Computersrhem. Engng, Vol. 13, No. IO,pp. 1175-1181, 1989 0098-1354189 53.00 + 0.

00
Printed 1x1Great Britain. All rights reserved CopyrIght C 1989 Persamon Press plc

A COMPUTER PROGRAM FOR THE DYNAMIC

SIMULATION OF A SEMI-BATCH SUPERCRITICAL
FLUID EXTRACTION PROCESS

G. CESARI,’M. FERMEGLIA,‘~.I. KIKIC’ and M. POLICASTRO~

‘Istituto di Chimica Applicata e lndustriale and ‘Dipartimento di Elettronica, Elettrotecnica ed
Informatica, University of Trieste, via Valerio 2, I-34127 Trieste, Italy

(Receioed 27 October 1987; final revision received 6 April 1989; received forpublication 28 April 1989)

Abstract-The mathematical model of a semi-batch supercritical fluid extraction process has been derived;
its implementation is divided into two distinct stages. Firstly, the differential equations that represent the
dynamic behavior of the process are derived. Secondly, a thermodynamic model suitable for representation
of phase behavior is selected and the algebraic equations and the parameters which describe the process
are identified.
The proposed algorithm is used to simulate the fractionation of ethanol from a water mixture using
carbon dioxide, and the supercritical extraction of citral from lemon oil with carbon dioxide. The program
is applicable to the simulation of extraction from liquids and, with some modifications, from solids.

INTRODUCTION and pressure of the fluid. The supercritical zone of

usual interest is shown in the shaded area in Fig. 1,
Several studies have been made in recent years on
referred to as the extracting agent.
possible industrial applications of supercritical ex-
A supercritical extraction process consists, essen-
traction processes which offer some advantages over
tially, of two stages that correspond to the fundamen-
the classical methods of separation with extractive
tal steps that characterize the extraction process; the
solvents or by distillation, particularly in the areas of
solubilization of the solute in the supercritical solvent
the food industry (decaffeination of coffee, extraction
takes place in an extractor; the separation of the
of the active components of hops), the fine chemicals
solute from the supercritical solvent, in a separator.
industry, the cosmetics industry (extraction of fra-
From the study of the characteristics of a fluid
grant essences) and the oil industry.
in supercritical conditions and, in particular, from
Such studies are normally carried out experimen-
the analysis of its solubility and density, it can be
tally in pilot plants; the scope of the present work
observed that the two operations; extraction and
includes a proposed mathematical model of the
separation, can be achieved by simply varying the
supercritical extraction process and its computer
operating parameter’s temperature and pressure
implementations to simulate the behavior of the
which, by changing the density of the fiuid. also
process.
modify its dissolving power.
Papers concerning dynamic models of chemical
Figure 2 shows the operation schematically. In
plants have been published in recent years by various
the extraction stage, the temperature and pressure
authors, including Holland and Liapis (1983), Gallun
are selected to maximize solubility, whilst in the
and Holland (1982) and Feng et al. (1984). The
separation zone the level of solubility is cut back-by
present work represents an extension of such studies
varying only the pressure, only the temperature or
to a problem where thermodynamic modelling is
both-thus ensuring the precipitation of the extract
important, since accounting for pressure and temper-
and recycling of the solvent.
ature sensitivity close to the critical is essential.
Generally, the mixture A + B is fed discontin-
Furthermore, the numerical solution of stiff systems
uously, semi-continuously or continuously into the
of algebraic-differential equations requires special
extractor and from this A is extracted by the action
methods as proposed by Michelsen and by Gear
of the solvent (C) and conveyed to the subsequent
(Finlahsen, 1980).
separation section; there precipitation of the solute
The separation process carried out using a super-
takes place and the solvent is reintegrated and re-
critical extraction plant exploits the solvent proper-
cycled to the first stage. Extraction can be effected
ties that a fluid exhibits in what is known as the
from either a liquid or a solid substratum; the
supercritical state. This state is defined, strictly speak-
structure of the process as well as the study of the
ing, as that field where temperatures and pressures
equilibrium properties are different in the two cases.
are equal to or greater than the critical temperature
In a supercritical extraction process, there are two
degrees of freedom: pressure and temperature, for
tAuthor to whom all correspondence should be addressed. each stage. Although the extraction conditions are

1175
 1176

/
1 2
\
Q,
G. CESARI er cd

P, T
p,
,c 98

F, H,
------------J

Fig. 3. The extraction process

Fig. 1. Operative conditions for supercritical fluid extraction
referred to the thermodynamic properties of the extract&
agent.
limited by thermodynamic consldcrations, the condi-
tions of separation can be varied over a wide range
since the solute can be made to precipitate by modi-
fying the pressure and/or the temperature. The choice
(often a combination of the two actions) has direct
effects on the operation of the cycle; in any case, it is
convcnicnt to limit the difference in pressure between
the two stages as this leads not only to a reduction
in pumping costs but also to the possibility of con-
densing the supercritical fluid (commonly COz) with
water rather than with a refrigeration cycic.
As far as the operation and regulation of the plant
is concerned, the controls to be provided are simple.
The pressure in the extractor is maintained around
the operating value by a control system consisting of
a sensing element, a regulator and an electro-valve.
This valve can he of the on-off type. which opens and
closes according to the pressure reached inside the
extractor, or of the continuously regulating type. The
temperature in the two sections can be kept constant
by thermostatic control of the heating fluid.
initially with the mixture to be treated; the supercrit-
ical solvent is then fed continuously to the extractor.
Within the extractor, the solvent becomes enriched
with the solute and in the separator of the solvent is
separated from the solute. To indicate the presence of
the control system, the diagram shows a pressure
regulating valve and the thermal power Q that is
required to be exchanged externally for stabilizing the
extractor temperature.
The problem of purely dynamic simulation of the
equipment consists, as all the input variables are
known, in establising the characteristics of the evolu-
tion with time of the other, internal and output.
variables.
Input variables ClyC i- 2j
F, (i = 1 . NC) = molar flowrate of the individual
components in the feed; in the general case provision
is made for working with NC components: in most
cases, only the supercritIcal solvent is present in the
input:
h, = molar enthalpy of the feed.
Q = thermal power supplied to the unit

THE MATHEMATICAL MODEL

Internal variables an&or ~~ariables ussociated n*irh
output J40,0.~(2 x NC + 3)
In order to describe a supercritical extraction plant
p, (i = 1 NC) = number of moles of the com-
it is not necessary to consider all the various parts
ponent i in the high.density
which constitute the plant; for studying the dynamics
phase,
of the process, it is sufficient to analyze the behavior
q, (i = 1 NC) = number of moles of the com-
of the extractor and the separator and to investigate
ponent i in the low density phase.
the temperature and the pressure control actions.
P = pressure.
A schematic representation of the extraction pro-
T = temperature,
cess is shown in Fig. 3. The extractor is loaded
G = molar flowrate at outlet.
To define the model (2 x NC + 3) equations are
AiC
necessary. These can be given by:
NC material balances, one per component,
NC equilibrium relationships.
1 energy balance,
1 fluid-dynamic relationship,
1 volumetric congruency relationship.
Bl II. These (2 x NC + 3) equations describe the
Fig. 2. The semi-batch supercritical fluid extraction process. behavior of the extractor. The separator can be
 Program for a semi-batch fluid extraction process 1177

analyzed by introducing another set of NC variables substitution was made on the basis of a test consider-
which represent the integral with time of the inputs ing the variation to which the values are subject in
to the separator itself, qint,: time.
The compressibility factors and the instantaneous
d/d?@;) + d/dt(q,) = F, - Gq,/Zq, (i = 1, NC),
separation factors were calculated at each integration
(1) step and compared to the previous value: if the
variation of at least one of them was bigger than 1%
d/dr(qint,) = Gq;/Zq,(i = 1,. . NC), (2) they were updated and the stability of the phases was
d/dt (h,EP, + h, Zy,) = II,CF, - h, G + Q (3) checked in order to maintain the proper phase distri-
bution, particularly in the critical region. The stability
q,/Cq, = kjpi/Epj(i = 1, . . NC), (4) test included in the algorithm was that developed by
V,,, = (RTIP)(ZqEQi + Zq%j), (5) Michelsen (1982).
The model used for describing the thermodynamic
G = IP~VI(P~Z,)‘~2111(P2 - J’~v)/pmll behavior of the mixture must be considered here as
x [(L/D)1 - 2 ln(P,,/P]}t’2. (6) an example of rigorous equation of state suitable for
thermodynamically describing the system. Any model
Equation (6) is derived from an energy balance
can be used; it must be able to calculate fugacity
around the pressure reduction valve assuming turbu-
coefficients and their derivatives for all the com-
lent flow (resistance number 1 constant) and com-
ponents in all the phases present at equilibrium in the
pressibility factor constant in the interval of time
mixture (Bertucco et al., 1986).
considered.
An attempt was made to find operating conditions
The enthalpies that appear in the energy balance
that guaranteed, throughout the simulation, separa-
equation (3) are functions of pressure, temperature
tion into a high-density phase and a low-density one.
and composition as indicated in the following equa-
Finally, control by a valve of the proportional type
tions for the high-density and low-density phases,
has been used in the fluid-dynamic relationship, that
respectively:
is:
h, = h,(p, T, P), (7)
P = B0 + k,(P - PO)> (11)
h, = h, (9, T, P). (8) where: PO is the set-point pressure, POis the coefficient
The thermal power Q exchanged externally of opening calculated in static conditions.
can be expressed as a function of temperature, as The thermodynamic model used for the calculation
follows: of the fugacity coefficients, and consequently the
instantaneous separation factors, is the key point of
Q = k(G - T), (9) the performance of the simulation, since in dynamic
where r, is the set-point temperature. processes the error is cumulative.
The equilibrium relationships (4), the volumetric The literature reports many equations of state
congruency relationship (5) and the fluid-dynamic capable of describing multicomponent mixtures of
relationship (6) contain the equilibrium constants ki different kinds. Among the others we have selected
and the compressibility coefficients Z, and Z, of the and included in the program the following models:
high- and low-density phases. Since the use of the Soave-Redlich-Kwong (SRK) (Soave, 1972),
term “equilibrium constant” may be confusing in Percus-Yevick-Redlich-Kwong (PYRK) (Mathias et
time domain, from now on the term “instantaneous al., 1986), and Skjold Joergensen’s GCEOS (Skjold
separation factors” will be used. In effect: Joergensen, 1984). The models have been chosen in
order to include in the program well-known (indus-
k, = k,(P, 9. T PI.
trially applied) as well as more complicated but more
2, = Z,(P, T, P). accurate equations. In particular, GCEOS has been
selected for its capability of describing polar systems
z, = z, (cl, r, P), (10)
and ill-defined molecules (key components in natural
that is to say, even these coefficients are functions of products).
the variables of the system of equations and, conse-
quently, are time dependent. However, they have NUMERICAL SOLUTION
been considered constant over intervals of time: in
The method of integration
fact, since it is not possible to obtain a simple
analytical expression for them to be included into the The system to be integrated is a system of differen-
system of equations to be integrated (at least not tial equations coupled with a system of algebraic
without artificially simplifying the thermodynamic equations which present the characteristics of being
models), they have been recalculated periodically stiff. It was integrated using the LSODI routine from
using the equation of state and the values obtained the ODEPACK package developed at the Lawrence
have been substituted, as appropriate, in the System Livermore Laboratory (Hindmarsh, 1981). This pro-
of equations during the integration course. The gram provides for the utilization of both Gear’s
117x G. CESARl 6?!Cd.

method, which is based on the so-called BDF formu- obtain even this partial result 600 s of CPU time wcrc
lae (backward differentiation formula) and of Adams’ needed.
implicit method. Gear’s method is one of the most The program has also been implemented on an
efficient for the intcgratlon of stiff systems and has IBM AT (Intel 80286’80287) system with a Microsoft
thus been selected. In order to improve efficiency, the Fortran 4.01 compiler. The total computer time for
Jacobian of the system was calculated analytically. the above example was of 450 s using Gear’s method
The program resolves systems written in the tbl- of integration. In this case graphic output of the
lowing form: dependent variable vs time is also provided. The
figures reported in the paper are directly plotted by
A(t. v)v = G(r, v) (12)
the program.
where A IS a squared matrix, v and G are vectors and
I the independent variable. If lr, is not singular the
RESLILTS
system of equations is purely differential; if it is
singular, the system is algebraic--differential. In that The implementation of the algorithm considered
case it IS necessary to calculate, in addition to the has allowed the study of the dynamics of the process.
initial conditions (i.e. the values of the variables at that is to say, the cvolutlon in time of the character-
the initial instant). the values of the derivatives at the istic properties of the process. i.e. the concentration
initial instant. of the components in the phases. pressure. tempera-
The initml conditions. that is the compositions of ture and flow. ‘The program in the present version
the phases in the extractor at time LeTo, were calcu- treats fluid phases, provided the equation of state is
lated with the equation of state considering the applicable to that kind of mixture. Modifications
composltion of the loaded material and the pressure would be necessary to handle pure solids since the
reduction valve completely closed. The initial deriva- solid species normally exist in pure (single-compo-
tives were calculated using Dini’s Thsorem. nent) phases and therefore the K-factor definition
[equation (4)1 does not apply. Another crucial prob-
lem limiting the applicability in the case of cxtractinn
The simulation program known as CISO (com- from solici materials is the availability of equilibrium
puter simulation of implicit systems). which implc- data, and consequently, of the model parameters.
ments the mathematical model of the plant, has a Two examples have been considered in order- to
modular structure. It consists of three parts: the main demonstrate the capabilities of the program.
program defines the system of equa!ions to bc intc-
grated: the LSODI routine carries out the integration
of the system; the thermodynamic routines calculate
the equilibrium properties and periodically updates The first example has been selected for the com-
their values for use in the system of equations. ‘This plete availability of experimental equilibrium da&,
modular structure provides a flexible instrument model parameters and experimental extraction data.
which can be tailored to various situations. This example should be interpreted as a demonstra-
One point which has been gone into in some detail tion of the program rather than a real feasibility study
is that relating to the computer time necessary to of the separation. The fugaclty coetticients defining
carry out a simulation cxtcnded over a sufficiently the K-values are calculated by means of the <KWX
long interval of time (of simulation). To establish a model, which takes into account the polarity of the
direct comparison, simulation has been performed mixture.
using two different methods of integration, one suited Comparison has been made with cxpcrimental
to stifi systems and one to nonstiff systems: that is to results obtained from an actual plant. The simulation
say. the LSODI integration routine has been utilized conditions have then been chosen to be as near as
selecting first Gear’s method of Integration and then possible to the operating ones. The feed is a mixture
Adams’. The results are significant; the computer of water and ethanol (rich in water) from which the
times requjl-cd by Adams’ method are greater than alcoholic component should be extracted, utilizing
those required by Gear’s, which justifies the u5e of the carbon dioxide as a supercrltical solvent. The charac-
selected method that has, indeed, resulted to be the teristics data of the plant are shown in Table 1.
only one efficient for problems of this kind. The first result obtained was the determination of
The diff’crence in performance between the two the parameters necessary for the regulation of the
methods can be illustrated by the following result: valve in the efflux pipe from the extractor. the values
attempts were made to simulate operation of the of enthalpy of the feed and of thermal power to be
plant for 900 s on a CDC Cyber 730 computer; with exchanged externally m order to keep the tempera-
Gear’s method the integration was achieved in 60.461 ture constant. These values were determined using the
CPU s: with Adams’ method, instead, integration of tuning tnethodulogq proposed by Ziegler and
the system could not be achieved over the required Nichols.
interval of time and. in fact, It was not possible to The proportional control strategy for main-
extend the integration beyond a period of 160 s. To taining the pressure set-point was found to he. in
 Program for a semi-batch fluid extraction process 1179

Table I Characteristic data for Example 1 Mole numbers

Data Units Numericalvalue
A” m2 1.4X IO_’
F, mols- ’ 0.064
h, calmol- - 2000
LID 200
P F” MPa 5.06
PO MPa 17.2
Pm 44.0
R Pam-‘mol--‘Km’ 8.3143
TO K 308
v,,, In’ 300X 10-e
Bo 0.6038
1 0.016

Timz (set)

concordance with the actual experiments, very deli-
cate, given the large difference in the pressures up-
stream and downstream of the valve, which tend to
push the system to the limit of instability, thus
provoking rapid oscillation around the values of
regime (see Fig. 4). Ziegler and Nichols’ tuning was
used for the pressure controller; the controller gain
used was K, = 0.2; no saturation of the pressure
control valve was encountered during the simulation
(Cesari, 1986).
The temperature was found to be notably depen-
dent on the enthalpy of the feed and possessed
considerable inertia. This confirms the experimental
results; in general, it is necessary to wait a long time
to reach the set-point temperature before beginning
the extraction process. It is also necessary to use a
thermostatic bath held at a temperature about 20”
higher than that of the set-point and to pay particular
attention to the thermal characteristics of the feed
flow. The controller gain used was K, = 50.
It has been observed that the greatest variation of
the K-factors with time occurs at the beginning of the
extraction process when variations in the concen-
trations of the components in the two phases inside
the extractor are at their greatest. Whilst the K-fac-
tors vary notably with time, the compressibility co-
efficients remain practically constant. Consequently
the thermodynamic/stability routine is called rather
Pa
1;
169.
Fig. 4. The pressure as function of time in the extraction
stage of the process (Example 1).
Fig. 5. Number of moles of CO, and ethanol in the
high-density phase (HD) and number of moles of H,O and
ethanol in the low-density phase (LD) vs time. Big circles are
experimental data.
often at the beginning of the simulation. The stability
test is performed when the control action is substan-
tial (i.e. compressibility factors or K-factors relative
variation bigger than 1%) in order to maintain the
proper phase distribution. If a phase is eliminated by
the controllers the simulation is carried on with a
pseudo-phase. The pseudo-phase is generated by the
density calculation routine of the thermodynamic
model.
Figure 5 shows the variations in the number of
moles in the high-density (HD) and in the low-density
phase (LD). The high-density phase remains preva-
lently constituted by H,O; however, whilst the num-
ber of moles of CO, in the high-density phase is
practically constant with time, the number of moles
of C, H, OH diminishes drastically as extraction pro-
ceeds. The low-density phase consists prevalently of
CO,; moreover, whilst the number of moles of H,O
present in the low-density phase remains constant,
the number of moles of C? H,OH diminishes notably.
The agreement between the calculated results and
the experimental data obtained in the pilot plan
(Bartolomeo et al., 1986) is satisfactory as it can be
clearly seen in the figures where the experimental
extraction data are shown with a circle. The experi-
mental data were obtained by performing extractions
with the same operative conditions at different total
extraction times. The analyses were made by gas
chromatography. Experimental data for the low-
density phase were not available because an on-line
analyzing device could not be placed at the outlet of
the extractor. The agreement is not perfect because of
the experimental difficulty in control and measure-
ment of the flowrate of CO1 in the input to the
extractor and because mass transfer phenomena are
not considered.
Figure 6 reports the results of the simulation in
terms of mol fractions (option in the program) so as
to provide more suitable information for scaling up
the process.
1180 G. CESARI
Mole fractions
Fig. 6 Mel fractions of CO, and ethanol in the high-density
phase (HD) and mol fractions of I-1,0 and ethanol in the
low-density phase (1-D) YStime. Big circles are experimental
data
Example 2 separation of lemon oil by CO2
The supercritical fluid extraction may have attrac-
tive aplications in the food and drug industry since it
offers a separation process for low-volatility heat-
labile substances where conventional distillation can-
not be applied. Lemon oil contains about 38
components, but according to Mathias et al. (1986)
it IS reasonable to treat it as a mixture of three key
components: limonene. citral and a sesquiterpene.
The valuable component in the oil is citral and the
goal of the process is to increase the citral concentra-
tion from 3 to 30%.
Figure 7 reports the calculated x, J and E (fraction
of citral extracted with respect to its initial amount)
profiles vs Q (amount of high-density phase in the
extractor at any time/initial amount of feed). The
simulation is performed at the following conditions:
solvent flowrate 5.56 x lo-’ kmol s-‘, initial solvent
free basis load IO 3kmol, initial concentration of
citral 3% by weight. set-point temperature 323.15 K,
set-point pressure 20 MPa. All the variables are cal-
culated on a weight basis. The fugacity coefficients
Fig. 7. X, y and E vs Q (defined in the text) on weight basis
for extraction of lemon oil: the reference component is
citral.
et al
are calculated by the PYRK model since this model
is suitable for describing the system (Mathias ef al.,
1986). The parameters of the model have been deter-
mined from equilibrium data (Kalra et al.. 1987). By
considering the experimental data one can already
detect a relatively low selectiGty of citral with respect
to limonene.
It is evident from Fig. 7 that a single-stage extrac-
tion is not suitable due to the necessity of recovering
the product of interest as raffinate. In fact. 8576 ofthe
loaded mixture should be extracted in order to obtain
a 10% citral concentration. Unfortunately, no exper-
imental data are available for the comparison of the
dynamic simulation.
CONCLUSIONS
The system of differential and algebraic equations
representing the model of the process has been inte-
grated numerically using the LSODl routine; the
characteristics which must be possessed by a numer-
ical method for resolving problems of this type have
thus been brought to light.
A study has been made of the characteristics of
the extraction of the alcoholic component from a
water+thanol mixture and of the extraction of citral
from lemon oil by means of compressed carbon
dioxide as the supercritical solvent: it has been possi-
ble to analyze, in detail. the fundamental problems of
the extraction process and to reproduce. on the
computer, some situations which have been found to
be critical during operation of the experimental plant.
The difficulties encountered in the simulation were
generated essentially by the model used for describing
the fugacity coefficients and not by the numerical
method used for the integration of the system of
equations.
For this reason the program has been developed
considering the possibility of adding new thermo-
dynamic models. In fact. better models and better
characterization procedures for ill-defined mixtures
(such as natural products) are needed for a more
comprehensive and realistic anallsrs of extraction
processes. In conclusion, the program presented is
able to describe the dynamic effects and to perform
a feasibility study if the thermodynamic mode! IS
appropriate.
The program developed can simulate the process
and examine different operating conditions Calcu-
lations has been done for the first example, showing
how the process performs at different temperature,
pressure and carbon dioxide flowrate. .4n accurate
analysis of this is beyond the scope of this paper and
should be done in the study of the particular process,
since the sensitivity of the process lo the operating
conditions depends upon the thermodynamic? of the
process itself.
Computer programs and details on the derivation
of equations are available from the authors upon
request (Cesarl, 1986).
 Program for a semi-batch fluid extraction process 1181

Acknowledgements--The authors thanks Professor Friedli Creek letters

of ETH Zurich for making LSODI routines available, Dr A. p = Coefficient of valve opening
Bertucco of the University of Padova for the helpful discus- PO= Coefficient of opening calculated in static con-
sions and MPI, Minister0 della Pubblica Istruzione for ditions
financial support. i = Resistance number

REFERENCES
NOMENCLATURE
Bartolomeo G., A. Bertucco, G. 3. Guarise and S. A.
A, = Cross-sectional area of the outlet from the extrac- Rienzi, Estrazione con fluidi in condizioni supercritiche:
tor. impianto sperimentale. &a. Comb. 15, 185 (1986).
E = Fraction of c&al extracted with respect to its Bertucco A., M. Fermeglia and I. Kikic, Modified Carna-
initial amount (Fig 7). ban-Starling-Van der Waals equations for supercritical
F, = Extractor feed flow in moles of component i fluid extraction. Chem. Engng J. 32, 21 (1986).
G = Molar Sow of the low-density phase at outlet &sari G. ~odelli~~one, simulazione e controllo di
h, = Enthalpy per molar unit of feed flow impianti chimici: appficazione all’ estrazione supercritica
h, = Enthalpy per molar unit of dense phase present Thesis, University of Trieste (1986).
in extractor Feng A., C. D. Holland and S. E. Gallun, Development and
h, = Enthalpy per molar unit of low-density phase comparison of a generalized semi-implicit BungeeKutta
present in the extractor method with Gear’s method for systems of coupled
K, = Instantaneous separation factor of component i differential and algebraic equations. Comput. them. Engng
k, = Controiler gain 8, 51 (1984).
L/L) = Equivalent length of tube Finlahsen G., Nvniinear Analysis in Cbemicut Engineering.
NC = Number of components present in the extractor McGraw-Hill, New York (1980).
Q = Amount of high-density phase in the extractor at Gallun S. E. and C. D. Holland, Gear’s procedure for the
any time/initial amount of feed (Fig 7). simultaneous solution of differential and algebraic equa-
P = Pressure inside the extractor tions with application to unsteady state distillation prob-
P,, = Pressure downstream of the valve lems. Cornput. ehrm. Engng. 6, 244 (1982).
p, = Number of moles of component i in dense phase Hindmarsh A. C., ODEPACK, Lawrence Livermore Na-
present in extractor tional Laboratory (1981).
PO= .%-point pressure Holland C. D. and A. I. Liapis, Computer Methods fir
p” = Molecular weight of supercritical solvent Sotuing Dynamic Sepwution Problems. McGraw-Hill,
Q = Thermal power exchanged between extractor and New York (1983).
its ambient Kalra H., S. Y. K. Chung and C. J. Chen, Phase equilibrium
.q, = Number of moles of component i in the extractor data for supercritical extraction of lemon flavors and
(low density phase) palm oils with carbon dioxide. Fluid Phase Eq~~ljbriu 36,
qint, = Integral of component i at outlet from extractor. 263 (1987).
R = Universal gas constant Mathias P. M., Copeman T. W. and Prausnitz J. M., Phase
T = Tem~rature equilibria for supercritical extraction of lemon flavors and
r, = Set-pomt temperature palm oil with carbon dioxide. Florid Phase Equilibria 29,
V,,, = Volume of extractor 545 (1986).
x = Weight fraction of citral in the high-density phase Michelsen M. The isothermal Aash problem. Part 2: phase
(Fig. 7) split calculation. Fluid Phase ~q~i~ib~iu 8, 21 (1982).
~1= Weight fraction of citral in the low-density phase Skjold Joergensen S., Gas solubility calculations. 11. Appli-
(Fig. 7) cation of a new group-contribution equation of state.
2, = Coeflicient of compressibility of high-density Fluid Phase Eq~~~ibT~a16, 317 (1984).
phase Soave G., Equiiibrium constants from a modified
Zs = Coefficient of compressibility of low-density Redtich-Kwon8 equation of state. C&m. Engng Sci. 37,
phase 463 (1972).