there is a small amount of water lost in the overhead from column C2. A water makeup stream is used to control the liquid level in the base of column C3. This makeup flow is very small compared to the water circulation, so the base of column C3 must be sized to provide enough surge capacity to ride through disturbances. Figure 7A gives the responses of the process to 20% changes in the feed flow rate. Figure 7B gives responses to changes in the feed composition. Effective plantwide control is achieved. A comparison of Figure 4 with Figure 7 shows that the two processes have dynamic responses that are quite similar. A direct comparison of some of the key variables is shown in Figure 8. The control structures developed provide stable base-level regulatory control for large disturbances. The purity of the TAME product is held quite close to its specification. 5. Conclusion The design and control of two alternative separation sections in the reactive distillation process to produce TAME have been studied. Both processes are capable of producing high-purity TAME. The process that uses extractive distillation is much more attractive in terms of steady-state economics. Capital investment and energy costs are very significantly lower. The two processes have essentially the same satisfactory dynamic controllability. Their plantwide control structures have similarities and differences.

Nomenclature AC ) heat-transfer area of the condenser (m2) AHX ) heat-transfer area of the extract-water cooler (m2) AR ) heat transfer area of the reboiler (m2) Area ) area (m2) Bn ) bottoms flow rate from column n (kmol/h) Cn ) column n D ) diameter of the vessel (m) Dn ) distillate flow rate from column n (kmol/h) HX ) heat exchanger L ) length of the vessel (m) QC ) condenser heat removal (MW) QR ) reboiler heat input (MW) TAC ) total annual cost (106 $/year) ∆F ) change in the feed flow rate ∆z ) change in the feed composition

Literature Cited (1) Al-Arfaj, M. A.; Luyben, W. L. Plantwide control for TAME production using reactive distillation. AIChE J. 2004, 50 (7), 1462. (2) Subawalla, H.; Fair, J. R. Design guidelines for solidcatalyzed reactive distillation systems. Ind. Eng. Chem. Res. 1999, 38, 3693. (3) Stichlmair, J. G.; Fair, J. R. Distillation: Principles and Practices; Wiley: New York, 1998. (4) Luyben, W. L. Use of dynamic simulation to converge complex process flowsheets. Chem. Eng. Educ. 2004, 38, 2. (5) Luyben, W. L.; Tyreus, B. D.; Luyben, M. L. Plantwide Process Control; McGraw-Hill: New York, 1999.

Received for review January 12, 2005 Revised manuscript received April 15, 2005 Accepted May 4, 2005 IE058006Q