Chemical Engineering Science 59 (2004) 4339 – 4347 www.elsevier.com/locate/ces
Time optimal start-up strategies for reactive distillation columns Frauke Reepmeyer∗ , Jens-Uwe Repke, Günter Wozny Department of Process Dynamics and Operation, Technical University Berlin, TU Berlin, Str. d. 17. Juni 135/KWT-9, 10623 Berlin, Germany Received 1 July 2003; received in revised form 20 May 2004; accepted 17 June 2004
Abstract In this contribution, a rigorous process model to simulate the start-up of a cold and empty reactive distillation (RD) column is developed and experimentally validated with a transesterification process. Strategies for time optimal start-up of an RD column are presented. The mostly used strategy for conventional distillation of total reflux for RD is only recommendable with limitations. New, alternative strategies, like the recycling of the off-spec bottom and top product or the initial charging with product, to minimize the necessary start-up time are presented. Suitable strategies can save up to 82% of the needed time for the column start-up. 䉷 2004 Elsevier Ltd. All rights reserved. Keywords: Reactive distillation; Start-up; Dynamic simulation; Ethyl acetate; Transesterification
1. Introduction Reactive distillation (RD) can be defined as the integration of chemical reaction and thermal separation in one process step. Significant savings of investment and capital costs can be realised by reactive distillation. Further advantages like overcoming equilibrium limitations, achieving higher selectivities, easier separation of azeotropic mixtures or use of reaction heat for the distillation call attention for RD as an interesting process alternative for the chemical industry. Industrial applications are, for example, esterifications, like ethyl acetate or etherifications like the production of MTBE. For an extensive overview on industrial applications of RD, please refer to chapter one of (Sundmacher and Kienle, 2003). The integration of reaction and separation increases the complexity of the process. This requires a better knowledge of the process and the kinetic activities. Various published papers can be found on design studies and steady-state simulation of RD processes. A comprehensive overview is
∗ Corresponding author. Tel.: +49-30-314-26905; fax: +49-30-31426915. E-mail address:
[email protected] (F. Reepmeyer).
0009-2509/$ - see front matter 䉷 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2004.06.029
given by Doherty and Buzad (1992) and Taylor and Krishna (2000). Sources of the dynamic simulation of RD processes, which gained attention just recently are sparse. Information can be found in Alejski and Duprat (1996) or Scenna et al. (1998) for the production of acetates with different alcohols. Fuel additives like MTBE and TAME have been analysed by Abufares and Douglas (1995) or Sundmacher (1995). But all these papers focus on the dynamic behaviour near a known operating point based on a known steady state. The start-up of each distillation column and especially an RD column is a very complex process that is difficult to control. In contrast to distillation without reaction, the offspec product produced during start-up is difficult to recycle. Also, control of a start-up process is very challenging. Almost all process variables change rapidly in value; control variables like heating power, reflux ratio or feed flow have to be switched at least once, but mostly several times to reach the correct steady state. Different start-up strategies can yield different, undesired steady states as Blagov et al. (2000) and Güttinger (1998) show. Scenna and Benz (2003) describe multiple steady states for different initial conditions as starting points. In these publications, the column does not start from a cold and empty state, but from a defined state, with the trays filled, warm and in-phase equilibrium. But
4340
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347
to have consistent and physically sound starting values for a process, it is necessary to simulate the complete startup process. Also, to have a precise prediction of start-up time and cost of the start-up process, it is crucial to have a model for the complete process. This model also needs to be experimentally tested. In this paper at first the pilot plant is presented, where experiments for an RD process have been conducted. Then the developed rigorous model to simulate the start-up process of an RD column from the cold and empty state is explained. Then the model is applied to two examples, a transesterification and an esterification process.
2. Pilot plant for RD processes Fig. 1 shows the set-up of the pilot plant at the Technical University Berlin. Experimental data has been collected for a transesterification process. The temperature in the column is measured via PT100 elements on every second tray and the pressure at the top and bottom are documented. The plant also has six liquid sampling stations over the column height. The two feeds can be separately preheated to a defined temperature. The
Methylester 80°C, 6 l/h
column is completely automated with a Freelance 2000 process station. All other details are listed in Table 1. For more construction details, please refer to Bock (1998), where a steady-state model has been validated. Here the dynamic start-up model is investigated in detail. The sample reaction is a transesterification of a fatty methylester with isopropanol to isopropylester and methanol. All details of the process as well as the experimental data are presented in Section 6.1. This example is very complex and has not been found in the literature until now. Steady state as well as dynamic investigations have been carried out. Dynamic data from a cold and empty start-up of a reactive distillation column is presented here for the first time.
3. Rigorous start-up model For developing suitable guidelines for starting up reactive distillation columns, a rigorous model to simulate the complete start-up process from a cold and empty state is needed. It is too cost intensive to conduct several pilot plant experiments before actually starting a reactive distillation column. Therefore, a model has been developed and validated for a transesterification process, which can simulate the complete start-up process of a reactive distillation column from the cold and empty state. The dynamic behaviour of the trays is described for a rigorous model with the following assumptions: • • • • •
each tray can be modelled as a theoretical plate, reaction only in the liquid phase, kinetic description of reaction via rA =−kh cA cB +kr cC cD , vapour phase behaves ideal, vapour and liquid phase are in thermodynamic equilibrium, when both phases exist and • column at start-up is cold and empty (very small liquid holdup (for computational reasons) and 298 K). In order to describe the dynamics of the column from a cold and empty start-up three conditions for switching of equations are considered. These conditions for a change in the equation system are:
Isopropanol 30°C, 4 l/h
Fig. 1. Pilot plant at TU Berlin.
• liquid leaves the tray, if level is higher than weir height (Francis formula). It is assumed that there is no weeping through the tray holes, because the holes in the special bubble cap trays are very small, stage II in Fig. 2. • vapour leaves the tray, if pressure on the tray is higher than the liquid inlet pressure (tray pressure plus pressure drop through liquid holdup) on tray above, stage V in Fig . 2. • phase equilibrium is assumed, if the bubble conditions of the liquid on the tray are reached. Before that, there is no vapour phase present, stage IV in Fig. 2.
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347
4341
Table 1 Pilot plant data Parameter
Value
Parameter
Value
Number of trays Tray diameter (mm) Tray height (mm) Weir height (mm) Weir length (mm) Upper feed tray Upper feed stream (l/h) Upper feed temperature (◦ C) Upper feed composition
28 107 180 31 10 4 3.5 80 Educt ester mixed with catalyst
Top pressure (bar) Maximum heat duty (kW) Holdup in Reboiler (l) Reflux stream (l/h) Lower feed tray Lower feed stream (l/h) Lower feed temperature (◦ C) Lower feed composition
1.02 6 5.7 2.3 27 2.4 130 Educt alcohol
I
Feed
II
III
IV
p = pbub p =1bar T = 298K T = Tfeed T > Tfeed L V L V L
VI
V
L V steady state state
V p >pin V
L V
L
Fig. 2. Different stages during start-up of a tray.
I F level weir height T H EN Fout,liq = length ELSE Fout,liq = 0
(level−weir height) cf w l
(2)
Also the phase equilibrium equations are switched. Eq. (3) shows the used replacements for the fill-up process.
Fig. 2 illustrates the different stages of an arbitrary feed tray during the start-up process. In stage I, the tray is empty at 1 bar (or any other arbitrary pressure), feed is entering the tray. In stage II, the tray is filled with liquid, temperature is the same as feed temperature (assuming no heat loss), and a liquid stream is leaving the tray. If the reboiler is turned on, vapour starts to ascend in the column until in stage III, the vapour from the stage below enters the tray. With this stream the tray is heated up until bubble conditions are reached in stage IV. In this stage, the phase equilibrium equation is activated. Now there is vapour accumulating on the tray until the pressure is higher than on the tray above and a vapour stream is leaving as shown in stage V. If the reflux is entering the column, the trays above the feed tray are filled and in stage VI all entering and leaving streams are greater than zero. In these modelling assumptions, the pressure on the tray as well as vapour and liquid holdup are integration variables and not fixed to a certain value. The pressure at the column top is fixed at 1 bar for the presented examples, and can also be fixed at any arbitrary value or controlled. The describing equations besides the energy, mass and component balances are the Francis weir equation for the leaving liquid stream (Eq. (1)) and a momentum balance for the leaving vapour stream as in Eq. (2):
I F ptray pin T H EN Fout,vap = f ptray − pin ELSE Fout,vap = 0
1,5
(1)
I F pbub ptray T H EN vap yi ptray = xi i pi vap ptray = pbub = xi i pi i
Vtray = Mtot,vap v v + Mtot,liq v l ELSE yi = x i ptray = 1 bar Mtot,vap = 0
(3)
For a detailed description of the whole model, please consult Reepmeyer et al. (2003). With the mentioned switches in the described equations the DAE system is not smooth, but shows discontinuities that introduce problems with the numerical solution. The integrator step size decreases drastically before an equation switch. Also, the variable bounds have to be chosen carefully as well as the initial conditions. These are for all trays as follows: Mtot,liq = 0.001 mol, xi = xfeed , T = 298 K, p = 1 bar.
(4)
The liquid holdup is very small (0.0003% of steady-state value) and the initial liquid composition has feed composition, but this is an arbitrary choice. The pressure on all trays is ambient pressure. All calculations have been conducted with the commercial simulation tool gPROMS◦ from PSEnterprise. The model has to be validated with experimental data from the department’s pilot plant (Fig. 1), for the results please refer to Section 5.1.2.
4342
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347
4. Column start-up
Table 2 Kinetic data for transesterification for 1 mass% catalyst concentration
4.1. Conventional strategies The start-up procedure for distillation columns without reaction has already been examined. Kister (1979) describes problems with column start-ups. Eden et al. (2000) developed a procedure for generating start-up sequences utilizing process knowledge for heat-integrated columns. Löwe and Wozny (2001) developed an optimized strategy for the startup of an energy integrated column system and validated the results. Ruiz et al. (1988) and Gani et al. (1998) as well as Irbarren and Chiotti (1991) published various papers on the dynamic simulation of distillation columns with the total reflux strategy. The strategy that is recommended for most of the column start-up processes is the total reflux operation. For the start-up process from the cold and empty state for reactive distillation, no guidelines or “good practice” advice can be found in the open literature. From distillation without reaction, several strategies are known. Conventional distillation column start-up can be distinguished into four strategies: • conventional: set all control variables to steady-state values and wait, • total reflux: column is run in loop operation, no distillate removal, • total distillate removal: exact opposite to before mentioned strategy, column is run without reflux and • time optimized (developed for heat-integrated columns): heating duty and reflux are set to 1.3 their steady-state values. For an extensive description of the start-up strategies for distillation without reaction, please refer to Flender (1999). The measure for reaching steady state is the so-called MX function MXtop = |(xi,actual − xi,steady state )| (5) i
developed by Yasuoka and Nakanishi (1982). The function is the sum of differences between the actual composition in the top product and the steady-state composition of all components. If the MX function is permanently below 0.001, the desired steady state is defined to be reached. For conventional distillation without reaction, the commonly used strategy is the total reflux strategy as Kister (1992) recommends. The results from the simulation of the two-sample reactive distillation processes suggest that the strategy of choice to start up a reactive distillation column should be the conventional strategy, because it takes the shortest time to reach steady state. All four strategies have one thing in common, only the control variables reflux and heating duty are manipulated during the start-up period. For conventional distillation without reaction Flender (1999) found that savings up to 80% of the start-up time
Forward reaction Backward reaction
Frequency factor (m3 /mol h) 33 15.49
Activation energy (J/mol) 34 800 31 200
can be achieved if a column is started up utilizing the total distillate removal strategy. 4.2. Alternative start-up strategies As can be seen from the resulting start-up times for the two-sample processes (Tables 2 and 4) for the conventional strategies, the savings are rather small. The two limiting cases, total reflux and total distillate removal, do not raise the hope of significant savings in start-up time using any of the reflux strategies in combination. To further shorten the start-up time, an understanding of the complete process and technological knowledge is necessary. Alternative startup strategies have been tested with the validated rigorous model. These are, for example: • recycling of the off-spec bottom and top product during the start-up period and • initial charging of the product to the column.
4.2.1. Recycle of bottom product During the start-up phase, the bottom and top product are off specification; they can only be discarded or have to be separated in a further separation step. Therefore, it would be helpful to recycle the stream to the column, if the starting up period is not significantly prolonged. For simulation, the bottom product stream is split and then a part of the stream is mixed with the feed stream and fed to the column. Different schemes have been analysed: • No recycle is the base case with conventional settings of all control variables. • Split factor = recycle stream/bottom stream = 1 means that as much product is recycled as is withdrawn from the bottom (this is done for a duration of 5000 and 10 000 s). • Split factor = 10, 10 times as much is recycled to the column as is withdrawn. The recycle times are chosen because it is two times/four times the residence time in the column. 4.2.2. Initial charge of product As a second alternative, the initial charges of product or educt in excess have been tested. Scenna and Benz (2003) show in their paper that this will infer multiple steady states to the process. For the specific ethyl acetate and transesterification case in this paper, this has not been observed. Five
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347
4343
schemes have been implemented to the rigorous model for the following two sample processes (ethyl acetate and transesterification):
Table 3 Calculated start-up time for transesterification process Strategy
Start-up time (min)
• initial charge with feed (acetic acid/ethanol or isopropanol/methylester), • initial charge with high boiling liquid feed component in excess (acetic acid or methylester), • initial charge with low boiling liquid feed component in excess (ethanol or isopropanol), • initial charge with liquid resembling the steady-state bottom product and • initial charge with liquid resembling the steady-state top product.
Conventional Total reflux Total distillate removal Initial charge feed Initial charge bottom product Initial charge top product Initial charge methylester Initial charge isopropanol
309.8 305.3 348.8 308.1 316.4 313.9 305.2 306.3
90
5. Application examples
mole [%]
Initial charge here means that liquid holdup up to weir height (no liquid flow leaving the tray) on all trays at 1 bar and ambient temperature of 298 K. 45
MeOH exp
MeOH sim
IPA exp
IPA sim
ME exp
ME sim
IPE exp
IPE sim
5.1. Transesterification of a fatty methylester The developed model is now applied to an RD process example. In the presented pilot plant, experiments have been conducted and dynamic process data have been collected. The comparison between experimental data and simulation results is presented in Section 5.1.2. 5.1.1. Process description The first example is a transesterification process. In the presented pilot plant, the reaction of isopropanol with a fatty methylester to methanol and isopropylester has been conducted. For this transesterification process the differences in boiling points are very high. The reaction is homogeneously catalysed with sulphuric acid, which leaves the column with the bottom product. The catalyst behaves like an inert and is therefore not included in the balances as a separate component. The column has two feed streams (compare Fig. 1), one at the top of the column and one at the bottom. The ester is premixed with the catalyst so that it yields 1 mass% of acid in the ester feed. The mixture of acid and ester is then preheated to 80 ◦ C and fed to the top tray. The alcohol is prevapourized and then inserted at the bottom tray. This will yield a countercurrent flow of the educts in the column. With a holdup of 0.33 l per tray, the residence time of the ester is about 1.8 h. The reaction is captured with a kinetic approach. The kinetic data for this special transesterification reaction is gathered in Table 3. During start-up, the temperature profile has been continuously recorded. Liquid samples have been withdrawn from four trays over the column height as well as from the top product and reboiler approx. every 40 min for dynamic validation. The samples have been analysed with a gas chromatograph.
0 top
tray 9
tray 13
tray 18
tray 22
reboiler
Fig. 3. Steady-state model validation—liquid mol fraction (%).
5.1.2. Simulation results The transesterification process has been simulated with the model presented in Section 3. The results of the simulation are then being compared to the measured data from the pilot plant for the steady-state point as well as the dynamic start-up profiles for temperature and concentration. At first Fig. 3 shows the comparison of experimental steady-state composition profile and simulated results. The simulation matches the experiments pretty well in the steady-state mode, the highest deviation is about 8 mol% for the reboiler methylester concentration. All other values are within 5% precision. For the simulation of the start-up process, the exact description of the dynamic behaviour of the RD column is important. With sulphuric acid as catalyst, the conversion is not as high, but what is interesting here is the dynamic behaviour of the compositions during the start-up process. Fig. 4 gives a dynamic temperature profile, experiment and simulation starting from the cold and empty state at time zero. Due to the manual control of the reboiler heat duty (no PI controller implemented) the experimental curve shows some peaks in the reboiler temperature. In the model this has been taken into account by giving directly the heat duty in kW to the reboiler. The comparison with the simulation results shows that the dynamics and the time constant for vapour flow and reboiler temperature are very well predicted. The first drop in the reboiler temperature profile is the time when the reflux reaches the bottom. The
4344
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347 Table 4 Calculated start-up times for conventional strategies
440
390
T [K]
reboiler simulation stage 1 experiment
Strategy
Start-up time (min)
Conventional Total reflux Total distillate removal Time optimized
175 225 183 191
340 reboiler experiment
p = 1 bar
stage 1 simulation
290 0
120
240 time [min]
360
1
480
reflux ratio 10
Fig. 4. Dynamic model validation—temperature profile from cold column. Acetic Acid Ethanol Water
dynamic concentration profile bottom product
mol [%]
100
ME exp
ME sim
IPA exp
IPA sim
IPE exp
IPE sim
0.534 mol/sec 0.517 mol/sec 0.024 mol/sec
dcolumn= 0.6 m
6
11 50
TC set point: 366K
0.868 mol/sec
0
75
175
275
375
Fig. 6. Plant sketch for ethyl acetate process as in Lee and Dudukovic (1998).
time [min] Fig. 5. Dynamic model validation—concentration profile for bottom product.
time when the vapour reaches the condenser is also very well predicted. To trust the prediction of a dynamic model, more than anything else for reactive distillation, it is not just important to match the temperature profile but also the composition profile. Fig. 5 shows the experimental data in dots and the simulation data as lines. The error bars indicate a 5% error. Here as well a good prediction precision can be stated. The previous section showed a very good match between the simulation model and the experimental pilot plant data. Therefore, in a next step the different start-up strategies presented in Section 4 are implemented and the time consumption to reach the same steady state has been collected. 5.1.3. Start-up behavior During start-up, the manipulated variables are the reboiler duty and the reflux ratio. For the other strategies, the column has been prefed with material at 298 K and 1 bar. The results for the transesterification are gathered in Table 4. For the transesterification process, the recycle of product streams during start up did not show a significant influence
on the needed start-up time. Therefore, the results are neglected. The data in Table 4 show that the newly developed strategies cannot yield significant savings in the start-up time for the transesterification process. The fastest strategy is the initial charging of the high boiling educt, but the savings compared to the conventional strategy are only about 1.5% (4.6 min total). 5.2. Esterification of ethanol with acetic acid 5.2.1. Process description As a second example the esterification of ethanol and acetic acid to ethyl acetate is analysed and the steady-state operating point is compared to the literature data. This process has been chosen because it is widely known and it represents a major group in RD applications, the esterification processes. Fig. 6 displays the characteristics of this sample process. It is a purely theoretical example and no experimental is available in the open literature. All process parameters are documented in Appendix A. 5.2.2. Simulation results This example has also been simulated with the dynamic start-up model presented in Section 3. Fig. 7 compares the
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347
4345
1.24
370
MX function - top product [-]
start up time in h
T [K]
360
350 Lee, 1998 own simulation 340 1
3
5
7
9
11
no recycle
0.62
no recycle
8.0
splitfactor 1, 1.3 h
7.11
splitfactor 1, 2.6 h
8.88
splitfactor 10
9.5
SF 1, 1.3h SF 1, 2.6 h
13
tray SF 10
Fig. 7. Steady-state temperature profile–comparison: simulation–literature.
0 0
2.5
5
7.5
10
time [h]
mole fraction
0.6
Fig. 9. MX function for recycle of bottom product—ethyl acetate. Acetic Acid - (Lee,1998)
Ethyl Acetate (Lee,1998)
Acetic Acid -own simulation
Ethyl Acetate - own simulation
tion processes, the total reflux, should not be considered for reactive distillation processes, because the start-up time will be prolonged by 28%.
0.4
0.2
0 1
3
5
7
9
11
13
tray Fig. 8. Steady-state concentration profile–comparison: simulation– literature.
calculated temperature profile at the steady-state operating point with the literature data from Lee and Dudukovic (1998). The match between simulation results and the literature data is very good. The highest deviation is about 3 K at the top stage. Fig. 8 compares the steady-state concentration profile from the Lee and Dudukovic (1998) paper to the steady-state values calculated with the rigorous start-up model. As Fig. 8 shows, there is a good match between the steadystate values of the rigorous model and the literature data. Dynamic data for validation of the simulated start-up process are not available anywhere in the open literature. 5.2.3. Start-up behavior 5.2.3.1. Conventional strategies. With the dynamic model presented in Section 3, start-up strategies developed in Section 4 have been simulated. Table 2 compares the calculated start-up times for the strategies known from non-reactive distillation processes. As can be seen, there are no significant savings in the start-up time, if only the reflux and heating rate is manipulated. The recommended strategy for conventional distilla-
5.2.3.2. Product recycling. The alternative strategies are also simulated with the rigorous start-up model presented in Section 3. Fig. 9 depicts the MX function (see Eq. (5)) as a function of time for the different schemes for the ethyl acetate process. If the MX function reaches zero, a steady state is established. A high split factor (SF = 10) yields a steep descent in the MX function, but after stopping of the recycling stream, it rises again and takes the longest time to reach steady state. With a moderate recycle, part of the necessary start-up time (about 11%) can be saved. A recycle of top product during start-up does not result in any acceleration; neither different split factors nor recycle durations show an influence. 5.2.3.3. Initial charging. The strategies with initial charging of product as explained in Section 4.2.2 have also been applied to the ethyl acetate example. The dynamic characteristics of the MX function of the top product are presented in Fig. 10. As can be seen from Fig. 9 the longest start-up time is needed, if acetic acid (high boiler) is initially charged. To heat up the high boiler takes the longest time because the only heat flow into the column is to the reboiler, so the ascending vapour stream heats the above trays. More energy is lost while vapourizing the high boiler acetic acid. The more low the boiler is on the tray (ethanol and top product), the faster the steady state is reached. For initial charge of the top product to the column, the start-up time can be reduced by 82% from 6.24 to 1.07 h (the first number in the table in Fig. 4 gives the duration for MXtop being steady state, the second for MXbottom ). For the ethyl acetate it is useful to keep the top product from the last charge to prefeed the column before the next start-up.
4346
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347 2 start up time in h
MX function - top product [-]
initial charge feed
6.24
bottom product
6.30
top product
0.79 / 1.07
acetic acid
9.26
ethanol
4.81
feed
1 ethanol
bottom product
acetic acid
top product 0
0
6
3
9
time [h]
Fig. 10. MX function for initial charging—ethyl acetate.
case can save approximately 12% of the start-up time for the ethyl acetate case. Higher recycling rates prolong the duration to reach steady state. Recycling of the top product stream does not have significant influence on the start-up time neither in the ethyl acetate nor in the transesterification case. With initial charging of the light boiling component in the column, the start-up time can significantly be reduced. For the ethyl acetate case, initial charging with ethanol in excess gives 23% savings in time and initial charging with the top product (ethyl acetate, ethanol mixture) will decrease the start-up time by 82%. For the transesterification case, no savings in start-up time could be observed for the charging strategies, because the very slow reaction does not depend on the excess of reactant but more on the residence time in the column. For a general conclusion incorporating all RD processes, different types of systems have to be classified and also be analysed with the developed model in the future.
5.3. Comparison of both examples The main difference between the esterification and the transesterification process is the reaction velocity. The ethyl acetate reaction is rather fast and the transesterification is very slow. Another difference is the Damköhler number, which is 0.019 for the transesterification process and 0.036 for ethyl acetate. For the exact benefit calculation, the loss in product to sell should be offset from the savings in start-up time and the off-spec product during that period. Different processes with different components and reaction kinetics should be analysed, so influences of these parameters can be concluded and used for recommendations for reactive distillation column start-up.
6. Conclusion In this paper, a rigorous model to simulate the start-up process of a reactive distillation column from the cold and empty state has been presented. The model has been validated using two different processes as example, one taken from the literature and one using the experimental data from a transesterification process in a 10 cm pilot plant column with 28 trays. The steady state as well as the dynamic behaviour of an RD column for the two-sample processes can be captured by the developed model. With this model, different start-up strategies have been tested. It has been found that the strategy of total reflux recommended for conventional distillation without reaction is not suitable for reactive distillation, but takes the longest start-up time. New, alternative strategies have been developed and applied to the ethyl acetate process and a transesterification process. Recycling of the off-spec product streams during the startup period has been investigated. The results show that a moderate recycling of bottom product for the ethyl acetate
Notation
A cA , c B , cC , cD cf w , c w Fout,liq , Fout,vap kh , kr Mtot,liq , Mtot,vap MXtop MXbottom vap pi , pbub rA reboiler set R SF T , ptray vl , vv Vtray xi , yi , xfeed
free area for vapour throughput, m2 concentration of components, mol/m3 resistance coefficient in down comer/valves leaving liquid/vapour stream, mol/s forward and backward frequency factor, m3 /mol s complete liquid/vapour hold up on tray, mol MX function for top/bottom product stream, dimensionless vapor pressure for component I/bubble pressure on tray, bar reaction rate of A, mol/s set point for reboiler temperature controller, K reflux ratio, dimensionless split factor for recycling stream, dimensionless temperature (K) and pressure (bar) on tray liquid/vapour volume, m3 /mol complete tray volume, m3 liquid/vapour mole fraction of component I, feed, mol/mol
Greek letters
p l
pressure drop between two trays, bar liquid density, kg/m3
F. Reepmeyer et al. / Chemical Engineering Science 59 (2004) 4339 – 4347 Table 5 Data for ethyl acetate process as in Lee and Dudukovic (1998) Parameter
Value
Parameter
Value
Number of trays Feed tray Tray diameter (m) Holdup condenser (l) Holdup reboiler (l) Top pressure (bar) Forward reaction (m3 /mol h)
13 6 0.6 3 10 1 483.33
Weir height (m) Weir length (m) Tray height (m) Reflux ratio Feed stream (mol/s) Bottom product (mol/s) Backward reaction (m3 /mol h)
0.05 0.457 0.34 10 1.076 0.208 123
Acknowledgements The authors gratefully acknowledge the financial grant from the BMWi through the AIF (Arbeitsgemeinschaft industrieller Forschungsvereinigungen), Grant No. 12897N/1. Appendix A Data for ethyl acetate process can be seen in Table 5. References Abufares, A.A., Douglas, P.L., 1995. Mathematical modelling and simulation of an MTBE catalytic distillation process using SPEEDUP and ASPENPlus. Institute of Chemical Engineers 73, 3–12. Alejski, K., Duprat, F., 1996. Dynamic simulation of the multicomponent reactive distillation. Chemical Engineering Science 51, 4237–4252. Blagov, S., Bessling, B., Schoenmakers, H., Hasse, H., 2000. Feasibility and multiplicity in reaction–distillation processes for systems with competing irreversible reactions. Chemical Engineering Science 55, 5421–5436. Bock, H., 1998. Theoretische und experimentelle Analyse der Reaktivrektifikation, VDI Verlag GmbH, Düsseldorf, ISBN 3-18358803-x. Doherty, M., Buzad, G., 1992. Reactive distillation by design. Chemical Engineering Research and Design: Transactions of the Institution of Chemical Engineers. Part A 70, 448–457.
4347
Eden, M., Koggersbol, A., Hallager, L., 2000. Dynamics and control during startup of heat integrated distillation column. Computers and Chemical Engineering 24, 1091–1097. Flender, M., 1999. Zeitoptimale Strategien für Anfahr- und Produktwechselvorgänge an Rektifizieranlagen. VDI Verlag GmbH, Düsseldorf, ISBN 3-18-361003-5. Gani, R., Jepsen, T., Perez-Cisnero, S., 1998. A generalized reactive separation unit model. Modeling and simulation aspects. Computers and Chemical Engineering 22 (Suppl.), S363–S370. Güttinger, T.E., 1998. Multiple steady states in azeotropic and reactive distillation. Ph.D. Thesis, ETH Zürich. Irbarren, O., Chiotti, O., 1991. Simplified analytical prediction of distillation column startup time at total reflux. Canadian Journal of Chemical Engineering (69), 377–382. Kister, H., 1992. Distillation operation. McGraw Hill, New York. Kister, H.Z., 1979. When tower startup has problems. Hydrocarbon Processing 50 (2), 89–94. Lee, J., Dudukovic, M., 1998. A comparison of the equilibrium and nonequilibrium models for a multicomponent reactive distillation column. Computers and Chemical Engineering (23), 159–172. Lowe, K., Wozny, G., 2001. A new strategy for product switchover and startup for a heat- and mass-integrated distillation system. Chemical Engineering and Processing 40, 295–302. Reepmeyer, F., Repke, J.-U., Wozny, G., 2003. Analysis of the start-up process for reactive distillation. Chemical Engineering and Technology 26, 81–86. Ruiz, C.A., Cameron, I.T., Gani, R., 1988. A generalized dynamic model for sitillation column - III. Study of startup operations. Computers and Chemical Engineering 12 (1), 1–14. Scenna, N.J., Benz, S.J., 2003. Start-up operation of reactive columns with mulitple steady states: the ethylene glycol case. Industrial Engineering and Chemical Research (42), 873–882. Scenna, N., Ruiz, C., Benz, S., 1998. Dynamic simulation of start-up procedures of reactive distillation columns. Computers and Chemical Engineering (22), S719–S722. Sundmacher, K., 1995. Reaktivdestillation mit katalytischen Füllkörperpackungen - ein neuer Prozess zur Herstellung der Kraftstoffkomponente MTBE. Clausthal-Zellerfeld: Papierflieger. Sundmacher, K., Kienle, A. (Eds.), 2003. Reactive distillation. Status and future directions. Wiley-VCH, Weinheim, ISBN 3-527-30579-3. Taylor, R., Krishna, R., 2000. Modelling reactive distillation. Chemical Engineering Science 55, 5183–5229. Yasuoka, H., Nakanishi, E., 1982. A convenient algorithm for the online optimum startup operation of binary distillation. Memoirs of the Faculty of Engineering, Kobe University (29), 161–169.