You are on page 1of 12

Chemical Engineering Science 71 (2012) 166–177

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

Shortcut design of reactive distillation columns


Akram Avami a, Wolfgang Marquardt b, Yadollah Saboohi c,n, Korbinian Kraemer b
a
Mechanical Engineering Department, Sharif University of Technology, Tehran, Iran
b
Aachener Verfahrenstechnik-Process Systems Engineering, RWTH Aachen University, Germany
c
Sharif Energy Research Institute (SERI), Sharif University of Technology, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: Shortcut design methods can be valuable tools for rapid screening of different separation process
Received 21 August 2011 alternatives to assess feasibility and determine minimum energy demand. This work presents a variant
Received in revised form of the feed angle method (Kraemer et al., 2011), which applies to single- and double-feed reactive
12 December 2011
distillation columns. The proposed method relies on pinch point analysis and determines the minimum
Accepted 13 December 2011
energy demand from the calculation of a tray at the feed pinch. It is fully algorithmic and insensitive to
Available online 24 December 2011
impurities in product specifications. Its validity and performance is illustrated by a variety of case
Keywords: studies covering multi-component and multi-reaction systems, mixtures with potential liquid–liquid
Chemical processes phase split, and double-feed columns. The results show that the method is sufficiently accurate and
Distillation
computationally efficient, even for highly non-ideal systems and complex configurations.
Reaction engineering
& 2011 Elsevier Ltd. All rights reserved.
Reactive distillation
Pinch points
Energy demand

1. Introduction during the early stages of the design process are facilitated.
Although there are several approaches in literature for non-
Combining reaction and separation in a single unit can reduce reactive distillation columns, shortcut design of RDC still remains
both operating and capital costs. The direct removal of products a largely open problem. In the following, the most important
or intermediates in reactive distillation columns often results in shortcut methods reported in literature and their applicability to
higher reaction conversion compared to configurations compris- the conceptual design of RDC are briefly reviewed.
ing interconnected reactors and separators. However, the inter-
action of equilibrium phase separation and reaction increases the
complexity of the design problem. Although there is a significant 1.1. Boundary-value methods
body of literature on simulation case studies of reactive distilla-
tion (Taylor and Krishna, 2000; Sundmacher and Kienle, 2002), The boundary-value method (BVM) is a well known shortcut
only little work has been done on model-based design optimiza- method, which considers tray-by-tray calculations for both recti-
tion using rigorous models (Ciric and Gu, 1994; Cardoso et al., fying and stripping sections starting from the column ends
2000; Stichlmair and Frey, 2001; Gangadwala and Kienle, 2007). (Barbosa and Doherty, 1988a). When rectifying and stripping
The solution of these complex models is computationally profiles intersect, the specified separation is considered to be
demanding. Often, convergence to good local optima can only feasible. The procedure can also provide an assessment of mini-
be achieved if good initial guesses are available. This situation mum energy demand (MED).
motivates the development of shortcut methods for the concep- The BVM has been extended to apply to double-feed RDC
tual design of reactive distillation columns (RDC), which allow for operating close to simultaneous phase and chemical equilibrium
a fast screening of design alternatives and provide good initializa- on every tray (Barbosa and Doherty, 1988b). In double-feed RDC,
tions for rigorous optimization (Marquardt et al., 2008). Shortcut one feed stream controls the MED, while the tray location of the
design methods provide valuable information on feasibility and other feed stream is a design degree of freedom. Therefore,
energy demand without the need for detailed specifications. the location of one of the feed trays has to be specified a priori
Hence, targeting and screening of a large number of alternatives if the BVM method is applied to double-feed RDC. Since the
intersection of the concentration profiles has to be checked by
visual inspection, the application of the BVM is effectively limited
n
Corresponding author. Tel.: þ982166085198. to separations involving two or three independent components
E-mail address: saboohi@sharif.ir (Y. Saboohi). (cf. Ung and Doherty, 1995a). In addition, its high sensitivity to

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.12.021
A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177 167

trace components demands tedious iterations. Methods based on Recently, Kraemer et al. (2011) presented a new shortcut
pinch point analysis avoid this problem. method for heteroazeotropic distillation columns called the feed
Dragomir and Jobson (2005) introduce the concept of stage angle method (FAM). The FAM extends the RBM by adding a
composition lines (SCL) for conceptual design of kinetically vertex to the rectification bodies in form of the tray above or
controlled RDC. The SCL of each column section form a surface below the feed pinch in the non-pinched section in order to
for a range of tray hold-ups, which correspond to a range of account for the curvature of highly non-ideal separation profiles.
extents of reaction. A feasible separation can be expected if these For direct or indirect splits, the only necessary condition for a
surfaces intersect and if the constraint on the desired overall column to operate at MED is that the feed pinch, the saddle pinch
extent of reaction is met. The intersection point is detected by of the non-pinched rectification body, and the tray above or
visual inspection, hence restricting the applicability of the below the feed pinch in the non-pinched section are collinear.
method. Dragomir and Jobson (2004) apply the SCL to multiple- Thus, in contrast to the RBM, the FAM captures the curvature of
feed columns with reactive and non-reactive sections. One feed the concentrations profiles in the vicinity of the feed pinch and
stage is assumed to be located at the interface of reactive and thus improves the quality of approximating MED enormously. As
non-reactive sections. At least one non-reactive SCL has to a consequence, a fairly accurate prediction of MED is possible at
intersect the reactive surface for a feasible separation. little computational effort even for highly nonideal separations
such as heteroazeotropic distillation. Since the method does not
rely on numerous tray-to-tray calculations and visual inspection,
1.2. Pinch-based methods it is not only computationally efficient but can also be part of a
constrained optimization problem.
Bausa (2001) extends the rectification body method (RBM) to In this work, we extend the FAM such that an efficient shortcut
RDC aiming at the determination of MED. The reactive pinch evaluation of RDC can be accomplished. The new method is
points in the rectification and stripping sections are connected illustrated for the evaluation and design of different types of
linearly to construct the rectification bodies, which are supposed RDC involving homogeneous and heterogeneous mixtures, multi-
to approximate the set of composition profiles resulting from component and multi-reaction systems as well as complex
closely related product specifications. The intersection of these column configurations. The main concept of the present metho-
rectification bodies indicates separation feasibility and provides dology is explained in Section 2. Industrial application of reactive
an estimate of MED. However, the quality of the linear approx- distillation often involves more complicated situations such as
imation is often not sufficiently accurate (Bausa, 2001; Urdaneta the occurrence of multiple reactions, liquid phase split in the
et al., 2004). column, and the presence of more components. In order to
Lee et al. (2003) propose a shortcut method for kinetically address these requirements, we analyze the application of the
controlled single-feed RDC, which combines the RBM for non- FAM to more complex mixtures in Section 3. Finally, double-feed
reactive distillation with the analysis of eigenvectors of reactive columns are widely used to improve the reaction conversions in
pinch points. Although this method provides insight into the the distillation columns. Consequently, the extension of the FAM
design of a class of RDC, this graphical approach cannot be widely to these column configurations is discussed in Section 4. Section 5
applied and cast into a systematic procedure, which is easy to use. concludes the paper with a short summary.
Urdaneta et al. (2004) present a method for equilibrium-con-
trolled RDC, which is based on continuous distillation boundaries
(CDB). Their method combines pinch tracking as in the RBM and 2. Feed angle method for single-feed reactive distillation
tray-to-tray calculations as in the BVM starting from saddle pinches columns
to overcome the limitations of both methods. The method is
illustrated by the design of homogeneous and heterogeneous RDC. This section introduces first the basic concepts of the FAM as
Due to the need of visual inspection for checking the intersection of introduced by Kraemer et al. (2011) for non-reactive distillation
composition profiles, the method is limited to two-dimensional and extends it to reactive separations.
graphical representations of reactive separations.
2.1. Basic concepts

The FAM is based on pinch point analysis. Pinch points are the
1.3. Shortest stripping line method singular points of the tray-to-tray recursion. They are computed
from mass (and energy) balances around the rectifying and the
Lucia et al. (2008) propose the shortest stripping line method stripping sections assuming a pinched concentration profile at the
(SSLM) to assess feasibility and to determine MED of RDC. This end of a section. The reactive pinch points in a column always lie
method relies on the observation that the shortest stripping line on the reaction equilibrium surface even for kinetically controlled
corresponds to MED of the column and, hence, minimizes the reactive separations (Chadda et al., 2000; Hauan et al., 2000).
distance along the tray-to-tray stripping profiles. The authors The pinch equations for a balance envelope (Fig. 1) around the
apply their method to the production of MTBE, an established fuel rectifying section in the presence of reactions are given by
additive. Though this case study involves a simple column where Nr
X Nc
X
the reaction occurs only in the rectifying section, it should be 0 ¼ V rp Lrp D þ Ej ni,j , ð1Þ
extensible to more general settings, too. The simplifying assump- j¼1 i¼1
tions, i.e. constant molar overflow and negligible heat of reaction,
may lead to an inaccurate assessment of feasibility and MED, but Nr
X
can easily be dropped to generalize the SSLM. 0 ¼ V rp yi,rp Lrp xi,rp DxD þ Ej ni,j , i ¼ 1,:::,N c , ð2Þ
j¼1
In summary, current shortcut design methods applicable to RDC
are limited by computational effort, simplifying assumptions, sensi- v l
tivity to impurities, or the dependency on visual inspection. In 0 ¼ V rp hrp ðyrp ,T rp ,pÞLrp hrp ðxrp ,T rp ,pÞDhD þ Q D , ð3Þ
addition, the extension of these methods to multi-reaction, multi-
feed, or multi-product systems is by no means straightforward. 0 ¼ yi,rp ki,rp ðxrp ,T rp ,pÞxi,rp , i ¼ 1,:::,Nc , ð4Þ
168 A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177

X
Nr
0 ¼ Lrp xi,rp V n yi,n Bxi,B þ En,j ni,j , i ¼ 1,:::,Nc ,n ¼ nf þ 1, ð9Þ
j¼1

l v
0 ¼ Lrp hrp V n hn BhB þ Q B , n ¼ nf þ 1, ð10Þ

0 ¼ yn kn ðxn ,T n ,pÞxn , i ¼ 1,:::,Nc ,n ¼ nf þ 1, ð11Þ

X
Nc
0¼ xi,n 1, n ¼ nf þ 1, ð12Þ
i¼1

X
Nc
Fig. 1. Balance envelope for the rectifying section. 0¼ yi,n 1, n ¼ nf þ 1, ð13Þ
i¼1
Nc
X
0¼ xi,rp 1, ð5Þ 0 ¼ r j ðxn ,T n ,pÞ, j ¼ 1,:::,N r ,n ¼ nf þ1: ð14Þ
i¼1
Similarly, one rectifying tray above the feed pinch is calculated
if the feed pinch is the stable node of the stripping section.
X
Nc
For an estimation of the MED using the FAM, the tray below
0¼ yi,rp 1, ð6Þ
i¼1
(or above) the feed pinch has to point toward the relevant saddle
pinch of the non-pinched section (Kraemer et al., 2011). There-
fore, the angle between the line connecting the feed pinch with
0 ¼ r j ðxrp ,T rp ,pÞ, j ¼ 1,:::,Nr : ð7Þ the saddle pinch and the line connecting the feed pinch with the
tray below (or above) the feed pinch is minimized. This angle is
These equations are based on the assumption of simultaneous calculated from
phase and chemical equilibrium, which is made throughout this
paper. They describe the total mass balance (Eq. (1)), the compo- ðxf p xtr Þðxf p xsap ÞT
cosðaÞ ¼ ð15Þ
nent mass balances (Eq. (2)), the energy balance (Eq. (3)), the :xf p xtr :2 :xf p xsap :2
phase equilibrium conditions (Eq. (4)), closure conditions (Eqs.
and should become zero at MED.
(5) and (6)), and the chemical equilibrium condition (Eq. (7)). The
In many reactive separations, a liquid phase split may occur on
solutions of Eqs. (1)–(7) determine the pinch points of
one or more trays of the RDC. The FAM can easily be extended to
the rectifying section, which correspond to the fixed points of
this kind of reactive mixtures by extending the methodology
the tray-to-tray recursion. The equations for the stripping section
presented by Kraemer et al. (2011) for non-reactive to reactive
can be formulated in a similar way. An efficient numerical
three-phase distillation systems. The feed angle method carries
algorithm, which relies on a reformulation of the pinch equations
over to the reactive case if the pinch equations reported in this
and has been introduced by Bausa (2001), is used in this work.
paper are extended such that they account for a liquid phase split
The pinch points can be classified according to the stability
(cf. Kraemer et al., 2011). In particular, the stability of the liquid
properties of the fixed point. They are either classified as stable
phase has to be assessed as part of the calculation of pinch points
nodes (all the concentration profiles approach the pinch point),
and tray compositions. To this end, an efficient and robust phase
unstable nodes (all the concentration profiles leave the pinch
stability test originally proposed by Bausa and Marquardt (2000)
point), or saddle points (the concentration profiles pass by the
is employed. In case an unstable liquid is detected (either as part
pinch point). The locations of the reactive pinch points only
of the pinch or the tray calculations) by means of this phase
depend on the specification of the product specifications and
stability test, the set of vapor–liquid equilibrium equations is
the energy duty. For a given separation with specified product
replaced by a set of vapor–liquid–liquid equilibrium equations in
compositions xD and xB, the position of the pinch points in the
the pinch or tray equations systems.
compositional space is a function of the energy duty of the
column (Chadda et al., 2000; Hauan et al., 2000).
The first step in the application of the FAM constitutes the 2.2. An illustrative example
identification of the relevant pinch points and their classification
as either stable or unstable nodes or saddles. This task can be The extension of the FAM to reactive separations sketched in
carried out efficiently by means of the RBM (Bausa et al., 1998; the previous section is illustrated next by means of a simple
Bausa, 2001), which determines the relevant pinch points as the reactive separation problem, which has been extensively studied
vertices of the rectification bodies of the rectifying and the in literature before (Agreda et al., 1990; Bausa, 2001; Huss et al.,
stripping sections. An interpretation of the results of the RBM 2003). In particular, we look at the quaternary reactive mixture
facilitates also the identification of the reactive feed pinch. occurring in the production of methyl acetate (C3H6O2) from
The second step of the FAM refers to the calculation of a tray acetic acid (C2H4O2) and methanol (CH4O). Acetic acid reacts with
adjacent to the feed tray in the non-pinched section in addition to methanol to form methyl acetate and water. In the methyl acetate
the relevant pinch points. In case the feed pinch belongs to the synthesis, a liquid phase split occurs between water and methyl
rectifying section, the tray nf below the feed pinch is calculated. acetate. However, Bausa (2001) shows that the heterogeneous
The equations for the calculation of the tray n ¼nf þ1 below the region does not touch the reactive equilibrium surface and is
feed pinch in the stripping section are given by therefore not controlling the design. Huss et al. (2003) show that
the dehydration of methanol to DME has no significant effect on
Nr
X Nc
X the process. Therefore, the main reaction in the liquid phase is
0 ¼ Lrp V n B þ En,j ni,j , n ¼ nf þ 1, ð8Þ
j¼1 i¼1 C2H4O2 þCH4O"C3H6O2 þ H2O. (16)
A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177 169

The reaction equilibrium constant Keq is computed from Methanol Water


64.5°C s2 100°C
782:98K B, s3
lnðK eq Þ ¼ 0:84156þ ð17Þ 1
T
Column profile r1
as suggested by Huss et al. (2003). Since the phase split is not
r2
relevant for column design, the Wilson gE-model can be used for 0.8
the calculation of liquid phase activity coefficients, although it F
does not allow for the calculation of liquid phase split. The Stripping section
equilibrium model parameters are taken from ASPEN Properties s1
0.6
and are listed in the Appendix for better reference. Rectifying section
We consider a single-feed column and a stoichiometric feed to
D,r3 1 s2,Sad
produce a mixture of methyl acetate and water as distillate and
0.4 Tray below
pure water as bottoms products. The column pressure is set to feed
1.013 bar. The number of degrees of freedom for a reactive system Azeotrope 55.9°C 0.95
can be computed as Nc þNI–Nr–1 where Nc and NI refer to the r1,Feed pinch
number of reactive and inert components and Nr is the number of 0.2
0.9
reversible reactions. Since the number of degrees of freedom is 0.65 0.7
reduced by one for each reaction, two degrees of freedom remain
in this example. Likewise, since simultaneous chemical and phase 0
equilibrium is assumed, transformed variables (Ung and Doherty, 0 0.2 0.4 0.6 0.8 1
1995a) can be introduced to reduce the dimension of the Methyl acetate Acetic acid
concentration simplex and to facilitate graphical representation 57.1°C 118°C
of the separation. The design specifications for the reactive
Fig. 2. Illustration of the RBM, the FAM, and the column profiles at the MED for
separation are given in the upper part of Table 1 in transformed
methyl acetate production.
coordinates.
The RBM is applied to compute the pinch points and the
120
rectification bodies of the reactive separation (Bausa, 2001). For
methyl acetate production, the RBM results in a rectification body 100
Angle (Degree)

of the rectifying section formed by the pinch D at the distillate 80


composition, the saddle pinch r2, and the feed pinch r1, which
intersects with the rectification body of the stripping section 60
formed by the pinch B at the bottoms product, the saddle point s2, 40
and the stable node s1 in a single point at the MED. The pinch
20
points and the rectification bodies are plotted together with the
column profiles in Fig. 2. 0
17.6 17.8 18 18.2 18.4 18.6
Comparable to a non-reactive direct or indirect split, the
column profile passes through the feed pinch point and then Reboiler duty (MJ/kmol F)
enters the non-pinched section passing the saddle pinch to reach Fig. 3. Relation between reboiler duty and target angle for methyl acetate
the desired product specification. In particular, in case of methyl production.
acetate production, the feed pinch point r1 is the stable pinch of
the rectifying section (cf. in Fig. 2). adjacent to the pinch points. For methyl acetate production, the
The angle between the line connecting the feed pinch r1 and optimization problem reads as
the composition on the tray below the feed pinch and the line
min z ¼ a
connecting the feed pinch r1 and the saddle pinch s2 is consid-
ered to determine the MED. More precisely, the estimate of the subject to :
MED is determined as the solution of an optimization problem, Eqs:ð1Þð7Þ, p A frp,spg,
which minimizes the angle a according to Eq. (15) subject to the Eqs:ð8Þð14Þ, n ¼ nf þ1,
equality constraints (1)–(7), which fix the relevant pinch points of
Eq:ð15Þ: ð18Þ
the rectifying and stripping sections, and the constraints (8)–(14),
which comprise the mass and energy balances of the trays This optimization problem (and all the others in this paper) is
solved on a standard PC with a 2.66 GHz Intel core2Quad CPU
Table 1 relying on the NLP solver SNOPT and the optimization platform
Design specifications (in transformed concentrations) and MED
GAMS. The RBM and the tray-to-tray calculations are performed
estimated by different shortcut methods for reactive distillation of
acetic acid, methanol, methyl acetate, and water.
in MATLAB. Phase equilibrium calculations are performed by
external functions connected to MATLAB and GAMS in order to
Specifications improve the computational robustness and efficiency.
The solution of this optimization problem yields a reboiler
XF 0.5/0.75/  0.25
duty of 17.97 MJ/kmol feed. This value is slightly higher than the
XD 0/0.5/0.5
XB 1/1/  1 MED determined by the RBM and the BVM, which both give the
D/F 0.5 same result (cf. Table 1). The CPU time for the calculation of the
MED is less than 1 s. A comparison of the results using the FAM
MED, QB,min/F[MJ/kmol] and the RBM shows that the linear approximation of the RBM
gives a good approximation of the profiles in the case of methyl
RBM 17.95 acetate production. In Fig. 3, the angle calculated by the FAM is
BVM 17.95
FAM 17.97
plotted for different values of the reboiler duty. The angle
increases if less or more energy than the MED is supplied.
170 A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177

3. Application and extension to more complex systems Table 2


Design specifications (in transformed coordinates) and MED computed with the
RBM and the FAM for the reactive distillation of formic acid, water, methanol, and
In this section, the FAM is applied to a number of case studies.
methyl formate.
Some of them not only involve multi-component but also multi-
reactive systems with possible liquid–liquid phase split. While Specifications
the basic concepts of the FAM introduced in Section 2.1 carry over
to these cases, the feed angle is not measured between two lines XF 0.5/0.5/0
XD 0.1/0/0.9
in case the system has to be represented in a simplex of XB 0.99/1/  0.99
transformed coordinates with dimension 3 or higher. Rather, the D/F 0.56
feed angle is now defined as the line connecting the feed pinch
with the tray above or below the feed pinch and the hyperplane MED, QB,min/F[MJ/kmol]
defined by the feed pinch and the relevant saddle pinches of the
RBM 21.5
non-pinched section has to be considered. Again, the relevant FAM 21.04
saddle pinches are identified using the RBM as an initialization
procedure. In higher dimensions, Eq. (15) has to be replaced by

nsap,f p ðxf p xtr ÞT


cosðaÞ ¼ , ð19Þ
:nsap,f p :2 :xf p xsap :2

where nsap,fp is the normal vector of the hyper plane defined by


the relevant saddle pinch points. The MED is then estimated by
the solution of an optimization problem similar to Eq. (18). The
objective function is substituted by

z ¼ ðcosðaÞÞ2 ð20Þ
where the angle is determined by Eq. (19).
The performance of the FAM is demonstrated next in compar-
ison to established methods for MED estimation. The first case
study (cf. Section 3.1) covers a system with four components and
one reaction. The mixture in the second case study (cf. Section
3.2) exhibits five components and two reactions. In the third
example (cf. Section 3.3), a system with five components and one
reaction resulting in three degrees of freedom is considered. Since
the FAM does not rely on graphical inspection, the MED can be
determined efficiently also for the higher-dimensional examples.

3.1. Production of methyl formate

Methyl formate (C2H4O2) is produced by the esterification of Fig. 4. Illustration of the rectification bodies and the FAM for methyl formate
formic acid (CH2O2) and methanol (CH4O) according to the production.
reaction:

CH2O2 þ CH4O"C2H4O2 þH2O. (21)


100
The design of a RDC for this chemical system has been studied
before by Bessling et al. (1997), Bausa (2001), and Urdaneta et al. 80
(2004). The UNIQUAC model is used to calculate the activity
Angle (°)

coefficients in the liquid phase. The chemical equilibrium constant 60


is calculated by means of Gibbs energies. All the phase equilibrium
40
model parameters are taken from ASPEN Properties. They are
summarized in the Appendix for better reference. A single-feed 20
column at atmospheric pressure and a stoichiometric feed ratio is
considered. The column produces a mixture of methyl formate and 0
methanol as distillate product and water as bottoms product. The 19 20 21 22 23 24 25
design specifications and the resulting MED are given in Table 2. Reboiler duty (MJ/kmol F)
The RBM results in two triangular rectification bodies in
transformed coordinates as shown in Fig. 4. The rectification body Fig. 5. Effect of the reboiler energy duty on the target angle for methyl formate
production.
of the rectifying section, formed by the distillate product D, the
stable node r1, and the saddle pinch r2, intersects the rectification
body of the stripping section, formed by the bottoms product B,
the feed pinch s1, and the saddle pinch s2, at the feed pinch. The the feed pinch. Since this composition profile is only slightly
RBM yields a MED of 21.5 MJ/kmol feed. curved, the linear approximation of the RBM yields a good
Fig. 4 also shows the illustration of the FAM, which results in estimate of the MED in this example. The relation between the
an estimate of the MED of 21.04 MJ/kmol feed, which is slightly reboiler energy duty and the feed angle of the FAM is shown in
less than the value determined by the RBM. At this energy Fig. 5. The solution of the minimization problem converges
demand, the rectification bodies do not intersect due to the rapidly such that the optimal solution can be found in less than
curvature of the composition profile in the rectifying section at 1 s CPU time.
A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177 171

Table 3 Benzene m-Xylene


Design specifications (in transformed coordinates) and MED estimated with RBM, 78.8 °C 138.7 °C
BVM, and FAM for the reactive distillation of ditert-butyl-benzene, meta-xylene,
tert-butyl-meta-xylene, tert-butylbenzene, and benzene. D
Stripping section
Rectifying section
Specifications s1
Column profile
XF 0.647/0.059/0.294
XD 1/  1/1
XB 0/2/ 1
D/F 0.647
F r2
MED, QB,min/F[MJ/kmol]
tert-
RBM 74.0
tert-Butyl-
-benzene
BVM 68.5 m-Xylene
168.06 °C
FAM 69.0 r1
99.5 °C
s2

3.2. Alkylation of xylenes


-0.2
Feed
The design of an RDC for the alkylation of xylenes has been pinch, r1

studied before by Ung and Doherty (1995a, 1995b) and Bausa B -0.3 Tray
(2001). The chemical system comprises of five components and
two reactions: di-tert-butyl-benzene (C14H22) reacts with meta- ditert-Butylbenzene s2,Saddle pinch
-0.4
xylene (C8H10) to produce tert-butyl-meta-xylene (C12H18) and 232.8 °C 0.65 0.7
tert-butyl-benzene (C10H14). In the second reaction, tert-butyl-
benzene (C10H14) and meta-xylene (C8H10) form tert-butyl-meta- Fig. 6. Illustration of the rectification bodies, the column profiles, and the FAM for
the alkylation of xylenes.
xylene (C12H18) and benzene (C6H6). These alkylation reactions
can be formalized as
80
C14H22 þC8H10"C12H18 þC10H14, (22)
60
C10H14 þ C8H10"C12H18 þ C6H6. (23)
Angle (°)

40
The reaction system has two degrees of freedom if simulta-
neous chemical and phase equilibrium is assumed. The Wilson 20
model is used to calculate the activity coefficients in the liquid
phase; the parameters are taken from Aspen Properties and are 0
listed in the Appendix for better reference. The coefficients in the 64 66 68 70 72
Antoine equation for the vapor pressure and the chemical Reboiler duty (MJ/kmol F)
equilibrium constants (Keq ¼0.6 for the first reaction and
Keq ¼0.16 for the second reaction) are taken from the work of Fig. 7. Effect of reboiler energy duty on the target angle for the xylene mixture.

Ung and Doherty (1995a). A single feed column with the speci-
fications given in transformed compositions in the upper part of Table 4
Table 3 is used to produce pure m-xylene as distillate product and Design specifications and MED estimated by the RBM, the FAM, and by rigorous
di-tert-butyl-benzene as bottoms product at atmospheric simulation for the reactive distillation of acetic acid, methanol, methyl acetate,
water, and sec-butyl acetate.
pressure.
The MED estimate determined by the RBM and the BVM is Specifications
74 MJ/kmol feed and 68.5 MJ/kmol, respectively (cf. Table 3). As
shown in Fig. 6, the rectification body in the rectifying section is xF 0.25/0.5/0/0.15/0.1
xD 0/0.5/0.5/0/0
spanned by the saddle pinch r2, the feed pinch r1, and the
xB 0/0/0/0.8/0.2
distillate D, while the rectification body of the stripping section D/F 0.5
involves the saddle pinch s2, the stable node s1, and the bottoms
product B. The MED is estimated as 68 MJ/kmol feed by means of MED, QB,min/F[MJ/kmol]
the FAM, which is in good agreement with the results of the BVM.
RBM 50
The column profiles shown in Fig. 6 indicate that the process is
Rigorous simulation 26.82
feasible at the MED determined by the FAM. Apparently, the RBM FAM 26.88
overestimates the MED significantly because of the curvature of
the stripping column composition profile at the feed pinch. The
FAM model is executed in about 1 s CPU time. Fig. 7 plots the
angle of the FAM over the reboiler duty for better illustration. number of degrees of freedom. In particular, there are Nc þNI–NR–
1¼4þ1–1–1¼3 degrees of freedom for this reactive system. In
3.3. Production of methyl acetate in the presence of an inert such cases, shortcut methods relying on graphical representations
are at least difficult to apply. The thermodynamic specifications
The esterification of acetic acid with methanol to produce are the same as those explained in Section 2.2.
methyl acetate, considered in Section 2.2, is here studied in the In this work, sec-butyl acetate is added to the mixture, which
presence of an entrainer as suggested previously in the patent is fed into the column at atmospheric pressure to obtain an
literature by Yeomans (1982). This example is considered here to equimolar stream of methyl acetate and methanol at the top of
illustrate that the FAM can be applied to systems with a higher the column and a mixture of water and butyl acetate as the
172 A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177

4 BuOH Water
117.7 °C 100.1 °C
3 1
Angle (°)

Tray above r2 D,r3


2 0.9 feed pinch

1 0.8
0 0.7
5
.7

81

82

85

.9
Stripping section
.8

.8

81
26

26
.8 0.6

.8

.8
26

26

.8
26

26

26
s1,Feed pinch
26
Reboiler duty (MJ/kmol F) 0.5 s2
Feed
Fig. 8. Effect of reboiler energy duty on the target angle for methyl acetate
0.4
production in the presence of an inert. Miscibility gap
0.3

0.2
Rectifying section
Table 5 0.1
Design specifications (in transformed coordinates) and the MED estimated with r1
the RBM and the FAM for the reactive distillation of levulinic acid, butanol, butyl
B,s3
0
levulinate, and water. 0 0.2 0.4 0.6 0.8 1
Specifications Butyl Levulinate Levulinc acid
232 °C 245 °C
XF 0.5/0.5/0
XD 0.98/1/  0.98 Stable/unstable pinches Saddle pinches
XB 0/0/1
D/F 0.5 Fig. 9. Illustration of the rectification bodies and the FAM for butyl levulinate
production.
MED, QB,min/F[MJ/kmol]
20
RBM 63
FAM 47 16
Angle (°)

12

bottoms product. The specifications are shown in the upper part 8


of Table 4 in real compositions.
An analysis of the results of the RBM reveals that the rectifying 4
section is shaped as a tetrahedron while the stripping section 0
forms a triangle with a stable node as the feed pinch. The MED
5

.3

.1

47

46

.7
.0

.7

.4

determined by the RBM is 50 MJ/kmol feed. The application of the


48

47

45
51

48

48

FAM requires the construction of the hyperplane formed by the Reboiler duty (MJ/kmol F)
relevant saddle pinches of the rectifying section and the feed
pinch. The target angle is calculated between the line connecting Fig. 10. Effect of the reboiler energy duty on the target angle for butyl levulinate
the feed pinch with one tray above the feed and the normal vector production.

of the hyperplane. The FAM yields a MED of 26.88 MJ/kmol feed,


which differs significantly from the value obtained by the RBM. In
order to check the quality of these estimates, we have performed by the reaction
intensive simulation studies (using Aspen Plus) to estimate the
C5H8O3 þC4H10O"C9H16O3 þH2O. (24)
MED by rigorous simulation using a column with 50 trays to
mimic a column with an infinite number of stages. A tedious
manual search involving the value of the reboiler energy demand This case study has been chosen to show that the FAM can be
has been performed to reach the desired product compositions. successfully applied to reactive distillation involving heteroge-
The results obtained by rigorous simulation are in good agree- neous mixtures. The vapor–liquid–liquid equilibrium is described
ment with those found by the FAM (cf. Table 4). The CPU time for by the NRTL model with parameters from the work of Harwardt
the calculation of the MED with the FAM is less than 1 s. et al. (2011). The mixture exhibits a large miscibility gap between
The feed angle is very sensitive to changes in the reboiler water and butanol. According to Bart et al. (1994), the reaction
energy duty close to the optimal value of the target angle as equilibrium constant Keq can be computed from
illustrated in Fig. 8.
702:97K
lnðK eq Þ ¼ 2:9275 : ð25Þ
T
3.4. Production of butyl levulinate An equimolar stream of levulinic acid and butanol is fed to the
column operated at ambient pressure to produce the ester at the
The mixture involved in butyl levulinate production exhibits a bottom and water at the top of the RDC. The design specifications
liquid–liquid phase split, which influences the column design. are given in the upper part of Table 5 in transformed coordinates.
Butyl levulinate (C9H16O3), an ester of levulinic acid (C5H8O3) and The MED determined by the RBM amounts to 63 MJ/kmol feed.
butanol (C4H10O), is a promising fuel candidate, which can be As shown in Fig. 9, the feed pinch s1 belongs to the stripping
produced from biomass (Werpy and Petersen, 2004) and shows section. Thus, the FAM method minimizes the angle between the
excellent combustion properties (Janssen et al., 2010). It is formed line connecting the feed pinch s1 with the saddle pinch r2 and the
A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177 173

line connecting the feed pinch s1 with the tray above the feed for balance envelope (A) are identical to Eqs. (1)–(7). The pinch
pinch. The FAM determines a MED of 47 MJ/kmol feed, which is equations for balance envelope (B) can be formulated in a similar
significantly lower than the one estimated with the RBM. At the way and are not presented for the sake of brevity. The pinch
value of the MED determined by the FAM, the rectification bodies equations for balance envelope C include the following equations:
do not intersect (cf. Fig. 9). This is due to the strong curvature of
X
Nr X
Nc
the composition profiles, which cannot be well approximated by 0 ¼ V mp þ F U Lmp D þ Ej ni,j , ð26Þ
the linear rectification bodies used in the RBM. The FAM, on the j¼1 i¼1
other hand, estimates the MED correctly, since the tray above
the feed pinch is located very close to the saddle pinch of the X
Nr

rectifying section. The relation between the target angle and the 0 ¼ V mp yi,mp þ F U xU Lmp xi,mp DyD þ Ej ni,j , i ¼ 1,:::,Nc , ð27Þ
j¼1
reboiler energy duty is plotted in Fig. 10.
v l
0 ¼ V mp hmp ðymp ,T mp ,PÞ þF U hU Lmp hmp ðxmp ,T mp ,PÞDhD þ Q D , ð28Þ
4. Extension to double-feed reactive distillation columns
but have to be combined with Eqs. (4)–(7) to compute the pinch
points of the middle section.
Double-feed reactive distillation columns often offer more
Barbosa and Doherty (1988b) stated that for a fixed feed ratio Fr,
favorable compositions on the column trays with respect to
defined as the ratio of the upper feed flow rate to the lower feed
conversion and selectivity compared to single-feed columns. For
flow rate, and for fixed compositions of the feed streams (xFU, xFL),
example, Ciric and Gu (1994) have shown that the distribution of
the bottoms product (xB), and the distillate (yD), a feasible double-
the reactant feed along the column height results in an economic-
feed column covers a continuous pass through rectifying, middle,
ally optimal design in case of ethylene glycol synthesis. Similar
and stripping sections. They determined the MED by an extended
findings have been reported later by Bessling et al. (1998), Chin
version of the BVM that requires one additional degree of freedom.
and Lee (2005), and Gadewar et al. (2007). Shortcut methods
Thus, the location of the feed tray is an exogenous parameter in their
should therefore cover these economically relevant configurations
analysis. In this work, this problem is avoided by applying pinch
of reactive distillation columns. Consequently, it is shown next
point theory. Previously, Levy and Doherty (1986) showed that in
how the FAM can be extended to double-feed RDC.
double-feed columns one feed stream controls the MED: above the
critical value, the concentration profiles cross each other, while they
4.1. Methodological extension do not intersect below. At MED, two out of the three column
sections are pinched in the vicinity of the feed controlling pinch
To determine the MED of double-feed RDC, we introduce three point, which is typically the stable pinch of the pinched concentra-
balance envelopes, namely around the rectifying section (A), the tion profiles. Levy and Doherty (1986) introduced an algebraic
rectifying and the middle section (C), and the stripping section approximation criterion to determine the MED in double-feed
(B) as indicated in Fig. 11. Pinch equation systems have to be columns. This corresponds to a configuration where the feed pinch
derived for all the three balance envelopes. The pinch equations point, the saddle pinch point, and the controlling feed composition
are collinear. However, this approximation suffers from low accu-
racy due to constant molar overflow. In this work, the FAM is
extended to consider the nonlinear behavior in active pinch zones
and an exact formulation is applied to determine the relevant
points. In analogy to the FAM for single-feed RDC introduced in
D
Section 2, the MED for double-feed RDC is calculated from the pinch
composition and those on a tray below or above the controlling feed
pinch in the non-pinched section adjacent to the pinched section.
These three compositions are checked for collinearity, which should
be established at MED as in case of single-feed RDC.
The application of the FAM requires a proper identification of
the feed pinch first, which can be accomplished by means of the
RBM as above. Depending on the controlling feed pinch, different
schemes have to be employed to calculate the MED accounting for
the type of split and the feed ratio as shown by Barbosa and
Doherty (1988b) in the context of the BVM. However, in contrast
to this work, the FAM applied to double-feed RDC does not
require a priori knowledge of the feed tray location.
If, for example, the stable pinch of the stripping section is the
feed pinch controlling MED (cf. the example treated in the next
subsection), the saddle pinch point in the non-pinched (middle)
section and one tray above the feed pinch in the middle section have
to be calculated to estimate MED by a minimization of the feed
angle. The balance equations for the tray calculations are given by
X
Nr X
Nc
0 ¼ V sp þ F U Ln D þ Ej ni,j , n ¼ nf 1 ð29Þ
j¼1 i¼1

Nr
X
0 ¼ V sp yi,sp þF U xU Ln xi,n DyD þ Ej ni,j , i ¼ 1,:::,Nc ,n ¼ nf 1
Fig. 11. Balance envelopes for the rectifying section (A), the stripping section (B),
j¼1
and the rectifying and middle sections (C) in a double-feed reactive distillation
column. ð30Þ
174 A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177

v l (4)–(7) representing the saddle pinch point in the middle section,


0 ¼ V sp hsp ðysp ,T sp ,PÞ þ F U hU Ln hn ðxn ,T n ,PÞDhD þ Q D , n ¼ nf 1:
Eqs. (29)–(31) describing the mass and energy balances for the
ð31Þ
tray above the feed pinch in the middle section, and Eq. (15) to
For the particular case considered, the angle between the line compute the target angle. A similar problem can be set up in a
connecting the feed pinch in the stripping section with the saddle similar way for other pinch point topologies.
pinch in the middle section and the line connecting the feed pinch
with one tray above the feed pinch in the middle section has to be
minimized to determine MED. The minimization problem reads 4.2. Illustrative example
as
The methodology is illustrated by the reactive separation of a
min z ¼ a quaternary system arising in the production of ethyl acetate
subject to : (C4H8O2) from acetic acid (C2H4O2) and ethanol (C2H6O), which
ð1Þð7Þ, rp-sp, has been extensively studied before (cf. Barbosa and Doherty,
ð26Þð28Þ, 1988b; Kenig et al., 2001). Acetic acid reacts with ethanol to form
ð4Þð7Þ, rp-mp, ethyl acetate and water by
ð29Þð31Þ, n ¼ nf 1,
C2H4O2 þC2H6O"C4H8O2 þH2O. (33)
ð15Þ: ð32Þ
The equality constraints are formed by Eqs. (1)–(7), which We consider a separation in a double-feed RDC with the
calculate the feed pinch in the stripping section, (26)–(28) and specifications given in Table 6 at atmospheric pressure.
The Wilson gE-model with coefficients from the work by
Suzuki et al. (1970) is used to calculate the liquid phase activity
coefficients. Ideal phase behavior is assumed for the gas phase
Table 6 though hydrogen bonding between the acid molecules is known
Design specifications (in transformed coordinates) and MED estimates based on
the BVM and the FAM for reactive distillation of acetic acid, ethanol, ethyl acetate,
to give rise to non-ideal gas phase behavior. Though the assump-
and water. tion is questionable, it is retained in this methodologically
oriented work to allow for a comparison of our results with those
Specifications of previous authors who also have used this assumption (cf.
Suzuki et al., 1970, or Barbosa and Doherty, 1988c). The chemical
XL 0/1/0
XU 1/0/0 equilibrium constant Keq is computed from
YD 0.99/0.9874/0.9774
450:382K
XB 0.8761/0.1239 lnðK eq Þ ¼ 0:8135 : ð34Þ
D/F 0.1265 T
according to Barbosa and Doherty (1988b).
Results
QB,min/FL [MJ/kmol] by FAM 304.92
The relevant pinch points used by the FAM to find the MED are
rmin by BVM 8.23 shown in Fig. 12. While the rectification body of the stripping
rmin by FAM 8.20 section is formed by the pinch B at the bottoms product, the
saddle pinch s2, and the feed pinch s1, the rectification body of

EtOH Ethyl acetate


78 °C 77.1 °C
1
FL D
0.9
Rectifying section
0.8

0.7

0.6

0.5

0.4 Tray above feed


s1, Feed pinch of
stripping section pinch in middle section
0.3 0.03
m2, Saddle pinch
0.2 of middle section 0.025
Stripping section m1, Unstable pinch
of middle section
0.1 0.02
r, Pinch point of
s2 B rectifying section
0 FU 0.015
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.98 0.99 1
Water Acetic Acid
100 °C 118.1 °C

Fig. 12. Illustration of the relevant pinches for the FAM for ethyl acetate production.
A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177 175

the rectifying section is formed by a straight line near the ethyl- 5. Conclusions
acetate–acetic acid edge (the path D-r). Since the lower feed
stream controls the problem, the angle between the line connect- Available shortcut methods for the design of reactive distilla-
ing feed pinch s1 to the saddle pinch m2 and the line connecting tion columns are either restricted by visual inspection required to
the feed pinch s1 to the tray above the feed controlling pinch check feasibility and by the sensitivity of the results to trace
approaches zero at MED. A MED of 304.92 MJ/kmol feed is components in the product specifications (BVM) or by inaccurate
determined by the FAM. A feasible double-feed RDC contains a estimates, which are due to highly curved composition profiles in
continuous pass from distillate to bottoms such that the rectifying the column sections (RBM). The present work proposes an
and the stripping profiles both intersect the middle-section extension of the feed angle method (FAM) for the estimation of
profile. In this work, the unstable pinch of the middle section the minimum energy demand for non-reactive distillation, devel-
(m2) intersects the rectifying section (D-r) at the value of MED oped recently by Kraemer et al. (2011), to highly non-ideal
determined by the FAM. The MED corresponds well to the results reactive distillation columns. The FAM only requires the calcula-
reported by Barbosa and Doherty (1988b) determined from an tion of the relevant pinches and the compositions on a tray above
adapted version of the BVM. The CPU time required to find the or below the feed pinch. Due to its algorithmic nature, the FAM is
MED using the FAM is less than 1 s. not only perfectly suited for the design of multi-component
systems represented in higher-dimensional composition spaces,
but can also be used as a simple column model in flowsheet
optimization. The novel method has been demonstrated by
various examples covering multi-component and multi-reaction
Table A.1 systems, mixtures with liquid–liquid phase split, and double-feed
M volumes (mixtures in Section 3.2). columns. In all cases, the FAM provides an assessment of feasi-
bility and an accurate estimation of the minimum energy demand
Component Vi
at very low computational effort. Future work will concentrate on
ditert-Butyl-benzene 242.356
further study of columns with multiple feeds. In particular, the
m-Xylene 122.479 approaches of Levy and Doherty (1986) obtained with the BVM
tert-Butyl-m-xylene 209.558 and those of Brüggemann and Marquardt (2004) using the RBM
tert-Butylbenzene 154.455 can be combined with the FAM proposed here for assessing the
Benzene 88.5091
critical feed ratio and the upper and lower bounds on energy
demand, first for non-reactive and then for reactive separations.

Table A.2
Parameters aij in Wilson model (mixtures in Sections 2.2 and 3.3).
Nomenclature
Acetic Methanol Methyl Water S-butyl
acid acetate acetate B bottoms product flow rate, kmol/s
D distillate stream flow rate, kmol/s
Acetic acid 0.0  3.4550e  1 0.0 0.0 0.0
Methanol 3.455e  1 0.0 0.0  2.0302 0.0 E extent of reaction, kmol/s
Methyl 0.0 0.0 0.0 5.4988 0.0 F feed flow rate, kmol/s
acetate h enthalpy, MJ/kmol
Water 0.0 4.90e  2 6.5378 0.0 0.0
k phase equilibrium constant
S-butyl 0.0 0.0 0.0 0.0 0.0
acetate
K chemical equilibrium constant
L liquid flow rate, kmol/s
n normal vector
p pressure, bar
Q energy duty, MW/kmol
Table A.3 T temperature, K
Parameters bij in Wilson model (mixtures in Sections 2.2 and 3.3).
V vapor flow rate, kmol/s
Acetic acid Methanol Methyl Water S-butyl x liquid mole composition in real compositional space
acetate acetate X liquid composition in transformed coordinates
z objective function
Acetic acid 0.0 3.3114e2 0.0 0.0 0.0 y vapor composition
Methanol  1.6086e2 0.0  2.3417e2 4.4868e2 0.0
Methyl 0.0  1.5765e2 0.0  2.8996e3 0.0
acetate Greek symbols
Water 0.0  2.1974e1  2.6745e3 0.0 0.0
S-butyl 0.0 0.0 0.0 0.0 0.0
acetate
u stoichiometric coefficient
a angle

Table A.4
Parameters bij in Wilson model (mixtures in Section 3.2).

ditert-Butyl-benzene m-Xylene tert-Butyl-m-xylene tert-Butyl-benzene Benzene

ditert-Butyl-benzene 0  591.5004  34.4702 80.9438  91.2584


m-Xylene 941.5922 0.0 484.7771  23.01521  544.3757
tert-Butyl-m-xylene  4.0670  369.5255 0.0  508.3478 65.0314
tert-Butyl-benzene  95.7633  47.0708 746.0344 0.0 408.7681
Benzene 10.3426 823.0789  185.8796  320.5157 0.0
176 A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177

Subscripts rp rectifying pinch points


sap saddle pinch points
B bottoms product sp stripping pinch points
D distillate product tr one tray above or below feed pinch
eq chemical equilibrium U upper feed
F feed stream
fp feed pinch Superscripts
i component index
j reaction index l liquid stream
L lower feed v vapor stream
nf feed pinch tray
Nc number of components
NI number of inert components
Acknowledgment
Table A.5
Parameters aij in UNIQUAC model (mixtures in Section 3.1). The authors gratefully acknowledge the financial support of
the ‘‘Iran Ministry of Science, Research, and Technology’’ (A.A.)
Formic acid Methanol Methyl formate Water
and the Cluster of Excellence ‘‘Tailor-Made Fuels from Biomass’’
Formic acid 0.0 0.0 0.0 0.0 funded by the Excellence Initiative of the German federal and
Methanol 0.0 0.0 0.0  1.0662 state governments (K.K. and W.M.).
Methyl formate 0.0 0.0 0.0 0.0
Water 0.0 6.437e  1 0.0 0.0

Nr number of reactions Appendix A. Physical property data

Table A.6 This appendix summarizes the phase equilibrium parameters


Parameters bij in UNIQUAC model (mixtures in Section 3.1). of the case studies, which are not available in the open literature.
All parameters are given in SI standard units.
Formic acid Methanol Methyl formate Water The Wilson model is used in the form
Formic acid 0.0 0.0 0.0 1.5496e2 0 1
Methanol 0.0 0.0  2.1582e1 4.3287e2
XNc XNc
Aji xj
lnðgi Þ ¼ 1ln@ xj Aij A PN c , ðA  1Þ
Methyl formate 0.0  2.6054e2 0.0  1.6715e2
j¼1 j¼1 k ¼ 1 Ajk xk
Water 2.3347e2  3.2213e2  2.0098e2 0.0

where
mp middle section pinch points
lnðAij Þ ¼ aij þ bij =T: ðA  2Þ

Table A.7 In the original formulation of Wilson model, the coefficients aij
Parameters r and q in UNIQUAC model (mixtures in Section 3.1). in the latter equation are calculated from
 
Components r q (¼ q0 ) Vj
aij ¼ ln , ðA  3Þ
Vi
Formic acid 1.49901 1.45376
Methanol 1.4311 1.432 for the cases in Section 3.2. The molar volumes are given in Table
Methyl formate 2.143 2.036
Water 9.2e  1 1.4
A.1. The coefficients aij and bij in the binary interaction para-
meters Aij are listed in Tables A.2–A.4.

Table A.8
Coefficients in extended Antoine equation (mixtures in Sections 2.2 and 3.3).

Components C1 C2 C3 C4 C5 C6 C7 C8 C9

Acetic acid 5.327e1  6.3045e þ3 0 0  4.2985 8.8865e  18 6.0 2.8981e2 5.9195e2


Methanol 8.2718e1  6.9045e3 0 0  8.8622 7.4664e  6 2.0 1.7547e2 5.1250e2
Methyl acetate 6.1267e1  5.6186e3 0 0  5.6473 2.1080e  17 6. 1.7515e2 5.0655e2
Water 7.3649e1  7.2582e3 0 0  7.3037 4.1653e  6 2.0 2.7316e2 6.4710e2
S-butyl acetate 5.2601e1  6.0979e3 0 0  4.2398 2.1506e  18 6.0 1.7415e2 5.6100e2

Table A.9
Coefficients in extended Antoine equation (mixtures in Section 3.1).

Components C1 C2 C3 C4 C5 C6 C7 C8 C9

Formic acid 5.0323e1  5.3782e3 0 0  4.2030 3.4697e  6 2.0 2.8145e2 5.88e2


Methanol 8.1768e1  6.8760e3 0 0  8.7078 7.1926e  6 2.0 1.7547e2 5.1264e2
Methyl formate 7.7184e1  5.6061e3 0 0  8.3920 7.8468e  6 2.0 1.7415e2 4.872e2
Water 7.3649e1  7.2582e3 0 0  7.3037 4.1653e  6 2.0 2.7316e2 6.4713e2
A. Avami et al. / Chemical Engineering Science 71 (2012) 166–177 177

The UNIQUAC model is used in the following representation: Brüggemann, S., Marquardt, W., 2004. Shortcut methods for nonideal multi-
component distillation: 3. Extractive distillation columns. AIChE J. 50,
F z y X 0 FX
lnðgi Þ ¼ ln i þ qi ln i q0i ln t 0i q0i yj tij =t0j þ li þq0i  i xl, 1129–1149.
xi 2 Fi j
xi j j j Cardoso, M.F., Salcedo, R.L., Azevedo, S.F., Barbosa, D., 2000. Optimization of
reactive distillation processes with simulated annealing. Chem. Eng. Sci. 55,
ðA  4Þ 5059–5078.
Chadda, N., Malone, M.F., Doherty, M.F., 2000. Feasible products for kinetically
where controlled reactive distillation of ternary mixtures. AIChE J. 46, 923–936.
z ¼ 10, ðA  5Þ Chin, J., Lee, J.W., 2005. Rapid generation of composition profiles for reactive and
extractive cascades. AIChE J. 51, 922–930.
X Ciric, A.R., Gu, D., 1994. Synthesis of nonequilibrium reactive distillation processes
t 0i ¼ y0k tki , ðA  6Þ by MINLP optimization. AIChE J. 40, 1479–1487.
k Dragomir, R.M., Jobson, M., 2005. Conceptual design of single-feed kinetically
controlled reactive distillation columns. Chem. Eng. Sci. 60, 5049–5068.
z Dragomir, R.M., Jobson, M., 2004. Conceptual design of reactive distillation
li ¼ ðr q Þ þ 1r i , ðA  7Þ
2 i i columns with non-reactive sections. In: Barbosa-Póvoa, A., Matos, H., (Eds.),
X Proceedings of the 14th European Symposium on Computer-Aided Process
Fi ¼ ri xi = r k xk , ðA  8Þ Engineering, 37th European Symposium of the Working Party on Computer-
k Aided Process Engineering, Lisbon, pp. 385–390.
Gadewar, S.B., Malone, M.F., Doherty, M.F., 2007. Feasible products for double-feed
X reactive distillation columns. Ind. Eng. Chem. Res. 46, 3255–3264.
y0i ¼ q0i xi = q0k xk , ðA  9Þ Gangadwala, J., Kienle, A., 2007. MINLP optimization of butyl acetate synthesis.
k Chem. Eng. Process. 46, 107–118.
X Harwardt, A., Kraemer, K., Rüngeler, B., Marquardt, W., 2011. Conceptual design of
yi ¼ qi xi = qk xk , ðA  10Þ a butyl-levulinate reactive distillation process by incremental refinement.
k Chin. J. Chem. Eng. 19, 371–379.
Hauan, S., Westerberg, A.W., Lien, K.M., 2000. Phenomena-based analysis of fixed
lnðtij Þ ¼ aij þ bij =T: ðA  11Þ points in reactive separation systems. Chem. Eng. Sci. 55, 1053–1075.
Huss, R.S., Chen, F.R., Malone, M.F., Doherty, M.F., 2003. Reactive distillation for
The parameters aij and bij in Eq. (A-11) are given in Tables A.5 methyl acetate production. Comput. Chem. Eng. 27, 1855–1866.
and A.6 while parameters r and q are given in Table A.7. Janssen, A., Muether, M., Pischinger, S., 2010. Potential of cellulose-derived
biofuels for soot free diesel combustion. SAE Int. J. Fuels. Lubr. 3, 70–84.
The vapor pressure is computed from an extended Antoine Kenig, E.Y., Bäder, H., Górak, A., Bessling, B., Adrian, T., Schoenmakers, H., 2001.
equation given as Investigations of ethyl acetate reactive distillation process. Chem. Eng. Sci. 56,
6185–6193.
C 2i
lnðpni Þ ¼ C 1i þ þ C 4i T þC 5i ln T þ C 6i T C 7i for C 8i r T r C 9i : Kraemer, K., Harwardt, A., Skiborowski, M., Mitra, S., Marquardt, W., 2011.
T þ C 3i Shortcut-based design of multicomponent heteroazeotropic distillation. Chem.
ðA  12Þ Eng. Res. Des. 89, 1168–1189.
Lee, J.W., Brüggemann, S., Marquardt, W., 2003. Shortcut method for kinetically
The parameters for the mixtures in Sections 2.2 and 3.3 are controlled reactive distillation systems. AIChE J. 49, 1471–1487.
given in Tables A.8 and A.9. Levy, S.G., Doherty, M.F., 1986. Design and synthesis of homogenous azeotropic
distillations 4. Minimum reflux calculations for multiple-feed columns. Ind.
Eng. Chem. Fundam. 25, 269–279.
References Lucia, A., Amale, A., Taylor, R., 2008. Distillation pinch points and more. Comput.
Chem. Eng. 32, 1342–1364.
Marquardt, W., Kossack, S., Kraemer, K., 2008. A framework for the systematic
Agreda, V.H., Partin, L.R., Heise, W.H., 1990. High-purity methyl acetate via
design of hybrid separation processes. Chin. J. Chem. Eng. 16, 333–342.
reactive distillation. Chem. Eng. Prog. 86, 40–46.
Stichlmair, J., Frey, T., 2001. Mixed-integer nonlinear programming optimization
Barbosa, D., Doherty, M.F., 1988a. Design and minimum-reflux calculations for
of reactive distillation processes. Ind. Eng. Chem. Res. 40, 5978–5982.
single feed multicomponent reactive distillation columns. Chem. Eng. Sci. 43,
Sundmacher, K., Kienle, A., 2002. Reactive Distillation—Status and Future Direc-
1523–1537.
Barbosa, D., Doherty, M.F., 1988b. Design and minimum-reflux calculations for tions. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.
double-feed multicomponent reactive distillation columns. Chem. Eng. Sci. 43, Suzuki, I., Komatsu, H., Hirata, M., 1970. Formulation and prediction of quaternary
2377–2389. vapor–liquid equilibria accompanied by esterification. J. Chem. Eng. Jpn. 3,
Barbosa, D., Doherty, M.F., 1988c. The simple distillation of homogeneous reactive 152–157.
mixtures. Chem. Eng. Sci. 43, 541–550. Taylor, R., Krishna, R., 2000. Review: modelling reactive distillation. Chem. Eng.
Bart, H.J., Reidetschlager, J., Schatka, K., Lehmann, A., 1994. Kinetics of esterifica- Sci. 5, 5183–5229.
tion of levulinic acid with n-butanol by homogeneous catalysis. Ind. Eng. Ung, S., Doherty, M.F., 1995a. Vapor–liquid phase-equilibrium in systems with
Chem. Res. 33, 21–25. multiple chemical-reactions. Chem. Eng. Sci. 50, 23–48.
Bausa, J., Watzdorf, R.V., Marquardt, W., 1998. Shortcut methods for non-ideal Ung, S., Doherty, M.F., 1995b. Synthesis of reactive distillation systems with
multicomponent distillation: 1. Simple columns. AIChE J. 44, 2181–2198. multiple equilibrium chemical reactions. Chem. Eng. Sci. 34, 2555–2565.
Bausa, J., Marquardt, W., 2000. Quick and reliable phase stability test in VLLE flash Urdaneta, R.Y., Bausa, J., Marquardt, W., 2004. Minimum energy demand and split
calculations by homotopy continuation. Comput. Chem. Eng. 24, 2447–2456. feasibility for a class of reactive distillation columns. In: Barbosa-Póvoa, A.,
Bausa, J., 2001. Näherungsverfahren für den konzeptionellen Entwurf und die Matos, H. (Eds.), Proceedings of the 14th European Symposium on Computer-
thermodynamische Analyse von destillativen Trennprozessen. Ph.D. Thesis. Aided Process Engineering, 37th European Symposium of the Working Party
VDI Verlag, Düsseldorf. on Computer-Aided Process Engineering, Lisbon, pp. 517–522.
Bessling, B., Schembecker, G., Simmrock, J.H., 1997. Design of processes with Werpy, T., Petersen, G., 2004. Top Value Added Chemicals from Biomass. NREL,
reactive distillation line diagrams. Ind. Eng. Chem. Res. 36, 3032–3042. PNNL Report.
Bessling, B., Löning, J., Ohligschläger, A., Schembecker, G., 1998. Sundmacher K. Yeomans, B., 1982. Process for the Production of Methyl Acetate by Esterifying
Investigations on the synthesis of methyl acetate in a heterogeneous reactive Methanol with Acetic Acid, EP19820301331.
distillation process. Chem. Eng. Technol. 21, 393–400.

You might also like