ARTICLE IN PRESS

BIOSYSTEMS ENGINEERING 97 (2007) 209– 217

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/issn/15375110

Research Paper: PH—Postharvest Technology

Oil extraction rates of soya bean using isopropyl

alcohol as solvent

S. Setha, Y.C. Agrawala, P.K. Ghoshb, D.S. Jayasb,, B.P.N. Singha

a
Department of Post-Harvest Process and Food Engineering, G.B. Pant University of Agriculture and Technology, Pantnagar 263145, India
b
Department of Biosystems Engineering, University of Manitoba, Winnipeg, Man., Canada R3T 5V6

art i cle info The extraction rates of soya flakes and soya brokens were determined using azeotrope of
isopropyl alcohol (IPA) at 4.15, 6.35, and 7.75 ml/min flow rates. Prior to the IPA extraction,
Article history: soya flakes and soya brokens were hydrated to the optimum moisture content of 13.4% and
Received 30 March 2006 12.6% wet basis (wb), respectively. Oil recovery from soya flakes and soya brokens was
Accepted 10 March 2007 increased up to 8.4% and 10.2% with increasing flow rates in 16 and 24 h, respectively. The
Available online 7 May 2007 maximum oil recovery in soya flakes was 95.6% of total oil compared to 79.1% in soya
brokens at the highest flow rate. IPA resulted in about 1.5–4.7% higher oil recovery from
soya flakes and about 4.8–13% higher oil recovery from soya brokens than n-hexane
extraction. The IPA extraction rates were compared to n-hexane extraction and
mathematical models were developed. The proposed modified model better predicted the
extraction of oil from the soya flakes whereas the extraction of oil from the soya brokens
was better fitted to the log model.
& 2007 IAgrE. All rights reserved. Published by Elsevier Ltd

1. Introduction
Conventionally, soya beans are subjected to direct solvent
extraction because of the low oil content (about 20%) and
relatively lower oil recoveries in mechanical expression
systems. Solvent extraction, in general, implies the removal
of soluble material from an insoluble solid phase by dissolu-
tion in a liquid solvent. Soya beans, like other oilseeds, are
pre-treated before extraction to release the oil. The pre-
treatment includes such operations as dehulling, cracking,
and flaking. Absolute n-hexane, a petroleum-derived product,
has been extensively used as a solvent for the oil extraction
from soya beans and other oilseeds because of its low
vapourisation temperature (boiling point 63–69 1C), high
stability, low corrosiveness, low greasy residual effect, and
better odour and flavour productivity for the milled products
(Johnson & Lusas, 1983; Becker, 1978). However, n-hexane is
listed as a hazardous air pollutant by the US Clean Air Act and
its use in the oilseed extraction plants can adversely affect
worker’s central nervous system (Becker, 1978; Johnson &
Lusas, 1983; Lusas et al., 1991; Gandhi et al., 2003). Narain and
Singh (1988) reported that the use of n-hexane in small-
capacity plants makes the process expensive because of high
operating losses (typically ranging from 0.75 to 7.5 l/t of seed
processed). Further, n-hexane, being a petroleum product,
faces occasional scarcity and fluctuation in price depending
upon supply and demand of gasoline (Johnson & Lusas, 1983).
Isopropyl alcohol (IPA, isopropanol, 2-propanol, C3H8O) is
an efficient and advantageous extraction solvent for soya
beans and other oilseeds as an attractive alternative to
extraction-grade n-hexane (Sullivan et al., 1982; Baker &
Sullivan, 1983; Lusas et al., 1991). The IPA extraction produces
high-quality oil which requires less refining, produces high-
quality meal which requires less toasting, uses less energy

Corresponding author.
E-mail address: digvir_jayas@umanitoba.ca (D.S. Jayas).
1537-5110/$ - see front matter & 2007 IAgrE. All rights reserved. Published by Elsevier Ltd
doi:10.1016/j.biosystemseng.2007.03.008
 ARTICLE IN PRESS
210 BIOSYSTEMS ENGINEERING
Nomenclature
A, k, k1, n model parameters
E fractional residual oil in meal for extraction time
t, decimal
M moisture content of the sample, % wet basis (wb)
and is safer and less toxic (Baker & Sullivan, 1983; Gandhi
et al., 2003). Although IPA has a higher vapourisation
temperature (boiling point 82 1C) than the n-hexane, only a
small portion of the total IPA in the system requires
vaporisation and energy savings should result. Owing to the
significant amount of trypsin inhibitor inactivation achieved,
the desolventising–toasting (D–T) cycle is considerably short-
er, producing a high-quality meal. The process can be
retrofitted to existing plants (Baker & Sullivan, 1983). The
temperature dependence of oil solubility in IPA forms the
basis of IPA extraction process, permitting oil recovery by
cooling and phase separation rather than distillation. Water
miscibility of IPA confers certain operating advantages: oil
extraction from high moisture content flakes is satisfactory,
water migration in the flake bed is not a problem, and dough
balls do not form in D–T cycle. Both the higher flash point
(18 1C) and water miscibility of IPA contribute to safer
operation. The disadvantage of higher latent heat of vapour-
isation for IPA (approximately 206 cal/g) in D–T has been
offset through process modifications in D–T (Sullivan et al.,
1982; Gandhi et al., 2003). Furthermore, IPA provides for a
much safer work environment in extraction plants since its
threshold limit value of the time weighted average (TLV-TWA)
is 400 ppm as compared to 100 ppm for n-hexane (Baker &
Sullivan, 1983). Moreover, IPA is a by-product from hydration
of propylene and is used primarily in making of acetone. It
can become surplus and easily available since alternate
methods for acetone production are being employed (Baker
& Sullivan, 1983). Baker and Sullivan (1983) further reported
that neutral oil values, peroxide values, and sensory qualities
of the oils from IPA extraction were acceptable and were not
significantly affected by the IPA concentration. Gandhi et al.
(2003) also obtained highest degree of purity soya bean oil for
extractions with IPA and reported that IPA extraction process
is equally effective when compared with n-hexane. Baker and
Sullivan (1983) and Gandhi et al. (2003) further reported that
the use of azeotropic mixtures is preferable to the absolute
solvent systems in extracting the soya oil because the organic
solvent requirement is low, nutritive profile of the oil is at par
with that extracted from conventional n-hexane process,
presence of anti-nutritional factors in oil are minimum, and
the meal contains least residual oil. The major nutritional
constituents of crude soya oil obtained from azeotrope of IPA
(IPA+water) were phosphorus (8.0 ppm), peroxide value
(2.1 meq/kg), free fatty acids (0.06%), neutral oil (99.3%), total
volatile matter (0.20%), water (0.17%) (Gandhi et al., 2003).
The research on IPA extraction has been limited to pilot
plant process development for soya bean (Sullivan et al., 1982;
Baker & Sullivan, 1983) and for cottonseeds (Harris &
97 (2007) 209 – 217
Q oil extracted for extraction time t, % of moisture-
free sample
Qo total oil content, % of moisture-free sample
t time of extraction, h
Y oil recovered in 8 h, % of total sample weight on
moisture-free basis
Hayward, 1950; Lusas et al., 1991). The extraction rate of oil
is a major influential factor for designing a large-scale solvent
extraction system. Coats and Wingard (1950) observed that n-
hexane extraction rate was largely dependent on the particle
size. Soya beans, cottonseed, flaxseed, and peanuts were used
as the study materials and it was found that cracked particles
(grits) could be extracted more easily than the flakes of
equivalent thickness but their extraction rate was more
dependent upon thickness. Othmer and Agarwal (1955), while
using batch counter-current extraction of soya flakes with n-
hexane, reported that the oil extraction rate was proportional
to 3.97 power of flake thickness. Becker (1978), while
extracting oil from soya flakes with n-hexane as solvent,
observed direct influence of extraction temperature and flake
thickness on the oil extraction rate. Duggal (1987) modified
the simple diffusion model to obtain a better fit to the
extraction rates from soya flakes.
Further investigation and optimisation is required into the
effect of soya bean moisture content on oil recovery. There-
fore, the objectives of this study were to determine experi-
mentally the extraction rates of soya flakes using IPA as
solvent and compare these with that of n-hexane, and to
determine the extraction rates of soya brokens using IPA in
order to establish if the preparatory operations could be
eliminated. A further objective was to develop a mathema-
tical model for the extraction rates using IPA.
2. Materials and methods
Commercial grade soya bean, PS-564 variety, was obtained
from the University Farm, Pantnagar, India. Soya beans were
cleaned and dehulled and were broken into 6–8 pieces with
the help of an adjustable plate grinder (Christy Hunt Engg Ltd,
UK). The soya brokens passing through 2.44 mm screen but
retained on 2 mm screen were used in the extraction process.
Initial moisture content of soya brokens was 7.8% wet basis
(wb), which was measured following the procedure of ASAE
Standards (2003) for soya beans. In addition, two kinds of soya
flakes were used in this study—laboratory-prepared soya
flakes of 0.87 mm thickness that were obtained by flaking PS-
564 variety soya beans and relatively thin (0.63 mm) commer-
cial soya flakes from PS-262 variety soya beans, which were
obtained from Soybean and Vanaspati Udyog Ltd, Halduchaur,
Nainital, India. For preparing soya flakes, soya brokens were
conditioned to 18–20% moisture content by adding calculated
amount of water and tumbling. The conditioned soya brokens
were then tempered in a hot air oven at 78 1C for 30 min
(Norris, 1964). Tempered beans, while still hot, were passed
 ARTICLE IN PRESS
BIOSYSTEMS ENGINEERING
through the corrugated rolls, 0.5 mm apart, of the flaking
machine (CIAE, Bhopal, India) to achieve 0.87 mm thick flakes.
Average thickness was calculated using 20 flakes, with each
flake being measured for thickness at five different points.
The initial moisture content of laboratory-prepared and
commercial soya flakes was 13.5% and 10.7% wb, respectively,
as measured following the procedures of ASAE Standards
(2003). The initial oil contents were 22.7%, 22.9%, and 25.7%
moisture-free basis in soya brokens, laboratory-prepared soya
flakes, and commercial soya flakes, respectively, which were
measured following the procedures of AOCS (1997a). Both the
soya brokens and the two kinds of soya flakes were stored in
sealed low-density polyethylene bags in a refrigerator until
used.
IPA of 99.7% purity having a boiling point in the range of
81.5–82.5 1C (for 95% evaporation) and specific gravity of
0.784–0.786 at 20 1C was used for extraction. The IPA was
diluted with distilled water to 90.5% weight-by-weight (w/w)
basis concentration in order to produce an azeotropic
mixture. Absolute n-hexane of 99.9% purity, having a boiling
point in the range of 65–70 1C (for 95% evaporation) and
specific gravity of 0.656–0.665 at 20 1C, was used.
2.1. Experimental plan
Experiments were conducted in two phases. In the first phase,
both the laboratory-prepared and commercial soya flakes as
well as the soya brokens with known moisture contents
(8–16% wb), were first extracted for 8 h (about 50% oil
extraction) using 90.5% (w/w) IPA to optimise the moisture
content for maximum oil recovery. The range of moisture
content investigated was chosen because 9.5–11% moisture
was proposed in the pilot plant process (Sullivan et al., 1982;
Baker & Sullivan, 1983). In the second phase, both the soya
flakes and soya brokens were conditioned to the optimum
moisture content and the oil was extracted using both IPA
and conventional n-hexane to compare the oil extraction
rates.
2.2. Experimental procedure
In the first phase, for each replication of an experiment, both
the soya flakes and soya brokens were conditioned to the
desired moisture levels by mixing calculated amounts of
water to 700 g samples in 2.5 l glass bottles and tumbling. The
moisture contents of the conditioned soya brokens and soya
flakes were measured following the procedures of ASAE
Standards (2003) and AOCS (1997b), respectively, at the
beginning of each oil extraction procedure. A moist sample
weighing 10 g was then extracted simultaneously for oil in the
Soxhlet apparatus in which 150 ml flat-bottom flasks were
filled with 100 ml of solvent. Flasks were heated at 110 1C for
IPA to attain the solvent flow rate of 7.75 ml/min. As the
solvent vaporised and condensed in the 60 ml butt tube,
the extraction occurred at the condensation temperature of
the solvent. The extraction temperature was about 82 1C for
IPA. A solvent flow rate of 155 drops/min was measured by
counting the number of drops of solvent condensing in the
butt tube. The samples were then loaded into the thimbles of
size 22 mm  80 mm. The extractions were continued for 8 h.
97 (20 07) 20 9 – 217 211
Thereafter, the flask was removed and placed in a hot air oven
at 100 1C for desolventising until a constant weight of flask
was reached. Desolventising took about 1 h. Flasks were then
cooled in a dessicator and weighed for oil. Oil recovery, for
each condition, was determined by calculating the difference
in recorded weights of flask. Oil recovery was calculated on a
moisture-free basis. All experiments were done in triplicates.
In the second phase, both the soya flakes and soya brokens
were conditioned to the optimum moisture content as
determined from the first phase of experiment. The IPA
extraction rates of soya flakes and soya brokens having
optimum moisture contents were determined by extracting
10 g samples in a Soxhlet apparatus for 4, 8, 12, 16, 20, and
24 h (only up to 16 h for two kinds of soya flakes) at three
solvent flow rates of 4.15, 6.35, and 7.75 ml/min (83, 127, and
155 drops/min). Preliminary experiments showed that oil
recovery beyond 16 h in soya flakes and 24 h in soya brokens
were not significant. The solvent flow rates of 4.15 and
7.75 ml/min were the minimum and maximum achievable in
the Soxhlet apparatus in view of the high boiling point of the
solvent. For each experimental condition, three replications
were run simultaneously. Similar data were collected for n-
hexane extraction for comparing oil extraction rates. Use of
the same moisture content in n-hexane extraction was
justified because the effect of moisture content over 10–14%
on oil extraction from soya brokens and soya flakes was
reported to be statistically insignificant (Duggal, 1987). The
flat-bottom flasks of the Soxhlet apparatus were heated at
temperatures of 98, 101, and 110 1C for IPA extractions and at
70, 75, and 80 1C for n-hexane extractions to attain solvent
flow rates of 4.15, 6.35, and 7.75 ml/min, respectively. Deso-
lventising, in case of n-hexane, was done at 100 1C for 30 min
(AOCS, 1997a).
3. Results and discussion
3.1. Optimum moisture content for maximum oil recovery
The oil recoveries from soya flakes and soya brokens in 8 h IPA
extraction, as affected by the moisture content, were about
15.8–21.4% and 4.9–7.9% (moisture-free total sample weight
basis), respectively (Fig. 1). This was expected because cell
walls were not well ruptured in soya brokens. The following
equations predicted the effect of moisture content on oil
recovery very well within a maximum error of about 0.65% for
both the soya flakes and soya brokens:
for soya flakes
Y ¼ 15:14 þ 5:43M  0:20M2 (1)
for soya brokens
Y ¼ 9:96 þ 2:11M  4:45  103 M3 , (2)
where Y is the oil recovered in 8 h, in % of total sample weight
on moisture-free basis; and M is the moisture content of the
sample, in % wet basis (wb).
The optimum moisture content for maximum oil recovery
was found by partially differentiating Eqs. (1) and (2) and
equating them to zero. This resulted in the optimum moisture
 ARTICLE IN PRESS
212 BIOSYSTEMS ENGINEERING
24
20
Oil recovery, %
16
12
8
4 extraction compared to 77.8%, 85.8%, respectively and 94.2%
8 10 12 14 16
Moisture content, % wb
Fig. 1 – Oil recovery on moisture-free sample weight basis
from soya flakes and soya brokens as affected by moisture
content: E, mean observed soya flakes; —, predicted soya
flakes; m, mean observed soya brokens; - - -, predicted soya
brokens.
contents of 13.4% and 12.6% wb for soya flakes and soya
brokens, respectively.
3.2. Isopropyl alcohol and n-hexane extraction rates of
soya beans
The data on oil extraction of optimally moistened soya flakes
and soya brokens for different times of extraction using IPA
and n-hexane are presented in Tables 1 and 2, respectively, at
different solvent flow rates for both laboratory-prepared and
commercial soya flakes as well as soya brokens. Tables 1 and 2
also show the extracted oil as percent of total oil and the
fractional residual oil in meal.
3.3. Effect of solvent flow rate on isopropyl alcohol
extraction of soya beans
Increasing solvent flow rate increased the oil extraction from
both kinds of soya flakes and the soya brokens with time
(Tables 1 and 2). The average oil yields from laboratory-
prepared soya flakes in 16 h were 88.2%, 91.6%, and 95.6% of
total oil at the solvent rates of 4.15, 6.35, and 7.75 ml/min,
respectively. This indicated that the increase in solvent flow
rate, within the range investigated, increased the mass
transfer coefficient resulting in higher extraction rates from
soya flakes. This behaviour was similar in the case of
commercial soya flakes. However, the oil extraction rate, as
well as oil recovery, was slower and low in case of soya
brokens. The average oil yields from soya brokens in 24 h were
71.8%, 74.8%, and 79.1% of the total oil, at the solvent flow
rates of 4.15, 6.35, and 7.75 ml/min, respectively. The max-
imum oil recovery in soya brokens was 79.1% of total oil
compared to 95.6% in soya flakes at the highest solvent flow
rate. This was expected because of the lack of cell wall
rupture in soya brokens.
97 (2007) 209 – 217
3.4. Comparison of isopropyl alcohol extraction with n-
hexane extraction
The use of IPA as solvent, in general, resulted in higher
extraction rates and average oil recoveries than n-hexane, at
all the solvent flow rates in both laboratory-prepared and
commercial soya flakes as well as soya brokens. In laboratory-
prepared soya flakes, about 88.2%, 91.6%, and 95.6% of total oil
were extracted with IPA in 16 h at solvent flow rates of 4.15,
6.35, and 7.75 ml/min, respectively, compared to 84.2%, 90.2%,
and 92.6% with n-hexane. In commercial soya flakes, these
amounts were 84.2%, 90.1%, and 98.7% of total oil in IPA
in n-hexane extraction. Total oil yields from the soya brokens
were 71.8%, 74.8%, and 79.1% with IPA in 24 h compared to
66%, 66.2%, and 75.5% with n-hexane. The results were in
general agreement with those of Baker and Sullivan (1983)
who reported a lower residual oil content of 0.7% in IPA (90.5%
concentration) extracted, compared to 1.3% in n-hexane
extracted meal. The higher extraction rates with IPA could
be attributed to higher extraction temperature for IPA than for
n-hexane. In this study, extraction took place at about the
vaporisation temperatures of the solvent, i.e. 81.5–82.5 1C for
IPA compared to 65–70 1C for n-hexane.
3.5. Mathematical modelling
Oil extraction data (Tables 1 and 2) of both laboratory-
prepared and commercial soya flakes as well as soya brokens,
were fitted to the log model (Parekh, 1964) and the modified
model proposed by Duggal (1987). These models, respectively,
are:
Q
E¼1 ¼ Aekt , (3)
Qo
Q n
E¼1 ¼ ek1 t , (4)
Qo
where E is the fractional residual oil in meal for extraction
time t , in decimal; Q is the oil extracted for extraction time t,
in % of moisture-free sample; Qo is the total oil content, in %
of moisture-free sample; and A, k, k1, n are the model
parameters.
The model fitting was done by performing regression
analysis on the following linearised forms of Eqs. (3) and (4):
loge E ¼ loge A  kt, (5)
loge ðloge EÞ ¼ loge k1 þ n loge t. (6)
3.6. Isopropyl alcohol extraction of soya beans
The values of model parameters A and k in the log model [Eq.
(3)], and k1 and n in the modified model [Eq. (4)] at different
flow rates are tabulated in Tables 3 and 4 for both laboratory-
prepared and commercial soya flakes, as well as soya brokens,
along with the correlation coefficients. The values of correla-
tion coefficient were all greater than 0.96 and 0.98 in case of
Eqs. (3) and (4), respectively, for both laboratory-prepared and
commercial soya flakes. In case of soya brokens, for Eq. (3),
 ARTICLE IN PRESS
BIOSYSTEMS ENGINEERING 97 (20 07) 20 9 – 217 213

Table 1 – Oil extraction data from soya flakes and soya brokens along with the residual oil in meal using isopropyl alcohol
at different extraction times

Soya type Solvent flow Extraction Oil extracted, % Oil extracted, % of Fractional residual oil
rate, ml/min time, h moisture-free sample total oil 100 Q/Qo in meal, E ¼ 1Q/Qo
(Qo)

Laboratory- 4.15 4 14.970.5 65.072.2 0.470.02

prepared soya 8 18.170.3 78.771.1 0.270.01
flakesa 12 19.770.1 85.870.3 0.170.00
16 20.270.1 88.170.2 0.170.00
6.35 4 15.170.6 66.072.5 0.370.02
8 18.970.6 82.572.6 0.270.03
12 20.270.2 88.270.9 0.170.01
16 21.070.1 91.670.7 0.170.01
7.75 4 16.570.4 72.071.8 0.370.02
8 19.570.1 84.970.4 0.270.00
12 21.270.2 92.470.7 0.170.01
16 21.970.4 95.671.7 0.070.00

Commercial soya 4.15 4 14.170.2 55.070.8 0.570.01

flakesb 8 18.770.1 72.770.4 0.370.00
12 20.370.5 79.171.8 0.270.02
16 21.170.2 82.270.8 0.270.01
6.35 4 15.570.5 60.271.9 0.470.02
8 20.770.1 80.770.3 0.270.00
12 21.970.2 85.370.9 0.170.01
16 23.270.2 90.170.8 0.170.01
7.75 4 16.270.1 62.970.4 0.470.00
8 21.570.5 83.571.8 0.270.02
12 24.470.2 94.970.9 0.170.01
16 25.470.2 98.471.2 0.070.01

Soya brokensc 4.15 4 4.070.2 17.671.0 0.870.01

8 7.470.4 32.571.9 0.770.02
12 11.870.1 52.070.7 0.570.01
16 13.370.2 58.770.9 0.470.01
20 15.070.1 66.270.5 0.370.01
24 16.370.3 71.871.4 0.370.01
6.35 4 4.570.4 19.871.8 0.870.02
8 8.170.2 35.670.7 0.670.01
12 13.870.1 60.770.5 0.470.00
16 14.170.1 62.470.3 0.470.00
20 15.470.4 69.572.7 0.370.03
24 17.070.2 74.871.0 0.370.01
7.75 4 5.070.1 22.270.5 0.870.01
8 8.470.2 37.070.8 0.670.01
12 14.070.2 61.670.8 0.470.01
16 15.370.1 67.570.6 0.370.01
20 15.173.2 75.371.2 0.270.01
24 17.970.2 79.170.7 0.270.01

a
Moisture content (M.C.), 13.5% wb; oil content Qo ¼ 22.9%.
b
M.C.,13.4%, Qo ¼ 25.7%.
c
M.C.,12.4%, Qo ¼ 22.7%.

the values of correlation coefficient were all greater than 0.97.
The high correlation coefficients indicate that both soya
flakes and soya brokens extraction with IPA could be
described fairly well using the log model. The higher values
of correlation coefficients for soya flakes indicated a better
degree of fit with the modified model than with the log model.
Representative predicted extraction curves to those observed
are shown in Figs. 2 and 3, for laboratory-prepared soya flakes
and commercial soya flakes, respectively. It is observed from
the model predictions for soya flakes of all the conditions that
the log model predicted the extracted oil as a percent of total
oil within about 1–4% error whereas the error amount was
about 1–2% in case of modified model. However, in case of
soya brokens, for Eq. (4), the values of correlation coefficients
were 0.99. Fig. 4 shows a representative plot for model
predicted extraction curve of soya brokens using IPA at the
solvent flow rate of 7.75 ml/min. Further, it was observed that
the modified model predicted the extracted oil as a percent of
 ARTICLE IN PRESS
214 BIOSYSTEMS ENGINEERING 97 (2007) 209 – 217

Table 2 – Oil extraction data from soya flakes and soya brokens along with the residual oil in meal using n-hexane at
different extraction times

Soya type Solvent flow Extraction Oil extracted, % Oil extracted, % of Fractional residual oil
rate, ml/min time, h moisture-free sample total oil 100 Q/Qo in meal, E ¼ 1Q/Qo
(Qo)

Laboratory- 4.15 4 14.070.0 60.970.2 0.470.00

prepared soya 8 16.670.4 72.371.7 0.370.02
flakesa 12 18.070.0 78.770.1 0.270.00
16 19.370.4 84.271.6 0.270.02
6.35 4 14.670.5 63.572.1 0.470.02
8 18.170.1 78.770.4 0.270.00
12 19.570.4 85.171.7 0.170.02
16 20.770.7 90.272.9 0.170.03
7.75 4 15.970.5 69.372.3 0.370.02
8 18.370.4 79.971.9 0.270.02
12 20.170.4 87.471.8 0.170.02
16 21.270.3 92.671.4 0.170.01

Commercial soya 4.15 4 13.470.4 51.972.1 0.570.02

flakesb 8 17.270.3 66.971.2 0.370.01
12 19.370.3 75.271.1 0.270.01
16 20.070.0 77.870.1 0.270.00
6.35 4 14.770.4 57.271.5 0.470.01
8 19.670.2 76.070.9 0.270.01
12 21.070.2 81.570.9 0.270.01
16 22.170.1 85.870.4 0.170.00
7.75 4 15.570.0 60.470.4 0.470.00
8 20.670.3 80.171.2 0.270.01
12 23.170.4 89.771.4 0.170.01
16 24.270.4 94.271.4 0.170.01

Soya brokensc 4.15 4 3.370.2 14.470.7 0.970.01

8 6.370.1 27.770.6 0.770.01
12 8.770.5 38.272.3 0.670.02
16 11.470.4 50.171.8 0.570.02
20 13.270.1 58.370.5 0.470.00
24 15.070.2 66.070.8 0.370.01
6.35 4 4.370.2 18.871.0 0.870.01
8 7.470.4 32.771.7 0.770.02
12 9.770.2 42.670.7 0.670.01
16 11.970.2 52.270.7 0.570.01
20 13.870.1 60.870.6 0.470.01
24 15.070.2 66.270.9 0.370.01
7.75 4 4.670.3 20.071.4 0.870.01
8 7.970.1 34.770.7 0.770.01
12 10.970.2 48.170.8 0.570.01
16 12.670.5 55.772.0 0.470.02
20 14.870.2 65.470.8 0.370.01
24 17.170.2 75.570.8 0.270.01

a
Moisture content (M.C.), 13.5% wb; oil content Qo ¼ 22.9%.
b
M.C.,13.4%, Qo ¼ 25.7%.
c
M.C.,12.4%, Qo ¼ 22.7%.

total oil within about 3% error compared to 1.5–2.5% error in
prediction by the log model. It appears that, unlike soya
flakes, the log model resulted in better accuracy in the case of
soya brokens.
3.7. Absolute n-hexane extraction of soya beans
The model parameters for both the log and the modified
models, at different solvent flow rates, are presented in Tables
3 and 4, respectively, for both laboratory-prepared and
commercial soya flakes as well as soya brokens. The high
values of correlation coefficients indicate that the soya flakes
and soya brokens extraction with n-hexane were described
fairly well using the log model. It was further observed that
the error produced by the log model prediction of extracted oil
as a percentage of total oil was within about 1–3% for soya
flakes and 1–2% for soya brokens. The values of correlation
coefficients were in the range of 0.97–0.99 for the modified
 ARTICLE IN PRESS
BIOSYSTEMS ENGINEERING 97 (20 07) 20 9 – 217 215

Table 3 – Values of parameters of the log model [Eq. (3)] for soya flakes and soya brokens for isopropyl alcohol and n-
hexane extraction

Soya types Solvent flow rate, ml/min Model parameters Correlation coefficient

A k, h1

IPA extraction
Laboratory-prepared soya flakes 4.15 0.4687 0.0914 0.98
6.35 0.4886 0.1147 0.97
7.75 0.5330 0.1604 0.97
Commercial soya flakes 4.15 0.5587 0.0764 0.96
6.35 0.5563 0.1098 0.96
7.75 1.4588 0.2959 0.96
Soya brokens 4.15 0.9998 0.0542 0.99
6.35 0.9519 0.0576 0.97
7.75 0.9950 0.0682 0.99

Absolute n-hexane extraction

Laboratory-prepared soya flakes 4.15 0.5183 0.0748 0.99
6.35 0.5460 0.1100 0.96
7.75 0.5055 0.1189 0.98
Commercial soya flakes 4.15 0.5878 0.0653 0.97
6.35 0.5561 0.0894 0.97
7.75 0.7454 0.1631 0.98
Soya brokens 4.15 1.0471 0.0463 0.99
6.35 0.9660 0.0442 0.99
7.75 1.0330 0.0569 0.99

Table 4 – Values of parameters of the modified model [Eq. (4)] for soya flakes and soya brokens for isopropyl alcohol and n-
hexane extraction

Soya types Solvent flow rate, ml/min Model parameters Correlation coefficient

k1, h1 n

IPA extraction
Laboratory-prepared soya flakes 4.15 0.5136 0.5248 0.99
6.35 0.4805 0.6001 0.98
7.75 0.5009 0.6586 0.98
Commercial soya flakes 4.15 0.3762 0.5657 0.98
6.35 0.3863 0.6546 0.98
7.75 0.2101 1.0798 0.98
Soya brokens 4.15 0.0448 1.0696 0.99
6.35 0.0540 1.0457 0.98
7.75 0.0567 1.0682 0.99

Absolute n-hexane extraction

Laboratory-prepared soya flakes 4.15 0.4790 0.4793 0.99
6.35 0.4393 0.5996 0.98
7.75 0.5265 0.5613 0.97
Commercial soya flakes 4.15 0.3534 0.5376 0.99
6.35 0.3819 0.6003 0.99
7.75 0.2973 0.8176 0.99
Soya brokens 4.15 0.0343 1.0802 0.99
6.35 0.0574 0.9247 0.99
7.75 0.0544 0.9992 0.99

model to predict oil yields from soya flakes and the predicted
values were within 1–2%. However, in case of soya brokens,
the values of correlation coefficients were better for the log
model than the modified model while error in prediction with
the modified model was about 3%.
4. Conclusions
Optimum moisture contents for soya flakes and soya brokens
were determined to maximise the oil recovery using isopropyl
 ARTICLE IN PRESS
216 BIOSYSTEMS ENGINEERING 97 (2007) 209 – 217

100 alcohol (IPA), which were 13.4% and 12.6% wb, respectively.
Oil extraction was performed on the optimally moistened
Oil extracted, % total oil

80 samples using azeotrope of IPA (IPA+water) and absolute n-

hexane and the extraction rates and oil recovery data were
60 compared between these two extraction processes. Isopropyl
alcohol resulted in higher extraction rates and oil recoveries
40 from soya flakes as well as soya brokens than n-hexane.
Increasing solvent flow rate increased the extraction rates
20 and oil recoveries. Extraction was slower and oil recoveries
were lower from the soya brokens than from the soya flakes.
0 Nonetheless, extraction from the soya brokens could find
0 4 8 12 16 20
application in small decentralised batch extractions. Mathe-
Extraction time, h
matical models for IPA, as well as n-hexane, extractions were
Fig. 2 – Representative observed and predicted oil extraction developed to determine the oil extraction rates. Extraction of
fit for laboratory-prepared soya flakes with isopropyl soya flakes was better predicted by the proposed modified
alcohol at the solvent flow rate of 4.15 ml/min: B, observed model whereas extraction of soya brokens was better fitted to
replication 1; &, observed replication 2; n, observed log model. Results from this research are expected to assist in
replication 3; —, predicted log model; - - -, predicted the process optimisation, equipment design and process
modified model. control.

Acknowledgements
100
Oil extracted, % total oil

This research was conducted in the Department of Post-

80
Harvest Process and Food Engineering of the G.B. Pant
University of Agriculture and Technology, Pantnagar, India.
60 We thank the United Nations Development Programme,
Indian Council of Agricultural Research, and Natural Sciences
40 and Engineering Research Council of Canada for funding this
study.
20
R E F E R E N C E S
0
0 4 8 12 16 20
Extraction time, h AOCS (1997a). The official methods and recommended practices
Fig. 3 – Representative observed and predicted oil extraction of the American Oil Chemists’ Society. Method Ac 3-44.
American Oil Chemists’ Society, Champaign, Illinois
fit for commercial soya flakes with isopropyl alcohol at the
AOCS (1997b). The official methods and recommended practices
solvent flow rate of 6.35 ml/min: B, observed replication 1; of the American Oil Chemists’ Society. Method Ba 2a-38.
&, observed replication 2; n, observed replication 3; —, American Oil Chemists’ Society, Champaign, Illinois
predicted log model; - - -, predicted modified model. ASAE Standards (2003). Method S352.2. Moisture measurement
—unground grain and seeds. St Joseph, Michigan
Baker E C; Sullivan D A (1983). Development of a pilot-plant
100 process for the extraction of soyflakes with aqueous isopropyl
alcohol. Journal of American Oil Chemists’ Society, 60(7),
Oil extracted, % total oil

80 1271–1277
Becker W (1978). Solvent extraction of soybean. Journal of
American Oil Chemists’ Society, 55(11), 754–761
60 Coats H B; Wingard M R (1950). Solvent extraction. III. The effect
of particle size on extraction rate. Journal of American Oil
40 Chemists’ Society, 27(1), 93–96
Duggal A (1987). Solvent extraction characteristics of dehusked
20 soy cotyledons and soyflakes. M.Tech. Thesis, Department of
Post-Harvest Processing and Food Engineering, G.B. Pant
University of Agriculture and Technology, Pantnagar, India
0 Gandhi A P; Joshi K C; Jha K; Parihar V S; Srivastav D C;
0 5 10 15 20 25
Raghunadh P; Kawalkar J; Jain S K; Tripathi R N (2003). Studies
Extraction time, h on alternative solvents for the extraction of oil-I soybean.
Fig. 4 – Representative observed and predicted oil extraction International Journal of Food Science and Technology, 38,
369–375
fit for soya brokens with isopropyl alcohol at the solvent
Harris W D; Hayward J W (1950). Solvent extraction of cottonseed
flow rate of 7.75 ml/min: B, observed replication 1; &, oil and isopropanol. Bulletin of the Agricultural and Mechan-
observed replication 2; n, observed replication 3; —, ical Engineering, Series 5, Vol. 6(9). Texas A&M University,
predicted log model; - - -, predicted modified model. College Station, Texas
 ARTICLE IN PRESS
BIOSYSTEMS ENGINEERING
Johnson L A; Lusas E W (1983). Comparison of alternative
solvents for oil extraction. Journal of American Oil Chemists’
Society, 60(2), 229–241
Lusas E W; Watkins L R; Koseoglu S S (1991). Isopropyl alcohol to
be tested as solvent. INFORM, 2(11), 970, 972–973, 976
Narain M; Singh B P N (1988). Characteristics of solvent extraction
plant and machinery being made in India. Proceedings of the
Second National Seminar on Soybean Processing and Utiliza-
tion, G.B. Pant University of Agriculture and Technology,
November 18–19, Pantnagar, India
97 (20 07) 20 9 – 217 217
Norris F A (1964). Extraction of fats and oils. In: Bailey’s Industrial
Oil and Fat Products (Swern D, ed), 3rd edition. Inter Science
Publishers, New York
Othmer D F; Agarwal J C (1955). Extraction of soybean: theory and
mechanism. Chemical Engineering Progress, 51(8), 372–378
Parekh H V (1964). Basic Principles of Solvent Extraction and its
Application, 2nd edn. Examiner Press, Fort Bombay, India
Sullivan D A; Campbell B D; Conway M F; Grimsby F N (1982).
Isopropanol extraction of oilseeds soybeans and other oil
seeds. Cotton Gin Oil Mill, 83, 16–18