Professional Documents
Culture Documents
Recent advances in technologies based on process intensification offer different design alternatives
for chemical processes including reaction and separation units. An excellent example of process
intensification is reactive distillation, which combines reaction and separation units in one piece
of equipment. This paper studies the steady-state economic optimum design of a chemical process
with a generic exothermic reversible reaction A + B T C + D for two different process
flowsheets: a conventional multiunit reactor/separator/recycle structure and a reactive distillation
column. Each system is optimized in terms of the total annual cost for a wide range of chemical
equilibrium constants, KEQ. In the conventional system, the design optimization variables include
the reactor temperature, reactor size, and recycle flow rate. In the reactive distillation system,
the design optimization variables include the pressure, number of reactive trays, and number
of total trays. The two systems are designed for identical feeds and identical products. Results
show that reactive distillation is significantly less expensive (by a factor of up to 3) than the
conventional process for all values of the chemical equilibrium constant.
recycle systems, they can have worse dynamics, and conventional multiunit reactor/separator/recycle struc-
their controllability can be more difficult. For a given ture and (2) a reactive distillation column. Both of the
feed flow rate, a reactive distillation column has the flowsheets are designed to achieve the steady-state
same control degrees of freedom as a single conventional economic objective of minimum total annual cost for a
column (four control valves give four control degrees of wide range of chemical equilibrium constants, KEQ.
freedom). A conventional multiunit flowsheet has four Systematic design procedures for both flowsheets are
control degrees of freedom for each column, plus the two developed as three-dimensional optimization problems
control valves on the reactor (coolant flow rate and with some heuristic rules and engineering assumptions.
reactor effluent). The effects of design parameters on the optimum design
One of the most fundamental differences between variables and total annual costs are studied for both
reactive distillation and a conventional flowsheet is the process designs. The two flowsheets have identical feeds
selection of operating temperatures. In the conventional and produce identical products.
system, the reactor temperature can be set at an The vapor-liquid equilibrium used in this paper is
optimum value, and the distillation temperatures can ideal with constant relative volatilities. The rationale
be independently set at their optimum values (by for considering ideal systems is to start with a simple
adjusting column pressures). In reactive distillation, system in which the complexities of nonideality do not
these temperatures are not independent. Therefore, cloud the issues and the results and some general
reactive distillation has fewer design degrees of freedom conclusions are possible. The occurrence of azeotropes
to be used to optimize the flowsheet. is very specific to the particular chemical system
The application of reactive distillation in the chemical considered. We wish to strip away such complexities so
and petroleum industries has increased rapidly in the that the fundamental differences between the two
past decade.2 Following the pioneering paper3 published flowsheets can be fairly compared. Future work will
by Eastman Chemical discussing the reactive distilla- explore the impact of temperature-dependent relative
tion of methyl acetate, many other papers and patents volatilities on the two processes. It is expected that there
have explored the use of this process for other chemical will be a significant effect on the comparison because
operations. A number of specific chemical systems have of the coupling of the distillation temperatures and the
been studied in the literature. Chapter 10 of the recent reaction temperatures in the reactive distillation pro-
textbook by Doherty and Malone4 presents an excellent cess. In the conventional flowsheet, the reactor temper-
summary of 61 different reactive distillation systems. ature and the column temperatures can be set indepen-
Another list of industrial applications of reactive distil- dently.
lation with 75 examples is given in the first chapter of
a recent book on reactive distillation.5 This work states 2. Process Studied
that the most common applications of reactive distilla- The basic process considered consists of a reversible
tions are etherification and esterification reactions. liquid-phase reaction
Although reactive distillation can be effectively used for
systems with a wide variety of chemical equilibrium A+BSC+D (1)
constants, the normal range of values for KEQ is between
1 and 50. The forward and backward specific reaction rates,
Because a reactive distillation column has more following the Arrhenius law, are given by
design variables than a conventional column, the design
of such systems is more difficult. The column pressure, kF ) aFe-EF/RT (2)
the catalyst holdup on each tray, and the positions of
the feed trays are important design considerations. In kR ) aRe-ER/RT (3)
addition to stripping and rectifying sections, there is one
more column section (the reactive zone). Also, the The rate law is based on concentrations in mole frac-
assumption of constant molar overflow is not valid if tions and liquid holdups in kilomoles. The forward
the reactions are not thermally neutral. Several papers reaction rate is specified as 0.008 kmol s-1 kmol-1 at
have optimized reactive distillation columns by mini- 366 K. Taking (KEQ)366 equal to 2 as a base case, the
mizing the objective function in terms of total annual reverse reaction rate at this temperature is varied by
cost (TAC).6-10 All of these papers proposed different selecting a range of (KEQ)366 values between 0.5 and 50.
optimization methods for reactive distillation systems
and illustrated their methods with examples of specific (kF)366
chemical systems. (kR)366 ) (4)
(KEQ)366
Only a few papers have appeared in the literature
that compare reactive distillation flowsheets with other Therefore, for different values of (KEQ)366, different
process flowsheets.11-13 Siirola11 claimed that both the values of the preexponential factor aR are calculated.
capital and energy costs of the reactive/extractive distil- Both reaction rates are temperature-dependent, and the
lation column design for methyl acetate production are reverse reaction rate is different for each case of (KEQ)366
5 times lower than those of the conventional design. The selected. Note that the ratio of kF to kR is not equal to
results in terms of TAC reported by Chiang et al.12 for (KEQ)366 at temperatures other than 366 K because the
the amyl acetate process indicated that reactive distil- activation energies are different. Rather, because the
lation is 4 times more efficient than the coupled reactor/ reaction is exothermic, the reverse reaction rate is more
separator system. However, Stitt13 claimed that reactive temperature-dependent than the forward reaction rate.
distillation for toluene disproportionation did not offer Ideal vapor-liquid equilibrium is assumed with con-
significant economic benefits. stant volatilities. The volatilities of the components are
The purpose of this paper is to quantitatively compare such that RC > RA > RB > RD. Hence, the reactants are
the designs of two different process flowsheets: (1) a intermediate boiling between the two products, which
Ind. Eng. Chem. Res., Vol. 43, No. 10, 2004 2495
tion columns, which means that neither energy balances search with the following optimization variables: (1) the
nor total balances are needed on the trays for steady- molar holdup in the reactor VR, (2) the composition of
state calculations. Other assumptions are constant reactant B in the reactor zB, and (3) the reactor
relative volatilities, isothermal reactor operation, theo- temperature TR.
retical trays, saturated liquid feed and reflux, and total Also, several different specifications for assumptions
condensers and partial reboilers in the columns. Ad- ii-v are used to investigate the effects of product
ditional assumptions and specifications are as follows: quality, conversion and recycle impurities on the eco-
(i) In all cases, the net production rates are set by nomically optimal steady-state design. However, these
fixing the fresh feed flow rates of pure components A variations do not affect the general structure of the
and B at F0A ) 12.60 mol/s and F0B ) 12.60 mol/s, design procedure.
respectively. A grid-search optimization strategy is used in this
(ii) The amount of reactant A lost in product stream work to find the optimum values of the three design
D1 is constant at Aloss ) 0.63 mol/s (xD1,A ) 0.05 mole optimization variables. Other methods, such as gradient
fraction). No B or D is going overhead, i.e., xD1,B ) xD1,D nonlinear programming techniques, could be used.
) 0. However, the grid method is more robust because a
(iii) The bottoms product stream B2 contains some numerical programming method can easily drive the
amount of component B as an impurity, Bloss ) 0.63 process into an infeasible region in which the specified
mol/s (xB2,B ) 0.05 mole fraction). No A or C is going purities and production rates cannot be achieved.
out the bottoms, i.e., xB2,A ) xB2,C ) 0. 3.2. Steady-State Design Procedure. The following
(iv) The impurity of product C in recycle stream D2 is steps in the design procedure are employed utilizing the
xD2,C ) 0.05. material balances and specifying necessary variables.
(v) The impurity of component D in recycle stream The production rates of components C and D (RC and
D2 is xD2,D ) 0.05. RD) are given by
(vi) As a result of the previous two assumptions and
the fact that C and D have identical stoichiometric RC ) RD ) VR(kFzAzB - kRzCzD) (5)
coefficients, the concentrations of components C and D
in the reactor, zC and zD, are equal for all cases (zC ) (i) Fix the value of reactor temperature TR at a small
zD) but vary in value from case to case. value.
(vii) The column pressures are set using the vapor (ii) Fix the value of reactor holdup VR at a small value.
pressures, PS, of pure components and the liquid (iii) Specify the value of reactor composition zB.
compositions in the reflux drum, xD,j, at 320 K (so that (iv) Using the assumption of equal compositions zC
cooling water can be used in the condenser). and zD in the reactor, calculate the concentration zA by
(viii) In each column, the reflux ratio, RR, is 1.2 times rearranging eq 5 to give
{
the minimum reflux ratio, RRmin, calculated via the
Underwood equations.
1 kF
(ix) In each column, the number of stages (or trays), zA ) 4 z + 2(1 - zB) (
2 kR B
]}
NT, is twice the minimum number of stages, NTmin,
x[ ] [
calculated via the Fenske equation. kF 2 RC
(x) Kirkbride’s method is used to find the optimal feed 4 zB + 2(1 - zB) - 4 (1 - zB)2 + 4 (6)
tray location, NF.
kR VRkR
Through these assumptions, nine of the design de-
grees of freedom are used as follows: one for the (v) Calculate the product concentrations zC and zD in
production rate, two for product impurities, two for the reactor as
recycle impurities, and four for total tray numbers and
optimum feed tray locations. This reduces the steady- 1 - zA - zB
zC ) z D ) (7)
state economic optimization to a three-dimensional 2
Ind. Eng. Chem. Res., Vol. 43, No. 10, 2004 2497
D1 ) RC + Aloss (8)
B2 ) RD + Bloss (11)
(ix) Use the total mass balance and component 4. Reactive Distillation Design
balances around the first column to calculate the
The reactive distillation alternative is shown in
bottoms flow rate and compositions for that column
Figure 2. The column is fed with two pure reactant fresh
B1 ) F - D1 (22) feed streams: F0A and F0B. The column has three zones.
There are NS trays and a partial reboiler in the stripping
Fzj - D1xD1,j section. Above this section, there is a reactive zone with
xB1,j ) (23) NRX reactive trays. The third section is the rectifying
B1 section with NR trays and a total condenser.
Light reactant A is fed to the bottom tray of the
(x) Now that the feed, bottoms, and distillate flow reactive zone, while heavy reactant B is introduced at
rates and compositions are known for both columns, use the top of the reactive section. Light product C leaves
the Fenske equation to calculate the minimum number in the distillate, while heavy product D is removed from
of trays. Set the actual number of trays equal to twice the bottoms. Because the light reactant goes up through
the minimum. Then use the Underwood equations to the column after being fed on the bottom tray of the
calculate the minimum reflux ratio. For sizing purposes, reactive zone, very little of component A is present in
set the actual reflux ratio equal to 1.2 times the the bottoms. Likewise, heavy reactant goes down through
minimum value. the column after being fed on the top tray of the reactive
2498 Ind. Eng. Chem. Res., Vol. 43, No. 10, 2004
are equal to LS. Analogously, all of the vapor rates, Vi, Table 3. Sizing and Economic Basis
beginning from the top feed tray throughout the rectify- parameter units value
ing section and total condenser are VNT, and all of the
reboiler
liquid rates, Li, are equal to LR. heat-transfer coefficient, UR kJ/(s‚K‚m2) 0.568
The vapor-liquid equilibrium is assumed to be ideal. temperature difference, ∆TR K 34.8
Column pressure P is optimized for each case. With condenser
given pressure P and tray liquid composition xi,j, the heat-transfer coefficient, UC kJ/(s.K.m2) 0.852
temperature Ti, and the vapor composition yi,j can be temperature difference, ∆TC K 13.9
calculated. This is a bubble-point calculation and can energy cost $/106 kJ 4.7
payback period, βpay year 3.0
be solved by a Newton-Raphson iterative convergence
method. can be easily observed from the continuously increasing
NC vapor boilup, and the simulation can be restarted in
P) ∑
j)1
S
xi,jPi,j(T) (32)
such cases with new optimization variables.
rectifying trays over a range, and repeat steps iii-xvi. (iv) The column height is calculated assuming a
(xviii) Finally, vary the value of the column pressure 0.61-m (2-ft) tray spacing and allowing 20% more height
over a wide range, and repeat steps ii-xvii for each for base-level volume.
pressure value. Select the design with the minimum
TAC as the economically optimum steady-state design LC ) 0.731 52NT (43)
for the given (KEQ)366 value.
Depending on the values of optimization variables (v) The heat-transfer areas of the reboiler and con-
chosen, the column can leave the feasible region. This denser are calculated using the steady-state vapor flow
2500 Ind. Eng. Chem. Res., Vol. 43, No. 10, 2004
Figure 5. Effects of optimization variables VR and zB for the base case of (KEQ)366 ) 2.00 with the constant temperature TR ) 367 K.
Figure 6. Effects of optimization variables VR and zB on design Figure 7. Effects of optimization variable TR for the base-case
costs for the base case of (KEQ)366 ) 2.00 with the constant (KEQ)366 ) 2.00 design.
temperature TR ) 367 K.
sures are set using the vapor pressures, PS, of the pure
For each kinetic case, the shapes of the curves are components at 320 K and the liquid compositions in the
similar, but they have minima at different tempera- reflux drum, xD,j. In a conventional multiunit process,
tures. For both low and high temperatures, the increase the column temperatures can be set independently so
of these flow rates occurs because of the higher reactant as to optimize column efficiency. The reactor tempera-
compositions in the reactor effluent, as shown in Figure ture can then be independently adjusted to its optimum.
7 for the base case, (KEQ)366 ) 2. As mentioned before, This is not the case in reactive distillation because both
higher temperatures give smaller reactor holdups VR reaction and separation are occurring in the same
and lower equilibrium constants KEQ, which give higher vessel, operating at a single pressure.
reactant concentrations. However, low temperatures
give small specific reaction rates, meaning that more The effect of the conversion design value is explored
reactant leaves the reactor. Therefore, higher vapor by increasing the conversion from 95 to 99%. Increasing
boilup is required in the columns to obtain the product the conversion means that the product purities increase
streams with specified purities. (less unreacted A and B are lost in the product streams)
Optimization results for the conventional design for and the optimum values of all of the design parameters
all cases are summarized in Table 4. These are the change. Figure 10 gives results for all of the (KEQ)366
optimum designs in terms of the reactor temperature, cases. The vapor boilup required to achieve 99% conver-
the reactor holdup, and the composition of component sion increases, as does the flow rate of the recycle
B in the reactor. Optimal operating conditions for the stream, D2. The reactor holdups are almost the same
same cases are reported in Table 5. The column pres- for both conversions. These results illustrate that, as
2502 Ind. Eng. Chem. Res., Vol. 43, No. 10, 2004
Figure 8. Effects of optimization variable TR on design costs for the base-case (KEQ)366 ) 2.00 design.
Figure 13. Effect of number of reactive stages for the base case Figure 15. Economically optimum results for three different
of (KEQ)366 ) 2 with constant number of separation stages NR ) kinetic cases.
NS ) 5.
optimum pressure is more important for small values
of (KEQ)366 than for large values. This figure shows that
the optimum pressures for (KEQ)366 ) 1, 2, and 5 are
6.5, 8.0, and 9.5 bar, respectively.
Figure 15 gives the economic optimum steady-state
design results for three different values of (KEQ)366. For
each case, the optimum pressure that minimizes the
vapor boilup is used, and the number of reactive trays
is varied over a range to find the minimum total annual
cost TAC. The upper graph shows that there is a
minimum in the TAC curve at some optimum number
of reactive stages. The optimum number of reactive
stages decreases with increasing values of (KEQ)366. The
column costs and total annual cost decrease as the value
of (KEQ)366 increases.
Results for five different kinetic cases are reported
in Table 6. The table includes the optimum design
parameters and costs in terms of the following optimi-
zation variables: (i) number of separation trays, (ii)
Figure 14. Effects of pressure on the vapor boilup for three
different kinetic cases.
number of reactive trays, and (iii) the column pressure.
Also, the temperatures at the top and bottom of different
from leaving in the bottoms. Thus, there is an optimum sections are included in this table.
number of reactive trays for each pressure that mini- Figure 16 shows the composition profile of the opti-
mizes the vapor boilup, energy cost, and heat exchanger mum design for the base case of (KEQ)366 ) 2. The
cost. highest composition of reactant A is at the bottom of
The left graph of Figure 13 shows that there is a reactive zone where A is fed. The other reactant, B, has
minimum in the total annual cost curve at a certain its highest composition at the top of reactive zone, which
number of reactive trays because of the tradeoff between is also its feed tray. While the composition of A
increasing column cost and decreasing vapor boilup. For decreases up through the reactive zone, the composition
this case, nine reactive trays are the optimum for all of product C increases. The reverse occurs for reactant
pressures. This figure also indicates that the optimum B and product D through the reactive zone. The rest of
pressure is 8 bar for this specific case. the column operates as a separation unit. Thus, the
Figure 14 shows the effect of operation pressure on composition of heavy product D increases down through
vapor boilup. These results are for the three different the stripping section, and the composition of light
kinetic cases (KEQ)366 ) 1, 2, and 5 with the number of product C increases up through the rectifying section.
reactive trays that is optimum for each case. The graph Figure 17 shows the temperature profiles of the
shows that there is an optimum pressure for all three optimum designs for the three different kinetic cases
cases that minimizes the vapor boilup. The higher (KEQ)366 ) 1, 2, and 5. The temperature profiles of the
values of (KEQ)366 require lower vapor boilups and three cases are similar, with higher temperatures for
operate at higher optimum column pressures. The higher values of (KEQ)366. Fairly significant temperature
higher equilibrium constant pushes the reaction to the breaks occur around tray 4 for all cases, and these tray
right. The equilibrium constant decreases with increas- temperatures can be used in control schemes to infer
ing temperature for an exothermic reaction. Therefore, bottoms purity. A similar break occurs at different tray
the column can operate at higher pressures when numbers near the top of the column for each case. These
(KEQ)366 is larger. It is possible to conclude from this temperatures could be used to infer distillate purity. For
figure that the sensitivity to pressure increases as the all kinetic cases, the temperatures show little change
value of (KEQ)366 decreases. Therefore, operating at the in the reactive zone.
Ind. Eng. Chem. Res., Vol. 43, No. 10, 2004 2505
Figure 16. Composition profile of the optimum design for the Figure 18. Effect of conversion for the base case of (KEQ)366 )
base case of (KEQ)366 ) 2.0. 2.0.