Optimization of Supercritical Fluid Extraction of Isoflavone from

Soybean Meal

Kanjana Lummaetee,1 Hong-Ming Ku,1* Wongphaka Wongrat2 and Ali Elkamel3

1. Chemical Engineering Practice School, Department of Chemical Engineering, King Mongkut's University of Technology Thonburi, Bangkok
10140, Thailand
2. Department of Food Engineering, Kasetsart University, Nakornprathom 73140, Thailand
3. Department of Chemical Engineering, University of Waterloo, Waterloo, ON, N2L 3G1, Canada

This study aims at developing a mathematical model to predict the yield of isoflavone from soybean meal in a supercritical extraction process using
carbon dioxide and aqueous methanol as a co-solvent and to optimize the process using a genetic algorithm. In the model, a partial differential
equation based conservation of mass was solved to predict the yield of isoflavone extraction. The model parameters such as densities of carbon
dioxide and co-solvent methanol, the mixture viscosity, the binary diffusion coefficient of isoflavone in the supercritical solvents, the film mass
transfer coefficient, effective diffusivity, and axial dispersion coefficient were estimated using available correlations, and the solubility was estimated
using the Mohsen-Nia-Moddaress-Mansoori equation of state. The model was successfully validated with experimental data. In the optimization,
the operating conditions of the isoflavone extraction process were identified as decision variables and a profit function was maximized. The
optimum was found under the condition in which the carbon dioxide flow rate was 5.88 kg/h and the particle diameter was 0.68 mm, when the
temperature was 323.15 K, the pressure was 59.45 MPa, and the extraction time was 283 min. The maximum profit found under these optimum
conditions was 46.18 $ per batch.

Keywords: isoflavone, supercritical extraction, mathematical modelling, genetic algorithm, optimization

INTRODUCTION MODELLING
A fixed bed extractor was used in the isoflavone extraction

I
ndustrial extraction of isoflavone from soybeans has increased
in demand because of the many health benefits associated with
isoflavone consumption. The health benefits of isoflavone are
derived from the fact that isoflavone behaves like estrogen.[1]
Moreover, isoflavone can prevent cancer and osteoporosis by
increasing bone mineral density and decreasing the bone
resorption marker.[2,3] Supercritical carbon dioxide (SCCO2) is
the most popular method used to extract isoflavone and other
bioactive natural products, such as pentacyclic triterpene a,
b-amyrin from rosemary leaves, seed oil from red prickly pears,
and Illicium verum volatile oil, in the industry.[4,5,6] The reason is
that this method can be considered a green separation process
since SCCO2 is a “green solvent” that is easy to remove, non-toxic,
and non-flammable. Therefore, the extracted products are clean
and the process operation is safe.[7] Over time, the isoflavone
extraction process has been improved to increase targeted
isoflavone extraction by adding aqueous methanol, which is
polar, as a co-solvent.[1] However, supercritical fluid extraction
incurs a high cost due to its high operating condition; therefore, it
is desirable to determine the most economical condition for this
operation. The genetic algorithm has been used as a search or
optimization technique to determine the optimum operating
condition that gives the maximum profit.[8] This method is widely
used because it is capable of finding the global optimal solution
instead of being stuck in a local minimum, even though it does not
guarantee global optimality. In this study, a mathematical model is
proposed to solve the optimization of a supercritical isoflavone
extraction process. This optimization problem involves varying
operating conditions, namely the temperature and the pressure,
while maintaining the solvent flow rate and the particle diameter
at a constant value.
process. Soybean meal is present as a pack of solids, while
SCCO2 and aqueous methanol present as solvents pass through
the packed bed axially. The assumptions of this model[5] are:
(1) Axial dispersion is significant;
(2) Radial concentration gradients are neglected in the fluid
phase;
(3) The system is isothermal and isobaric;
(4) The physical properties of the supercritical fluid are constant
along the bed;
(5) There is local equilibrium at the interface of the fluid and the
solid phase;
(6) The bed void fraction and the particle porosity do not change
during extraction;
(7) For small particles, it was assumed that the particle solute
concentration is independent of their radii and is a function
of time only (lumped system);
(8) The solute concentration in particles is independent of the
coordinates F and u.
Applying the conservation of mass, the following equation can
be written for the solute concentration in the bulk phase:
* Author to whom correspondence may be addressed.
E-mail address: hongming.ku@gmail.com
Can. J. Chem. Eng. 9999:1–9, 2017
© 2017 Canadian Society for Chemical Engineering
DOI 10.1002/cjce.22786
Published online in Wiley Online Library
(wileyonlinelibrary.com).

VOLUME 9999, 2017 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 1

 where Psub
i is the sublimation pressure of a pure solid at the system
@Ci 1 @ 2 Ci @Ci ð1
eÞ kf L temperature (MPa), P is the system pressure or the total pressure

þ
3 ½Ci
Ci;s  ¼ 0 ð1Þ (MPa), R is the universal gas constant (8.314  10
6
@t Pe @Z2 @Z e v Rp
MPa  m3/(gmol  K)), T is the system temperature (K), vsi is the
pure solid molar volume (m3/gmol), and fsi is the fugacity
The boundary and initial conditions of the bulk phase are: coefficient of component i in solid solute.
scf
The fugacity of a supercritical fluid (f i ) can be estimated from
At Z ¼ 0
; Ci ¼ 0 ð2aÞ the following equation:
@Ci
At Z ¼ 0þ ; ¼ Pe ðCi
0Þ ð2bÞ scf
f i ¼ y i fi P
scf
ð12Þ
@Z

@Ci where yi is mole fraction of component i in the supercritical

At t ¼ 0 ; ¼0 ð3Þ scf
phase, and fi is the fugacity coefficient of component i in the
@Z
supercritical fluid phase.  
At t ¼ 0 ; Ci ¼ 0 ð4Þ Substitute the fugacity f si ; f scf
i from Equations (11) and (12)
into Equation (10):
Similarly, applying the conservation of mass, the following Z !
P
equation can be written for the solute concentration in the vsi
i fi exp
Psub s
dP ¼ yi fscf
i P ð13Þ
solid phase: Psub
i
RT

dqi kf L
 Rearrangement:
¼
3 Ci;s
Ci ð5Þ
dt v Rp !
Z
i fi
P
Psub s
vsi
yi ¼ exp dP ð14Þ
The initial condition of the solid phase is: Pfscf
i Psub
i
RT

At t ¼ 0 ; qi ¼ qi;0 ð6Þ
In general, it can be assumed that the solid is pure because the
supercritical fluid cannot dissolve in the solid phase; therefore, vsi
where q and Ci,s are related by assuming equilibrium condition as:
does not depend on the system pressure. Moreover, fsi can be
assumed to be unity (fsi  1) because Psub is very low; thus
qi ¼ KCi;s ð7Þ i
Equation (14) can be written as shown below:
where K stands for the distribution coefficient.  !
Psub vsi P
Psub
By applying the conservation of mass law for the whole yi ¼ scf exp
i i
ð15Þ
extractor system, the extraction yield (F) can be described as Pf RT
i
follows:
In order to estimate the solid solubility (yi ) in Equation (15), the
dF LAe solid molar volume (vsi ), the sublimation pressure (Psub i ), and
¼ Cz¼1 ð8Þ scf
dt n0 the fugacity coefficient in the supercritical fluids phase (fi Þ are
needed for the estimation. The calculation flow diagram of solid
where the initial condition is: solubility is shown in Figure 1.

At t ¼ 0 ; F¼0 ð9Þ Conventional van der Waals mixing rule

The conventional van der Waals mixing rule is used to estimate the
mixing parameters for MMM-EOS; am and bm are defined as
AUXILIARY EQUATIONS follows:

Solid Solubility Xnc Xnc

am ¼ yya
j¼1 i j ij
ð16Þ
The solubility of solid in supercritical fluids can be estimated using i¼1

phase equilibrium between the solid phase and the supercritical !

fluids phase in terms of fugacity.[9] 1 Xnc Xnc Xnc
bm ¼ 3 yi yj bij þ yi bii ð17Þ
4 i¼1 j¼1 i¼1
scf
f si ¼ f i ð10Þ
where nc is the number of components in the mixture and aij and
scf
where f si and f i are the fugacity of component i in solid and bij are the cross parameters which can be obtained from a set of
supercritical fluid, respectively. combining rules as follows:
The fugacity of solid (f si ) can be estimated from the following
 pffiffiffiffiffiffiffiffiffiffi
equation: aij ¼ 1
kij aii ajj ð18Þ
Z !
P
vsi ðbi þ bj Þ
f si ¼ Psub
i fi exp
s
dP ð11Þ bij ¼ ð1
lij Þ ð19Þ
Psub
i
RT 2

2 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 9999, 2017

Figure 1. Solubility calculation flowchart.

Mohsennia-Modarres-Mansoori equation of state Model Parameters

The MMM-EOS is based on a statistical mechanics model of Under the assumption that co-solvents are incompressible fluid,
repulsive and attractive forces.[10] This EOS is suitable when the LKP-EOS and PR-EOS are used to estimate the density of carbon
aromaticity index or AI is less than or equal to 0.3.[11] dioxide (nonpolar) and aqueous methanol (polar) respectively.
The molar fractions of carbon dioxide and aqueous methanol
RTðv þ 1:3191bm Þ am which are co-solvents (yCO2 and yCS ) are required in order to
P¼
0:5 X ð20Þ
v
bm T vðv þ yi bii Þ estimate the mixture density.
i

rm ¼ yCO2 rCO2 þ yCS rCS ð24Þ

R2 T 2:5
ci
ai ¼ 0:48748 ð21Þ
Pci
Because this system operates at the supercritical state, the
0:064662RT ci American Petroleum Institute-Technical Data Book (API-TDB)[12]
bi ¼ ð22Þ recommends the following equations for estimating the viscosity
Pci
(m) of gas at high pressure.
2 2 X XX
3ð2 j yj bij
y y b Þ þ bii
fM 4 4 i j i j ij For non-polar gas: ðm
ma Þj ¼ 1:08x10
4 ½expð1:439rr Þ
i ¼ exp 2:3191
4ðv
bm Þ  
20 X 1
exp
1:11rr 1:858  ð25Þ

 2 j yj aij
bm am 4@Xb A

ln 1

ln Z þ X i


v RT 1:5 i yi bi yb
i i i
am For polar gas: ðm
ma Þj ¼ 1:656x10
4 rr 1:111 for rr  0:1
X ! ##
yb bi ð26Þ
i i i X
ln 1 þ
ð23Þ
v vþ yi bi
i

VOLUME 9999, 2017 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 3

 coefficient can be estimated from the correlation shown in
ðm
ma Þj ¼ 6:07x10
4 ð9:045rr þ 0:63Þ1:739 for 0:1  rr  0:9 Equation (34).
ð27Þ
Sh DAB

 kf ¼ ð34Þ

4 dp
log10 4
log10 ðm
ma Þx10 j ¼ 0:6439
0:1005rr

for 0:9  rr  2:2 ð28Þ where Sh is the Sherwood number which can be calculated from
the following equations:
where j is a viscosity parameter which is a function of M, Tc, Wakao and Funazkri correlation:[13] Sh ¼ 0.38Re0.83Sc0.33
and Pc:
When 0:5 < Sc < 10 000; 3 < Re < 3000 ð35Þ
j ¼ T c1=6 M
1=2 ð0:987Pc Þ
2=3 ð29Þ
Tan et al: Correlation:½14 Sh ¼ 0:38Re0:83 Sc0:33 When 2 < Sc < 40
where m and ma are the viscosity at the desired condition and the
and 2 < Re < 40 ð36Þ
viscosity at 101 kPa (1 atm), respectively. Also, r, rc , and rr
are the density at the desired condition, the critical density, and the
critical reduced density (r/rc, respectively. ma can be estimated for Mongkholkhajornsilp et al. correlation:[7] Sh¼0.135Re0.5Sc0.33
nonpolar and polar gases.
When 6 < Sc < 25; 0:1689 < Re < 1:2918 ð37Þ

5
For nonpolar gases : ma j ¼ 34:0x10 T 0:94
r for T r  1:50;
ð30Þ The binary diffusion coefficient of solutes (A) in supercritical
solvents (B) can be estimated from the correlation proposed by He
ma j ¼ 17:78x10
5 ½4:58T r
1:675=8 for T r > 1:50; ð31Þ et al.[15]
rffiffiffiffiffiffiffi
T
10
For polar gases, ma j is a function of Tr and Zc, where Zc is the DAB ¼ ðV kB
23Þ 10 ½0:61614 þ 3:0902
MA
compressibility factor at the critical point (Zc ¼ PcVc/RTc). In order
to estimate ma j for polar gases, hydrogen bonding is effective in pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!#

0:87756 MB V C;B
determining ma j. exp ð38Þ
PC;B
For polar gases in which the hydrogen bonding is present:

ðma jÞZc5=4 ¼ ½7:55T r
0:55x10
5 ð32Þ where k ¼ 1 for rr 1:2 ð39Þ

pffiffiffiffiffiffiffi
k ¼ 1 þ ðrr
1:2Þ= MA for rr  1:2 ð40Þ
For polar gases in which the hydrogen bonding is not present:
where A and B are solute and solvent, respectively. PC,1 is the
ðm jÞZc2=3 ¼ ½1:90T r
0:294=5 x10
4 ð33Þ critical pressure of solvents (bar), VC,B is the critical volume of
solvents (cm3/mol), VB is the volume of solvents (cm3/mol), and
The film mass transfer coefficient describes the mass transfer MA and MB are the molecular weight of solute and solvent,
resistance from the particle surface to the solute. The mass transfer respectively.

Figure 2. Procedure of the distribution coefficient estimation.[16]

4 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 9999, 2017

 Table 1. Model parameters

dp Density Viscosity (104) DAB (109) DL (105) Cs kf (105)

P (MPa) T (K) Q (kg/h) (mm) (kg/m3) (kg/ms) (m2/s) (m2/s) Re Pe (g/g) (m/s) K

30 313.15 5.88 0.68 884.45 2.02 8.223 1.18 5.87 136 0.0689 5.977 250.15
40 313.15 5.88 0.68 922.64 2.23 7.570 1.15 5.31 134 0.0777 5.302 231.12
50 313.15 5.88 0.68 951.54 2.41 7.110 1.13 4.91 132 0.0852 4.839 186.24
60 313.15 5.88 0.68 975.08 2.57 6.756 1.11 4.60 131 0.0917 4.491 177.34
50 323.15 5.88 0.68 926.66 2.26 7.623 1.14 5.23 135 0.0857 5.280 201.36
50 333.15 5.88 0.68 901.51 2.12 8.174 1.15 5.57 137 0.0864 5.762 246.03
50 343.15 5.88 0.68 876.08 2.00 8.766 1.16 5.93 139 0.0873 6.288 270.03
50 313.15 3.92 0.68 951.54 2.41 7.110 7.83 3.97 147 0.0852 3.854 152.39
50 313.15 7.84 0.68 951.54 2.41 7.110 2.06 7.40 112 0.0852 6.798 187.19
50 313.15 9.80 0.68 951.54 2.41 7.110 2.81 9.25 102 0.0852 8.182 207.91
50 313.15 5.88 0.48 951.54 2.41 7.110 8.48 3.92 203 0.0852 5.681 215.27
50 313.15 5.88 1.19 951.54 2.41 7.110 3.00 9.71 57 0.0852 4.868 241.94

MODELLING RESULTS

One of the parameters used for tuning the results from the
mathematical model to fit experimental data is the distribution
coefficient (K). The value of K can be determined by minimizing
the sum of the squares of the difference between the experimental
extraction yield and the predicted extraction yield. The criterion is
described in the following equation:
X
Objective ¼ ðF model
F exp Þ2 ð41Þ

The algorithm for the distribution coefficient calculations is shown

Figure 3. Comparison of model results with experimental data at
in Figure 2. In this case, a value for K was guessed and then solved in
T ¼ 40 8C, Q ¼ 5.88 kg/h, and dp ¼ 0.68 mm.
the mathematical model to be used to calculate the objective function
in Equation (41). The optimization will stop when the objective
function reaches a minimum value; otherwise the procedure
continues.
The mathematical model requires four parameters, namely the
solubility (Cs), the Peclet number (Pe), the film mass transfer
coefficient (kf), and the distribution coefficient (K), all of which
were estimated from transport properties. The model used 100 g of
soybean meal with 1 L of the extractor. The ratio of carbon dioxide
and aqueous methanol (80 %v/v) is 92.2:7.8 wt%. The transport
properties and the model parameters as a function of the operating
condition are shown in Table 1.
The mathematical model was validated by comparing the yield
of isoflavone with the experimental data from Zuo et al.[17] The
temperature (T), the pressure (P), the solvent flow rate, and the Figure 4. Comparison of model results with experimental data at
solid particle size (dp) of the soybean meal were varied at different P ¼ 50 MPa, Q ¼ 5.88 kg/h, and dp ¼ 0.68 mm.
operating conditions. The yield of isoflavone from the experimen-
tal data was collected every 30 min, and the total extraction time of
the experimental data was 210 min. The results of cumulative
yield of isoflavone from the mathematical model and experimental
results within 210 min are shown in Figures 3–6.
The average absolute relative deviation (AARD) was used to
show the different values of the cumulative yield between the
mathematical model and the experimental data. This mathemati-
cal model can be used to predict the yield of isoflavone with a
maximum AARD of 17.79. The AARD of each operating condition
is shown in Table 2.

PROCESS OPTIMIZATION
The Genetic Algorithm (GA) inside MATLAB’s Global Optimiza- Figure 5. Comparison of model results with experimental data at
tion Toolbox was used in the optimization of this study. This P ¼ 50 MPa, T ¼ 40 8C, and dp ¼ 0.68 mm.

VOLUME 9999, 2017 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 5

Figure 6. Comparison of model results with experimental data at
P ¼ 50 MPa, T ¼ 40 8C, and Q ¼ 5.88 kg/h.
Figure 8. Effect of pressure on the profitability at T ¼ 40 8C, Q ¼ 5.88 kg/h,
and dp ¼ 0.68 mm.

Table 2. The error between experimental data and mathematical

model results The objective function of this study is profitability which is more
comprehensive than the cumulative isoflavone production rate.
Pressure Temperature Solvent flow Particle AARD The profit of the extraction process consists of four parts as shown
(MPa) (K) rate (kg/h) diameter (mm) (%) in the following equations:[18]
30 313.15 5.88 0.68 13.74
40 313.15 5.88 0.68 11.94 Profit ¼ C1
C2
C3
C4 ð42Þ
50 313.15 5.88 0.68 11.03
60 313.15 5.88 0.68 9.28 C1 ¼ yield  total isoflavone  Cost1 ð43Þ
50 323.15 5.88 0.68 10.95
50 333.15 5.88 0.68 12.66 C2 ¼ Pwc  time  Cost2 ð44Þ
50 343.15 5.88 0.68 15.73
50 313.15 3.92 0.68 17.79 C3 ¼ Cost3  Qm  time ð45Þ
50 313.15 7.84 0.68 8.89
50 313.15 9.80 0.68 6.87 C4 ¼ Cost4  wt ð46Þ
50 313.15 5.88 0.48 15.51
50 313.15 5.88 1.19 13.90
where Cost1 is the isoflavone price which was found to be 228.3
$/g. Pwc is the real compressor work that provides the supercritical
fluid and circulates it.
" Y
1 #
method starts with a randomly generated initial population whose 1 Y pffiffiffiffiffiffiffiffiffi P2 Y N s
individual fitness (the maximization of profit) is calculated. PWC ¼ z1 z2 N s P1 V 1
1 ð47Þ
440 Y
1 P1
Solutions in the population with smaller fitness are removed from
the population. This elimination process is used to ensure that the
fitness value of the population keeps increasing. Reproduction is Cost2 is the price of electricity (i.e. in Ontario, Canada), which is
used to make a new generation or “offspring” in which two 11.9 cent/kWh. Cost3 is the price of solvents; the price of liquid
individuals selected from the parent population are made to CO2 was found to be 0.3 $/kg. Cost4 is the price of soybean meal at
crossover, and one individual is selected for mutation. The 0.512 $/kg (USDA Market News, 2013).[19]
reproductive method is used to ensure that it can produce diverse The effects of the operating parameters, namely the pressure,
solutions. The process is repeated until a certain termination the temperature, the solvent flow rate, and the size of solid particle
condition is satisfied as shown in Figure 7. diameter, on the profitability are presented in Figures 8–11.

Figure 7. Simplified flowchart of GA optimization.

6 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 9999, 2017

 Table 3. Operating variables and their bounding in maximum
extraction yield

Independent variables Lower–upper bounded variable

Pressure (MPa) 30
60

Temperature (K) 313.15
343.15
CO2 flow rate (kg/h) 5.88
Particle size (mm) 0.68
Extraction time (min) 200
320

Figure 9. Effect of temperature on the profitability at P ¼ 50 MPa,

Q ¼ 5.88 kg/h, and dp ¼ 0.68 mm.

Figure 12. The profitability at the operating temperature of

313.15–343.15 K and the operating pressure of 30–60 MPa under
250 min of extraction time.

Figure 10. Effect of carbon dioxide flow rate on the profitability at

P ¼ 50 MPa, T ¼ 40 8C, and dp ¼ 0.68 mm. Table 4. GA optimum values for isoflavone extraction

Quantity T, P, and time

Optimum temperature (K) 323.13

Optimum pressure (MPa) 59.45
Optimum time for extraction (min) 283.00
Maximum yield (wt%) 90.926
Maximum profit ($/batch) 46.18
Number of generations 51
Number of function evaluations 1040
Optimization terminated Average change in the
fitness value less than
options

Figure 11. Effect of particle diameter on the profitability at P ¼ 50 MPa,
T ¼ 40 8C, and Q ¼ 5.88 kg/h.
The study on the effects of the operating conditions (the
temperature, the pressure, the carbon dioxide flow rate, and the
particle diameter) on the profitability shows that profitability
increases when the pressure and the carbon dioxide flow rate
increase. For the particle diameter, the profitability decreases
when the particle size increases and will reverse at the particle
size of 0.622 mm which provides the maximum profitability. On
the other hand, the profitability decreases when the temperature
increases, but after a certain point in time, the profitability starts
to decrease.
The optimum operating conditions which provide the maxi-
mum profitability of this extraction process were determined using
GA in MATLAB. The bounds for the operating variables are shown
in Table 3.
The profitability within the bounds of the operating temperature
and the operating pressure under 250 min of extraction time are
shown in Figure 12. The optimization results from GA are shown
in Table 4. The optimum operating conditions which give the
maximum profit are 323.13 K, 59.45 MPa, and 283 min. About
90.1 wt% (0.901 g/g) of isoflavone from soybean meal was
obtained at this condition, which yields 46.48 $/batch of profit.
CONCLUSIONS
Isoflavone offers many health benefits such as cancer and
osteoporosis prevention and as a substitute for estrogen. Isofla-
vone is found in soybeans and can be extracted using supercritical
carbon dioxide. In this research, a mathematical model was

VOLUME 9999, 2017 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 7

proposed to predict the yield of isoflavone from soybean meal in a rs solid density (kg/m3)
supercritical extraction process, after which it was validated with t time of process (min)
published experimental data. The proposed mathematical model tr time of process, reference value (min)
can predict the isoflavone extracts from a supercritical fluid v total system volume per mol in vapour phase (m3/mol)
extraction process with a maximum AARD value of 17.79. Genetic vf interstitial velocity of SCCO2 through the process vessel
algorithm in MATLAB was then used to determine the optimum (m/s)
operating condition which provides maximum profitability. The vs solid molar volume (m3/mol)
maximum profitability of this process was 46.18 $/batch at the u dimensionless time (tDe/R2)
temperature of 323.15 K, the pressure of 59.5 MPa, and the f fugacity coefficient
optimum extraction time of 283 min. In future work, the v acentric factor
equilibrium constant should be modified based on theoretical
equations to make this model more realistic. Moreover, other
parameters such as the particle diameter and the solvent flow rate Subscripts
should also be included as decision variables in the optimization. a apparent
Finally, the configuration of the extractors in the form of parallel c critical
and series units should also be considered in order to make the exp experimental
simulation model reflect more closely what really takes place in f fluid phase
industry. g gas phase
p particle
s surface of particle, solid
NOMENCLATURE
o at time zero
Csat saturated solute concentration in SCCO2 (mol/m3)
CS co-solvent concentration (wt%)
dp particle size (mm) Superscripts
DAB binary diffusion coefficient of solutes (A) in supercritical v vapour phase
solvents (B) (m2/s) l liquid phase
De effective diffusivity (m2/s) sat saturation condition
DL axial dispersion coefficient (m2/s)
fi fugacity of component i
fis fugacity of component i as a pure solid phase REFERENCES
fiscf fugacity of component i in supercritical fluid phase
kf film mass transfer coefficient (m/s) [1] M. Nakada, M. Imai, I. Suzuki, J. Food Eng. 2009, 95, 564.
K distribution coefficient [2] P. Siwarasak, W. Wirivutthikorn, A. Chaiyasat, Warasan-
Pc critical pressure (MPa) WichaiLaeFukOpromSathaban Technology Ratchamong-
Pe Peclet number khon 2001, 5, 112.
Pisub sublimation pressure of the pure solid (MPa) [3] P. Wei, M. Liu, Y. Chen, D. C. Chen, Asian Pac. J. Trop. Med.
Pr reduced pressure (P/Pc) 2012, 5, 243.
q solid phase concentration (mol/m3) [4] O. Bensebia, B. Bensebia, Kh. Allia, D. Barth, Separ. Sci.
qo initial solid phase concentration (mol/m3) Technol. 2016, 51, 2174.
Q solvent mass flow rate (kg/h)
[5] R. Scopel, C. Silva, A. Lucas, J. Gracez, A. Santo, R. Almeida,
Re Reynolds number
E. Cassel, R. Vargas, Fluid Phase Equilibr. 2016, 417, 203.
Sc Schmidt number
SCF supercritical fluid [6] M. Koubaa, H. Mhemdi, F. Barba, A. Angelotti, F. Bouaziz,
Sh Sherwood number S. Chaabouni, E. Vorobiev, J. Sci. Food Agr. 2016, 97, 613.
t time (min) [7] D. Mongkholkhajornsilp, S. Douglas, P. L. Douglas, A.
T temperature (K) Elkamel, W. Teppaitoon, S. Pongamphai, J. Food Eng.
Tc critical temperature (K) 2005, 71, 331.
Tr reduced temperature (T/Tc) [8] T. Hatami, M. A. A. Meireles, G. Zahedi, J. Supercrit. Fluid.
v interstitial fluid velocity (m/s) 2010, 51, 331.
Vc critical volume (m3/mol) [9] B. E. Poling, J. M. Prausnitz, J. P. O’Connell, The
Vr reduced volume (vPc/(RTc)) Properties of Gases and Liquids, 5th edition, McGraw-
x0 initial mass fraction of solute in solid phase Hill, New York 2001 p. 213.
yield mass of extract per mass of raw material [10] M. Mohsen-Nia, H. Modarres, G. A. Mansoori, Fluid Phase
z dimensionless axial Equilibr. 2003, 206, 27.
Z compressibility factor (PV/RT)
[11] A. Ajchariyapagorn, P. L. Douglas, S. Douglas, S. Pongam-
phai, W. Teppaitoon, American Journal of Food Technology
Greek Letters 2008, 3, 275.
Dz dimensionless length of axial coordinate increment (1/h
1) [12] M. R. Fenske, API Technical Data Book - Petroleum
e void fraction in bed Refining. Additional Data and Correlations for the Revision
ep porosity of the solid of Chapter 11 - Viscosity, Report No. API-4-68, Pennsylvania
mm mixture viscosity (kg/m  s) State University, State College 1968.
rm mixture density (kg/m3) [13] N. Wakao, T. Funazkri, Chem. Eng. Sci. 1978, 33, 1375.

8 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 9999, 2017

[14] C. S. Tan, S. K. Liang, D. C. Liou, Chem. Eng. J. 1988, 38, 17.
[15] C. H. He, AIChE J. 1997, 43, 2944.
[16] M. A. A. Melreles, G. Zahedi, T. Hatami, J. Supercrit. Fluid.
2009, 49, 23.
[17] Y. B. Zuo, A. W. Zeng, X. G. Yuan, K. T. Yu, J. Food Eng. 2008,
89, 384.
[18] G. Zahedi, A. Elkamel, A. Lohi, J. Food Eng. 2010, 97, 127.
[19] U. Raghunathan, USDA Market News, "Outlook for US
Dairy," Agricultural Outlook Forum, 2014, accessed on 24
October 2015, https://www.ams.usda.gov/market-news.

Manuscript received August 7, 2016; revised manuscript received

October 24, 2016; accepted for publication October 30, 2016.

VOLUME 9999, 2017 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 9