Steady-State Reactive Distillation Simulation

Using the Naphtali-Sandholm Method

Ram Singh, M. V. Pavan Kumar and Nitin Kaistha*
Department of Chemical Engineering, Indian Institute of Technology, Kanpur 208016 India

The steady-state simulation of reactive distillation columns using the conventional Naphtali-Sandholm (NS) method is studied. Results for methyl
acetate and MTBE columns show that the method effectively solves the highly non-linear MEH equations. In the MTBE system, three distinct
steady states are obtained as the Newton step factor is decreased from high to low values. The versatility of the method in handling different
column specifications and its application for preliminary control studies is also demonstrated. A novel feature of the work is the inclusion of
reaction conversion as a possible column specification. The specification of reaction conversion and product purity provides a direct means for
obtaining operating conditions for high purity and high conversion column operation.

On a étudié la simulation en régime permanent des colonnes de distillation réactive à l’aide de la méthode de Naphtali-Sandholm (NS) classique.
Les résultats pour les colonnes d’acétate de méthyle et de MTBE montrent que la méthode résout efficacement les équations MEH hautement non
linéaires. Dans le système MTBE, trois régimes permanents distincts sont obtenus alors que le facteur d’étapes de Newton diminue, passant de valeurs
élevées à des valeurs faibles. On démontre également la polyvalence de cette méthode pour traiter des spécifications de colonnes différentes ainsi
que son application pour des études de contrôle préliminaires. Une nouvelle caractéristique du présent travail réside dans l’inclusion de la conversion
de la réaction comme une spécification possible de colonne. La spécification de la conversion de la réaction et la pureté des produits fournissent un
moyen direct pour obtenir les conditions opératoires pour un fonctionnement de colonne avec une pureté et une conversion élevées.

Keywords: reactive distillation, Naphtali-Sandholm, rating simulation

INTRODUCTION Our recent interest in the control of reactive distillation

columns led us to explore the available techniques for solving

I
n the last two decades, reactive distillation (RD), which
the governing steady-state equations. A key solver requirement
combines reaction and separation in a single column, has
was flexibility in the specifications used to satisfy the column
emerged as a promising technology for process intensifica-
degree of freedom as each specification corresponds to specific
tion. Industrial applications of the technology include esterifica-
control structures (Luyben, 1996). Of the popular commercial
tion systems such as methyl acetate (Agreda and Partin, 1984)
simulators, the RADFRAC module in Aspen Plus was found to be
and etherification systems such as MTBE (Smith, 1994), ETBE
extremely robust in converging to a solution. However, flexibility
(Thiel et al., 1997) and TAME (Hickey and Adams, 1994). For
in the set of variables that can be specified was limited. On the
these, the RD processes have replaced conventional “reaction
other hand, it was extremely difficult to converge an RD simula-
followed by separation” processes at a fraction of the capital and
tion using the solver in HYSYS. Other authors have also reported
operating costs (Siirola, 1995).
a similar experience using HYSYS for solving RD problems
The steady-state performance simulation of an RD column is an
(Luyben, 2002).
integral part of all key steps in the successful implementation of
The quest for a flexible, yet robust RD solver reveals a vast
the technology. For example, steady-state simulation is needed in
body of literature that recommends the use of computationally
design studies where the proposed column design is validated for
intensive tools such as homotopy continuation and relaxation
the reaction conversion and product purity achieved; in operabil-
methods. Chang and Seader (1988) use homotopy continuation
ity studies where the steady-state column response to changed
operating conditions and disturbances is evaluated; in control
studies where the steady-state disturbance rejection using differ- * Author to whom correspondence may be addressed.
ent column control strategies (structures) is assessed, etc. E-mail address: nkaistha@iitk.ac.in

VOLUME 85, FEBRUARY 2007 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 75

to solve an ethyl-acetate RD column. Singh et al. (2005) apply account for reaction on a reactive tray. Top-down tray numbering
continuation with respect to catalyst mass (or reactive tray hold- is assumed. The feed stream conditions (flow, temperature and
up) to obtain converged solutions for MTBE and methyl acetate composition), tray energy streams and side draw flows (if any)
RD columns. Al-Arfaj and Luyben (2000) apply a relaxation are assumed to be specified. Neglecting reaction in the vapour
technique to an ideal, a methyl acetate and an ETBE column, phase and assuming that the vapour is in equilibrium with the
where the tray-by-tray dynamic material and energy balances liquid, the mole balances, equilibrium relations and heat balance
are integrated until steady state. Early literature also alludes to must be satisfied for all the trays, reboiler and partial condenser.
the use of relaxation methods for RD systems (Komatsu, 1977). For a C component mixture with N trays in the column section,
The need for such sophistication is attributed to the high non- the governing MEH equations for the jth tray are thus:
linearity due to the combination of reaction and separation in a
single column. The non-linearity can also result in multiple Mole balances
steady states, now accepted as routine phenomena in RD systems NRX
(Hauan et al., 1995; Jacobs and Krishna, 1993; Mohl et al., M i , j = li , j(1 + s j ) + vi , j(1 + S j ) - li , j -1 - vi, j +1 - fi, j - ∑ υi,nrj,n = 0 (1)
n =1
1999). Even though these techniques eventually converge to the
desired steady state, the computational burden is substantial and Equilibrium relations
can be prohibitive. C C
It is obvious that established solution methods for ordinary
distillation columns such as the Naphtali-Sandholm method
η j K i , j li , j ∑ vk , j (1 - η j )vi , j +1 ∑ vk , j
Ei , j = k =1 - vi , j + k =1 =0 (2)
(Naphtali and Sandholm, 1971) and the inside-outside method C C
(Boston and Sullivan Jr., 1974), can be easily adapted to account ∑ lk, j ∑ vk, j +1
for reaction on a tray and used for solving steady-state RD k =1 k =1
problems. Surprisingly, there are very few reports in the literature Heat balance
on the application of these techniques to RD columns. Simandl C C C
and Svrcek (1991) demonstrate the application of these techniques
to an ethyl acetate RD system. Lee and Dudukovic (1998) compare
H j = h L j (1 + s j ) ∑l i, j + hVj (1 + S j ) ∑v i, j - hL j -1 ∑l i , j -1
i =1 i =1 i =1 (3)
C C
the traditional Naphtali-Sandholm method to homotopy continua-
tion for an ethyl-acetate system to conclude that convergence with - hVj + 1 ∑v i , j +1 - h Fj ∑f i, j - Qj = 0
i =1 i =1
the former is difficult while the latter is much more robust,
though computationally intensive. These reports, however, study In the above, i = 1 to C and j = 0 to N+1 so that there are a
the mildly non-linear ethyl acetate column with a reaction conver- total of (N+2).(2C+1) MEH equations. The component genera-
sion of around 50 mol%. Also, only problems with standard tion term in the mole balance equations is non-zero for the
column specifications of reflux ratio and distillate rate are solved reactive trays and zero otherwise. Treating the partial reboiler
so that the suitability of these methods for other reasonable (Tray N+1) and condenser (Tray 0) as equilibrium trays, a
column specifications remains in question. square system of equations results with the N+2 tray tempera-
In this work, the application of the traditional Naphtali- tures and (N+2).2C component flows corresponding to the
Sandholm method for solving RD problems for a variety of specifi- vapour and liquid streams leaving a tray, constituting the
cation sets is demonstrated for the industrially relevant methyl (N+2).(2C+1) unknown iteration variables.
acetate and MTBE columns. The utilization of the flexibility in The above formulation assumes that the condenser and
specification sets for preliminary control studies such as sensitiv- reboiler duties are specified since these are tray heat specifica-
ity analyses to identify potential input-output pairings and tions for the condenser (Tray 0) and the reboiler (Tray N+1),
evaluating the steady-state disturbance rejection is illustrated. The respectively. Alternative independent specifications can be
work also demonstrates the possibility of using the method for accommodated by replacing the enthalpy balances for the
obtaining converged solution for a specified product purity and condenser and reboiler, respectively, by the appropriate specifi-
reaction conversion. These specifications are extremely important cation equations (see Seader and Henley, 1998). In addition to
for obtaining operating conditions such as reflux ratio and reboiler the ordinary distillation specifications such as a tray temperature
duty for high purity/conversion column operation. or composition, for an RD column, the reaction conversion X can
The article is organized as follows. The Naphtali-Sandholm be specified. The corresponding replacement equation is:
method is briefly introduced, followed by algorithm performance
results on the MTBE and methyl acetate columns. Results are
in
Fltr (1 − X ) − ( lltr , N +1 + vltr ,0 ) = 0 , ltr: Limiting Reactant (4)
also reported for a variety of possible column specifications to
demonstrate the power of the method. We briefly touch upon the It states that an overall mole balance on the limiting reactant
utilization of the choice in column specifications for designing should agree with the specified reaction conversion.
control structures based purely on steady-state analyses. Last, Given a set of column specifications, the square system of
but not least, results for the specification set of column conver- equations can be solved using the standard Newton-Raphson
sion and product purity are reported. The article ends with the technique. The iterations are repeated until the squared sum of
conclusions that can be drawn from this work. the discrepancies is below a tolerance. To provide a reasonable
initial guess for the iteration variables, the tridiagonal algorithm
(Seader and Henley, 1998) with the reaction term in the mole
THE NAPHTALI-SANDHOLM METHOD balance set to zero is used. During the iterations, some of the
A detailed treatise on the method as applied to ordinary distilla- component flows can become negative. The exponential
tion columns can be found in standard textbooks, e.g. Seader transforms suggested by Naphtali and Sandholm are used to
and Henley (1998) or the original article by Naphtali and ensure that these negative flows are mapped to small positive
Sandholm (1971). Here we focus only on the modifications to values. Also, to ensure a reasonably scaled system of equations,

76 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 85, FEBRUARY 2007

the M and H equations are scaled by the total feed flow rate and
an average latent heat of vaporization, respectively.
A steady-state RD simulator in MATLAB has been developed
implementing the Naphtali-Sandholm algorithm. For VLE
calculations, the Wilson equation is used for the liquid phase
activity coefficients while the vapour phase is modelled as ideal.
For calculating liquid and vapour phase enthalpies, the DIPPR
method is used. The extended Antoine equation is used for
calculating the pure species saturation pressures. Partial deriva-
tives for the Jacobian are obtained analytically. The next section
briefly describes the methyl acetate and MTBE RD column
simulation examples used here.
SIMULATION EXAMPLES
Commercially important examples of a methyl acetate and MTBE
RD columns are used in this work for demonstration. Schematics
of the two columns with important design and operating
conditions are shown in Figure 1. These design and operating
parameters are taken from existing literature reports (Wang et
al., 2003; Al-Arfaj and Luyben, 2002). The methyl acetate
column operates at 126.65 kPa pressure with a reflux ratio of
Figure 1. Schematic of example RD columns with base case operating
conditions – (a) MTBE, (b) Methyl Acetate
1.875. Pure acetic acid is fed near the top of the column (Tray 8)
and a stoichiometric amount of pure methanol is fed near the
bottom of the column (Tray 25). For the MTBE system, a slight
excess of methanol is fed on Tray 9 and fresh isobutene is fed on
Tray 10. A constant column pressure of 1114 kPa is assumed and
a reflux ratio of 7.5 is used.
For modelling the two systems, activity-based kinetic expres-
sions are used for the reactions as given in Table 1. The extended
Antoine equation is used to calculate the pure component
saturation pressure. The Antoine parameters are taken from
Aspen Plus 10.2. For the methyl acetate system, acetic acid
dimerizes in the vapour phase. Marek’s method (Marek, 1955)
is used to account for its impact on the VLE.
STEADY-STATE SIMULATION RESULTS
The convergence results for the different specification sets for the
example columns are reported in this section. The simulation is
considered converged if the sum of squares of the discrepancies
in the scaled MEH equations is below 5 × 10–12. The two
columns are first simulated for the standard specifications of the
reflux ratio (L/D) and the flow rate of the product stream of
interest, which is the distillate stream for the methyl acetate
column and the bottoms stream for the MTBE column. These
specifications are also indicated in Figure 1. Convergence is
achieved in 55 iterations for the methyl acetate column while the
MTBE column converges in 52 iterations. The converged product
compositions and reaction conversion for the two systems is
reported in Table 2. For the methyl acetate column, almost
complete reaction conversion of 97.92 mol% is achieved. The
distillate methyl acetate purity is 94.76 mol% while the bottoms
is 96.68 mol% water. For the MTBE column, the reaction conver-
sion is 91.07 mol%. The bottoms is 99.35 mol% MTBE. The inert
n-butene, unreacted methanol and isobutene leave from the distil-
late. These results are in agreement with literature reports (Wang
et al., 2003; Al-Arfaj and Luyben, 2002) from where the column
designs are taken. The results have also been validated with the
RADFRAC module in Aspen Plus 10.2 (data not shown).
Having obtained a base-case converged column, different
specification sets are simulated. The value chosen for the specifi-
cation variable is set equal to the converged value for the
standard specification simulated earlier to ensure that the
specification is feasible. Table 3 reports the convergence results
for representative specification sets for the two columns. Note
that converged results are obtained for all the cases reported.

Table 1. Reaction kinetics for example RD systems

System MTBE Methyl Acetate

ZZZ
C 4 H 8 + CH3OH + I Y X
ZZZ C5 H12O + I , ( Inert , I = n − C4 ) ZZZ
CH3COOH + CH3OH Y X
ZZZ CH3COOCH3 + H 2O
Reaction
(i - C 4 ) ( Methanol ) ( MTBE ) ( AcOH ) ( MeOH ) ( MeOAc ) ( Water )

⎛ a MeAc awater ⎞
M c kf ⎜ a AcOH a MeOH - ⎟
⎡ a a MTBE ⎤ ⎜ K eq ⎟
⎝ ⎠
Rate Expression rMTBE = M c kf ⎢ IB − ⎥ rMeAc =
(1 + K AcOH aAcOH + K MeOH a MeOH + K MeAc a MeAc + Kwater awater )
2
⎢⎣ a MeOH K eq a 2 MeOH ⎥⎦
( K HOAc = 3.18, K MeOH = 4.95, K Water = 10.5, K MeOAc = 0.82 )

Kinetic
k f = 13.21 × 10
12
exp ( )
-11110

T
kmol / ( kg catalyst . h )
⎛ -52275.93 ⎞
kf = 69.42 × 109 exp ⎜
⎝ RT ⎟ kmol /(kg catalyst. h )
⎠
Parameters
⎛ 6820 ⎞ ⎛ 782.98 ⎞
K eq = 8.3368 × 10 −8 exp ⎜ ⎟ K eq = 2.32 exp ⎜ ⎟
⎝ T ⎠ ⎝ T ⎠

VOLUME 85, FEBRUARY 2007 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 77

 Table 2. Converged solutions for MTBE and Methyl Acetate RD systems for base case specifications (MTBE: L/D – B , Methyl Acetate: L/D – D)
System MTBE Methyl Acetate
Design Specifications Reflux Ratio (L/D =7.5) and Reflux Ratio (L/D = 1.875) and
Bottoms (B = 648.28 kmol/h) Distillate (D = 310 kmol/h)
B, kmol/h 648.28 290
DISTILLATE BOTTOMS

Composition xIB = 0.0002, xMeOH = 0.0001, xAcOH = 0.0215, xMeOH = 0.0117,

(mole fraction) xMTBE = 0.9935, xn-C4 = 0.0062; xWater = 0.9668, xMeOAc = 0.0000;
D, kmol/h 1383.90 310
Composition xIB = 0.0460, xMeOH = 0.0480, xAcOH = 0.0001, xMeOH = 0.0091,
(mole fraction) xMTBE = 0.0002, xn-C4 = 0.9058; xWater = 0.0432, xMeOAc = 0.9476;
Conversion, X 91.07% 97.92%
Reboiler Duty, MW 40.52 4.628

that the algorithm does not diverge. Also, limiting the tray temper-
Table 3. Converged results for MTBE and Methyl Acetate RD
columns for various sets of specifications
ature change between iterations to within ±5ºC substantially
improves the convergence properties of the algorithm.
Specification set and Value t No. of Iter. In order to assess the robustness of the algorithm for the
L/D 7.5 0.8 52 different specification sets, the value each of the specification
variables was varied about its base case while keeping the other
B, kmol/h 640.8
one fixed. The range of specification values for which the
L/D 7.5 0.5 302 algorithm converges is tabulated in Table 4. For the methyl
Qr, MW 40.52 acetate column, convergence is achieved for a large range of
values for the specification variables. Thus, for example, for the
L, kmol/h 10 379.25 0.4 69 standard L/D-D specification set, for a fixed distillate rate of 310
T12, K 385.92 kmol/h, the algorithm converges as L/D is varied between 0.1 to
L, kmol/h 10 379.25 0.4 748
2.3 times its base case value of 1.875. Similarly for a fixed L/D
of 1.875, the algorithm converges as the D is varied between 100
Qr, MW 40.52 kmol/h and 590 kmol/h.
L/D 7.5 0.5 49 For the MTBE column, the range of specification values for
MTBE

which the algorithm converges is much smaller. For the L/D-B

T12, K 385.92
specification set, the L/D range is between 4.00 and 14.70 for a
L/D 1.875 0.5 55 fixed bottoms flow rate of 648.28 kmol/h. In contrast, the range
D, kmol/h 310 of values for B is between 493 and 752 kmol/h at a fixed L/D of
7.5. Also, low values of the Newton step factor must be used to
L/D 1.875 0.5 48
prevent the algorithm from diverging. These results indicate that
Qr, MW 4.628 the MTBE system is highly non-linear.
L, kmol/h 581.25 0.5 53 A consequence of the high non-linearity in the MTBE system is
the existence of multiple steady states as has been reported in
Qr, MW 4.628 many articles. An interesting observation during the course of our
D, kmol/h 310 0.5 58 simulations was that as the Newton step correction factor was
Qr, MW 4.628
varied from 1 to 0.2 for the L/D-B specification set, the algorithm
converged to different low conversion steady states. The conversion
L/D 1.875 0.5 50 achieved for the three steady states are 91.07 mol%, 74.7 mol%
T34, K 369.38 and 0.047 mol%, respectively. The temperature profiles correspond-
ing to these three different steady states are shown in Figure 2. The
D, kmol/h 310 0.5 46
profile near the bottom is very steep for the high conversion steady
T34, K 369.38 state. Note that it is always the high conversion steady state that is
of practical interest. In the context of multiple steady states, it is
Methyl Acetate

L, kmol/h 581.25 0.5 40

T34, K 369.38
xMeOAc, D 0.9476
Qr, MW 4.628
For the methyl acetate column, convergence is achieved in
about 40–300 iterations, while for the MTBE column, convergence
can take up to 1000 iterations depending on the specification set.
The large number of iterations compared to ordinary distillation
columns that typically converge within 20 iterations highlights
the high non-linearity in the system of governing equations. In
many cases, especially for the MTBE system, the Newton step
correction factor must be reduced from 1 to about 0.5 to ensure
highlighted that the homotopy continuation technique is prefera-
ble as it provides a more complete picture of the solution space,
0.5 64 particularly for column design (Singh et al., 2005).
Another consequence of the high non-linearity in the MTBE
system is that the algorithm tends to diverge for some specifica-
tion sets such as L/D-D. Most of these specification sets correspond
to specifications towards the same end of the column. Given a
converged steady-state solution, these specifications are useful in
evaluating the column steady-state response to small changes in
the operating conditions, e.g. a small change in D at fixed L/D. In
contrast, for specifications at the opposite ends of the column,
e.g. L/D-B, a converged steady-state solution is easily obtained
even for this highly non-linear system of equations.

78 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 85, FEBRUARY 2007

 Table 4. Ranges of different specifications for which convergence is achieved

System Specification 1 Specification 2

Variable Base case Range Variable Base case Range
L/D 7.5 5 – 9.7 Qr, MW 40.52 36.47 – 52.67
” ” 4.00 – 14.70 B, kmol/h 648.28 493 – 752
” ” 1.2 – 11.2 T12, K 385.92 385.92 – 409.26
MTBE

L, kmol/h 10379.25 9342 – 14012 ” 385.92 385.92 – 402.59

” ” 4671 – 15049 B, kmol/h 648.28 499.17 – 790.81
L/D 1.875 0.1 – 2.3 D, kmol/h 310 100 – 590
” ” 1.0 – 3.2 T34, K 369.38 358.83 – 377.16
” ” 0.9 – 5.0 Qr, MW 4.628 1.39 – 14.34
L, kmol/h 581.25 175 – 1336 ” ” 0.93 – 8.79
Methyl Acetate

D, kmol/h 310 31 – 589 ” ” 1.71 – 12.04

” 310 306.9 – 313.1 T34, K 369.38 361.60 – 374.94
L, kmol/h 581.25 465 – 697.5 T34, K ” 363.82 – 374.94
XMeOAc, D 0.9476 0.92 – 0.9476 ” ” 363.85 – 374.90

Figure 2. Temperature profiles for different steady states Figure 3. Tray temperature sensitivity with respect to fresh feeds
in the MTBE RD column for the MTBE column

The above results clearly establish the power of the Naphtali-
Sandholm method as a robust solver for reactive distillation
problems. In addition, the method provides immense flexibility in
choosing the column specifications. Since each specification set
corresponds to a particular column operating policy, the method
can be usefully employed for preliminary control structure design
studies. This is demonstrated in the next section.
CONTROL STRUCTURE DESIGN
There are two primary tasks in the regulation of double-feed RD
columns that are operated with near stoichiometric feeds. First,
the column internal flow rates should be such that the separa-
tion load is handled and the desired product purity is maintained.
Second, the feeds should be maintained in stoichiometric
balance as a large excess of any reactant would cause it to
eventually end up in one (or both) the product streams. Thus,
two control loops, for adjusting the column internal flow and for
stoichiometric feed balancing are needed. Assuming that the
column is operated at a fixed reflux ratio and the reflux rate and
bottoms are used to regulate the condenser and reboiler levels
respectively, the valves available for control are the two fresh
feeds and the reboiler duty.
Sensitive trays for temperature/composition measurements
are needed so that by maintaining them at their set points, the
column product purity and conversion are maintained for distur-
bances such as production rate changes and feed composition
changes. For the sake of brevity, we focus only on the control
structure design for the MTBE column.
Figure 3 plots the tray temperature sensitivities with respect
to the two fresh feeds and the reboiler duty for the MTBE
column. Tray 12 in the stripping section is the most sensitive
with respect to the reboiler duty so that they may be paired in a
loop. Another loop is needed for balancing the two feeds into the
column. Use of reactive tray temperatures leads to negative
Niederlinski Index values indicating instability due to interac-
tion between the two loops.
An alternative measurement is the isobutene composition on
the trays. Figure 4 plots the isobutene composition sensitivities
with respect to the two fresh feeds with Tray 12 temperature
being held constant. There are thus two candidate control
structures (CS1 and CS2) with reboiler duty controlling Tray 12

VOLUME 85, FEBRUARY 2007 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 79

 temperature and one of the fresh feeds controlling the isobutene
composition on a reactive tray (Tray 3 or Tray 10). The fresh
butenes feed is used for composition control in both control
structures. Note that for these control structures, the flow rate of
the fresh feed not used for composition control (Methanol) acts
as the production rate handle.
For further screening, the steady-state relationship that the
composition controller will see for production rate changes given
that the temperature controller keeps Tray 12 temperature fixed
at its set point as plotted in Figure 5 for CS1 with Tray 3 composi-
tion being controlled and for CS2 with Tray 10 composition being
controlled. For CS2, the relationship is reasonably linear so that
the feeds are more or less kept in stoichiometric balance. On the
other hand for CS1, input multiplicity in the relationship is seen.
In fact, for a small production rate change of +5%, the base case
composition set point is infeasible. Therefore, only CS2 with Tray
Figure 4. Tray iso-butene mole fraction sensitivity with respect to fresh 10 isobutene composition that is controlled using the fresh
feeds with fixed Tray12 temperature butenes feed is found to be a reasonable control structure for this
column. The steady-state product purity and reaction conversion
(a) for ±20% changes in the production rate handle set point and
for ±3% changes in the fresh isobutene composition in the
butenes feed is given in Table 5. The control structure is able to
maintain the column at its design purity and conversion for these
principal disturbances into the column.
In the above analysis, the open-loop temperature sensitivities
(Figure 3) are generated using the L/D-Q specification set. The
tray composition sensitivities (Figure 4), the steady-state relation-
ship that the composition controller sees (Figure 5) and the
steady-state column performance metrics for production rate and
feed composition changes (Table 5) are generated using the L/D-
T12 specification set. The flexibility in choosing the specification
set thus plays a pivotal role in guiding the preliminary control
structure screening exercise. The Naphtali-Sandholm method
provides such flexibility and is a valuable tool for control
structure design.
(b)

REACTION CONVERSION SPECIFICATION

As alluded to earlier, the reaction conversion can be used as a
specification variable by replacing the condenser/reboiler energy
balance equation with the overall material balance constraint.
Since the RD process is designed to achieve a certain reaction
conversion and product purity, one would like to use the reaction
conversion and component composition as specifications to
simulate the column and calculate the operating conditions
required. Using the simulator, converged results were obtained
for both the systems for high purity and high conversion specifi-
cations as in Table 6. The range of conversion and purity specifi-
cation values for which convergence is achieved is, however,
very small. Also, a very small value (~0.05) of the Newton step
factor must be used.
Figure 5. Open loop steady-state input output relationship for the In view of the difficulties in converging the purity-conversion
composition loop with fixed Tray 12 temperature – (a) CS1: Tray 3 specification set, an alternative is to specify one of the column
isobutene mol fraction vs. fresh butenes feed, (b) CS2: Tray 10
streams instead of the product purity. Thus, for the methyl
isobutene mol fraction vs. fresh butenes feed
acetate column, specifying the distillate rate to be slightly more
than the fresh acetic acid (or methanol) feed rate and setting the
Table 5. Conversion and purity of MTBE of MTBE RD column with
±20% production rate change and ±3% IB composition change reaction conversion specification to nearly one should result in
a converged solution that gives a distillate stream with high
± 20% production rate* ± 3% IB composition change* methyl acetate purity. The advantage of this specification set is
Conversion Purity Conversion Purity that the simulation is more likely to converge since one of the
column streams is held constant between the iterations. Table 7
89.16 99.36 92.20 99.33
reports the converged results for the two columns when reaction
(93.25) (99.34) (97.71) (99.39) conversion is used as a specification variable. The other specifi-
*Values in brackets correspond to negative change cation variable is the distillate rate and the bottoms rate for the

80 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 85, FEBRUARY 2007

 Table 6. Converged results for MTBE and Methyl Acetate RD systems with product purity and high reaction conversion as specifications

System Specification set and Values t # Iter* L/D D kmol hr-1 Qr MW

MTBE xMTBE,B 0.9935 0.05 809 7.5 1383.9 40.52
Conversion, % 91.07
Methyl Acetate xMeOAc,D 0.9476 0.05 492 2.0721 310 6.779
Conversion, % 97.92
* Number of iterations for convergence

Table 7. Converged results with reaction conversion as a specification

System Specification set and Values t No. Iter L/D Purity of Product* D Q
L/D 7.5 0.5 72 7.5Ψ 0.9935 1383.9 40.52
MTBE X, % 91.07
T12, K 385.92 0.4 75 5.5156 0.9925 1383.3 28.40
X, % 91.07
X, % 97.92 0.3 249 1.8755 0.9478 309.9 4.628
Methyl Qr, MW 4.628
Acetate D 310 0.2 179 1.8745 0.9476 310.0Ψ 4.628
X, % 97.92
Ψ Same as specified value
* Product leaves in the bottoms for the MTBE column and in the distillate for the methyl acetate column

methyl acetate and MTBE columns, respectively. The high E equilibrium relation
product purity achieved for both the cases shows that a reactive f feed of a component (kmol/h)
distillation column can indeed be operated to achieve high F feed (kmol/h)
purity and reaction conversion. Once such a converged solution hF specific enthalphy of feed (MJ/kmol)
is obtained, incremental changes to the column operating hL specific enthalpy of liquid (MJ/kmol)
conditions can be made to fine tune the product purity and hV specific enthalpy of vapour (MJ/kmol)
reaction conversion. The above demonstration highlights the Hj heat balance equations for stage j
utility of the conventional Naphtali-Sandholm simulator in K adsorption constant
obtaining column-operating conditions for high reaction conver- kA dimerization constant for acetic acid
sion and purity. Keq equilibrium constant
kf forward rate constant (kmol/kg-catalyst . h)
l liquid component flow (kmol/h)
CONCLUSIONS
L reflux flow rate (kmol/h)
This work has demonstrated the use of the Naphtali Sandholm L/D reflux ratio
method for solving the governing steady-state equations of a M mole balance equation
reactive distillation column. The simulation results for the highly Mc mass of catalyst (kg-catalyst per tray)
non-linear methyl acetate and MTBE RD systems show that for N number of stages
reasonably chosen specification sets, the algorithm shows good NRX number of reations
convergence properties. The method is extremely flexible in P pressure (kPa)
handling a range of specifications with ease. The flexibility in the Q heat input to a stage (MW)
choice of the specification variables can be exploited for the Qr reboiler duty (MW)
design of control structures using purely steady-state analysis. r rate of reaction (kmol/h)
The work also demonstrates the use of reaction conversion as a s fractional liquid side draw (kmol/h)
specification variable for obtaining high purity and high conver- S fractional vapour side draw (kmol/h)
sion column designs that are of practical interest. t Newton step factor
T temperature (K)
ACKNOWLEDGEMENT v component vapour flow rate (kmol/h)
x liquid mole fraction
The authors gratefully acknowledge the financial assistance of X conversion
the Initiation Grant Scheme, Indian Institute of Technology, y vapour mole fraction
Kanpur, India.
Greek Symbols
NOMENCLATURE υ stoichiometry coefficient
η tray efficiency
a activity
B bottom flow rate (kmol/h) Superscript
C number of components
in inlet
D distillate (kmol/h)

VOLUME 85, FEBRUARY 2007 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 81

Subscripts Seader, J. D. and E. J. Henley, “Separation Process Principles,”
John Wiley & Sons, New York, U.S.A. (1998).
B bottoms
Siirola, J. J., “An Industrial Perspective on Process Synthesis,”
c catalyst
AIChE Symp. Ser. 91, 222–233 (1995).
D distillate
Simandl, S. and W. Y. Svrcek, “Extension of the Simultaneous
i component
Solution and Inside-Outside Algorithms to Distillation with
j stage number
Chemical Reactions,” Comp. Chem. Eng. 15, 337–348
k component
(1991).
ltr limiting reatant
Singh, B. P., R. Singh, M. V. P. Kumar and N. Kaistha, “Steady
MeOAc methyl acetate
State Analysis of Reactive Distillation Using Homotopy
MTBE methyl tert-butyl ether
Continuation,” Chem. Eng. Res. Des. 83, 959–968 (2005).
IB iso-butene
Smith, L. A., “Chemical Research and Licensing Company,”
U.S.A. Patent 4,307,254 (1994).
REFERENCES Taylor, R. and R. Krishna, “Modeling of Reactive Distillation,”
Agreda, V. H. and L. R. Partin, “Reactive Distillation Process Chem. Eng. Sci. 55, 5183–5229 (2000)
for the Production of Methyl Acetate,” U.S.A. Patent, Thiel, C., K. Sundmacher and U. Hoffmann, “Synthesis
4,435,595 (1984). of ETBE: Residue Curve Maps for the Heterogeneously
Al-Arfaj, M. A. and W. L. Luyben, “Effect of Number of Catalysed Reactive Distillation Process,” Chem. Eng.
Fractionating Trays on Reactive Distillation Performance,” J. 66,181–191 (1997).
AIChE J. 46, 2417–2425 (2000). Wang, S. J., D. S. H. Wang and E. K. Lee, “Effect of Interaction
Al-Arfaj, M. A. and W. L. Luyben, “Comparative Control Study Multiplicity on Control System Design for a MTBE Reactive
of Ideal and Methyl Acetate Reactive Distillation,” Chem. Distillation Column,” J. Proc. Contr. 13, 503–515 (2003).
Eng. Sci. 57, 5039–5050 (2002).
Boston, J. F. and S. L. Sullivan Jr., “New Class of Solution
Manuscript received January 17, 2006; revised manuscript received
Methods for Multicomponent, Multistage Processes,” June 6, 2006 accepted for publication August 28, 2006.
Can. J. Chem. Eng. 52, 52–63 (1974).
Chang, Y. A. and J. D. Seader, “Simulation of Continuous
Reactive Distillation by a Homotopy Continuation Method,”
Comp. Chem. Eng. 12, 1243–1255 (1988).
Hauan, S., T. Hertxberg and K. M. Lien, “Why Methyl-tert-
Butyl-Ether Production by Reactive Distillation may Yield
Multiple Solutions,” Ind. Eng. Chem. Res. 34, 987–991,
(1995).
Hickey, T. P. and J. R. Adams, “Multi-Purpose Catalytic
Distillation Column and Etherification Process Using Same,”
U.S.A. Patent 5,321,163 (1994).
Jacobs, R. and R. Krishna, “Multiple Solutions in Reactive
Distillation for Methyl-tert-Butyl Ether Synthesis,” Ind.
Eng. Chem. Res. 32, 1706–1709, (1993).
Komatsu, H., “Application of the Relaxation Method of Solving
Reactive Distillation Problems,” J. Chem. Eng. Japan 10,
200–205, (1977).
Lee, J. H. and M. P. Dudukovic, “A Comparison of Equilibrium
and Nonequilibrium Models for Multicomponent Reactive
Distillation,” Comp. Chem. Eng. 23, 159–172 (1998).
Luyben, W. L., “Design and Control Degrees of Freedom,”
Ind. Eng. Chem. Res. 35, 2204–2214, (1996).
Luyben, W. L., “Plantwide Dynamic Simulators in Chemical
Processing and Control,” Marcel Dekker, New York,
U.S.A. (2002).
Marek, J., “Vapor Liquid Equilibria in Mixtures Containing an
Associating Substance. II Binary Mixtures of Acetic Acid at
Atmospheric Pressure,” Coll. Czech. Chem. Commun. 20,
1490–1502 (1955).
Mohl, K., A. Kienle, E. Gilles, P. Rapmund, K. Sundmacher
and U. Hoffmann, “Steady-State Multiplicities in Reactive
Distillation Columns for the Production of Fuel Ethers
MTBE and TAME: Theoretical Analysis and Experimental
Verification,” Chem. Eng. Sci. 54, 1029–1043 (1999).
Naphtali, L. M. and D. P. Sandholm, “Multicomponent
Separation Calculation by Linearization,” AIChE. J. 17,
148–153 (1971).

82 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 85, FEBRUARY 2007