Vapor-Liquid Equilibrium of Monomer
Lisa Hedborn
Department of Chemical Engineering, Lund University, Sweden
May 2013
Abstract
The main goal with this master thesis work has been to experimentally generate vapor-liquid
equilibrium, VLE, data for chemicals included in production of component A. This data is to be used
in the Chemcad simulation model of the production process of component A in a plant positioned in
the UK. The existing model deviates from plant process data, and for this reason VLE and physical
property data are needed to be included in the simulations. Experiments performed in this work were
vapor pressure determination of component A and VLE measurements of four binaries. The
experimental data were evaluated and thereafter implemented and regressed to NRTL in Chemcad. A
comparison was made of the Chemcad model, including and excluding the experimental data
generated in this work, and plant process data from the production. This comparison was carried out
for a distillation column of the product purification section. The temperature profiles were chosen to
be compared and it was found that including the experimental VLE and vapor pressure data generated
in this work resulted in temperature profiles more similar to the plant process data from the
simulations.
Keywords: Vapor-Liquid Equilibrium, Experimental VLE data, Regression, NRTL
Introduction
Component A is a monomer with
excellent reactivity. It is commonly used for
production of polymers, polyols, and for
modification of resins and polymers [1, 2]. In
the production process of component A a
strong oxidizing agent is used to produce
peracetic acid, PAC, from acetic acid. The
PAC is thereafter used to oxidize component B
and the product of this reaction is component
A. The world-wide production of component A
is located on four sites, of which one of them is
in the UK, and is owned by the Perstorp group
[2].
Simulation as a tool is today commonly
used for calculations of for example mass and
energy balances, or to predict process
performances that could save large amounts of
money. Simulators such as Aspen plus or
Chemcad are very powerful for this purpose.
The development of simulators has made it
possible for a single engineer to perform
complex and comprehensive calculations in a
very short time [3]. Physical properties in
simulators are crucial. Inadequate or lack of
physical properties results in unreliable models
[3, 4, 5].
There is a Chemcad steady-state
simulation model of the process for production
of component A that includes all unit
operations present in the process. The purpose
of the model is to be used for optimization
work and process understanding. The existing
model deviates from the plant process data. For
example, temperature profiles of some
distillation columns in the product purification
section are inconsistent compared to the plant
process data. For this reason the Chemcad
model must be improved in order to use it for
its purpose. The deviations are most probably
due to incorrect vapor-liquid equilibrium,
VLE, data in the model. The VLE calculations
in the existing Chemcad model are mainly
predicative. For this reason, new experimental
VLE data and physical properties are needed to
be included in the Chemcad model. VLE data
for binaries related to component A are very
limited in the literature. Additionally, other
physical properties data such as vapor
pressure, heat of vaporization etc. are also
limited in literature.
Prior to the start of this master thesis work
an experimental setup was built for
determination of VLE. Experimental VLE
determinations are based on measurements of
the temperature and pressure combined with
liquid and vapor phase compositions [6]. In a
distillation column it is the two components
with closest volatility that mainly affect each
other in the separation even in a
multicomponent system; this is why VLE of
binaries, i.e. a mixture of two pure fluids, most
often are of interest [7]. Furthermore, even if
the system of interest contains more than two
components, binary parameters in activity
coefficient models can often be used
describing the behavior of multicomponent
systems [4].
Vapor-liquid equilibrium aims the
equilibrium between a liquid and its vapor
phase. For equilibrium to occur the
temperature, pressure and fugacity of a
component must be the same in all present
phases. There is no net-exchange of molecules
between the phases at equilibrium [6]. The
liquid phase fugacity is more complex to
theoretically describe compared to the vapor
fugacity, and activity coefficients are
commonly used to describe the liquid fugacity.
There are many different models present in the
literature that can be used to calculate activity
coefficients. Most of them include binary
parameters which need to be determined by
regression to experimental data [5, 6].
Vapor-liquid equilibrium
There are different approaches of how to
calculate VLE, and at low or moderate
pressures the gamma-phi approach is
commonly used. In this approach the vapor is
treated as an ideal gas and the liquid fugacity is
described by activity coefficients [5, 8]. The
equilibrium requirement of equal fugacities of
a component in the vapor and the liquid then
can be described according to equation 1:
(∫
)
(1)
If assumptions are made that the liquid is
incompressible and the temperatures are well
below the critical, equation 1 can be
approximated to equation 2:
(2)
Where γ is the activity coefficient, and depends
on the composition, temperature and pressure.
However,
commonly
the
temperature
dependence of γ can be neglected if the
temperature does not vary largely. For an ideal
system, γ is by definition equal to unity, and in
this case equation 2 is reduced to Raoults law
[5].
NRTL, non-random two liquid, is an
activity coefficient model based on local
compositions; i.e. the composition locally
around one molecule is different from in the
bulk due to intermolecular interactions [9]. The
NRTL equation for a multicomponent system
is presented in equation 3:
∑
∑
∑
∑
]
∑
(3)
Where:
( )
(
∑
)
T=temperature in Kelvin
[
Normally the three-parameter equation of
NRTL is used, in that case the parameters Bji,
Bji and αij are included. It is however possible
to include up to nine parameters [10].
Experimental determination of binary
VLE
A method has been developed for
experimental determination of VLE for the
experimental setup that was built prior to this
master thesis work was started. The method
used to generate VLE data was the circulation
method, which is based on continuous
separation of the liquid and vapor phases. The
liquid mixture is added to an equilibrium
vessel were it is brought to boil under
controlled pressure by supplying heat. The
vapor phase is separated, condensed, and
reintroduced into the vessel. Principle of the
circulation method is presented in figure 1 [6].
Figure 1: Principle of the circulation method [6]
The composition of the liquid and vapor
phases varies until the system has reached
steady-state, which represents the state of
equilibrium. In figure 2 the experimental setup
is presented. The numbers in the figure are
explained as follows:
(1) Equilibrium vessel
(2) Condenser
(3) Sampling point of the condensate
(4) Sampling point of the liquid
(5) Vacuum pump
(6) Oil-bath for cooling in the condenser
(7) Stirrer
(8) Sampling cylinders
Figur 2: Experimental setup
The temperatures for heating and cooling were
controlled in LabVIEW. In LabVIEW online
data was constantly available (i.e. temperatures
and pressure) and logged every minute. As can
be seen in figure 2 the system was well
insulated.
Initially for each experiment the binary
mixture was prepared by weighing in the
desired amount of each component, adding up
to a total volume of approximately 1 L. The
total volume of the vessel was 2 L; i.e. 50 % of
the vessel was filled. The liquid was
introduced in the equilibrium vessel by first
applying vacuum in the system and thereafter
connecting a tube to the top valve of the vessel,
thereafter the liquid was withdrawn in the
vessel by placing the tube in the liquid. The
liquid phase was constantly agitated. After the
liquid had started to boil it was left for 1.5 h to
assure representative equilibrium conditions.
Thereafter the condensate was sampled
followed by the liquid phase. Approximately
10 mL was totally removed from each phase
when sampling.
Equipment and method verification
The equipment and method was initially
tested and verified by boiling water and
ethanol-water, and the results were compared
to literature values. Valuable experience in
how to operate the equipment was obtained
from those tests which were used to improve
the method procedure. In most of the trials it
was found that the temperature of the liquid
and the vapor seldom showed the same value
even though boiling occurred. For this reason it
was difficult to assure boiling since the
equipment was made of stainless steel, hence
boiling could not be visually confirmed. It was
found that the liquid phase most probably was
superheated and the temperature measured in
the vapor represented the boiling/ bubble point.
Before and after regression to experimental data
Component A(1) -component B(2), 155 mbar
1
x=y
Experimental data
Chemcad old
Chemcad new
y1, vapor phase molar composition
0.7
0.6
0.5
Old
0.4
0.3
New
0.2
0.1
0
0
0.1
0.2
Component A(1) -component B (2), 155 mbar
150
Experimental xT-data
Liquid composition, Chemcad old
Vapor composition, Chemcad old
Liquid composition, Chemcad new
Vapor composition, Chemcad new
140
0.3
0.4
0.5
0.6
0.7
x1, liquid phase molar composition
0.8
0.9
Figure 3: x-y diagram of the binary A1- B2
1
Temperature, degree celcius
Experimental VLE data was developed for
four binaries, the components A-B, A-C, A-D
and B-C. Additionally vapor pressure data was
generated for component A. The results of the
vapor pressure data are however not presented
here.
Initially the experimental VLE data were
evaluated and outliers were identified.
Thereafter the data was used to regress BIPs to
NRTL in Chemcad. BIPs, i.e. Binary
Interaction
Parameters,
are
adjustable
parameters included in NRTL (see equation 3).
When regressing VLE data in Chemcad the
result, i.e. the obtained BIPs, are always
thermodynamic consistent since this is a built
in function in the program. In figure 3 and 4 xy and T-x-y diagrams are presented for the
binary A-B. In the figures the phase diagrams
that were used in the existing Chemcad model
are presented as “Chemcad old” and the result
from regression are presented as “Chemcad
0.8
Before and after regression to experimental data
130
VLE results
0.9
new”, additionally the experimental data are
included in the figures.
As can be seen in figures the experimental
data and the regressed curves deviates from
what has been used in the existing Chemcad
model.
120
110
Old
100
New
90
80
70
60
0
0.1
0.2
0.3
0.4
0.5
x and y
0.6
0.7
0.8
0.9
1
Figure 4: T-x-y diagram of the binary A1- B2
The regression result of the additional three
binaries showed good agreement with the
experimental data. The obtained BIPs are
presented in table 1.
Table 1: BIPs obtained from regression of
experimental data to NRTL
Binary
A1-B2
A1-C2
A1-D2
B1-C2
B12
297.8
272.03
2201.3
2335.6
B21
423.8
527.6
2413.4
262.8
α
0.401
0.610
0
1.785
A12
0
0
-8.29
0
A21
0
0
-4.38
0
Comparison to plant process data
A comparison of plant process
data and the Chemcad simulation model was
made, in which the temperature profile of a
distillation column was chosen to be
compared. The chosen distillation column is
part of the product purification section.
Operating conditions of a specific day and time
was recreated in Chemcad and simulated, and
the results were compared to plant process
data. The simulations were performed
including and excluding the new BIPs and
vapor pressure parameters developed in this
work. The result is presented in figure 5. In the
figure, the temperature profile obtained from
simulations when the new BIPs and vapor
pressure parameters was excluded is called
“The old Chemcad model” and the obtained
temperature profile from when they were
included is called “The new Chemcad model”.
It can be seen in figure 5 that when the new
BIPs and vapor pressure parameters were
included a temperature profile more similar to
the plant process data was obtained from the
simulation.
Temperature profile
Temperature [°C]
35
135
1
The old
Chemcad
model
Plant
process
data
3
5
Ideal
stage 7
9
The new
Chemcad
model
11
Figure 5: Temperature profile in a distillation column
Nomenclature
α
P
R
T
x
y
Subscripts
i
L
s
v
vp
binary interaction parameter
activity coefficient
fugacity coefficient
pressure [Pa]
ideal gas constant []
temperature [T]
molar composition of a
component in the liquid phase
molar composition of a
component in the vapor phase
component i
liquid
saturation
vapor
vapor pressure
References
[1] Tanaka et al, Process for producing
purified ε-caprolactone, United States Patent,
patent number: 5,994,565, Date of patent:
Nov.30, 1999
[2] SIDS (Screening Information Data Set)
ε-Caprolactone, CAS N°: 502.44-3, Initial
Assessment Report for SIAM 19, Berlin,
Germany 19-22 October 2004, section 2 and
ANNEX
[3] E. C. Carlson, Don’t gamble with
physical properties for simulations, Aspen
Technology, Inc. Chemical engineering
progress, October 1996
[4] P. Uusi-Kyyny, Vapor liquid
equilibrium measurements for process design,
Helsinki University of technology, Chemical
Engineering Report Series, No. 45, 2004
[5] B.E. Poling, J.M. Prausnitz, J.P.
O’Connel, The properties of gases and liquids,
5th edition, 2007, chapter 4, 7, 8 (p. 4.8-4.32,
p.7.1-7.7, p. 8.1-8.195) Appendix A
[6] G. Zacchi, Fasjämvikter för
kemitekniker, Lunds Tekniska Högskola,
Kemisk apparatteknik, augusti 2000
[7] Personal contact and Chemcad course
with Oleg Pajalic, February-May, 2013,
Perstorp group
[8] P. Haimi, Vapour liquid equilibrium
measurements with three methods: static total
pressure, circulation still and inert gas
stripping, Aalto University, Department of
Biotechnology and Chemical Technology,
Doctoral dissertations, 2012
[9] C. Gutiérrez-Antonio, A. BonillaPetriciolet, J.G. Segovia-Hernández, A.
Briones-Ramírez,
Data
correlation
in
homogeneous azeotropic mixtures using NRTL
model and stochastic optimization methods,
Mexico, ESAT 2009
[10] Chemcad version 6.5, Help and
Reference, BIP Regression, Chemstation