Академический Документы
Профессиональный Документы
Культура Документы
Salvador Hernandez-Castro
Facultad de Ciencias Quimicas, Universidad de Guanajuato, Col. Noria Alta S/N, Guanajuato, Gto.,
CP 36050, Mexico
This work proposes a simplified methodology for the analysis and design of reactive batch
distillation columns based on the McCabe-Thiele method for reactive continuous columns and
on the concept of a reactive difference point. To extend the application of the concept of a reactive
difference point for reactive batch distillation columns, expressions for the McCabe-Thiele
operating line and for the dynamics of the reboiler in the reactive case have been derived; two
cases are considered depending upon the phase in which reactions occur (i.e., liquid or vapor).
We also provide an approach to the derivation of an expression to calculate the molar turnover
flow rate by considering each reactive distillation plate as a combined system of a conventional
equilibrium stage linked to a chemical reactor (phenomena decomposition). Furthermore, we
present our derivation of the Underwood equations (minimum reflux) for continuous and batch
reactive distillation columns. Finally, five illustrative examples allow us to show that our results
present reasonable agreement with those obtained through BatchFrac of AspenPlus.
ξn
xn ) xD - (2)
D
{ [ ] }
liquid flow rate, and x is the liquid composition. In
dxi,B 1 V νi principle, the calculation of υ would require the use of
) (xi,B - xi,D) + - νTxi,B ξk,T (14)
dt B 1+R -νk column hydraulic analysis, which could provide a rela-
tionship between Lj-1 and υ.
Observe again that, when compared to the nonreactive The reactor outlet liquid stream, L′j, can be ex-
case for batch distillation, eqs 12 and 14 include an extra pressed in terms of the molar conversion fraction of the
term to account for changes due to reaction. Further, reactant, Cj:
notice that in eq 14 the separate effects of reaction and
separation are well defined. L′jx′k,j ) Lj-1xk,j-1(1 - Cj) (16)
3.1.2. Extent of the Reaction: Molar Turnover
Flow Rate. Hauan et al.7,11 provided a generalization so that eq 15 becomes
of the concept of a difference point for reactive cascades.
The authors assume that a reactive separation cascade υ(rk,j) ) Lj-1xk,j-1Cj (17)
4004 Ind. Eng. Chem. Res., Vol. 43, No. 14, 2004
V1 R V1 1 νi ξk,j
yi,j+1 ) xi,j + xi,D -
Vj+1 R + 1 Vj+1 R + 1 -νk Vj+1
(22)
where changes in the vapor flow rate due to reaction
are given by
dB V1
) νTξk,T -
dt 1+R
( )
L′jx′k,j equilibrium stage. The consideration of having a sepa-
νk
rk,j ) kj (18) ration stage connected to a CSTR in the case of reaction
υ0
in the vapor phase could seem inappropriate; however,
more than the need of a CSTR, our approach works by
where the term inside the parentheses represents the assuming that the compositions of reactive and nonre-
molar concentration of the reactant in the reactor and active streams are homogeneous as if the mixing on
kj is the kinetics parameter given by the Arrhenius them was perfect. A derivation similar to that of section
equation. υ0 is the volumetric flow rate entering the 3.1.2 results in the following expressions:
stage. υ0 can be calculated from
( Cj
)
n
υ0 ) Lj-1/F (19) ξk,j ) ∑
j)1
V′jy′k,j
1-C
(24)
j
where F is the density of the liquid stream. Knowing υ0
would also allow the evaluation of the residence time, where the conversion fraction of the reactant can be
τj ) υ/υ0; hence, if the liquid holdup is known or can be obtained from the implicit eq 25:
[ ]
evaluated, the use of υ0 or τj is equivalent.
When eqs 16-18 are combined, an implicit equation V′jy′k,j νk
for the molar conversion of the reactant can be obtained: V′jy′k,jCj ) υkj (1 - Cj) (25)
υ0
[
Lj-1xk,j-1(1 - Cj)
]
νk
3.3. Solution Approach. The resulting model is a
Lj-1xk,j-1Cj ) υkj (20) well-posed differential-algebraic system of equations
υ0
(DAE). Table 1 presents a summary of the model
Once eq 20 is solved to obtain Cj (analytically or specifications and the equations used for the calcula-
numerically for νk greater than 2), one can finally tions of the variables involved in each stage. The
calculate the molar turnover flow rate through the solution procedure for each time step of a reactive batch
expression distillation column is similar to the procedure for solving
the rectifying section of a continuous reactive column.
n The equations involved in the calculation could be solved
ξk,j ) ∑
j)1
(Lj-1xk,j-1Cj) (21) either simultaneously with an appropriate numerical
method or sequentially by using a method such as the
direct substitution method. We describe next the se-
3.2. Reactive Batch Distillation Columns: Reac- quential procedure for the case in which reaction occurs
tion in the Vapor Phase. If the reaction takes place in the liquid phase (constant molar flow in the vapor
in the vapor phase, derivations similar to the ones phase) in all of the stages but the reboiler; an iterative
presented in the previous subsection can be obtained. procedure for the calculation of the distillate composi-
Hence, the component and overall mass balances for tions has to be completed.
section 1 of the column presented in Figure 6 result in Considering a constant reflux policy, we assume that
the operating line for reactive batch distillation columns the number of stages (N), the reflux ratio (R), and the
(quasi-steady-state approximation): initial conditions (feed load and still compositions) are
Ind. Eng. Chem. Res., Vol. 43, No. 14, 2004 4005
given. Also, a typical specification in this case is the compositions are equal to the known values, then we
vapor boilup flow rate (V), which will remain constant proceed to the next time step; otherwise, we modify our
through the column. As one of the main limitations of guess for the distillate compositions and repeat the cal-
our approach, assumed values of the liquid holdup, υ, culations in the column. Notice that, with respect to the
and entering volumetric flow rate, υ0 (or residence time, number of degrees of freedom, the model is square be-
τ), are used in this work (see section 5 for an explana- cause the distillate compositions can be indirectly found
tion). from the independent equations used to calculate the
Hence, we start by assuming the values of the dis- still compositions (stage-by-stage procedure). Before one
tillate compositions. Then, by using a bubble-point cal- moves on to the next time step, integration of eqs 12-
culation, the temperature of the distillate can be com- 14 is needed to find the values of the amount of mixture
puted. When the temperatures of the reboiler (which in the still, the distillate collected, and the bottom com-
can also be computed through a bubble-point calculation positions for the following time step. N, R, and V remain
because the compositions in the reboiler are known) and as constants so that the simulation continues for the
the condenser are known, a linear temperature profile next time steps and ends when a specified stopping
is assumed. The temperature of each stage is required criterion is met.
because the computations of the kinetics parameter and
the relative volatility of the mixture depend on it. Then,
a mass balance in the condenser allows the calculation 4. Limiting Conditions for Reactive Distillation
of L0; for the condenser ξk,0 ) 0 (no reaction in the Columns
condenser) and eq 5 reduces to V ) L0 + D, which is
In this section we derive the expressions to calculate
equivalent to L0 ) V[R/(R + 1)]. Also notice that the
the minimum reflux (Underwood equations) for continu-
distillate compositions are equal to the compositions of
ous and batch reactive distillation columns; further-
the vapor flow entering the condenser (yk,1).With that,
more, a discussion regarding the Fenske equation (mini-
we proceed to perform the calculation on stage 1.
mum number of stages) for continuous reactive distil-
Equations 20 and 21 are used for the calculation of C1
lation columns is also provided.
and the molar turnover flow rate, ξk,1, respectively. The
liquid molar flow rate on stage 1, L1, is obtained from 4.1. Underwood Equations. To derive the expres-
eq 6. Equation 16 and the summation equation for the sion for the minimum reflux in the case of continuous
compositions are used to find the values of the liquid reactive distillation columns, a system such as the one
flow (L′1) and the compositions (x′k,l) entering the equi- shown in Figure 8 is assumed. Our approach is based
librium separation stage of Figure 5. Then, the equi- on the derivation for the nonreactive case presented by
librium curve (equilibrium in terms of relative volatility) King.14 The section with an infinite number of stages
and the operating line (eq 8) allow the calculation of is placed in the rectifying section above the feed plate.
the rest of the compositions for stage 1 (xk,1 and yk,2). A further assumption in our derivation is that there is
The procedure continues for the rest of the stages no reaction in the section with an infinite number of
moving down the column until we evaluate the composi- plates (to avoid depletion of the reactant).
tions in the reboiler. Those values provide the conver- The mass balance equation for component i around
gence criteria: If the computed values of the reboiler section 1 of Figure 8 is
[ ]
4006 Ind. Eng. Chem. Res., Vol. 43, No. 14, 2004
V-V
h )
NC
∑
(
Ri,∞ Dxi,D -
-νk
νi
ξk,T + Bxi,B
) (34)
where the index ∞ has been assigned to the streams i)1 Ri,∞ - φ
coming from the section with an infinite number of
plates. The overall mass balance for the condenser is Finally, by introducing the definition of q
V ) L0 + D (27) V-V
h ) F(1 - q)
It is also considered that the liquid-vapor equilibrium and the component mass balance equation for the whole
on the last stage of section 1 is given by column
[ ]
Ri,∞zi,F
NC
νi ξk,T
V -νk D
xi,D - 1-q) ∑
i)1 R -φ
(35)
yi,∞ ) (29) i,∞
D 1 L∞
1- which is the second of the Underwood equations. Equa-
Ki,∞ V
tion 35 can be used for calculating φ and eq 32 for
calculating Rmin for continuous reactive distillation
When the definition of relative volatility (with respect columns.
to the key component r) is incorporated: 4.1.1. Extension of the Underwood Equations to
( )
Reactive Batch Distillation Columns. A direct ex-
νi ξk,T tension to the Underwood equations derived above can
Ri,∞ xi,D -
V -νk D be done for the case of reactive batch distillation
yi,∞ ) (30) columns. An approach similar to the one proposed by
D 1 L∞ Diwekar13 serves that purpose (Figure 9).
Ri,∞ -
Kr,∞ V In Figure 9, a reactive batch distillation column is
represented as a continuous column where the feed
Here a new variable, φ ) (1/Kr,∞)(L∞/V), called the enters the column at the bottom stage with a temper-
absorption factor, is introduced. Also, because the ature equal to its bubble point and with a composition
summation of the compositions in the vapor stream equal to xi,B. Such a representation allows rewriting of
going into the section with an infinite number of stages eqs 32 and 35 as
[ ]
has to be 1, eq 30 reduces to eq 31. where NC is the
[ ]
NC Ri,∞xi,B
( )
νi ξk,T ∑
i)1 R -φ
)0 (36)
[ ]
Ri,∞ xi,D - i,∞
( )
V NC -νk D
D
) ∑
i)1 Ri,∞ - φ
(31)
Ri,∞ xi,D -
νi ξk,T
NC -νk D
number of components. Finally, the substitution of eq
Rmin + 1 ) ∑
i)1 Ri,∞ - φ
(37)
27 and the definition of the reflux ratio in eq 31 result
in
[ ]
which correspond to the Underwood equations for
( )
reactive batch distillation columns. Observe that the
νi ξk,T minimum reflux computation not only depends on the
Ri,∞ xi,D - distillate and bottom compositions, as it does for a
NC -νk D nonreactive column, but also depends on the extent of
Rmin + 1 ∑
i)1 Ri,∞ - φ
(32) the reaction. Therefore, a value of the extent of the
reaction has to be given in order to make eq 37 useful
in practical terms. It might be difficult to provide an
which is the first of the Underwood equations for approximation to the extent of the reaction a priori, and
calculating the minimum reflux. Following a similar a parametric search might be necessary to obtain a
procedure for section 2 (nonreactive stripping section) feasible value. For a specific reaction-separation sys-
of Figure 8, one can obtain tem, a reasonable value of the extent of the reaction can
[ ]
be estimated, for instance, by assuming that the molar
h
V NC R′i,∞xi,B turnover flow rate is equally distributed among the
-
B
) ∑
i)1 R′ - φ′
(33) reactive stages of the column and by defining the
i,∞ amount of product obtained through the reaction (using
a value of the molar conversion fraction based on the
Combining eqs 31 and 33 and taking into account that kinetics of the reaction involved and the amounts of the
φ ) φ′ and Ri,∞ ) R′i,∞ result in reactants initially fed to the column).
Ind. Eng. Chem. Res., Vol. 43, No. 14, 2004 4007
∫0tDxi,D dt
xji,D(t) ) (38)
∫0tD dt
When possible, the results were compared to those
Figure 9. Reactive batch column as a continuous column. obtained through the use of BatchFrac from AspenPlus.
The Antoine parameters (for the vapor-liquid equilib-
4.2. Comments on the Fenske Equation for rium) were taken from Reid et al.18 To keep the design
Reactive Distillation. The derivation of the Fenske method simple enough, we did not include column
equation for nonreactive continuous columns assumes hydraulic equations in our approach; that means that
operation at total reflux. Interesting conclusions can be values for the liquid holdup have to be assumed.
drawn with respect to the corresponding derivation in Furthermore, the kinetic parameters (frequency factor
the reactive case. Let us assume that an irreversible and activation energy) were provided and adjusted to
isomerization reaction (binary mixture) is taking place achieve an accumulated value of the extent of the
in the column and that the product of the reaction is reaction equal to the one obtained through AspenPlus
the most volatile component. simulations. Note that our goal in this work was not to
perform realistic calculations but to show that our
The case of total reflux is then the first one that we simulation results present reasonable agreement with
analyzed for the reactive case, as shown in Figure 10a. those of the state of the art process simulators; as a
A problem arises, however, in such a scheme. Because matter of fact, the assumed values for the liquid holdup
the reaction takes place regardless of whether products and the kinetic parameters were the same as the ones
are being obtained or not, as time and the reaction used in BatchFrac. Finally, we also assumed a constant
progress, the reactant will get depleted and the column value of the volumetric flow rate υ0; however, such a
behaves just like a chemical reactor. Hence, the mini- simplification can easily be removed by evaluating the
mum number of stages in a reactive distillation column liquid density on each stage as a function of temperature
does not necessarily correspond to total reflux operation. and composition and then using eq 19.
An outlet product stream, such as the distillate stream 5.1. Reaction Causes No Change in the Number
in Figure 10b, makes the problem worse because the of Moles. For a scenario in which the assumption of
light product will be continuously removed and the constant molar flow holds for both the liquid and vapor
column will become empty if no inlet stream is incor- phases, two cases were solved. In the first one, the
porated; a feed to the column is therefore required. reaction takes place in the liquid phase; in the second
Figure 10c shows a column similar to the scheme example, the reaction is assumed to take place only in
considered for reactive batch distillation columns in the the vapor phase.
derivation of the Underwood equations because the feed 5.1.1. Reaction in the Liquid Phase. The example
enters the column just above the reboiler. We believe corresponds to an isomerization reaction in the liquid
that such a configuration could be used for obtaining phase: o-xylene (C8H10) f ethylbenzene (C8H10). The
the Fenske equation in reactive distillation. However, lightest component is ethylbenzene. The data used for
this example are shown in Table 2.
our experience of using the methodologies for the
The McCabe-Thiele diagram for the initial time step
derivation of the Fenske equation reported by King,14
of this example is shown in Figure 11. As expected, the
Henley and Seader,15 and Hohmann16 is that the
contribution of the reaction is greater for stages closer
appearance of the molar turnover flow rate and the
to the reboiler (larger reactant concentrations). Figure
reflux ratio makes the equation more complicated as we 11 does not show the last of the operating lines because
move down the column. Also, most of the reasonable the difference point is negative. Figure 12 presents three
assumptions used to simplify it introduce significant instances of the calculation of the average distillate
errors in the numerical calculation. Moreover, the composition by changing the value of the reflux ratio.
summations involved do not seem to correspond to a Because the reaction shows high conversion values, the
converging series of any kind. This topic should be the effect of the reflux ratio is not significant. Observe also
subject of further investigation. Sundmacher and Kien- that, because of reaction, the average distillate composi-
le17 provide a simple expression for calculating the tion increases with time; this result is contrary to the
minimum number of stages for the case of a reaction behavior observed in nonreactive batch distillation
taking place in the reboiler of the column. columns. This effect can be explained by looking at the
4008 Ind. Eng. Chem. Res., Vol. 43, No. 14, 2004
Figure 14. McCabe-Thiele diagram for example 2. Figure 16. McCabe-Thiele diagram for example 3.
Figure 18. Comparison in example 3 for the composition of Figure 19. Average distillate composition for a reflux ratio equal
isobutyraldehyde. to 2 in example 4.
Table 5. Data for the Reaction-Separation of the
Multicomponent Mixture of Example 4 continuous reactive distillation columns was tested for
the case of the isomerization reaction isobutane f
parameter value n-butane. The system considered is a column with nine
no. of stages 2 + reboiler stages where the feed enters the column as a saturated
feed 100 kmol liquid in stage number 8. The column pressure is 1 atm.
feed composition of benzene 0.006 The feed is 1000 kmol h-1 with a composition of 0.04 of
feed composition of o-xylene 0.485
vapor boilup flow rate 140 kmol h-1 n-butane; the distillate composition of n-butane is 0.98.
column pressure 1 atm It is assumed that 380 kmol h-1 of n-butane are
batch time 1h produced in the column and that such a molar turnover
order of reaction 1.0 flow rate is equally distributed in the stages of the
activation energy 30 000 kJ kmol-1 column. Solving eqs 32 and 35 results in Rmin equal to
frequency factor 8 × 102 h-1 0.518. After the solution approach using the McCabe-
volumetric flow rate 2 m3 h-1
liquid holdup 4 × 10-4 m3 Thiele method is used for a range of values of the reflux
ratio, a value of 0.51 was found for the minimum value
consisting of the simultaneous calculation of NC - 1 of the reflux ratio that allows the reaction and separa-
components is needed. A further assumption is that the tion specified. Such a result indicates that the Under-
rate of reaction can be expressed in terms of the wood equations derived in this work present a reason-
concentration of just one of the reactants (if more than able agreement with the simulations.
one). Hence, the molar turnover flow rate for the rest
of the components in terms of the base component is 6. Concluding Remarks
given by a stoichiometric relationship
This paper provides the mathematical derivations
νi needed for the simplified analysis and design of reactive
ξi,j ) ξ (39) batch distillation columns. In our opinion, the main
νk k,j contributions of the paper are as follows: (1) The
development of the theoretical and mathematical tools
Also, equilibrium can be calculated in terms of the that allow the use of a McCabe-Thiele strategy for
equilibrium constant instead of relative volatilities as reactive batch distillation columns. This includes the
follows. derivation for the dynamics of the reboiler and the
operating line under the assumption of quasi steady
xi,j ) Ki,jyi,j (40) state in the plates. Furthermore, an approach to the
computation of the molar turnover flow rate of reactive
An example for the reaction-separation of multicom- distillation stages has been proposed. Such an approach
ponent mixtures is illustrated with the system involving is based on phenomena decomposition and is needed in
the reaction (no change in the number of moles) in the order to apply the concept of the reactive difference point
liquid phase: 2-toluene (C7H8) f benzene (C6H6) + to the case of batch distillation. (2) The derivation of
o-xylene (C8H10), where the order in relative volatility the Underwood equations for continuous and batch
is benzene > toluene > o-xylene. The data used for this reactive distillation columns.
case are shown in Table 5. The many assumptions considered in this work have
The McCabe-Thiele solution approach was used for been extensively explained, and we expect that our
benzene and toluene. Figure 19 shows the average numerical examples involving various reactive systems
distillate composition of the three components when the show the scope and confirm the usefulness of our
reflux ratio is equal to 2. This is an interesting case approach. Through our analysis and numerical ex-
because reaction presents a higher contribution than amples we have (1) shown that our results present a
separation for benzene and o-xylene, so that their reasonable agreement with those of the state of the art
distillate compositions increase with time. However, for process simulators for reactive batch distillation col-
the case of toluene, the trend is different and the umns, (2) emphasized the advantages of the combina-
composition decreases with time. tion of separation and reaction for batch distillation,
5.4. Minimum Reflux Calculation. The calculation where the product purity can even be improved as time
of the minimum reflux ratio by using the Underwood progresses, and finally (3) proved that the results
equations derived in this work (eqs 32 and 35) for obtained for Rmin by the derived Underwood equations
Ind. Eng. Chem. Res., Vol. 43, No. 14, 2004 4011
for reactive columns are approximated enough to the (8) Lee, J. W.; Hauan, S.; Lien, K. M.; Westerberg, A. W. A
results obtained through parametric search. Graphical Method for Reaction Design Reactive Distillation
Columns. 1. The Ponchon-Savarit Method. Proc. R. Soc. London
2000, 456, 1953.
Acknowledgment (9) Lee, J. W.; Hauan, S.; Lien, K. M.; Westerberg, A. W. A
Graphical Method for Reaction Design Reactive Distillation
M.A.H.-G. and V.R.-R. are thankful for the financial Columns. 2. The McCabe-Thiele Method. Proc. R. Soc. London
support provided by the Consejo del Sistema Nacional 2000, 456, 1965.
de Educacion Tecnologica (CoSNET) and by the Consejo (10) Lee, J. W.; Hauan, S.; Westerberg, A. W. Circumventing
Nacional de Ciencia y Tecnologia (CONACYT). V.R.-R. an Azeotrope in Reactive Distillation. Ind. Eng. Chem. Res. 2000,
and S.H.-C. also thank the Consejo de Ciencia y Tec- 39, 1061-1063.
nologia del Estado de Guanajuato (CONCyTEG). (11) Hauan, S.; Ciric, A. R.; Westerberg, A. W.; Lien, K. M.
Difference Points in Extractive and Reactive Cascades. I. Basic
Properties and Analysis. Chem. Eng. Sci. 2000, 55, 3145.
Literature Cited (12) Lee, J. W.; Hauan, S.; Lien, K. M.; Westerberg, A. W.
(1) Lee, J. W.; Westerberg, A W. Visualization of Stage Calcula- Difference Points in Extractive and Reactive Cascades. II. Gen-
tions in Ternary Reacting Mixtures. Comput. Chem. Eng. 2000, erating Designing Alternatives by the Lever Rule for Reactive
24, 639. Systems. Chem. Eng. Sci. 2000, 55, 3161.
(2) Lee, J. W.; Hauan, S.; Westerberg A. W. Graphical Methods (13) Diwekar, U. M. Batch Distillation; Taylor and Francis:
for Reaction Distribution in a Reactive Distillation Column. AIChE Washington, DC, 1995.
J. 2000, 46, 1218. (14) King, C. J. Separation Processes; McGraw-Hill: New York,
(3) Barbosa, D.; Doherty, M. F. A New Set of Composition 1980.
Variables for the Representation of Reactive Phase Diagrams. (15) Henley, E. J.; Seader, J. D. Equilibrium Stage Separation
Proc. R. Soc. London 1987, 413, 459. Operations in Chemical Engineering; Wiley: New York, 1981.
(4) Barbosa, D.; Doherty, M. F. Design and Minimum-Reflux (16) Hohmann, E. C. Analytical McCabe-Thiele Method (Smok-
Calculations for Single-Feed Multicomponent Reactive Distillation er’s Equation). AIChE J. 1980, B1.8, 42.
Columns. Chem. Eng. Sci. 1988, 43, 1523. (17) Sundmacher, K.; Kienle, A. Reactive Distillation; Wiley-
(5) Espinosa, J.; Scenna, N.; Perez, G. Graphical Procedure for VCH: Weinheim, Germany, 2003.
Reactive Distillation Systems. Chem. Eng. Commun. 1993, 119, (18) Reid, R. C.; Prausnitz, J. M.; Poling, B. E. The Properties
109. of Gases and Liquids; McGraw-Hill: New York, 1987.
(6) Perez-Cisneros, E. S. Modelling, Designing and Analysis of
Reactive Separation Processes. Ph.D. Thesis, Technical University Received for review August 11, 2003
of Denmark, Lyngby, Denmark, 1997. Revised manuscript received May 5, 2004
(7) Hauan, S.; Westerberg, A. W.; Lien, K. M. Phenomena Accepted May 5, 2004
Based Analysis of Fixed Points in Reactive Separation Systems.
Chem. Eng. Sci. 1999, 55, 1053. IE030658W