Design and Control of Refrigerated-Purge Distillation Columns

Design and Control of Refrigerated-Purge Distillation Columns. William L. Luyben†. Process Modeling and Control Center, Department of Chemical Engin...
0 downloads 0 Views 233KB Size
Ind. Eng. Chem. Res. 2004, 43, 8133-8140

8133

RESEARCH NOTES Design and Control of Refrigerated-Purge Distillation Columns William L. Luyben† Process Modeling and Control Center, Department of Chemical Engineering, Lehigh University, Bethlehem, Pennsylvania 18015

Because cooling water is much less expensive than refrigeration in removing heat, many distillation columns are designed so that reflux-drum temperatures are higher than the available cooling water temperature. This is achieved by adjusting the column pressure. However, if the light-key component in the distillate has a high vapor pressure (low-boiling component), the resulting column pressure is high. This usually makes the separation more difficult than that in lower-pressure operation and increases the energy consumption. This paper explores an alternative configuration called “refrigerated purge” in which two stages of condensation are used. The first condenser uses cooling water and condenses a portion of the vapor coming from the top of the column. The condensate is refluxed to the column. The remaining vapor flows to a small refrigerated vent condenser, operating at a lower temperature and resulting in lower pressure. Expensive refrigeration is required in this second condenser, but less energy is used in the column reboiler because the pressure in the column is lower. This paper compares the two alternative flowsheets in terms of both the steady-state design and the dynamic controllability. Results demonstrate that the refrigerated-purge configuration is more economical than the conventional one when the column feed contains small amounts (about 10%) of the lightkey component. However, dynamic control of this more complex process is more difficult. 1. Introduction Because cooling water is an inexpensive energy sink, many distillation columns are designed so that refluxdrum temperatures are higher than the available cooling water temperature. This is achieved by adjusting the column pressure. Reflux-drum temperatures are typically set around 120 °F so that reasonable condenser areas result with 90 °F inlet cooling water. The column pressure depends on the reflux-drum temperature and distillate composition. If the distillate is a liquid (total condenser), the condenser pressure is the bubblepoint pressure of the distillate liquid at 120 °F. If the distillate is a vapor (partial condenser), the condenser pressure is the dewpoint pressure of the distillate vapor at 120 °F. The distillate consists of mostly the light-key component in the feed. If this component has a high vapor pressure (low-boiling component), the resulting column pressure is high. In many systems, the difficulty of separation increases as the pressure increases (relative volatilities decrease with increasing pressure). Therefore, the high column pressure makes the separation between the light- and heavy-key components difficult, which results in high energy consumption. This paper explores an alternative configuration in which two stages of condensation are used, which permits lower pressure operation. The first condenser uses cooling water and condenses a portion of the vapor coming from the top of the column. The condensate temperature (120 °F) is fixed by the available cooling †

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

water temperature, but the composition of the condensate is not as high in the light-key component as it is in the final distillate product. This means that the pressure can be lower for the same temperature because of the larger amount of heavy component. The liquid from the first condenser is refluxed to the column. The remaining vapor flows to a small refrigerated vent condenser, operating at a lower temperature, from which the distillate product is produced. For a given distillate composition, the lower temperature attainable using refrigeration (chilled water or brine) results in a lower pressure. This reduces the energy consumption in the column. However, some of the cooling is done by refrigeration, which is much more expensive than cooling water. Thus, there is an engineering tradeoff between the reboiler energy cost and refrigeration cost. The amount of the light-key component in the column feed is a key parameter in assessing the economic viability of the refrigerated-purge flowsheet. This paper explores both the steady-state design and the dynamic controllability of this system and compares them to those of a conventional system. 2. Process Studied The numerical case used comes from an industrial example in a polyvinyl alcohol process. A feed stream is a ternary mixture of acetaldehyde (AA), methyl acetate (MeAc), and methanol. We want to remove the AA in a distillation column. The light-key component (AA) is significantly lighter than the heavy-key component (MeAc). The normal boiling points are 69.5 and

10.1021/ie040188l CCC: $27.50 © 2004 American Chemical Society Published on Web 11/10/2004

8134

Ind. Eng. Chem. Res., Vol. 43, No. 25, 2004 Table 1. Design Parameters for the Conventional Column composition (mol %) feed AA MeAc MeOH distillate AA bottoms AA reflux ratio heat duties condenser (106 Btu/h) reboiler diameter (ft) temperature (°F) reflux drum base pressure (psia) area (ft2) condenser reboiler 3 capital (10 $) shell heat exchangers energy (103 $/year) TAC (103 $/year)

Figure 1. Conventional water-cooled-purge column.

134.5 °F, respectively. The desired distillate purity is 98 mol % AA. We assume that the distillate can be removed as a vapor product, which means lower pressure than if it is produced as a liquid. The reflux-drum temperature is set at a conservative 126 °F. This means the condenser pressure must be 38.5 psia in a conventional column, as shown in Figure 1. The desired bottoms purity is 0.01 mol % AA in the mixture of MeAc and methanol leaving in the bottoms. The separation of this mixture is complicated by the existence of a minimum-boiling azeotrope. Extractive distillation or pressure-swing flowsheets can be used to produce the MeAc and methanol products of desired purity. This separation is not considered in this paper. A range of feed compositions is considered. In the base case, the feed is 1000 lb‚mol/h with a composition of 5 mol % AA, 50 mol % MeAc, and 45 mol % methanol. The commercial process simulation products from Aspen Technology are used in this study: Aspen Plus for steady-state design and Aspen Dynamics for dynamic control studies. The Wilson physical property package is used. 3. Steady-State Design The economic optimum steady-state designs of the two alternative flowsheets are developed in this section. Identical feed and product streams are produced by each process. However, equipment sizes (columns and heat exchangers), cooling costs, and energy consumptions are different. 3.1. Conventional Column. With the composition of the vapor distillate fixed (98 mol % AA) and the reflux-drum temperature fixed (126 °F), the condenser pressure is 38.5 psia. To establish the optimum number of trays, the column is simulated in Aspen Plus with a specified number of total stages and feed stages. The “design spec/vary” feature is used to fix the distillate and bottoms compositions by varying the distillate flow rate and reflux ratio. Cases are run with different numbers of stages. As more trays are used, the required reflux ratio decreases. The minimum reflux ratio corresponds to the situation where the reflux ratio no longer decreases as more trays are added. The minimum reflux ratio is 26. Using the heuristic that the economic optimum design has a reflux ratio 1.2 times the minimum, the optimum reflux ratio

2 53 45 98 0.01 83.0 17.7 21.2 5.98 126 190 38.5 5620 3070 600 714 1070 1506

5 50 45 98 0.01 31.6 16.9 20.5 5.88 126 189 38.5 5360 2930 589 693 1030 1460

10 45 45 98 0.01 15.4 16.5 20.6 5.88 126 190 38.5 5230 2980 589 689 1040 1462

20 35 45 98 0.01 6.60 14.2 19.0 5.63 126 191 38.5 4490 2800 562 639 959 1360

Table 2. Basis of Sizing and Economics water-cooled condenser heat-transfer coefficient ) 150 Btu/h‚°F‚ft2 differential temperature ) 126 - 105 °F capital cost ) 1557(area)0.65 refrigerated condenser (chilled water at 60 °F) heat-transfer coefficient ) 150 Btu/h‚°F‚ft2 differential temperature ) TD2 - 60 °F capital cost ) 1557(area)0.65 reboiler heat-transfer coefficient ) 100 Btu/h‚°F‚ft2 differential temperature ) 259 - TB °F capital cost ) 1557(area)0.65 column vessel capital cost ) 1917D1.066L0.802 energy cost: 65 psia steam ) $5.75 per 106 Btu refrigerant cost: $4.20 per 106 Btu TAC ) (capital cost/payback period) + energy cost payback period ) 3 years

is 31. This corresponds to a column with 50 trays (52 stages in Aspen terminology). The feed stage is varied in each of the cases considered above to find the location that minimizes the reboiler heat input. The optimum feed stage for the 52stage column is stage 10, which corresponds to a reboiler heat input of 20.5 × 106 Btu/h. The “tray sizing” feature of Aspen Plus is used to determine the diameter of the column. The base pressure in this 52-stage column is 46.15 psia, assuming a 0.15 psi pressure drop per stage, with a condenser pressure of 38.5 psia. The resulting base temperature is 189 °F with the MeAc/methanol mixture in the bottoms. We assume that low-pressure steam (65 psia) is used to heat the reboiler. Providing a 30 psi pressure drop over the steam control valve results in a steam pressure in the reboiler of 35 psia (259 °F). The differential temperature driving force in the reboiler is 259 °F minus the column base temperature, which varies with the column pressure and bottoms composition. An overall heat-transfer coefficient of 100 Btu/h‚ ft2‚°F is used to size the reboiler area. The cost of steam is assumed to be $5.75 per 106 Btu. The condenser area is calculated assuming an overall heat-transfer coefficient of 150 Btu/h‚ft2‚°F. The temperature differential driving force is 126 °F minus an average cooling water temperature of 105 °F. Table 1 gives the design parameters and the economics of the conventional column for several cases. Details of the sizing and economic assumptions are given in Table 2. The base case is 5 mol % AA in the feed. The column diameter is 5.88 ft, giving a vessel cost of

Ind. Eng. Chem. Res., Vol. 43, No. 25, 2004 8135

Figure 2. Refrigerated-purge column.

$589 000. The areas of the condenser and reboiler are 5360 and 2930 ft2, respectively, giving a total heatexchanger cost of $693 000. Total capital investment in the column and heat exchangers is $1 280 000. The cost of energy is $1 030 000/year. The resulting total annual cost (TAC) is $1 460 000/year, assuming a payback period of 3 years. As the concentration of AA in the feed increases from 2 to 20 mol % AA, the reflux ratio decreases drastically, the condenser heat duty decreases slightly, and the reboiler duty is essentially constant. Therefore, the diameter of the column changes very little, and capital and energy costs are very weak functions of the feed concentration. These results are for columns with the same number of trays (52 stages). To illustrate the effect of pressure on the ease of separation in the system, the pressure in the standard column is varied over a wide range as shown in Figure 3. As the pressure is reduced, the energy consumption is reduced. However, the required reflux-drum temper-

Figure 3. Effect of the pressure; 50-tray standard column.

ature is also lower. At some minimum pressure, cooling water can no longer be used in the condenser. 3.2. Refrigerated-Purge Column. In the refrigerated-purge column, the idea is to perform a partial condensation of the vapor from the top of the column, producing a vapor stream from the reflux drum that is not at the specified distillate product purity. This vapor will contain more heavy-key component, so at the same reflux-drum temperature, the pressure in the column will be lower. This makes the separation in the column easier. The vapor vent from the reflux drum can be handled in several ways. Most frequently a heat exchanger is mounted vertically above the reflux drum. This vent condenser is cooled by a refrigerant (chilled water or brine). The vapor enters the bottom of the tubes, and some condensation occurs. The liquid runs back down into the reflux drum, countercurrent to the rising vapor flow. Thus, some fractionation is achieved in the vent condenser. Another alternative is to use a rectifier. Figure 2 shows such a system, which is what is used in the simulation because the rectifier unit operation is available in Aspen Plus and Aspen Dynamics. A small fivetray column is used that has a refrigerated condenser. Reflux is returned to the top of the rectifier, producing separation. The vapor stream from the rectifier reflux drum is the distillate product from the system. Its composition is 98 mol % AA. The pressure in the rectifier reflux drum is assumed to be 2 psi lower than the pressure in the main column reflux drum. The temperature in the rectifier reflux drum is lower because of the higher AA concentration as well as the lower pressure. Thus, refrigerant must be used in the second condenser. There is an additional design degree of freedom in this system, which is selected to be the composition yD1 of the vapor stream from the first reflux drum. As the concentration in the stream is reduced from 98 mol % AA, the pressure in the system can be reduced (for the

8136

Ind. Eng. Chem. Res., Vol. 43, No. 25, 2004

Table 3. Effect of the Vapor Composition from the First Condenser in the Refrigerated-Purge Column (5 mol % Feed) composition (mol %) vapor from the first condenser (mol % AA) pressure (psia) first reflux drum refrigerated drum temperature (°F) first reflux drum refrigerated drum column base reflux ratio heat duties (106 Btu/h) reboiler first condenser refrigerated condenser diameter (ft) column rectifier rectifier bottoms flow (lb‚mol/h) area (ft2) first condenser refrigerated condenser reboiler capital (103 $) column rectifier heat exchangers total energy (103 $/year) steam (65 psia) refrigeration TAC (103 $/year)

50

60

70

80

20 18

22 20

24.5 22.5

28 26

126 83 156 5.48

126 89 160 7.37

126 95 165 9.68

126 103 171 10.2

15.6 11.6 2.04

16.2 12.4 1.05

16.8 13.2 0.748

17.3 14.1 0.511

5.65 2.04 113

5.66 1.78 81

5.69 1.57 59

5.65 1.38 42

3680 591 1510

3940 239 1640

4190 143 1790

4480 79 1970

565 30 604 1200

566 26 584 1180

569 23 594 1190

565 20 610 1190

786 75 1260

816 38 1250

846 27 1270

871 19 1290

same 126 °F temperature in the first condenser), but more vapor must be condensed in the refrigerated heat exchanger. Thus, there is a tradeoff between saving reboiler energy because of the easier separation and increasing the refrigerant load. The conditions shown in Figure 3 correspond to the optimum steady-state design of this type of system for the 5 mol % AA feed. Table 3 gives results for several different values of composition yD1 of the vapor from the first reflux drum. As the mole percent of AA (yD1) of the

Figure 4. Effect of the feed composition.

vapor decreases, the reboiler duty, column diameter, and reboiler and condenser areas all decrease in the main column. However, the condenser duty, condenser area, diameter, and refrigerant load all increase in the rectifier. The minimum TAC occurs with a first reflux-drum vapor composition of 60 mol % AA. The pressure in the column is reduced from 38.5 psia in the standard column to 22 psia. This decreases the reflux ratio from 31.6 to 7.34 and the reboiler heat input from 20.5 × 106 Btu/h in the standard column to 16.25 × 106 Btu/h. However, the refrigeration required in the second condenser is 1.05 × 106 Btu/h, which is much more expensive than that of cooling water. The TAC of the standard column is $1 360 000/year for the case with a 5 mol % AA feed composition. The TAC of the refrigerated-purge column is $1 250 000 for the same case, a 9% reduction. So, for small amounts of the light-key component in the feed, the refrigeratedpurge column is economically better. 3.3. Other Feed Compositions. The effect of the feed composition on the standard column has already been seen (Table 1) to be quite small. The effect of the feed composition on the refrigerated-purge column would be expected to be quite significant. To explore this, the process is designed for several values of the feed composition. See Figure 4. For each feed composition, the optimum value of vapor composition from the first condenser must be found (as illustrated in Table 3 for the 5 mol % AA case). For a composition of 30 mol % AA in the feed, the optimum vapor composition is 70 mol % AA and the TAC of the standard column is lower than that of the refrigerated-purge column. For a composition of 2 mol % AA in the feed, the optimum vapor composition is 40 mol % AA and the TAC of the refrigerated-purge flowsheet is $1 157 000/ year compared to the conventional column with a TAC of $1 506 000 (a reduction of 23%). So, the refrigerated-purge configuration is only attractive when the concentration of the light-key component in the feed is fairly small.

Ind. Eng. Chem. Res., Vol. 43, No. 25, 2004 8137

The capital investment in this system is $947 000, the energy cost is $484 000/year, and the refrigeration cost is $42 000/year, giving a TAC of $842 000/year. This represents a 21% reduction, which is more than twice that of the other chemical system for the same 5 mol % feed. This occurs because of the strong dependence in this hydrocarbon system of the relative volatilities on pressure. These results demonstrate the economic advantage of the refrigerated-purge configuration when the amount of the light-key component in the feed is fairly small. The next section compares the dynamics of the two alternative flowsheets. 4. Dynamic Control

Figure 5. Control scheme for the conventional column.

3.4. Another Chemical Example. The specific numerical example considered up to this point is the separation of AA from MeAc and methanol. To make sure the results are not limited to this system, results from another separation are generated. The industrially important separation of propane from isobutane and normal butane is considered. Results for the 5 mol % propane case are presented. The standard column operates at 200 psia to give a 110 °F reflux-drum temperature with a 98 mol % propane vapor distillate. The bottoms specification is 0.01 mol % propane. A 30-tray column is used. The required reflux ratio is 34.2, and the reboiler heat input is 14.2 × 106 Btu/h. The capital investment is $1 070 000, and the energy cost is $715 000/year, giving a TAC of $1 072 000/year. The refrigerated-purge column flowsheet operates at 125 psia with a vapor composition of 60 mol % propane leaving the first condenser at 110 °F. The rectifier has a reflux-drum temperature of 75 °F. The reboiler energy is reduced, because of the lower pressure, to 9.60 × 106 Btu/h. The refrigeration load is 1.15 × 106 Btu/h.

Figure 6. Control scheme for the refrigerated-purge column.

The base-case feed composition of 5 mol % AA is used to compare the dynamics of the two processes. The steady-state Aspen Plus file is exported into Aspen Dynamics as a pressure-driven simulation after all of the dynamic parameters have been specified. Column diameters are calculated by Aspen’s “tray sizing” function. Reflux-drum holdups and base holdups are specified to provide 5 min of liquid holdup when levels are at 50%. 4.1. Conventional Column. The control structure is a conventional scheme with the following loops (shown in Figure 5): 1. The pressure is controlled by the vapor distillate flow rate. 2. The condenser heat removal (cooling water) is held constant. 3. The reflux-drum level is controlled by reflux because the reflux ratio is very large. 4. The temperature of stage 4 is controlled at 155 °F by the reboiler heat input. This location is selected because it is where the temperature profile is changing the most. 5. The base level is controlled by the bottoms flow rate. The temperature loop has a 1-min dead time and is tuned by using the relay-feedback test to determine the ultimate gain (3.1 with a 100 °F temperature transmitter span) and the ultimate period (4.2 min). Then the

8138

Ind. Eng. Chem. Res., Vol. 43, No. 25, 2004

Figure 7. Feed-rate disturbances: with and without TC2.

Figure 8. Feed-rate disturbances.

Tyreus-Luyben tuning rules are used (KC ) 0.97 and τI ) 9.2 min). The default pressure controller settings from Aspen Dynamics are used. Both level controllers are proportional-only with a gain of 2. 4.2. Refrigerated-Purge Column. This process has more control degrees of freedom and more variables to be controlled. The principal new variable is the refrigeration load, which can be manipulated. Two alternative control structures are explored. The first keeps the refrigerant load constant. The second manipulates it to control a temperature in the rectifier reflux drum. For both structures, the following loops are as shown in Figure 6:

1. The pressure is controlled by the vapor distillate flow rate from the rectifier reflux drum. 2. The condenser heat removal (cooling water) in the first column is held constant. 3. The reflux-drum levels in both columns are controlled by their respective reflux flow rates. 4. The temperature of stage 2 in the first column is controlled at 136.4 °F by the reboiler heat input (TC1). This location is selected because it is where the temperature profile is changing the most. 5. The base levels in both columns are controlled by their respective bottoms flow rate. Note that the bottoms

Ind. Eng. Chem. Res., Vol. 43, No. 25, 2004 8139

Figure 9. Feed composition disturbances.

from the rectifier column is refluxed to the top of the first column. 6. In the first structure, the heat removal in the rectifier condenser (refrigeration) is held constant. In the second structure, the temperature of the rectifier reflux drum is controlled by manipulating the refrigerant load (TC2). Both temperature loops have 1-min dead times. The stage 2 temperature loop is tuned first to determine the ultimate gain (5.44 with a 100 °F temperature transmitter span), the ultimate period (3.6 min), and the Tyreus-Luyben settings (KC ) 1.7 and τI ) 8.0 min). When the rectifier reflux-drum temperature is controlled, it is tuned with the stage 2 temperature controller on automatic. The ultimate gain (12 with a 100 °F temperature transmitter span) and the ultimate period (1.8 min) give Tyreus-Luyben settings of KC ) 3.7 and τI ) 4.0 min. These settings were found to be too oscillatory, so this loop was detuned by reducing the gain and increasing the integral time (KC ) 2 and τI ) 8.0 min). 4.3. Dynamic Results. Disturbances in the feed flow rate and feed composition are made to both processes. Figure 7 shows the effect of using the rectifier refluxdrum temperature controller in the refrigerated-purge process. The disturbances are two step changes in the setpoint of the feed flow controller: feed is increased 20% at t ) 0.2 h and then decreased 40% at t ) 4 h. The composition of the vapor distillate product yD2 from the rectifier reflux drum is shown in the fourth graph, left column. With the TC2 on manual (refrigerant load constant), the product composition is not maintained at its specified value of 98 mol % AA. The temperature in the rectifier reflux drum changes. With TC2 on automatic, this temperature is maintained by manipulating the refrigerant load. The reflux flow in the rectifier (R2) changes, and the product composition is held very close to the desired value. Therefore, the second temperature controller is needed in this process.

Figure 8 gives a comparison between the conventional column and the refrigerated-purge process for the feed flow-rate disturbance. Both processes provide good dynamic responses for these large disturbances. Figure 9 compares the two systems when two step disturbances in the feed composition are made. The feed composition is changed at t ) 0.2 h from 5 mol % AA and 50 mol % MeAc to 8 mol % AA and 47 mol % MeAc. Then at t ) 4 h, the feed composition is changed to 2 mol % AA and 53 mol % MeAc. These very large changes produce large changes in the flow rate of the vapor distillate product, but both systems do a good job in maintaining stable operation with the product quality closely controlled. These results show that the refrigerated-purge process requires a more complex control structure (two temperature controllers) and is somewhat more difficult to tune. Interaction among the pressure loop, the column temperature loop, and the rectifier reflux-drum temperature controller requires a careful analysis of the dynamics of the system and the use of controller tuning techniques that take this interaction into account. 5. Conclusion The results of this study demonstrate that the refrigerated-purge process configuration should be considered in those situations where there is a relatively small amount of the light-key component in the feed and the separation is more difficult at higher pressures. The small amount of expensive refrigeration may be more than compensated for by the reduction in energy consumption. The refrigerated-purge process is more complex and is more difficult to control, but an effective control structure has been developed and tested with very large disturbances. The dynamic performance of the refrigerated-purge system is comparable to that of the standard process.

8140

Ind. Eng. Chem. Res., Vol. 43, No. 25, 2004

Nomenclature AA ) acetaldehyde B ) bottoms flow rate (lb‚mol/h) D ) distillate flow rate (lb‚mol/h) F ) feed flow rate (lb‚mol/h) KC ) controller gain MeAc ) methyl acetate MeOH ) methanol QC ) condenser heat removal (Btu/h) QR ) reboiler heat input (Btu/h) R ) reflux flow rate (lb‚mol/h) RR ) reflux ratio ) R/D

T1 ) temperature of the control tray in the first column (°F) TD2 ) temperature of the reflux drum in the rectifier (°F) yD ) composition of the vapor distillate product (mole fraction AA) yD1 ) composition of the vapor from the first column reflux drum (mole fraction AA) τI ) controller integral time (min) Received for review June 25, 2004 Revised manuscript received September 24, 2004 Accepted October 20, 2004 IE040188L