Isophorone

Chemical Engineering Journal 207208 (2012) 349367
Contents lists available at SciVerse ScienceDirect
Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej
Isophorone reactor: Modelling and performance enhancement

Pranay J. Darda , Vivek V. Ranade
Chemical Engineering and Process Development Division, National Chemical Laboratory, Pune 411 008, India
h i g h l i g h t s
" Production of isophorone incorporating multiple reactions was modelled.
" A dynamic non-equilibrium model of reactive distillation was developed.
" The model was used to understand inuence of various parameters on a key mass ratio.
" Simulated results will be useful for improving performance of isophorone reactor.
a r t i c l e
i n f o
Article history:
Available online 15 July 2012
Keywords:
Isophorone
Reactive distillation
Model
Performance enhancement
a b s t r a c t
The production of isophorone in a reactive distillation column was modelled. Mass transfer and heat
transfer accompanied by chemical reactions were incorporated in a dynamic non-equilibrium multistage
model. A base case simulation was identied and established. The base case simulation showed the value
of key performance indicator that is mass ratio of acetone consumed to isophorone produced as 1.54. The
computational model was then used to understand and to evaluate inuence of various key design and
operating parameters on this key performance indicator. Inuence of reactor volume, mass transfer coefcient and feed composition, location and temperature was evaluated. The simulated results indicate that
the key performance indicator can be improved (reduce acetone consumed per isophorone produced) by
manipulating acetone content in the feed and location of the feed. The presented results will be useful for
realising performance enhancement of isophorone and other similar reactors.
2012 Elsevier B.V. All rights reserved.
1. Introduction
Isophorone is an important industrial chemical having applications in inks, paints, lacquers, adhesives, vinyl resins, copolymers,
coatings, nishes, and pesticides. It is also used as a chemical intermediate for organic synthesis. Isophorone is commercially produced [1] by alkali catalysed condensation of acetone at high
pressure and high temperature. Aldol condensation of acetone is a
complex reaction [28] involving the feedstock: acetone, water
and hydroxide catalyst; several intermediate products (diacetone
alcohol, mesityl oxide), byproducts (oxylitone and heavies) and
the desired product of the reactionisophorone. Acetone condensation for isophorone production can be commercially feasible, if the
specicity of the reaction is controlled to achieve a high selectivity
towards isophorone. One of the key indicators of the plant performance in isophorone production is the overall mass ratio of
acetone consumed to isophorone produced. In this work, a mathematical model of isophorone reactor is developed and used for
Corresponding author. Tel.: +91 9096925842.

E-mail address: pranay.darda@gmail.com (P.J. Darda).
1385-8947/$ - see front matter 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cej.2012.06.137
evolving guidelines for enhancing performance of isophorone

reactors.
Isophorone is usually produced by carrying out acetone condensation in single or multiple distillation columns. The reactive distillation facilitates formation and separation of isophorone and
intermediate products of the reaction including diacetone alcohol,
mesityl oxide and xylitones. It is essential to understand the key
aspects of isophorone production process and of reactive distillation for manipulating selectivity towards isophorone and reducing
the acetone consumption. In this section we address the issues pertinent to isophorone production via reactive distillation from a
process perspective.
Issues relevant to isophorone production have been addressed in
various patents, which describe the reactor conguration, reaction
conditions and nature of contacting among various reactants for
catalysed condensation of acetone. Walton and Yeomans [1] outline
a process for isophorone production using a potassium hydroxide
catalyst in multiple distillation columns consisting of a reaction
and a hydrolysis column (Fig. 1). The temperature in the reaction
and hydrolysis columns is in the range of 150250 C and pressure
is in the range of 1055 bar. The feed consisting of 6585% w/w acetone, 3515% w/w water and 0.70.3% w/w potassium hydroxide is
350
P.J. Darda, V.V. Ranade / Chemical Engineering Journal 207208 (2012) 349367
Nomenclature
Total and component molar balance
MLj
total number of liquid phase moles on stage j (kmol)
MVj
total number of vapour phase moles on stage j (kmol)
MLi;j
total number of liquid phase moles of ith component on

stage j (kmol)
total number of vapour phase moles of ith component
on stage j (kmol)
MVi;j
Initial molar holdup and stage accumulation ratios

ILj
total initial number of liquid phase moles on stage j
(kmol)
IVj
total initial number of vapour phase moles on stage j
(kmol)
V
ft
time dependent functional multiplier for ratio of liquid
phase molar output ow to molar accumulation rate
(function of time)
L
ft
time dependent functional multiplier for ratio of vapour
phase molar output ow to molar accumulation rate
(function of time)
Output=Accumulation
LRatioj
ratio of output liquid molar ow to molar
accumulation rate on stage j ()
Output=Accumulation
VRatioj
ratio of output vapour molar ow to molar
accumulation rate on stage j ()
MaxVol=InitialVol
LRatioj
ratio of maximum liquid stage volume to
initial volume on stage j ()
MaxVol=InitialVol
VRatioj
ratio of maximum vapour stage volume to
initial volume on stage j ()
t
simulation time (h)
Molar ows
L(j)
total liquid phase molar ow originating from stage j
(kmol h1)
L(j1)
total liquid phase molar ow originating from stage
j 1 (kmol h1)
total vapour phase molar ow originating from stage j
V(j)
(kmol h1)
V(j+1)
total vapour phase molar ow originating from stage
j + 1 (kmol h1)
F Lj
total molar ow of liquid feed entering stage j
(kmol h1)
F Vj
total molar ow of vapour feed entering stage j
(kmol h1)
R
reux ratio ()
Gas liquid mass transfer
Adiff
parameter A corresponding to component i ()
i
Bdiff
i
parameter B corresponding to component i ()
Hdiff
i
W diff
i
parameter H corresponding to component i ()
MW(i)
molecular weight of component i (kg kmol1)
T Critical
i
critical temperature of component i (K)
Tr(i,j)
DL(i,j)
reduced temperature of component i on stage j ()

liquid phase diffusion coefcient of component i on
stage j (m2 s1)
packing area for structured packing/interfacial area per
unit liquid volume (m2 m3)
factor for structured packing ()
fractional liquid phase holdup on stage j ()
characteristic length: side dimension of the corrugation
cross section (m)
effective liquid velocity in the column on stage j (m s1)
a
CE
holdupL(j)
S(j)
ULE(j)
parameter W corresponding to component i ()
ULS(j)
supercial liquid velocity in the column on stage j

(m s1)
void fraction of the structured packing ()
channel ow angle of the structured packing ()
liquid phase viscosity on stage j (Pa s)
e
h
lLj
Mass Flowj liquid phase mass ow on stage j (kg h1)

V
Mass Flowj vapour phase mass ow on stage j (kg h1)
qLj
qVj
liquid phase density on stage j (kg m3)
Abubbling
j
bubbling area of the tray on stage j (m2)
vapour phase density on stage j (kg m3)
heightw(j) exit weir height of stage j (m)

pitch(j)
hole pitch of stage j (m)
Ahole
hole area of the sieve tray on stage j (m2)
j
L
heightj clear liquid height on stage j (m)

supercial vapour velocity on stage j (m s1)
uSV(j)
r(j)
surface tension of the liquid phase on stage j (N m1)
ahfj
interfacial area per unit bubbling area on stage j
(m2 m2)
Distribution
K i;j
gasliquid distribution coefcient of component i on
stage j ()
cLi;j
liquid phase activity coefcient of component i on stage
j ()
kL(i,j)
liquid side mass transfer coefcient of component i on
stage j (m s1)
kLa(i,j)
overall liquid side mass transfer coefcient of component i on stage j (h1)
GL
Ni;j
molar gasliquid mass transfer rate of component i on
stage j (kmol h1)
Reaction parameters
z(i,k)
stoichiometric coefcient of component i in the kth reaction ()
b(k)
base stoichiometric coefcient for the kth reaction ()
k0(k)
pre-exponential
factor
of
the
kth
reaction
((m3 kmol1)(order of reaction1) h1)
activation energy of the kth reaction (kJ kmol1)
E(k)
o(i,k)
order of the kth reaction for component i ()
k(k)
rate
constant
of
the
kth
reaction
((m3 kmol1)(order of reaction1) h1)
equilibrium constant of the k/k + 1 reaction ()
K(k,k+1)
Mole fractions
x(i,j)
liquid phase mole fraction of component i on stage
j ()
x(i,j1)
liquid phase mole fraction of component i on stage
j 1 ()
xInterface
interface liquid phase mole fraction of component i on
i;j
stage j ()
xinitial
initial liquid phase mole fraction of component i on
i;j
stage j ()
y(i,j)
vapour phase mole fraction of component i on stage j ()
y(i,j+1)
vapour phase mole fraction of component i on stage
j + 1 ()
yInterface
interface vapour phase mole fraction of component i on
i;j
stage j ()
yinitial
initial vapour phase mole fraction of component i on
i;j
stage j ()
zLi;j
liquid phase mole fraction of component i fed to stage
j ()
zVi;j
vapour phase mole fraction of component i fed to stage
j ()
hj
Molar volume and volumetric holdups

lLi
liquid molar volume of component i (m3 kmol1)
3
lVi
1
vapour molar volume of component

i (m kmol
3
AVj
V
T
T Interface
j
T LB
j
bubble point temperature of liquid phase on stage j (C)
T VD
j
dew point temperature of vapour phase on stage j (C)
total liquid volume of stage j (m )3
hj
V
Volj
total vapour volume of stage j (m )
eL
H
i;j
Maximum dened volumetric holdup on a stage

VmaxLj maximum liquid volumetric holdup on stage j (m3)
VmaxVj
maximum vapour volumetric holdup on stage j (m3)
Weir ow calculations
heightow(j) height of the clear liquid owing over the weir (m)
Q Lj
liquid phase volumetric ow originating from stage j
(m3 h1)
lengthw(j) weir length (m)
g
acceleration due to gravity (m s2)
L
dVolumej
dt
liquid phase volumetric accumulation on stage j

(m3 h1)
liquid phase heat transfer coefcient on stage j

(kJ h1 m2 C1)
vapour phase heat transfer area on stage j (m2)
vapour phase heat transfer coefcient on stage j
(kJ h1 m2 C1)
specic liquid phase enthalpy of component i on stage j
(kJ kmol1)
specic vapour phase enthalpy of component i on stage j
(kJ kmol1)
temperature (C)
gasliquid interface temperature on stage j (C)
Volj
351
eV
H
i;j
T
T Lj
reference temperature (C)

liquid phase bulk temperature on stage j (C)
T Vj
vapour phase bulk temperature on stage j (C)
C Li
Energy balance
ELj
liquid phase energy content of stage j (kJ)
C Vi
EVj
vapour phase energy content of stage j (kJ)
ELj
PTotal
j
total pressure on stage j (bar)
PSi;T
HLf
j
heat transfer rate across liquid phase/gasliquid

interface on stage j (kJ h1)
heat transfer rate across vapour phase/gasliquid
interface on stage j (kJ h1)
liquid phase enthalpy fed to stage j (kJ kmol1)
DHS
kAi
liquid phase heat capacity of component i (kJ

kmol1 C1)
vapour phase heat capacity of component i
(kJ kmol1 C1)
heat of solution (kJ kmol1)
heat of vapourisation of component i (kJ kmol1)
HVf
j
vapour phase enthalpy fed to stage j (kJ kmol1)
OPLj
liquid phase energy withdrawn from stage j (kJ h1)
vapour pressure
(mm Hg/bar)
Antoine constant
Antoine constant
Antoine constant
IPVj
vapour phase energy introduced to stage j (kJ h1)
EVj
QLostLj
QLostVj
ALj
liquid phase heat losses from stage j (kJ h
1
1
vapour phase heat losses from stage j (kJ h )

liquid phase heat transfer area on stage j (m2)
introduced to an intermediate point in the reaction column at such

a rate so as to maintain an acetone/water azeotrope reux having a
potassium hydroxide concentration in the range of 3001000 ppm
within the reaction column. The stream from the bottoms of the
reaction column is transferred to the hydrolysis column where acetone is separated overhead and returned to the reaction column,
removing the fraction containing water, isophorone and high boiling compounds from the bottom of the hydrolysis column and
thereafter recovering isophorone from the bottom fraction.
Isophorone production in a distillation column reactor has also
been described by Schmitt et al. [9], which consists of distilling
acetone under pressure in counter current contact with alkali solution in a reaction zone between points of introduction of alkali and
acetone to form acetone condensation reaction products. Acetone
distilled is recovered from the head of zone in the liquid form. Part
of the acetone condensed is reuxed to the head and remaining
recycled to the introduction point. The distillation column reactor
is divided into an upper distillation zone, median condensation
(reaction) zone and lower product purication zone. Alkali solution
is introduced at the boundary of the distillation/reaction zone;
while acetone is introduced at the boundary of the reaction/purication zone. The pressure in the reaction zone is adjusted to maintain a temperature between 140250 C. The concentration of
alkali in the alkali solution is in the range of 0.020.1% by weight
A(i)
B(i)
C(i)
of component i at temperature T
A of component i ()
B of component i ()
C of component i ()
Abbreviations
NC
number of components
RatioA/IPH key mass ratio of acetone consumed to isophorone
produced
AvgVolL average liquid stage volume
with the solution containing water or even lower alkanols. Typically a temperature gradient exists in the column so that the temperature in the lower purication section (231 C) exceeds the
temperature in the median reaction section (210 C), while temperature in the reaction section is higher than the temperature in
the upper distillation section (200 C). A pressure of 27 atm is
maintained in the column while a pressure of 1516 atm is maintained in the head of the distillation column reactor. Alkali solution
and acetone condensation reaction products formed as a consequence of the reaction are collected in the sump. The alkali solution
collects as the lower aqueous phase while the condensation reaction products collect in the sump as the upper organic phase. Removal of a fraction of water from the aqueous phase and loss of
alkali due to neutralisation may be made up by fresh alkali addition to the recycle stream introduced at the boundary of the reaction and distillation zones. Acetone consumed in the reaction zone
is also made up by fresh acetone introduced at the boundary of the
reaction and purication zones. The acetone condensation products consist of isophorone and high boiling reaction products.
The objective of the design is to minimise the formation of undesired products consisting of mesityl oxide, phorone, b isophorone
and xylitones. Alkali in lower phase of the sump aids in hydrolysis,
yielding isophorone and minor portion of the higher boiling over
condensates. The condensation reaction products recovered from
352
Reflux
Condenser
Recycle to
Reaction Column
Reflux
Feed
Liquid
Flow
Stage: Gas Phase

Stage: Liquid Phase
Height of Clear
Liquid Flowing
Over Weir
Weir
Weir
Gas
Liquid
Mass
Transfer
Gas
Flow
Hydrolysis
Column
(30 stages)
Reaction
Column
(11 stages)
Bottoms
Stage
Steam Feed
Product of
Reaction Column
Product of
Hydrolysis Column
Fig. 1. Schematic representation of the isophorone production process using a reactive distillation column (adapted from [1]).
the sump of the zone may be subject to vacuum distillation to recover pure isophorone.
Production of isophorone via alkali catalysed condensation of
acetone (for example see [1,9]) can lead to colour forming compounds and the unwanted isomer: b isophorone. Cane and Yeomans [10] address the issue of colour formation and claim a
process, which incorporates removal of a side-stream from hydrolysis section of the column. Colour forming compounds are concentrated at a point in the column where acetone content is stripped
to a relatively low level. The side stream is distilled to obtain acetone/water in the distillate product, while the residue is treated
with a strong mineral acid like toluene-p sulphonic acid, which
decomposes the contaminants into high boiling compounds. The
treated residue is further distilled to obtain isophorone in the distillate (containing 9% w/w of unwanted isomer: b isophorone).
Additionally the emulsied base product from the hydrolysis section is treated with a strong mineral acid like phosphoric acid to reduce the pH in the range of 6.57.5. Neutralisation of the crude
isophorone stream from the bottom of the hydrolysis section reduces the alkali metal ion content signicantly. The acid treated
stream is decanted and the isophorone containing phase is fed to
distillation column for water/crude isophorone separation. Crude
isophorone is further distilled to obtain pure isophorone in the distillate. The nal product contains about 0.9% w/w of b isophorone.
In contrast to multiphase (vapour/liquid) processes for isophorone production in reactive distillation columns, a process for
isophorone production in vapour phase via aldol condensation of
acetone has been claimed by Grebinoski et al. [11]. The objective
of their invention is to improve efciency in product yield and energy consumption. The process consists of feeding fresh acetone
mixed with various process recycle streams made up of acetone,
mesityl oxide and isophorone. Recycled by product streams
either equilibrate or convert to the desired productisophorone
thereby increasing the overall process yield of isophorone from
acetone. Vaporized and superheated stream is fed to the reactor
at 225325 C. The superheated feed reacts in vapour phase
converting 1035% of acetone. Initial reaction product consists of
420% of desired product: a isophorone. The catalyst used for the
process has been described by Schutz [12] and Schutz and Cullo
[13]. Unreacted acetone is separated from reaction products and
recycled to the reactor. Several distillation columns are used to
separate acetone/organic components/water, mesityloxide/water/
reaction products, phorone/b isophorone/mesitylene/a isophorone
and a isophorone/heavies.
The objective of improving process efciency in isophorone
production has also been investigated by Cook [14]. Cook [14]
highlights the strategies of reduce/reuse/recycle applied to the
isophorone production process. According to Cook [14], the
resources that can possibly be affected include: raw materials

(acetone and caustic catalyst), energy (natural gas, steam, electricity, demineralised water, cooling water, chilled water, instrument
air and nitrogen) and waste treatment (waste water and organic
waste). Strategies of reduce/reuse/recycle can make a nancial
contribution to the overall process, while reducing the burden on
infrastructure and communities. Cook [14] indicates that valuable
recommendations can be generated for any additional problems in
the process including: inefciencies in heat transfer systems and
aws in the mechanical systems of the process. Among the three
strategies, the concept of reduction is most sustainable. Reduction
leads to lower cost of production, improvement in the raw material
yield (inversely related to production rates) and optimisation of
utilities (via a data acquisition system). The strategy of reduction
as applicable to the isophorone production in terms of reducing
raw material and energy requirements is highlighted next.
Strategy 1.1: reducing the raw material requirements of the
isophorone process: The reduction in raw material can be achieved
by optimising the operating conditions of isophorone production
process. A typical isophorone process is operated around at least
two capacity levels: lower feed rate (e.g. 95% of nameplate capacity) and higher feed rate (excess design capacity by 14%). The strategy of reduction can lead to lowest specic production cost for
multiple raw material (acetone) feed rates. Various experiments
can be done by choosing appropriate variables including: (I) acetone feed rate, (II) energy input to the reboiler, (III) ow of dilute
caustic catalyst and (IV) concentration of the dilute caustic catalyst. Cook [14] also indicates that the approach of PlackettBurman
design of experiments can be applied in case of isophorone production process, which requires testing operating conditions at a high
and low limit of each variable. Each trial can constitute to a combination of the upper/lower limit of the operating variables. Samples can be collected at regular intervals of few hours, while
variables like: (1) ow, (2) temperature and (3) pressure can be updated using a continuous process data historian. Collected samples
can be analyzed for quantity of (1) isophorone in the crude organic
product from the separator and (2) the caustic concentration in the
wastewater stream. The experimental data can also be collected for
the following additional variables: (1) ow of waste water stream,
(2) feed of crude stream to extraction and (3) export steam ow.
The variables of the experimental study can be used as an input
to a cost model in order to analyze the impact of changes in the
variables on the specic cost of production. The specic variable
cost of isophorone production can be valued at: cost of acetone + cost of caustic + cost of natural gas revenue generated
from steam. Various trials in the study can be evaluated from the
cost model, and the trials with the lowest specic variable cost
can be assigned a higher ranking in terms of desired operating conditions. Reduction studies at the Degussa isophorone production
plant reported by Cook [14] have shown that reaction system
achieves higher yields at lower acetone feed rate, while annual cost
savings of up to 5.4% can be achieved. It is expected that numerous
possible operational problems including: (1) level control and
ooding in the reaction system, (2) pressure swings in the reux
drum and (3) problems in nned tube steam condenser can arise
in the process. These problems can be analyzed for generating design recommendations for enhancing the performance of the
isophorone production. The strategy of reduction can also be applied to energy consumption and this strategy is described next.
Strategy 1.2: reducing energy requirements of isophorone process: reducing energy requirements of the isophorone process
can lead to economical ways for maximising heat recovery while
minimising external utility requirements. A critical aspect of the
strategy is continuous feedback/interaction with the operators of
the unit/plant regarding the utility consumption. One of the possible outcome of reducing energy methodology is to analyze the
353
impact of actions of the operators at various capacity utilisation

(production rates), on the cost of the nal product. Utilities under
consideration include (1) natural gas, (2) steam, (3) electricity, (4)
cooling water, (5) chilled water and (6) nitrogen. The methodology
includes testing the following utilities: (1) natural gas, (2) steam
(high, medium and low pressure) and (3) electricity, since cost
associated with these utilities are relatively higher. The study of
energy reduction requires a continuous monitoring of the utility
consumption and hence continuous process data transfer from
DCS to data acquisition software is essential. The methodology of
reducing energy also includes comparison of the actual utility consumption to a standard/PFD consumption rate. Performance
indices can be created for each utility and each user/operator.
Overall utility performance index for each utility includes all the
equipment in the process and is dened as the ratio of total actual
consumption to total PFD consumption. Utility consumption rate
can be compared at various capacity utilisation rates in order to
capture its economical impact. When the utility consumption rate
is adjusted for utilisation rates, the plot of specic consumption as
a function of capacity utilisation gives an idea about the inefciencies in the process at lower utilisation rates and the start up/standby costs associated with the various utilities of the process.
Apart from reduction, strategies of reuse and recycle can also be
applied for recovering caustic catalyst and water from the waste
streams. Cook [14] indicates that Degussa studies explored the
possibility of using ash distillation process for recovery. However
the cost of process steam exceeded the saving potential. Favorable
costing and environmental considerations can possibly make the
recycle and reuse process of ash distillation more viable.
Implementation of strategies described by Cook [14] at the
commercial scale necessitates experimental trials at the actual
plant, which can be resource intensive (e.g. at high temperature/
pressure) and may not be feasible. A viable alternative to the plant
trials is development of computational models for simulating
isophorone reactive distillation process and using the model for
evolving performance enhancement guidelines. Modelling approach used in this work is described in the next section.
2. Modelling approach
Modelling of the reactive distillation process is fairly well established and is usually carried out by setting up equations which constitute a system of differential algebraic equations (DAEs). A
number of authors in the literature have addressed modelling issues
in reactive distillation. Non-equilibrium (NEQ) models of the reactive distillation process have been described by Taylor and Krishna
[15], Katariya et al. [16], Xu et al. [17] and Sundmacher and Hoffmann [18,19]. Alejski and Duprat [20] describe an equilibrium stage
model, while Hupen and Kenig [21] have developed a general rate
based model for absorption operations. Additional modelling works
include approximation of mixing effects on the distillation plate by
Alejski [22], minimisation method for solution of the reactive distillation system by Alejski et al. [23], steady state model by Qi et al.
[24] and a model used for control of reactive distillation column
by Gruner et al. [25]. Various modelling frameworks formulated in
the literature are classied in this study as non-equilibrium reactive
distillation models and other models and these are described next
along with the model developed in this study.
2.1. Non-equilibrium models of reactive distillation
Various models including the conventional non-equilibrium
(NEQ) model, NEQ model for homogenous and heterogeneous systems and the dynamic NEQ cell model have been described by Taylor and Krishna [15]. In the cell model, ow patterns on distillation
354
trays are modelled by choosing an appropriate number of cells in

each ow direction. Each cell is modelled essentially using the
NEQ model for heterogeneous systems. The bulk uid phases are
assumed to be completely mixed. Mass transfer resistances are located in lms near the vapourliquid and liquidsolid interfaces,
and the MaxwellStefan equations are used for calculation of the
mass transfer rates across each lm. Thermodynamic equilibrium
is assumed only at the vapourliquid interface. Taylor and Krishna
[15] indicate that the mass transfer inside the porous catalyst can
be described with the dusty uid model. The set of equations in the
model consists of dynamic liquid and vapour phase component
and overall molar balances, overall enthalpy balances, Maxwell
Stefan relations for mass transfer, energy transfer terms based on
temperature gradient and molar uxes, and vapourliquid equilibrium relations. Similar modelling frameworks of reactive distillation are described by Krishnamurthy and Taylor [26], Taylor and
Krishna [27], Baur [28] and Baur et al. [29] and Katariya et al. [16].
Katariya et al. [16] have used a non-equilibrium (NEQ) stage
model for reactive distillation by taking into account overall and
component liquid and vapour molar balances, overall liquid and
vapour energy balances and vapour liquid equilibrium at the interface. The mass transfer uxes are calculated using Ficks law, while
the NEQ model equations are implemented in large scale equation
oriented simulator: DIVA.
Three phase non-equilibrium models have also been used by Xu
et al. [17] and Sundmacher and Hoffmann [18]. Xu et al. [17] have
demonstrated the application of a dynamic three phase non-equilibrium (NEQ) model of catalytic distillation process (batch and
continuous) for simulating diacetone alcohol production. The three
phase NEQ model developed by Xu et al. [17] accounts for mass
transfer and heat transfer in vapour/liquid and solid/liquid phases
based on MaxwellStefan mass transfer theory. The reaction system studied is aldol condensation of acetone to produce diacetone
alcohol using Amberlite IRA-900 anion exchange resin as a catalyst.
In the aldol condensation reaction, diacetone alcohol can undergo
hydration to produce mesityl oxide and water. Aldol condensation
also has unfavourable equilibrium towards production of diacetone alcohol. Equilibrium conversion at the normal boiling point
of acetone (54 C) is only 4.3 wt%. The catalytic distillation process
greatly increases conversion by insitu removal of diacetone alcohol
from the reaction zone. The model consists of overall liquid/vapour
phase molar balance, overall liquid/vapour phase heat balance, and
vapour/liquid and solid/liquid phase heat and mass transfer rates.
Batch and continuous catalytic distillation were simulated using
the NEQ model. In case of continuous catalytic distillation, predicted and experimental mole fractions of acetone, diacetone alcohol and mesityl oxide were compared over time. Variation of liquid
phase temperature in reboiler with time was also investigated.
Sensitivity analysis was done by varying the kinetics and solid/liquid mass transfer in the three phase NEQ model. Dynamic response of diacetone alcohol/mesityl oxide liquid phase mole
fractions in the reboiler, to a negative step change in reboiler duty
were studied as a function of time in order to interpret the results
of sensitivity analysis. Sensitivity analysis indicated that diacetone
alcohol formation is controlled by solidliquid mass transfer, while
mesityl oxide formation is kinetically controlled under simulation
conditions.
Adding to the contribution of Xu et al. [17], Sundmacher and
Hoffmann [18] describe a detailed three phase non-equilibrium
model for packed catalytic distillation column. The model addresses heterogeneous chemical reaction accompanied by multicomponent mass transport phenomena between the liquid phase
and the catalytic packing elements. Mass transfer at the vapourliquid interface is characterised by the explicit solution of the generalised MaxwellStefan equations. The model also takes into
account liquid and vapour phase component balances with reac-
tion occurring in the liquid phase, and overall mass and enthalpy
balances. Vapourliquid interface mole fractions are related via
UNIQUAC model dened activity coefcients and distribution coefcients. Accumulation terms were introduced in the liquid and vapour phase component balances and the resulting system of
differential algebraic equations (DAEs) was solved using the solver:
LIMEX. Sundmacher and Hoffmann [19] also describe a steady
state non-equilibrium model for packed reactive distillation column. The model consists of molar balances for the condenser, column and the reboiler. Column balances consist of component and
overall vapour and liquid molar balances. Mass transfer in the liquid and vapour phase is driven by liquid and vapour phase mole
fraction differential. Equilibrium conditions were dened using
activity coefcient based distribution coefcients, with the activity
coefcients dened using UNIQUAC method. The model system
was solved using a relaxation method which gave rise to a system
of differential algebraic equations (DAEs). Resulting DAEs were
solved using extrapolation solver: LIMEX.
2.2. Other models
Amongst other models, the general rate based model applicable
for a broad spectrum of NOx absorption operations described by
Hupen and Kenig [21] consists of balances in the bulk liquid/gas
phases, liquid/gas lms, and in the gas area. The bulk liquid balance takes into account axial change along with mass transfer
and chemical reactions. The differential heat balance of the liquid
bulk takes into account heat ux and cooling effects. Balances
are also described for the gas bulk. A differential component balance in the liquid lm accounts for mass transfer with chemical
reactions resulting in an effective diffusivity description of the
component uxes. The differential energy balance for the lm accounts for heat ux including the conductive and convective terms.
In addition to liquid/gas bulk and lm balances, special gas area
equations describing the change in gas composition, reactions,
and energy balances are also incorporated in the model. The system of equations is discretised in axial and normal co-ordinates
to yield an algebraic system of equation which were implemented
in Aspen Custom Modeler.
Contrary to non-equilibrium models summarised in Section 2.1,
Alejski and Duprat [20] describe an equilibrium stage model, which
includes the component and total balances for condenser/partial
condenser, plates and reboiler. Total energy balance along with vapourliquid equilibrium and plate efciency complete the description of the model. The model was solved using Gears method for
stiff systems with automatically chosen order and length of the
integration step. In addition to the aforementioned equilibrium
model [20], the critical issues of approximating the mixing effects
on the distillation plate by a cascade of stirred cells and application
of minimisation method for solution of the reactive distillation
system were addressed by Alejski [22] and Alejski et al. [23]
respectively.
Relevant models of reactive distillation have also been described by Qi et al. [24] and Gruner et al. [25]. Qi et al. [24] describe
a steady state model for packed reactive distillation column, while
the problem of control in reactive distillation is addressed by Gruner et al. [25].
2.3. Model development
In this work, the dynamic non-equilibrium (NEQ) model was
developed for a multi-stage reactive distillation system including
multiple components, multiple chemical reactions and appropriate
heat transfer and mass transfer terms.
Although the interest of the present work was to study steady
state performance of reactive distillation system, a dynamic model
355
was chosen to have the possibility of carrying out transient simulations and also for numerical reasons. Solution of non-linear algebraic equations describing a steady state model, using a solver like
Sundials-2.4.0, requires good initial guesses for stable region of
convergence. Convergence issues can be better handled by dening
the model system as a time dependentdynamic model, which can
be solved using a suitable ordinary differential equation solver or a
solver for differential/algebraic equations in Sundials-2.4.0.
The reactive distillation model consisted of the following balances (Appendix A): (1) overall molar balances, (2) component molar balances, (3) component mole fractions, (4) liquid/vapour phase
volumetric hold up on each stage and (5) overall energy balances.
Additional balances equations were also formulated for: (1) equilibrium and forward reaction rates, (2) multi-component mass
and heat transfer rates across gasliquid interface using a lm
model, (3) dew point, bubble point and interface temperatures
and (4) temperature of the bulk liquid and vapour phases along
various stages using stage enthalpies. The various components
modelled in the reactive distillation system included: water, acetone, isophorone, hydroxide (catalyst), mesityl oxide, isoxylitone
and diacetone alcohol (7 components).
The input to the model consisted of the following parameters:
(1) liquid and vapour phase molar volumes of various components,
(2) initial liquid and vapour phase mole fractions of various components, (3) overall liquid side mass transfer coefcients, (4) liquid
and vapour phase input mole fractions, (5) pressure, (6) ratio of liquid and vapour phase capacity volume to initial volume, (7) liquid
output/accumulation ratio I, vapour output/accumulation ratio, liquid output/accumulation ratio II and the time instance at which
the ratio of liquid output/accumulation is switched from ratio I
to ratio II, (8) initial liquid and vapour phase moles on each stage,
(10) input liquid and vapour phase molar ow to the feed stage, (9)
location of feed stage and reux ratio, (10) Antoine constants A, B
and C for calculating vapourliquid distribution coefcients, dew
point and bubble point temperatures, (11) liquid heat capacity, vapour heat capacity, heat of vapourisation, heat of solution, input liquid and vapour phase enthalpies, reference temperature and
liquid and vapour phase heat transfer coefcients and heat transfer
area. Additionally the following inputs were set in the main program in order to complete the denition of the reactive distillation
system studied: (1) number of equations, (2) number of species, (3)
number of reactions, (4) number of stages, (5) overall simulation
time, (6) rate constants of various reactions involved in isophorone
production and (7) activity coefcients of various components.
The values of various inputs used in the simulation of the reactive distillation model are dened in the base case simulation (Tables 3a and 3b). Specic data and information were not readily
available for estimating many of these signicant parameters.
Since the objective of the present study was to evolve guidelines
for performance enhancement, suitable approximations were
made to estimate these parameters. The published information
on reaction mechanism, stoichiometry, order, forward rate constants and equilibrium constants and effect of temperature/catalyst concentration on the rate constants was used. The VLE
parameters of the considered system including activity coefcients
were estimated using UNIFAC group contribution method for various components [30]. Activity coefcients of acetone and water
obtained by UNIFAC group contribution method were tted for
various values of acetone/water liquid phase mole fractions. Azeotrope formation at higher acetone mole fraction was avoided in
simulations by operating in the lower range of acetone mole fraction. Heat and mass transfer coefcients were estimated from the
literature, while physicochemical properties of various components were obtained from the literature and from plant data. The
key output variables (results described in Section 3) of the model
included: (1) liquid and vapour phase mole fractions of various
components on each stage, (2) liquid and vapour phase molar ows
originating from each stage, (3) liquid and vapour phase volume on
each stage, (4) dew point and bubble point temperatures corresponding to the stage composition on each stage and (5) bulk liquid and vapour phase enthalpies and temperatures on each
stage. The issue of dening output liquid ows originating from
various stages of the reactive distillation column is discussed next.
2.3.1. Volume and liquid outow from a stage
One of the key issues in model development for reactive distillation is the description of the output liquid and vapour molar
ows originating from various stages. The output ows originating
from a stage also inuence the steady state volume attained at
each individual stage. In this work, since a dynamic and non-equilibrium model of reactive distillation was developed, it was essential to account for variable volume in conjunction with output
ows in the denition of the model equations (Appendix A: Eqs
A.2, A.3, A.11, A.12, A.19 and A.20). Various authors in the literature
have outlined different methodologies for addressing the issue of
characterising liquid and vapour ow in a distillation column and
these are described in this section.
Dalaouti and Seferlis [31] have proposed a modelling approach
that combines the rigorous non-equilibrium (NEQ) rate based
equations with the model order reduction properties of orthogonal
collocation on nite elements (OCFEs) approximation techniques.
These techniques can be employed to the optimal design, operation
optimisation and dynamic simulation of complex staged reactive
separation processes. The main feature of the OCFE formulation
is that material and energy balances resulting from the NEQ rate
based equations are satised exactly only at the chosen collocation
points the roots of the discrete Hahn family of orthogonal polynomials. Such a selection ensures that the collocation points coincide with the location of the actual stages in the limiting case that
the order of the polynomial becomes equal to the number of actual
stages for any given column section. Consequently, the complete
model can be fully recovered. Lagrange interpolation polynomials
were used within each nite element to approximate the liquid
and vapour component ow rates, as well as the liquid and vapour
stream enthalpies.
The non ideal ow behaviour of liquid phase has also been characterised by Noeres et al. [32] by implementing one point injection/ one point detection method to study liquid phase residence
time distribution for the catalytic structured packing MULTIPAK.
In their study, conductivity measurements were done under different liquid and gas ow rates. Measurements were evaluated using
the one dimensional axial dispersion (ADM) model and partial differential equation (PDE) model with axial dispersion and mass exchange between the dynamic and the stagnant liquid zone.
Wijn [33] and Jhon and Lee [34] cite the application of the Francis Weir equation for calculation of output liquid volumetric ow
from a stage, using the height of the clear liquid owing over the
weir and the weir length. Wijn [33] indicates that considerable
deviations from the Francis Weir equation can be expected,
depending on the ow regime, gas/liquid properties, operating
conditions and tray geometry. We use the Francis Weir equation
to have an order of magnitude comparison of the height of the clear
liquid owing over the weir (heightow) in simulations with the
experimental data of Wijn [33]. The Francis Weir equation used
for validating liquid ows is shown in Eq. (W1).
heightowj 1:43
Q Lj =3600 lengthwj 2
!1=3
g
Q Lj
6593:7572 lengthwj
!2=3
W1
356
In this work, the overall liquid molar ow originating from a stage

was calculated using the overall liquid molar accumulation rate at
any given stage. Linear time dependent functional multiplier and
ratio of output molar ow to stage accumulation rate were assumed
for calculation purposes (Appendix A: Eq. (A.2)). The expression
relating output volumetric ow to the stage volumetric accumulation rate was obtained by accounting for the molar volume of liquid
on a stage and the resulting expression is shown in Eq. (W2).
L
Q Lj
Output=Accumulation
L
ft
LRatioj
dVolumej
!
W2
dt
When the calculated value of the volumetric ow originating from a

stage in Eq. (W2) is substituted in the Francis Weir formula in Eq.
(W1), the resulting value of heightow in various simulations can
be expressed as a function of the stage volumetric accumulation
rate via Eq. (W3).
Output=Accumulation
heightowj
L
ft
LRatioj
!2=3
dVolumej
6593:7572 lengthwj
dt
!2=3
W3
Eq. (W3) indicates a linear relationship between heightow and the

liquid volumetric accumulation rate raised to the power of 2/3. In
this work the simulated values of the liquid volumetric ows were
used along with a value of 1.0 m for the length of weir (Table 2) as
inputs to Eq. (W1) and the resulting values of the implied heightow
were obtained. Apart from the values of implied heightow, dynamic
liquid stage volume can also indicate important information about
the hydraulics of the reactive distillation column. Fig. 2 shows the
plots of liquid stage volume for various stages in the reactive distillation column expressed as a function of the dimensionless time
(ratio of simulation time to residence time of the reactive distillation column) for the base case simulation (Tables 3a and 3b).
The stages 1, 11, 21, 31 and 41 in Fig. 2 correspond to the 41 reaction/hydrolysis stages, excluding 1 condenser and 1 bottoms stage
of the total 43 stage reactive distillation column. The liquid stage
volume is a function of simulation time and stage location in the
distillation column reactor. As steady state approaches on various
stages of the distillation column, the liquid stage volume increases
with time until steady state is attained. Fig. 2 shows that liquid
stage volume on various stages attains steady state for values of ratio of simulation time to residence time greater than 20. It can also
be seen from Fig. 2 that liquid stage volume is higher around the reux/ liquid feed region of the column and decreases with increasing
stage number towards the bottoms part of the column. Results of
Fig. 2 correspond to the denition of the liquid phase volumetric
ows depicted in Eq. (W2).
The dening expression for Eq. (W3) Francis Weir Eq. (W1) is
based on the premise that heightow is a function of the volumetric
liquid ow originating from a stage and the weir length. Simulation
results (base case simulation) indicate that volumetric liquid ow
originating from various stages increases with increase in the simulation time and decreases as the stage number increases from 1 to
41. This increase/decrease in the liquid volumetric ow originating
from a stage translates to corresponding increase/decrease in the
value of implied heightowin the simulation results. The Francis
Weir Eq. (W1) can also be applied to the isophorone distillation
column reactor plant data to obtain the values of the implied heightow. When the values of the output liquid volumetric ows of reaction and hydrolysis columns (Fig. 1) are used, the values of implied
heightow obtained are: reaction column output: 1.67 102 m and
hydrolysis column output: 8.88 103 m. Besides the volumetric
liquid ow, the values of the overall liquid side mass transfer coefcient also inuence the values of implied heightow. Steady state
results of the simulations with increasing values of overall liquid
side mass transfer coefcients including 1.25 (base case), 1.75,
2.25 and 3.50 h1 show that the value of implied heightow for various stages increases with increasing values of the overall liquid
side mass transfer coefcient. It is also important to note here that
the simulation values of the implied heightow (7.02 103
1.42 102 m) and plant data implied heightow values of
1.67 102 m (for reaction column output) and 8.88 103 m
(for hydrolysis column output) are in line with the experimental
results cited by Wijn [33] for sieve and valve trays indicating consistency in order of magnitude for implied heightow in simulations
with experimentally validated tray hydraulics. Contrary to volumetric accumulation rate dependent output liquid ow originating
from various stages, the description of the kinetics and reaction
rates in the model (Appendix A) is substantially rigorous and this
requires precise denition of multiple reaction rate parameters.
Various reactions involved in the aldol condensation of acetone
for isophorone production are described next along with the reaction rate parameters obtained from the literature and from the
simulation results.
1.20
Liquid
Volume:
Stage 1
1.00
0.80
Liquid
Volume:
Stage 11
Liquid Stage
0.60
Volume
[m3]
Liquid
Volume:
Stage 21
0.40
Liquid
Volume:
Stage 31
0.20
Liquid
Volume:
Stage 41
0.00
1.00E+00
1.00E+01
1.00E+02
1.00E+03
Dimensionless Time : Ratio of Simulation Time

to Residence Time
[]
Fig. 2. Variation of liquid volume with time.
1.00E+04
357
2.3.2. Reaction kinetics

Isophorone is produced via alkali catalysed condensation of acetone at high pressure and high temperature using multiple distillation columns, which aid in the formation and separation of
isophorone. The reaction network [3] is initiated by the condensation of two acetone (A) molecules to form diacetone alcohol (DAA)
(reversible reactions: (R1,R2)). Under most catalytic conditions
(using potassium hydroxide: KOH as catalyst), this alcohol rapidly
dehydrates to mesityl oxide (MO) (reversible reactions: (R3,R4)).
k1
2A
DAA
k2
k3
DAA
MO H2 O
k4
R1 ; R2
R3 ; R4
Mesityl oxide can condense with acetone to form a linear dione

phorone (PH) (reversible reactions: (R5,R6)), which rapidly cyclises
to form isophorone in two isomeric forms labelled a and b, and
water (H2O) (reversible reactions: (R8,R9)). Self condensation reaction of mesityl oxide also yields isophorone (IPH) and acetone (irreversible reaction: (R7)).
k5
A MO
PH
k6
k7
2MO ! IPH A
k8
PH
IPH H2 O
k9
R5 ; R6
R7
R8 ; R9
Since the cyclisation of phorone is a relatively fast reaction, in this

analysis it is assumed that condensation of acetone and mesityl
oxide leads directly to the formation of cyclised product: isophorone and water combining the original reversible reactions:
(R5,R6) and (R8,R9) into one net reversible reaction: (R5,R6) shown
below:
k5
A MO
IPH H2 O
k6
R5 ; R6
Cross condensation between various ketones: acetone and isophorone leads to the formation of decomposition product of isophorone:isoxylitone (IX) (reversible reactions: (R10,R11)). Also, two
molecules of mesityl oxide can further condense to form xylitones
(reversible reactions: (R12,R13)). This condensation pathway may
continue further to form heavier ketones.
k10
IPH A
IX H2 O
k11
k12
2MO
IX H2 O
k13
R10 ; R11
R12 ; R13
Various reversible reactions (R1R13) involved in the condensation

reaction of acetone lead to multitude of products including mesityl
oxide, diacetone alcohol, isoxylitone and the desired product of the
reaction: isophorone. The reaction mechanism for isophorone formation is a function of the hydroxide concentration. Chaubal [3] describes various mechanisms dependent on hydroxide catalyst
(KOH) concentration. Simulation scenarios with relatively lower
hydroxide concentration in the range of 0.010.1% can be described
with the reaction mechanism (originally dened for KOH concentration range of 0.10.3%) which consists of dening a subset of
various reactions in equations (R1R13) so that:
kk 0 for k 2; 4; 6 and 13
M1
The resulting 7 reactions: R1, R3, R5, R7, R10, R11 and R12 considered
in mechanism (M1) dene the reaction system used in the
simulations for isophorone production. Reaction terms for various

components in the reactive distillation model (Appendix A) were
derived using (1) stoichiometric coefcients, (2) reaction orders,
(3) base stoichiometric coefcients and (4) kinetic parameters corresponding to the reference components for various reactions.
These values used in calculating reaction terms in the reactive distillation model are represented in matrix form in Eqs. (I1)(I4) [35].
Stoichiometric Coefcient Matrix:
2
6
6
6
6
6
6
6
6
6
6
6
6
6
4
Components=Reactions
R1
R3
R5
R7
Water
1 1
Acetone
2
1
1
1 1
1
0
0
MesitylOxide
1 1 2
1 7
7
7
0 7
7
0 7
7
7
0 7
7
2 7
7
7
1 5
1
1
Isoxylitone
R12
1 1
Hydroxide
1 1
R11
Isophorone
DiacetoneAlcohol
R10
1
0
I1
Reaction Order Matrix:
2
6
6
6
6
6
6
6
6
6
6
6
6
6
4
Components=Reactions
R1
R3
R5
R7
R10
R11
R12
Water
0 7
7
7
0 7
7
0 7
7
7
0 7
7
2 7
7
7
0 5
Acetone
Isophorone
Hydroxide
0
0
0
0
0
0
0
0
1
0
0
0
MesitylOxide
Isoxylitone
DiacetoneAlcohol
1 2
0
I2
Base Stoichiometric Coefcient Matrix:
Reactions
BaseCoefficient
R1
R3
R5
R7
R10
R11
R12
1 1 1 1

I3
Reference Components for Base Stoichiometric Coefcient Matrix

and for Rate Constants:
Reactions
BaseComponent
R1
R3
DAA DAA
R5
R7
R10
R11
R12
IPH
IPH
IX
IX

I4
A number of authors in the literature have quantied the kinetic

parameters of various reactions involved in aldol condensation of
acetone. These kinetic parameters including the overall reaction orders, pre-exponential factors and activation energies are shown in
Table 1. In this study, kinetic set of parameters which led to plant
data comparable values of the specic mass ratio of acetone consumed (kg h1) to isophorone produced (kg h1) (1.54) and isophorone productivity (512.43 kg h1), were also identied in the
simulations (Table 3a).
The relevant kinetic parameters were estimated by taking multiples of a set of assumed rate constants and carrying out simulations of the reactive distillation column. Rate constants for
various reactions were classied into three sets of rate constants:
(1) DAA and MO formation: k1 and k3; (2) IPH formation: k5, k7
and k11; and (3) IX formation: k10 and k12. These three sets of rate
constants were increased/decreased in terms of percentage of the
assumed rate constants as a group. This methodology (aggregation
of relevant rate constants) involved investigating multiple solutions of differential algebraic equations (DAEs) for obtaining the
relevant kinetic parameters. Alternatively, the rate constants can
also be estimated using a rigorous algorithm in conjunction with
the differential algebraic equation (DAE) solver. However in this
work we focus on the kinetic parameters obtained by aggregating
the relevant rate constants and evaluate the isophorone production
358
Table 1
Pre-exponential factors and activation energies for various reactions involved in the aldol condensation of acetone.
Reaction
Rate constant/
equilibrium
constant
Overall
reaction
order
Inferred pre-exponential
factor k0(k)/equilibrium
constantb
Inferred/assumed
activation energy E(k)
(kJ kmol1 K1)
Reference
DAA formation from A

A formation from DAA
A formation from DAA
Equilibrium constant for A,
DAA equilibrium reaction
MO formation from DAA
DAA formation from MO and W
Equilibrium constant for DAA,
MO equilibrium reaction
Net A consumption
Phorone formation from DAA and MO
k1
k1
k1
k1
k1
k2
k2
K1,2
2
1
2
2
2
1
1
1,1
4.08 102
4.17 109
6.67 109
1.93 103
9.71 106
1.95 1017
3.37 109
2.26 102
14,650
34,400
71,057
34,400a
34,400a
97,540
34,400a
Kustrowski et al. [8]

Podrebarac et al. [6]
Thotla et al. [36]
Huang et al. [7]
Bakker et al. [5]
Thotla et al. [36]
Bakker et al. [5]
k3
k3
k3
k3
k4
K3,4
1
1
0
1
1
1,1
7.19 1015
9.68 1011
2.55 1012
4.89 1015
5.98 108
5.36 106
90,000
78,523
90,000a
90,000a
53,038
25,322

Kim and Hateld [37]
Huang, Yang et al. [7]
Bakker et al. [5]
Kim and Hateld [37]
Kim and Hateld [37]
0
2
5.35 102
3.19 1017
34,400a
93,259
4.52 1014
90,000a
Watanabe and Seto [38]

Lemcoff and Cunningham
[4]
Bakker et al. [5]
DAA and MO consumption via deactivation

leading to possible higher aldol
condensation products
a
Activation energy values assumed from Podrebarac et al. [6].

Values of pre-exponential factors for various reactions are given in terms of: kmol m3 h1 (for reaction order = 0), h1 (for reaction order = 1) and m3 kmol1 h1 (for
reaction order = 2).
b
process via sensitivity analysis described in Section 3. The base

case rate constant values used in simulations, and inferred values
of comparable rate constants from various literature sources are
shown in Table 3a.
In conjunction with the reaction kinetics, gasliquid mass
transfer dened in the model balances (Appendix A) is also expected to play an important role in the overall dynamics of the
isophorone production process. The issue of obtaining mass transfer parameters for the model, using suitable correlations outlined
in the literature is described next.
2.3.3. Mass transfer coefcients
Non-equilibrium modelling of reactive distillation, incorporating effects of multicomponent mass transfer via a lm model requires input of mass transfer coefcient for calculating the mass
transfer ux terms. In this study mass transfer coefcients have
been calculated for structured packings and for sieve trays using
correlations available in the literature [16,39,40]. It is assumed that
the calculated values of the mass transfer coefcients can be readily applied to the isophorone reactive distillation process. These
calculations and the relevant input parameters are shown in this
section and in Table 2.
Calculation of the liquid side mass transfer coefcient in gasliquid lm for reactive distillation while using structured packings
[16] requires knowledge of the following parameters: (1) liquid
phase diffusion coefcient of volatile components at specied
pressure and temperature conditions (DL(i,j)), (2) liquid phase effective and supercial velocity in the column (ULE(j), ULS(j)), (3) fractional liquid holdup (holdupL(j)), (4) void fraction of packing (e),
(5) characteristic length (S(j)), (6) specic factor (CE), (7) channel
ow angle (h), and (8) packing area for structured packing/interfacial area per unit liquid volume (a). Liquid side mass transfer coefcient for the non-reactive SULZER BX and reactive KATAPAK
structured packing is given by the following expression:
kLi;j

DLi;j C E U LEj 0:5
2
pSj
K1
Table 2
Parameters used for estimating mass transfer coefcients.
Mass transfer parameter
Mass
L
Flowj ;
Mass
V
Flowj
Description
Numerical values
Liquid and vapour phase

mass ows
Liquid phase:
2.341.15 kg h1
Vapour phase:
1.890.68 kg h1
Liquid phase:
587.0
746.0 kg m3
Vapour phase:
29.64
18.70 kg m3
1.0 m
1.50 m2
qLj ; qVj
Liquid and vapour phase

density
lengthw(j)
Weir length
Bubbling area of the tray
Abubbling
j
heightw(j)
Exit weir height, calculated

using the clear liquid height
Hole pitch
Hole area of the sieve tray
0.64 m
0.26 m
uSV(j)
Clear liquid height

calculated from simulated
average stage volume of
liquid phase and bubbling
area of the tray
Supercial vapour velocity
r(j)
Surface tension at 200 C
pitch(j)
Ahole
j
L
height j
0.20 m
0.29 m2
4.25
2.42 102 m s1
2.04 102 N m1
In the above expression (K1), liquid phase effective velocity is calculated using the following equation:
U LEj
U LSj
e holdupLj sin h
K2
The diffusion coefcient in the Eq. (K1) can be computed using the
correlation provided by Umezawa and Nagashima [41] for acetone,
benzene and alkanes in supercritical carbon dioxie using the Taylor
Dispersion method. The dependence of diffusion coefcient on
359
component molecular weights and reduced temperature is given by

Eq. (K3) as:
DLi;j
Adiff
i
exp
Bdiff
i
T ri;j
Hdiff
T ri;j 12
i

0:5 exp
W diff
i
W diff
i
!
K3
Where the dependence of various parameters dened in Eq. (K3), on

component molecular weight is given by:
14
Adiff
2:4348 106 exp4:0020 102 MW i
i 5:0529 10
1
2
4
2
Bdiff
i 1:1407 10 2:5974 10 MW i 1:8000 10 MW i
11
Hdiff
8:5820 104 MW i
i 2:3312 10
4
W diff
8:7540 107 MW i
i 2:7365 10
reactive distillation column and the correlation shown in Eq. (K6)

[39,40]:
kLi;j 0:26 105
T ri;j
T Lj
Critical
T i
K5
Numerical values of various parameters including those of structured packing [16] required for calculation of liquid side mass transfer coefcient (Eq. (K1)), liquid phase effective velocity (Eq. (K2)) and
for liquid diffusion coefcient (Eq. (K3)) are: (1) ULS(j) = 1.84
103 m s1, (2) holdupL(j) = 0.40, (3) e = 0.90 and 0.622; h = 60 and
45; a = 492.0 m2 m3 and 128.20 m2 m3 for Sulzer BX and KATAPAK structured packings respectively, (4) S(j) = 1.11 m, and (5)
CE = 0.90.
On substituting the numerical values of various parameters in
Eqs. K1, K2 and K3, the following values of the liquid phase diffusion coefcient and liquid side mass transfer coefcient for Sulzer
BX structured packing are obtained: DL(i,j) = 2.90 108 m2 s1,
kL(i,j) = 1.32 105 m s1. Using the value of packing area
(=492.0 m2 m3) for Sulzer BX structured packing, the value of
overall liquid side mass transfer coefcient including specic area
is: kLa(i,j) = 6.49 103 s1 = 23.36 h1. Similar calculations for the
structured packing: KATAPAK yield the values of liquid side mass
transfer coefcients: kL(i,j) = 1.75 105 m s1 and kL a(i,j) = 2.25
103 s1 = 8.10 h1.
Mass transfer coefcient values obtained for structured packings were followed by calculations for sieve trays in order to have
a better understanding of the mass transfer parameters used in the
model. Svandova et al. [39] have chosen four empirical correlations
(including correlations of [40]) for the estimation of the number of
transfer units or relations for the estimation of binary mass transfer coefcients for a sieve tray. In this work we estimate the mass
transfer coefcient and interfacial area based on the correlations of
Zuiderweg [40] shown in Eqs. (K6)(K14). These correlations are
applicable to sieve trays. We assume that the value of interfacial
area obtained from these correlations can be used in the present
study. The value of liquid side mass transfer coefcient can be calculated using the viscosity of liquid phase (1.61 104 Pa s) in the
lLj
0:25
m s1
K6
Eq. (K6) yields a value of 2.31 105 m s1/8.31 102 m h1 for
the mass transfer coefcient. Apart from the mass transfer coefcient, interfacial area is also required, to complete the denition
of the overall liquid side mass transfer coefcient. The interfacial
area per unit volume of the liquid phase was expressed as the
parameter ahf(j): interfacial area per unit bubbling area. Estimation
of this parameter depends on the regime of operation in the distillation column (spray regime/froth emulsion regime). Expressions
relating the interfacial area per unit bubbling area and various process parameters are shown in Eqs. (K6)(K14):
K4
It is possible to dene the dependence of diffusion coefcient on
molecular weight (Eqs. (K3) and (K4)) and conduct an extensive
analysis with different diffusion coefcients and mass transfer coefcients for various components. However in this study, acetone was
assumed to be the primary component for calculation of liquid diffusion coefcient using Eqs. (K3) and (K4).
The reduced temperature used in Eq. (K3) was calculated using
the average column temperature of 473.0 K and the critical temperature of carbon dioxide (=304.1 K). On substituting these values
of average column temperature and reduced temperature in Eq.
(K5), the value of reduced temperature for Eq. (K3) turns out to
be 1.56.
FPj
bj
Mass Flowj qVj
K7
V
L
Mass Flowj qj
lengthwj
K8
Abubbling
j
L
0:5
heightj 0:6 heightwj

bj
!0:5

0:25
pitchj FPj
bj
Ahole
j
bubbling
Aj
K10
L
FPj < 3:0 bj height j implies spray regime

FPj > 3:0
K9
L
bj height j
implies froth emulsion regime
K11
K12
Interfacial area per unit bubbling area (m2 m2) is given by:
Spray regime:
ahfj
40
b0:3
!0:37

2
L
uSVj qVj height j FPj
rj
K13
Froth emulsion regime:
ahfj
43
b0:3
!0:53

2
L
uSVj qVj height j FPj
rj
K14
Table 2 indicates the parameters used in the calculation of the interfacial area per unit bubbling area in Eqs. (K6)(K14). While using the
L
parameters of Table 2, the value of FPj 3:0 bj heightj turns out to
0.24, indicating presence of spray regime in the reactive distillation
column. Calculations of the interfacial area yield a value of
27.45 m2 m2 for ahfj . When the interfacial area per unit bubbling
area (1.50 m2) is adjusted for base case average liquid stage volume
(0.39 m3), the value of the interfacial area per unit liquid volume
takes a value of 106.11 m2 m3. This value of interfacial area per unit
volume of liquid results in the value of kL a(i,j) = 106.11
m2 m3 2.31 105 m s1 = 2.45 103 s1 = 8.81 h1.
Simulations of reactive distillation for isophorone production
indicate that steady state convergence is attained when the values
of kL a(i,j) used were in the range of 1.253.50 h1. Hence, for analysis purposes, a value of kLa(i,j) = 1.25 h1 was used as base case,
which is described in the Section 3. The maximum value of kLa(i,j)
set to 3.50 h1 in various simulations, was 43.20% and 39.71% of
the corresponding values for the KATAPAK structured packing
and sieve trays respectively. For purposes of sensitivity analysis
the base case kLa(i,j) was varied in the range of 1.253.50 h1 and
simulations were carried out to study the inuence of the overall
liquid side mass transfer coefcient on the mass ratio of acetone
consumed to isophorone produced in the reactive distillation
column.
360
3. Simulation results: base case simulation and sensitivity

analysis
The mathematical model for reactive distillation described in
Appendix A was used to simulate the isophorone production process. Simulations were carried out using the base case simulation
parameters described in Tables 3a and 3b. Simulation results
including acetone consumed and isophorone produced at each
stage of the reactive distillation column were validated with the
balance equations used in the model including dynamic molar
accumulation rates. Validated mass and energy balances along
with chemical reactions for each stage and for the entire column
were used to obtain the key mass ratio of acetone consumed to
isophorone produced, and the overall productivity of isophorone
in the reactive distillation column. The base case simulation results show isophorone productivity of 512.43 kg h1, total acetone
conversion of 90.23% and the mass ratio of acetone consumed to
isophorone produced of 1.54. Liquid and vapour phase mole fractions obtained in the base case simulation for various components
at steady state are shown in Figs. 3 and 4. While, Fig. 5 shows the
liquid/vapour and bubble point/dew point temperature proles in
the column for the base case simulation. The temperature proles
in Fig. 5 correspond to an average liquid phase temperature of
191.06 C, an average vapour phase temperature of 216.50 C and
bubble point/dew point temperatures of 184.86 C and 193.13 C
respectively. It is worthwhile to mention here that stages 141 in
Figs. 3, 4, 5 and 8 correspond to the 41 reaction/hydrolysis stages,
excluding 1 condenser and 1 bottoms stage of the total 43 stage

reactive distillation column.
The average liquid phase temperature obtained in the base case
simulation deviates from the objective temperature of 200.0 C
used for dening temperature dependent distribution (200.0 C)
and UNIFAC activity coefcients (473.0 K) [30] in Eq. (A.34) of
Appendix A. Simulation results of temperature showed strong
dependence on the liquid/vapour phase heat capacity values and
liquid enthalpy fed to the column. In this work we assume that a
set of liquid and vapour phase heat capacity parameters and/or
appropriate enthalpy withdrawal/introduction stage functions (Table 3b) exists for simulations which result in an average liquid
phase column temperature of 200.0 C. Hence a sensitivity analysis
was carried out to evaluate the inuence of various parameters on
the key mass ratio of acetone consumed to isophorone produced
while assuming an average liquid phase column temperature of
200.0 C. Part of the results of sensitivity analysis discussed at
the end of this section (Fig. 8) show that the average liquid phase
temperature can be easily manipulated while using liquid phase
feed enthalpy as the variable parameter.
Simulated results for isophorone production are strongly
dependent on: (1) kinetics of various reactions involved in isophorone production during aldol condensation of acetone, (2) overall
liquid side mass transfer coefcients within the distillation column
dening the interfacial mass transfer rates, (3) hydrodynamics and
transport rates in the reactor including liquid/vapour phase molar
ows and volumetric holdup/residence time of various stages, and
Table 3a
Base case simulation parameters.
Reaction
Rate constant and reaction order
Base case rate constant values used in simulations and inferred

comparable rate constant values from various literature sources
k1 (2)
k3 (1)
IPH formation from A and MO

IPH formation from MO
k5 (2)
k7 (2)
IX formation from IPH and A

IPH formation from IX and W
IX formation from MO
k10 (2)
k11 (2)
k12 (2)
1.39 101: simulation value

3.07 101: Huang et al. [7]
6.82 10+2: simulation value
2.93 10+2: Huang et al. [7]
2.06 10+3: Kim and Hateld [37]
2.47: simulation value
1.48 10+3: simulation value
1.60 10+7: phorone formation from DAA and MO: Lemcoff and
Cunningham [4]
5.20 10+4: higher aldol condensation products formation from
DAA and MO: Bakker et al. [5]
9.41 103: simulation value
Initial/input liquid phase mole fractions, xinitial

i;j
Initial/input vapour phase mole fractions, yinitial

i;j
Water
Acetone
Isophorone
Hydroxide catalyst
Mesityl oxide
Isoxylitone
Diacetone alcohol
0.72
2.78 101
1.20 103
1.30 103
0.0
0.0
0.0
1.0
0.0
0.0
0.0
0.0
0.0
0.0
Initial molar holdups
Initial liquid molar holdup:ILj 0:35 (kmol)

ow:F L2
Initial vapour molar holdup:IVj 0:17 (kmol)

Input vapour molar ow: F VNSTAGES 135:86 (kmol h1)
Input molar ows
Input liquid molar
Feed/reux stage and reux ratio
Liquid feed and reux stage = 2
Reux ratio: 1000
Mass transfer coefcient and column pressure
kLa(i,j) = 1.25 h1
P Total
15 bar
j
54:19 (kmol h
1
Component/Antoine constants
A(i)
B(i)
C(i)
Water
Acetone
Isophorone
Hydroxide catalyst
Mesityl oxide
Isoxylitone
Diacetone alcohol
8.14
7.63
8.59
0.0
10.98
0.0
7.91
1810.94
1566.69
2602.97
0.0
2991.21
0.0
2036.93
244.49
273.42
265.0
0.0
265.0
0.0
236.18
361
Table 3b
Simulation parameters dening the liquid and vapour phase energy transfer rates for base case simulation.
Liquid phase heat
Vapour phase heat
capacity: C Li
capacity: C Vj
(kJ kmol1 C1)
(kJ kmol1 C1)
kAi (kJ kmol1)
Water
Acetone
Isophorone
312.46
1007.31
2396.99
104.41
374.27
890.60
Hydroxide
catalyst
693.61
257.71
Component
Heat of
vapourisation:
Additional liquid and vapour phase

energy transfer rates dening parameters
Numerical value
4.17 104
6.08 104
1.45 105
Reference temperature: T0 (C)

Heat of solution: D HS (kJ kmol1)
19.69
900.0
41,267.93
4.19 104
Input vapour enthalpy: HVf

NSTAGE
66,468.60
(kJ kmol1)
Liquid/vapour side heat transfer
coefcient and heat transfer area:
10,000.0/21,600.0
Mesityl
oxide
1702.09
632.41
1.03 105
Isoxylitone
2396.99
890.60
1.45 105
Diacetone
alcohol
2014.62
748.53
1.22 105
Input liquid enthalpy:
HLf
1
(kJ kmol1)
ALj hj =AVj hj (kJ h1 C1)

Function dening liquid phase enthalpy
withdrawn from stage j (kJ kmol1)
Function dening vapour phase enthalpy
introduced to stage j (kJ kmol1)
(0.99 j2 9.17 j 1,797.61)

(1.60 103 j4 6.87 102 j3 + 5.65
101 j2 30.60 j + 645.25)
Liquid Phase Mole Fraction []
1
6
Water
11
16
Distillation
Column
21
Stages
[]
26
Isophorone
Acetone
Isoxylitone
Mesityl Oxide
Hydroxide
31
36
41
Fig. 3. Variation of liquid phase composition along the column at steady state.
(4) multicomponent vapour liquid equilibrium (VLE) of the reaction system under consideration.
A sensitivity analysis was carried out to study the inuence of
overall liquid side mass transfer coefcient and average liquid
stage volume on the performance of the isophorone reactive distillation column. Appropriate range of diffusion coefcient and overall liquid side mass transfer coefcient were identied (Section
2.3.3) and varied in the sensitivity analysis. Simulations were
carried out by increasing values of the overall liquid side mass
transfer coefcients in the range of 1.25 (base case value)
3.50 h1. This change in the overall liquid side mass transfer coefcient amounted to a maximum increase of +180% in the base case
value. In addition to the overall liquid side mass transfer coefcient, the average liquid volume of various stages was increased/
decreased by changing the ratio of stage liquid input/output molar
ows to corresponding liquid molar accumulation rates (Appendix
A: Eqs. (A.2) and (A.3)). The base case average liquid stage volume
of 0.39 m3 was consequently varied in the range of 32.88% and
+68.04%, and the resulting average liquid stage volume values in
the range of 0.26 m3 and 0.65 m3 were used in the sensitivity analysis. The liquid stage volumes for various stages were used in conjunction with the volumetric ows for the corresponding stages in
order to obtain stage wise and overall residence time values in the
distillation column reactor. Fig. 6 shows that the key mass ratio of
acetone consumed to isophorone produced increases with the
overall residence time in the distillation column reactor and with
overall liquid side mass transfer coefcient. Increase in the overall
liquid side mass transfer coefcient increases the average liquid
stage volume and overall consumption of acetone while leading
to higher value of the key mass ratio. Analysis of simulations
(including base case simulation) with similar volumetric ow
parameters indicates that an increase in the value of overall liquid
side mass transfer coefcient in the range of 1.253.50 h1 successively reduces the average relative values of liquid phase mole
fractions of acetone and mesityl oxide. This reduction causes
reduction in selectivity towards isophorone. In addition, increase
in mass transfer coefcient also increases liquid ow within the
column (at set value of reux ratio, R) and therefore increases
the average liquid stage volume. The combined effect of these
two factors results in increase in the mass of acetone consumed
per unit mass of isophorone produced with increase in the overall
liquid side mass transfer coefcient.
In addition to the design parameters of volume (residence time)
and overall liquid side mass transfer coefcient, vapour liquid
362
Vapour Phase Mole Fraction []
1
6
Water
11
Isophorone
16
Distillation
Column
Stages
[]
Acetone
21
Mesityl Oxide
26
31
36
41
Fig. 4. Variation of vapour phase composition along the column at steady state.
Liquid and Vapour Phase Temperature Profiles [ C]
1
6
11
Liquid Phase
Temperature
16
Vapour Phase
Temperature
Distillation
Column 21
Stages
[]
Bubble Point
Temperature
Dew Point
Temperature
26
31
36
41
Fig. 5. Temperature proles along the column.
equilibrium (VLE) parameters also signicantly inuence the key

mass ratio of acetone consumed to isophorone produced. Signicant differences were observed when UNIFAC VLE parameter results were compared to results obtained assuming ideality. The
results with ideal VLE are not discussed here. The results obtained
with the UNIFAC VLE parameters were used for drawing conclusions regarding the reactive distillation column performance.
Few simulations were also carried out in order to assess the impact of using different initial conditions on the key mass ratio. In
contrast to the initial conditions for base case simulation dened
in Table 3a, simulations were conducted by setting initial liquid
phase mole fractions of water, acetone, isophorone and hydroxide
catalyst each to 0.25 and, vapour phase mole fractions of water and
acetone to 0.5 each. These results conrmed that steady state results are independent of the initial conditions.
Apart from the key issues of mass transfer, average liquid stage
volume (residence time) and specic VLE parameters, addressing
the conguration and operational aspects of the process can also
help gain insights into making improvements in the performance
of the reactive distillation column reactor. The reactive distillation
modelling framework in this study was modied to study the inuence of conguration and operational aspects including (1) feed
composition, (2) feed location, and (3) liquid enthalpy introduced
in the column.
Figs. 7a and 7b show the inuence of variation in the feed composition and feed location on the relative values of the mass ratio
of acetone consumed to isophorone produced and average liquid
stage volume compared to the base case simulation.
When the feed acetone mole fraction is increased in the range
of: base case -20%, and equal water/acetone mole fraction of
363
1.30
Mass transfer coefficient = 3.5 (1/hour)

2.40
2.20
Mass transfer coefficient = 2.25 (1/ hour)

1.25
1.20
Influence of Feed Composition: Average Volume

Influence of Feed Composition: Ratio
1.15
Relative values
of the ratio of acetone
1.10
consumed to
isophorone produced
and average liquid stage 1.05
volume compared to
base case
1.00
[ ]
Ratio of
2.00
Acetone
Consumed to
Isophorone
Produced 1.80
[ ]
0.95
1.60
0.90
1
1.40
1.50
FC = 20% xA
1.75
2.00
2.25
2.50
FC = +0% xA
2.75
FC = + 20% x A
FC =Equal x A, xW
FC: Feed Composition
Residence Time [hours]

Fig. 7a. Inuence of feed composition.
Fig. 6. Inuence of mass transfer coefcient on ratio of acetone consumed to
isophorone produced.
(0.5 along with hydroxide catalyst mole fraction of 1.30 103

introduced in the feed), the average liquid stage volume signicantly increases while the key mass ratio is marginally inuenced
(Fig. 7a). Increasing the feed acetone mole fraction marginally increases the overall conversion of acetone in the multistage distillation column reactor. However, as the average mole fractions of
both acetone and isophorone increase in the reactive distillation
column, the overall mass ratio of acetone consumed to isophorone
produced increases marginally with increasing mole fraction of
acetone in the feed. Higher mole fraction of acetone in the feed also
leads to higher average liquid stage volume on account of larger
molar volume of acetone compared to water.
Contrary to the simulation results obtained by varying feed acetone mole fraction, when the feed location is varied as 2, 4, 6, 9 the
key mass ratio decreases, while the average liquid stage volume
excluding stages above feed stage, remains almost constant
(Fig. 7b). The denition of the feed location in simulations included
1 condenser, 41 reaction/hydrolysis stages and 1 bottoms stage. A
decrease in the feed location from stage 2 to stage 9 decreases the
overall liquid phase residence time in the distillation column reactor. Decrease in the liquid phase residence time reduces the net
quantities of intermediates-mesityl oxide and diacetone alcohol
and by product isoxylitone formed in the distillation column
reactor, thus reducing the mass ratio of acetone consumed to
isophorone produced.
The issue of energy transfer in the reactive distillation model
was addressed with the objective of obtaining an average liquid
phase column temperature of 200.0 C. The set of liquid and vapour
phase heat capacity parameters and appropriate enthalpy withdrawal/introduction stage functions are shown in Table 3b. Liquid
and vapour phase energy was introduced in the column via the liquid feed at the top and vapour feed at the bottom portion of the
column. In order to study the inuence of the feed liquid enthalpy
on the liquid phase temperature, the liquid phase enthalpy fed to
the column was varied in the range of 10% and +20% of the base
case simulation liquid feed enthalpy (Table 3b). Results of the sensitivity analysis: liquid phase temperature proles along the reactive distillation column with average liquid phase temperature in
the range of 188.75 C and 195.69 C are shown in Fig. 8 for various
simulations. Fig. 8 shows that the liquid phase temperature
increases for any given stage of the column, while increasing the
liquid phase feed enthalpy in the range of 37,141.14 kJ kmol1
49,521.52 kJ kmol1.
1.04
Influence of Feed Location: Average Volume

Influence of Feed Location: Ratio
1.02
1.00
Relative values
of the ratio of acetone
consumed to
isophorone produced
and average liquid stage
volume compared to
base case
[]
0.98
0.96
0.94
0.92
0.90
1
FL= 2
2
FL= 4
3
FL= 6
4
FL= 9
FL: Feed Loctaion

Fig. 7b. Inuence of feed location.
The temperature proles shown in the Fig. 8 for various values

of liquid phase feed enthalpy correspond to average bubble point
and dew point temperatures of 184.86 C and 193.13 C. Reactive
distillation results also indicate that the component liquid and vapour phase mole fractions along various stages of the reactive distillation column are independent of the liquid phase feed enthalpy.
This highlights the decoupling between heat transfer rates and
mass transfer/reaction rates in the reactive distillation model. It
was also observed that the average vapour phase temperature
was 216.5 C for various values of liquid phase feed enthalpy. Thus
it can be inferred from the sensitivity analysis that a desired and
independent change in the average liquid phase temperature can
be induced by varying the liquid phase feed enthalpy for a given
set of heat transfer parameters shown in Table 3b.
4. Summary and conclusions

The process of isophorone production in reactive distillation
column via aldol condensation of acetone using potassium hydrox-
364
Liquid Phase Temperature Profiles [ C]
1
6
Input Liquid
Enthalpy = 90%
of Base Case
11
Base Case Input

Liquid Enthalpy
16
Distillation
Column
21
Stages
[]
Input Liquid
Enthalpy = 110%
of Base Case
26
31
Input Liquid
Enthalpy = 120%
of Base Case
36
41
Fig. 8. Liquid phase temperatures proles along the column for four cases: (1) 90% of the base case feed liquid enthalpy, (2) base case, (3) 110% of the base case feed liquid
enthalpy and (4) 120% of the base case feed liquid enthalpy.
ide as a catalyst was modelled using liquid and vapour phase molar
and energy balances. The resulting system of differential algebraic
equations (DAEs) was solved using the DAE solver: sundials 2.4.0.
Reactive distillation model simulations were carried out for a 43
stage column with 7 components including: (1) water, (2) acetone,
(3) isophorone, (4) hydroxide catalyst, (5) mesityl oxide, (6) isoxylitone and (7) diacetone alcohol. The results of the simulations:
liquid/vapour phase mole fractions of various species and liquid/
vapour phase temperatures along various stages of the reactive
distillation column were plotted and analyzed. Mass and energy
balances including liquid and vapour ows for all the components
on various stages were attributed to the model balances. Numerical solution while using heat transfer, mass transfer and chemical
reactions on each stage indicated that acetone was consumed in
various stages and, products: isophorone, isoxylitone, mesityl
oxide, diacetone alcohol and water were formed as a result of 7
reactions incorporated in the reaction scheme. Simulations also
indicated that steady state was reached for the liquid phase volumetric holdup while using the stage input/output molar ow
parameters described in Eqs. (A.2) and (A.3).
In this work, a base case simulation was dened with net acetone consumption of 800 kg h1. The base case simulation mass
ratio of acetone consumed to isophorone produced was 1.54. The
model was then used to carry out sensitivity analysis by varying:
(1) average liquid stage volume/residence time (via stage input/
output molar ow parameters) and overall liquid side mass transfer coefcient, (2) feed composition, (3) feed location, and (4) liquid enthalpy fed to the column.
Sensitivity analysis results showed that: (1) the key mass ratio
of acetone consumed to isophorone produced (RatioA/IPH) increases
with increase in the overall reactive distillation column residence
time and overall liquid side mass transfer coefcient, (2) increase
in the mole fraction of acetone in the feed signicantly increases
the values of average liquid stage volume (AvgVolL), while the RatioA/IPH increases marginally, (3) an increase in the feed location
from 29 decreases the value of RatioA/IPH, while AvgVolL is independent of the feed location, and (4) increase in the liquid phase
feed enthalpy is reected in terms of increase in the stage depen-
dent liquid phase temperature and the average liquid phase column temperature.
The reactive distillation model presented here can be used to
provide design recommendations and suggest improvements
while providing appropriate details of the isophorone production
process along various stages of the column. The model presented
here will also be useful starting point for further extensions to include more rened approaches including multi-zonal models.
Appendix A. Model equations for reactive distillation
The model for reactive distillation system consists of a set of 10
differential algebraic equations (DAEs):
1. 6 ordinary differential equations (ODEs) and,
2. 4 algebraic equations (AEs).
The system of equations consists of the following balances: (1)
overall liquid molar balance (ODE), (2) overall vapour molar balance (ODE), (3) component liquid molar balance (ODE), (4) component vapour molar balance (ODE), (5) component liquid mole
fraction (AE), (6) component vapour mole fraction (AE), (7) overall
liquid volume (AE), (8) overall vapour volume (AE), (9) overall liquid phase energy balance (ODE) and (10) overall vapour phase energy balance (ODE). These individual molar balances are outlined
next along with dening equations for bulk liquid/vapour phase
temperatures and dew point/bubble point temperatures.
A.1. Overall liquid molar balance
Overall liquid phase molar balance for stage j:
L
dMj
dt
Lj1 Lj F Lj
0
NC
X
NGL
i;j
i1
1
iPNC Oi;k
i1
L
M
Ek
Y Oi;k C
j
Bzi;k k0k e RT
xi;j C
@ bk
A
h iPNC Oi;k 1
i1
i1 k1
L
V j
NC X
NR B
X
A:1
365
Output liquid phase molar ow for stage j:

L
Volj
If :
VmaxLj
<
A.3. Component liquid molar balance
! Lj 0
If : Volj P VmaxLj ! Lj
dMi;j
dt
Input liquid phase molar ow for stage j:
If :
L
Volj1
If :
L
Volj1
<
VmaxLj1
Liquid phase molar balance for component i on stage j:
A:2
Output=Accumulation dM j
L
ft
LRatioj
dt
Lj1 xi;j1 Lj xi;j F Lj zLi;j NGL

i;j
P
0
NC
h
i
O
i1 i;k
! Lj1 0
L
P VmaxLj1 ! Lj1
Output=Accumulation
L
ft
LRatioj1
dM j1
dt
A:3
MaxVol=InitialVol
ILj
NC
X
lLi xinitial
i;j
i1
Output=Accumulation
Vapour phase molar balance for component i on stage j:
A:4
dMi;j
Reux molar ow from stage 1 to stage j:

V
LRj ft
VRatio1
R
R1
dt
V
dM 1
A:5
dt
A:6
xi;j
Gasliquid mass transfer term for component i on stage j:

xInterface
xi;j M Lj
i;j
MLi;j
A:17
M Lj
A:7
Liquid phase interface mole fraction for component i on stage j:
xInterface
i;j
A:16
Liquid phase mole fraction for component i on stage j:
F Lj F Lj LRj
NGL
i;j kL ai;j
V j1 yi;j1 V j yi;j F Vj zVi;j NGL

i;j
A.5. Component liquid mole fraction
Total liquid phase molar ow fed to stage j:
A:15
A.4. Component vapour molar balance
Maximum liquid phase volumetric holdup on stage j:
VmaxLj LRatioj
NR B
MLj
Ek
X
Y Oi;k C
Bzi;k k0k e RT
xi;j C
@ bk
A
h iPNC Oi;k 1
i1
k1
L
V j
yInterface
i;j
K Distribution
i;j
A.6. Component vapour mole fraction

Vapour phase mole fraction for component i on stage j:
A:8
yInterface
yBulk
i;j
i;j
A:9
yi;j
M Vi;j
A:18
MVj
A.2. Overall vapour molar balance

A.7. Overall liquid volume
Overall vapour phase molar balance for stage j:
V
dM j
dt
V j1 V j
F Vj
Overall liquid phase volume on stage j:
NC
X

NGL
i;j
A:10
L
Volj
i1
NC
X
lLi MLi;j
A:19
i1
Output vapour phase molar ow for stage j:

V
If : Volj < VmaxVj ! V j 0

V
If :
V
Volj
VmaxVj
! V j
Output=Accumulation
V
ft
VRatioj
dMj
A:11
Overall vapour phase volume on stage j:
dt
Input vapour phase molar ow for stage j:
If :
V
Volj1
<
VmaxVj1
A.8. Overall vapour volume
Volj
NC
X
lVi MVi;j
A:20
i1
! V j1 0
V
Output=Accumulation
V
VRatioj1
If : Volj1 P VmaxVj1 ! V j1 ft
dMj1
dt
A:12
MaxVol=InitialVol
VmaxVj VRatioj
IVj
lVi yinitial
i;j
Overall liquid phase energy balance for stage j:

L
dEj
Maximum vapour phase volumetric holdup on stage j:

NC
X
A.9. Overall liquid phase energy balance
dt
Lj1
ELj1
M Lj1
Lj
ELj
MLj
L
L
L
F Lj HLf
j Ej OP j QLost j
A:13
A:21
i1
Total input vapour phase molar ow fed to the bottoms stage:
V Vj1 F Vj V VNSTAGES1 F VNSTAGES 0 F VNSTAGES F VNSTAGES

A:14
Energy transfer rate across the liquid lm on stage j:

NC

X
L
eL
ELj ALj hj T Interface
T LB
NGL
j
i;j H i;j
j
i1
A:22
366
S
L
PTotal
j yi;j ci;j xi;j P i;T
A.10. Overall vapour phase energy balance
log10 PSi;T
Overall vapour phase energy balance for stage j:

V
dEj
dt
EVj1
V j1 V
Mj1
EVj
V j V
M j
V
V
V
F Vj HVf
j Ej IP j QLost j
yi;j cLi;j
A:23
Energy transfer rate across the vapour lm on stage j:

NC

X
V
Interface
eV
EVj AVj hj T VD
N GL
j T j
i;j H i;j
A:24
i1
Interface temperature used in the calculation of liquid and vapour

lm energy transfer rates on stage j:
ELj
L
ALj hj
NC

X
Interface
eV
T VD
NGL
j T j
i;j H i;j
A:25
i1
L
T Interface
j
PNC
ALj hj
L
ALj hj
GL e V
i1 N i;j H i;j
L
ALj hj
T LB
j
V
AVj hj
PNC
AVj hj
L
ALj hj
GL e L
i1 N i;j H i;j
V
AVj hj
A:26
Partial molar liquid and vapour phase enthalpies of component i on

stage j:

e L C L T LB T 0 DHS xi;j
H
i
i;j
j

e V kA C V T VD T 0
H
i
i;j
i
j
A:27
A:28
A.11. Bulk liquid/vapour phase temperatures, and dew point/bubble

point temperatures
Bulk liquid and bulk vapour phase temperatures on stage j:
ELj
T Lj
T Vj
M Lj
DH s
T0
L
x
C
i1 i;j i
P
EVj
A
NC
i1 yi;j ki
M Vj
PNC
T0
V
y
C
i1 i;j i
PNC
A:29
A:30
Vapour pressure, distribution coefcient and dew point/bubble

point temperatures:
xi;j K Distribution
xi;j
i;j
PSi;200
C
A:33
!
A:34
PTotal
j
A:35
i1
xi;j yi;j
PTotal
j
A:36
cLi;j PSi;T VD
yi;j
S
L
i;j P i;T VD
A:37
PTotal
j
NC
X
yi;j 1
A:38
i1
yi;j xi;j
T VD
j
V
AVj hj
PTotal
j
A:32
NC
X
xi;j 1
i1
i1
V
AVj hj
PSi;T
cLi;j
K Distribution
i;j
NC
X
NC

X
V
eL
T Interface
T LB
NGL
j
i;j H i;j Ej
j
A:31
Bi
Ai
C i T
cLi;j PSi;T LB
A:39
PTotal
j
NC
X
xi;j cLi;j PSi;T LB PTotal
j
A:40
i1
The 10 sets of differential algebraic equations described in this section for reactive distillation column constitute to 4NC + 6 balances
(Table A.1) for each stage. Additional stage equations for liquid/vapour molar outow, gasliquid mass transfer, reaction terms and
calculations for energy balances are also required to complete the
description of the model (Table A.2). Thus for a reactive distillation
model consisting of NSTAGES, the total number of equations
dened as input to the solver amounts to NSTAGES x (4NC + 6). In
case of a 43 stage column (41 reaction/hydrolysis stages + 1 condenser + 1 bottoms stage) with 7 components, total number of
equations becomes 43 34 = 1462. This system of 1462 equations
was solved using the differential algebraic equation solver: Sundials
- 2.4.0 (https://computation.llnl.gov/casc/sundials/main.html). The
solution of the differential algebraic equations is obtained as time
dependent/stage wise liquid and vapour phase mole fraction and
temperature prole. The solver was initialized using the base case
parameters dened in Table 3a and solved in order to obtain steady
state in simulations. Additionally, the non linear equations: (A.37)
and (A.40) dening the dew point and bubble points of the vapour
and liquid phase mixtures were solved using the bisection method
for purposes of numerical solution (http://numericalmethods.eng.
usf.edu).
Table A.1
Stage balances used as input to the differential algebraic equation (DAE) solver for the reactive distillation model.
1
2
3
4
5
6
7
8
9
10
Stage balances
Number of balances
Type of balances
Equation number in Appendix A
Overall liquid molar accumulation

Overall vapour molar accumulation
Component liquid molar accumulation
Component vapour molar accumulation
Component liquid phase mole fraction
Component vapour phase mole fraction
Overall liquid volume
Overall vapour volume
Overall liquid energy accumulation
Overall vapour energy accumulation
Total number of balances
1
1
NC
NC
NC
NC
1
1
1
1
4NC + 6
ODE
ODE
ODEs
ODEs
AEs
AEs
AE
AE
ODE
ODE
Eq.
Eq.
Eq.
Eq.
Eq.
Eq.
Eq.
Eq.
Eq.
Eq.
(A.1)
(A.10)
(A.15)
(A.16)
(A.17)
(A.18)
(A.19)
(A.20)
(A.21)
(A.23)
367
Table A.2
Additional stage equations for the reactive distillation model.
1
2
3
4
5
6
7
8
9
10
Additional stage equations
Number of
equations
Type of
equation
Equation number in
Appendix A
Liquid molar output ow as a function of liquid molar accumulation

Vapour molar output ow as a function of vapour molar accumulation
Gasliquid mass transfer terms for various components
Reaction terms for various components and reactions
Bubble point calculation from liquid phase mole fractions
Dew point calculation from vapour phase mole fractions
Interface temperature calculation by equating liquid and vapour phase energy transfer rates across gas
liquid interface
Calculation of component specic liquid and vapour phase enthalpies used as input to calculation of energy
transfer induced by mass transfer
Overall energy transfer calculation from bubble point/dew point interface temperature differential, and
energy transfer induced by gas liquid mass transfer
Calculation of the liquid and vapour phase bulk temperatures from liquid and vapour phase energy holdup
1
1
NC
NC NR
1
1
1
AE
AE
AEs
AEs
Implicit AE
Implicit AE
AE
Eqs. (A.2) and (A.3)

Eqs. (A.11) and (A.12)
Eq. (A.7)
Eq. (A.1)
Eq. (A.40)
Eq. (A.37)
Eq. (A.26)
AEs
Eqs. (A.27) and (A.28)
AE
Eqs. (A.22) and (A.24)
AEs
Eqs. (A.29) and (A.30)
References
[1] J.R. Walton, B. Yeomans, Isophorone Production Using a Potassium Hydroxide
Catalyst. US Patent 3,981,918, 1976.
[2] G.S. Salvapati, K.V. Ramanamurty, M. Janardanarao, Selective catalytic selfcondensation of acetone, J. Mol. Catal. 54 (1989) 930.
[3] A. Chaubal, Simulation of Reactive Distillation (Case Study: Isophorone Plant),
M.Tech Project Report, Indian Institute of Technology, Mumbai, 1998.
[4] N.O. Lemcoff, R.E. Cunningham, Kinetics of diacetone alcohol conversion to
mesityl oxide catalyzed by ion exchange resin, J. Catal. 23 (1971) 8192.
[5] R. Bakker, G. Hangx, G. Kwant, H. Maessen, P. Markusse, Reaction and Catalyst
Deactivation Kinetics of the Aldol Condensation of Acetone, DSM Technical
Report, Intelligent Column Internals for Reactive Separations (INTINT) Project
No. GRD1 CT1999 10596, 2001.
[6] G.G. Podrebarac, F.T.T. Ng, G.L. Rempel, A kinetic study of the aldol
condensation of acetone using an anion exchange resin catalyst, Chem. Eng.
Sci. 5217 (1997) 29913002.
[7] C. Huang, L. Yang, F.T.T. Ng, G.L. Rempel, Application of catalytic distillation for
the aldol condensation of acetone: a rate-based model in simulating the
catalytic distillation performance under steady-state operations, Chem. Eng.
Sci. 5319 (1998) 34893499.
[8] P. Kustrowski, D. Sulkowska, L. Chmielarz, A. RafalskaLasocha, B. Dudek, R.
Dziembaj, Inuence of thermal treatment conditions on the activity of
hydrotalcitederived MgAl oxides in the aldol condensation of acetone,
Micropor. Mesopor. Mat. 78 (2005) 1122.
[9] K. Schmitt, J. Disteldorf, G. Baron, Production of Isophorone, US Patent
3,337,633, 1967.
[10] C. Cane, B. Yeomans, Process for the Production of Isophorone, US Patent
4,059,632, 1977.
[11] M.C. Grebinoski, D. Glassman, C.L. Elias, A.A. Schutz, Isophorone Process, US
Patent 5,352,839, 1994.
[12] A.A. Schutz, Basic Mixed Oxide, US Patent 4,970,191, 1990.
[13] A.A. Schutz, L.A. Cullo, Process for making Efcient Anionic Clay Catalyst,
Catalysts made thereby, and method of making Isophorone, US Patent
5,153,156, 1992.
[14] J.B. Cook, Industrial Application of Reduce Reuse Recycle Strategies in the
Production of Isophorone, MS Thesis, University of South Alabama, Mobile,
2006.
[15] R. Taylor, R. Krishna, Modeling of homogeneous and heterogeneous reactive
distillation processes, in: K. Sundmacher, A. Kienle (Eds.), Reactive
Distillation: Status and Future Directions, Wiley-VCH Verlag GmbH & Co.
KGaA, 2002, pp. 217240.
[16] A. Katariya, R. Kamath, K. Moudgalya, S. Mahajani, Non-equilibrium stage
modeling and non-linear dynamic effects in the synthesis of TAME by reactive
distillation, Comput. Chem. Eng. 32 (2008) 22432255.
[17] Y. Xu, Y. Zheng, F.T.T. Ng, G.L. Rempel, A three-phase nonequilibrium dynamic
model for catalytic distillation, Chem. Eng. Sci. 60 (2005) 56375647.
[18] K. Sundmacher, U. Hoffmann, Development of a new catalytic distillation
process for fuel ethers via a detailed nonequilibrium model, Chem. Eng. Sci.
5110 (1996) 23592368.
[19] K. Sundmacher, U. Hoffmann, Oscillatory vapourliquid transport phenomena
in a packed reactive distillation column for a fuel ether production, Chem. Eng.
J. 57 (1995) 219228.
[20] K. Alejski, F. Duprat, Dynamic simulation of the multicomponent reactive
distillation, Chem. Eng. Sci. 5118 (1996) 42374252.
[21] B. Hupen, E.Y. Kenig, Rigorous modelling of NOx absorption in tray and packed
columns, Chem. Eng. Sci. 60 (2005) 64626471.
[22] K. Alejski, Computation of the reacting distillation column using a liquid
mixing model on the plates, Comput. Chem. Eng. 155 (1991) 313
323.
[23] K. Alejski, J. Szymanowski, M. Bogacki, The application of a minimization
method for solving reactingdistillation problems, Comput. Chem. Eng. 128
(1988) 833839.
[24] Z. Qi, K. Sundmacher, E. Stein, A. Kienle, A. Kolah, Reactive separation of
isobutene from C4 crack fractions by catalytic distillation processes, Sep. Purif.
Technol. 26 (2002) 147163.
[25] S. Gruner, K.D. Mohl, A. Kienle, E.D. Gilles, G. Fernholz, M. Friedrich, Nonlinear
control of a reactive distillation column, Control Eng. Pract. 11 (2003) 915
925.
[26] R. Krishnamurthy, R. Taylor, A nonequilibrium stage model of multicomponent
separation processes. Part I: model description and method of solution, AIChE
J. 31 (1985) 449456.
[27] R. Taylor, R. Krishna, Modeling reactive distillation, Chem. Eng. Sci. 55 (2000)
51835229.
[28] R. Baur, Modeling Reactive Distillation Dynamics, PhD Thesis, Clarkson
University, Potsdam, 2000.
[29] R. Baur, A.P. Higler, R. Taylor, R. Krishna, Comparison of equilibrium stage and
nonequilibrium stage models for reactive distillation, Chem. Eng. J. 76 (2000)
3347.
[30] S. Sandler, Chemical, Biochemical and Engineering Thermodynamics, fourth
ed., Wiley, New Delhi, 2010.
[31] N. Dalaouti, P. Seferlis, A unied modeling framework for the optimal design
and dynamic simulation of staged reactive separation processes, Comput.
Chem. Eng. 30 (2006) 12641277.
[32] C. Noeres, A. Hoffmann, A. Gorak, Reactive distillation: non-ideal ow
behaviour of the liquid phase in structured catalytic packings, Chem. Eng.
Sci. 57 (2002) 15451549.
[33] E.F. Wijn, Weir ow and liquid height on sieve and valve trays, Chem. Eng. J. 73
(1999) 191204.
[34] Y.H. Jhon, T. Lee, Dynamic simulation for reactive distillation with ETBE
synthesis, Sep. Purif. Technol. 31 (2003) 301317.
[35] M. Banaji, P. Donnell, S. Baigent, P Matrix Properties, Injectivity and Stability in
Chemical Reaction Systems, University College London.
[36] S. Thotla, V. Agarwal, S.M. Mahajani, Simultaneous production of diacetone
alcohol and mesityl oxide from acetone using reactive distillation, Chem. Eng.
Sci. 62 (2007) 55675574.
[37] Y.K. Kim, J.D. Hateld, Kinetics and equilibrium data of the dehydration
hydration reaction between diacetone alcohol and mesityl oxide in phosphoric
acid, J. Chem. Eng. Data 30 (1985) 149153.
[38] H. Watanabe, J. Seto, Catalytic effect of barium ferrite magnetic particles on
aldol condensation of acetone, Appl. Catal. A-Gen. 134 (1996) 217224.
[39] Z. Svandova, J. Markos and L. Jelemensky, Impact of mass transfer on modelling
and simulation of reactive distillation columns, in: Mohamed ElAmin (Eds.),
Mass Transfer in Multiphase Systems and its Applications, InTech, 2011, pp.
649676.
[40] F.J. Zuiderweg, Sieve trays: a view on the state of the art, Chem. Eng. Sci. 3710
(1982) 14411464.
[41] S. Umezawa, A. Nagashima, Measurement of the diffusion coefcients of
acetone, benzene, and alkane in supercritical CO2 by the taylor dispersion
method, J. Supercrit. Fluid. 5 (1992) 242250.

Isophorone

Enviado por

Dados do documento

Descrição original:

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Isophorone

Enviado por

Direitos autorais:

Formatos disponíveis

Chemical Engineering Journal 207208 (2012) 349367

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Journal

Isophorone reactor: Modelling and performance enhancement

Corresponding author. Tel.: +91 9096925842.

evolving guidelines for enhancing performance of isophorone

total number of vapour phase moles on stage j (kmol)

total number of liquid phase moles of ith component on

Initial molar holdup and stage accumulation ratios

parameter B corresponding to component i ()

parameter H corresponding to component i ()

molecular weight of component i (kg kmol1)

critical temperature of component i (K)

reduced temperature of component i on stage j ()

parameter W corresponding to component i ()

supercial liquid velocity in the column on stage j

Mass Flowj liquid phase mass ow on stage j (kg h1)

Mass Flowj vapour phase mass ow on stage j (kg h1)

liquid phase density on stage j (kg m3)

bubbling area of the tray on stage j (m2)

vapour phase density on stage j (kg m3)

heightw(j) exit weir height of stage j (m)

heightj clear liquid height on stage j (m)

Molar volume and volumetric holdups

vapour molar volume of component

bubble point temperature of liquid phase on stage j (C)

dew point temperature of vapour phase on stage j (C)

total liquid volume of stage j (m )3

total vapour volume of stage j (m )

Maximum dened volumetric holdup on a stage

maximum vapour volumetric holdup on stage j (m3)

liquid phase volumetric accumulation on stage j

liquid phase heat transfer coefcient on stage j

reference temperature (C)

vapour phase bulk temperature on stage j (C)

vapour phase energy content of stage j (kJ)

total pressure on stage j (bar)

heat transfer rate across liquid phase/gasliquid

liquid phase heat capacity of component i (kJ

vapour phase enthalpy fed to stage j (kJ kmol1)

liquid phase energy withdrawn from stage j (kJ h1)

vapour phase energy introduced to stage j (kJ h1)

liquid phase heat losses from stage j (kJ h

vapour phase heat losses from stage j (kJ h )

introduced to an intermediate point in the reaction column at such

Stage: Gas Phase

resources that can possibly be affected include: raw materials

impact of actions of the operators at various capacity utilisation

trays are modelled by choosing an appropriate number of cells in

In this work, the overall liquid molar ow originating from a stage

When the calculated value of the volumetric ow originating from a

Eq. (W3) indicates a linear relationship between heightow and the

Dimensionless Time : Ratio of Simulation Time

2.3.2. Reaction kinetics

Mesityl oxide can condense with acetone to form a linear dione

Since the cyclisation of phorone is a relatively fast reaction, in this

Various reversible reactions (R1R13) involved in the condensation

simulations for isophorone production. Reaction terms for various

Reaction Order Matrix:

Base Stoichiometric Coefcient Matrix:

Reference Components for Base Stoichiometric Coefcient Matrix

A number of authors in the literature have quantied the kinetic